Studying Differences in Lipid Nanoparticles for Delivering Genetic Drugs Using Density Separation

Using microfluidic technology, LNPs of the standard formulation were prepared at a flow rate of 12 mL/min, with the N/P ratio set at three variable levels: 2, 4, and 6. The prepared LNPs were subsequently separated and purified by S-DGC to systematically investigate the effect of different N/P ratios on their physicochemical properties and transfection efficiency. As shown in Table 1, the particle size of LNP exhibits a significant decreasing trend with increasing N/P ratio. Specifically, the particle size at N/P = 2 (approximately 140 nm) is larger than that at N/P = 4 (approximately 120 nm), with the smallest size observed at N/P = 6 (approximately 90 nm). These results indicate that a higher N/P ratio can effectively modulate and reduce the particle size of LNPs.

Based on encapsulation efficiency, mechanistic insights, and separation analysis, significant differences in encapsulation efficiency were observed among LNPs prepared with different N/P ratios: N/P = 6 showed the highest efficiency, up to 99.5%, followed by N/P = 4, while N/P = 2 performed relatively poorly, only 75.4% (Figure 1A). Mechanistically, a higher N/P ratio increases the proportion of cationic lipid to nucleic acid, thereby elevating the net positive charge [21]. However, the overall physicochemical properties of LNPs across different formulations did not exhibit marked differences, which may obscure critical variations between formulations. To address this, S-DGC was employed to separate LNPs and analyze their density distribution profiles. S-DGC separation revealed that for all three N/P ratios (2, 4, and 6), a portion of the LNP components was distributed in the top 0% sucrose density layer (Figure 1C). Combined with agarose gel electrophoresis results (Figure 1B), it was observed that most components of N/P = 6 LNPs were uniformly distributed within the 0–20% sucrose density range. LNPs with N/P = 2 and 4 exhibited similar distribution patterns. Within this density interval, the particle size and zeta potential of the separated fractions remained stable across all three formulations, with N/P = 2 and 4 showing relatively low polydispersity index (PDI) values, with most values below 0.15, indicating more uniform particle size distributions. The three formulations exhibited a significant increase in particle size exceeding 300 nm within the 30% sucrose density layer (commencing around the 20th sample), accompanied by a marked decrease in zeta potential, which became negative (Figure 1D,E,H).

To further clarify the functional performance of LNPs with different N/P ratios, the transfection activity of each formulation was systematically evaluated before and after S-DGC separation (Figure 1F,G). Prior to separation, LNPs with N/P = 2 and 4 exhibited significantly higher reporter gene expression than those with N/P = 6. After S-DGC separation, plasmid DNA-containing fractions from all three N/P groups showed detectable reporter expression under fed (complete medium) conditions. Specifically, formulations N/P = 2 and N/P = 4 demonstrated particularly outstanding transfection efficiency in the sucrose density range of 0% to 20%, achieving rates exceeding 70, comparable to pre-separation transfection efficiency. In contrast, almost no expression was observed under serum-starved (−FBS) conditions. In summary, this study successfully applied S-DGC to characterize and visualize the heterogeneity of LNPs prepared with different N/P ratios. The results demonstrate that LNPs with N/P = 2, 4, and 6 can all form homogeneous and stable systems. Although lower N/P formulations (N/P = 2 and 4) exhibited higher transfection activity, they were associated with lower encapsulation efficiency and potential stability limitations compared to the higher N/P formulation.

The structural and physicochemical properties of LNPs are not determined by a single lipid component but by the integrated effect of the lipid combination. Based on the N/P = 6 formulation, we systematically investigated the specific effects of omitting individual lipid components on the structure and physicochemical properties of LNPs. As demonstrated by the overall physicochemical properties of the unseparated LNPs in Table 2, the absence of lipid components leads to an increase in particle size and a reduction in encapsulation efficiency. The most significant changes were observed in the N/P = 6–Chol–DSPC formulation (Figure 2A), which exhibited an encapsulation efficiency of only 13.7%. Helper lipids, such as DSPC and cholesterol, play crucial roles in LNP systems. Previous studies have confirmed that these lipids contribute to the formation of stable nucleic acid–lipid complexes at neutral pH, thereby ensuring efficient nucleic acid encapsulation [22].

To further validate the impact of lipid omission on LNP homogeneity, each formulation was analyzed by S-DGC. The results showed that most components of the N/P = 6–Chol formulation were evenly distributed within the 0–10% density region. In contrast, the majority of components from the N/P = 6–DSPC and N/P = 6–Chol–DSPC formulations accumulated in the 0% density layer, with the latter exhibiting poorer distribution uniformity due to its lower encapsulation efficiency (Figure 2B,C). Furthermore, after S-DGC separation, the particle size (average ~160 nm) and zeta potential (average ~−5 mV) of the fractions from the N/P = 6–Chol–DSPC formulation showed no significant variation across fractions. However, compared to the N/P = 6–Chol and N/P = 6–DSPC formulations, this formulation displayed larger overall particle size, lower zeta potential, and a higher PDI, further confirming its broader size distribution and poorer dispersion homogeneity (Figure 2D,E,H). In vitro transfection results demonstrated that both the N/P = 6–Chol and N/P = 6–DSPC formulations exhibited weak reporter gene expression under fed (complete medium) conditions. The expression level of N/P = 6–DSPC was higher than that of N/P = 6–Chol, reaching over 40%, with the highest transfection efficiency observed in the fractions from the 10% and 20% density layers. In contrast, no reporter gene expression was detected in any fraction of the N/P = 6–Chol–DSPC formulation, either before or after S-DGC separation (Figure 2F,G).

These results demonstrate that S-DGC can effectively separate LNPs with different lipid omissions based on their density differences. Furthermore, they confirm the critical role of cholesterol and DSPC in maintaining the structural stability of LNPs during storage and intracellular trafficking. The absence of these helper lipids not only significantly impairs the nucleic acid encapsulation capability of LNPs but also directly affects their intracellular transfection and reporter gene expression performance.

Microfluidic mixing is currently the mainstream technique for the preparation of LNPs, with various devices having been developed for their scalable production. This technology enables laminar flow mixing by precisely controlling the flow rate ratio of the two fluid phases, thereby producing LNPs with high reproducibility and uniform size [23]. Accordingly, we employed microfluidic mixing to prepare DNA-loaded LNPs based on the standard formulation (N/P = 4), using five different total flow rates: 8, 10, 12, 15, and 18 mL/min. As shown in Table 3, the overall physicochemical properties of the prepared LNPs did not exhibit a clear trend across the tested flow rates. Therefore, S-DGC was subsequently applied to further investigate the characteristics of LNP subpopulations separated by density.

S-DGC results indicated that LNPs prepared at different flow rates exhibited density distribution profiles similar to that shown in Figure 1C after separation: only a small fraction of LNP components remained at the top of the gradient, while the majority were uniformly distributed within the 0–20% sucrose layers, all of which are less than 200 nm (fractions 1–19). Within this region, both the zeta potential and particle size of the separated fractions remained stable (Figure 3A,B). In contrast, within the 30% sucrose layer, LNPs from all flow rates showed marked physicochemical abnormalities: a sharp drop in zeta potential accompanied by a pronounced increase in particle size and PDI (Figure 3A–C).

Transfection of HeLa cells with LNPs collected before and after S-DGC separation revealed that LNPs prepared at a flow rate of 12 mL/min yielded significantly higher reporter gene expression than those prepared at other flow rates (Figure 3D). In summary, S-DGC analysis demonstrated that LNPs prepared at flow rates of 8, 10, 12, 15, and 18 mL/min showed no clear differences in their overall physicochemical properties and could all form homogeneous systems, which supports the robustness of the microfluidic preparation process. However, this does not rule out the possibility of subtle structural variations among the formulations.

Polyethylene glycol (PEG) is a hydrophilic polymer widely used in the preparation and storage of LNPs. PEG-lipids stabilize the structure by providing a steric barrier that both drives self-assembly and prevents particle aggregation [24]. In this study, LNPs based on the standard N/P = 4 formulation were prepared using microfluidic mixing at a fixed total flow rate of 12 mL/min, with PEG-lipid concentrations set at 0.5%, 1.5%, and 2.5% to evaluate their influence on the physicochemical properties of LNPs. Table 4 presents the overall physicochemical properties results, demonstrating that the particle size of LNPs exhibits a significant decreasing trend with increasing PEG proportion. At a PEG concentration of 0.5%, the particle size reached 202 nm, whereas at 2.5% concentration, it decreased to 118.7 nm. Concurrently, the polydispersity index (PDI) progressively increased, indicating that the PEG-to-lipid ratio significantly influences the structure and physicochemical properties of LNPs.

Figure 4A shows the agarose gel electrophoresis of LNPs before S-DGC separation, indicating satisfactory encapsulation for all three formulations with different PEG-lipid concentrations. The primary role of PEG-lipids is to prevent particle aggregation. At a low PEG concentration (0.5%), insufficient surface coverage during LNP formation leads to particle fusion, resulting in larger particle sizes. After density-based separation by S-DGC, the distribution profiles (Figure 4C) and the corresponding characterizations of particle size, zeta potential, and PDI for each fraction (Figure 4D–F) revealed that LNPs with 1.5% and 2.5% PEG were uniformly distributed within the 0–20% density layers. In contrast, within the high-density 30% sucrose layer (starting around fraction 20), a significant increase in particle size and a sharp decrease in zeta potential were observed. From an overall trend perspective, the majority of components in the PEG 0.5% formulation LNPs are distributed in the upper layer of the gradient column. Most particles exhibit diameters exceeding 200 nm, with a decreasing trend in zeta potential and poor stability. Their PDI shows a gradual increase from the low-density layer to the high-density layer. Post-separation agarose gel electrophoresis (Figure 4B) further confirmed that at 0.5% PEG, considerable nucleic acid accumulation occurred in the 0% density layer, and compared with the lysed sample, some nucleic acid remained unencapsulated in the intact LNPs. To more directly visualize the density distribution, the absorbance of pre-separation samples and fractions 1, 5, 10, and 19 was measured using a full-wavelength scan on a microplate reader. Absorbance at 360 nm (Figure 4G) showed that lipid components of the 0.5% PEG formulation accumulated mainly in the upper gradient, whereas the 1.5% and 2.5% PEG formulations were evenly distributed within the 0–20% density layers and exhibited better encapsulation.

Previous studies have shown that PEG influences the circulation time and cellular interactions of LNPs, with its content significantly affecting in vitro transfection [25]. Therefore, we transfected HeLa cells with LNPs (before and after S-DGC separation) containing different PEG concentrations. The results (Figure 4H,I) showed that DNA-containing fractions from all three PEG formulations, both before and after separation, mediated high transfection efficiency under fed (complete medium) conditions, with the 1.5% PEG formulation performing the best, reaching approximately 70%. Combined with the S-DGC separation data, these findings indicate that at PEG concentrations of 1.5% and above, more stable and compact LNPs are formed. These LNPs are uniformly distributed in the 0–20% density layers after S-DGC, reflecting significantly improved homogeneity of density-based subpopulations. Although the overall PDI increased slightly, the consistency of the core functional subpopulation was enhanced, thereby contributing to superior transfection efficiency.

The field of mRNA vaccine development is advancing rapidly; however, the inherent instability of mRNA molecules and the physical stability challenges of mRNA formulations based on LNPs—such as particle aggregation, fusion, and nucleic acid leakage—remain significant concerns. Commercially available mRNA-LNP products still require frozen storage and transportation to maintain their biological activity, underscoring the importance of systematic stability evaluation for the industrial translation of LNP-based therapeutics [26,27]. In this study, the freeze–thaw stability of a standard LNP formulation (N/P = 4) was evaluated. The experiment included three sample groups: Group A was stored at 4 °C for one week, while Groups B and C underwent three freeze–thaw cycles to simulate potential temperature fluctuations during cold-chain transportation. The physicochemical property test results in Table 5 and Table 6 indicate that after one week of storage, the LNP samples in the 4 °C storage group (Group A) and the −80 °C freeze–thaw group with cryoprotectant (Group C) maintained particle sizes below 200 nm with virtually unchanged zeta potentials. In contrast, the freeze–thaw group without cryoprotectant (Group B) exhibited significant alterations: particle size increased to 1323 nm, zeta potential decreased to 2.633 mV and encapsulation efficiency decreased to merely 5.3%. This indicates structural disruption of the LNPs in this group, leading to leakage of the encapsulated nucleic acids.

After freeze–thaw treatment, LNPs in Group B exhibited clear signs of stability deterioration: the sample appeared as a turbid milky-white suspension with visible precipitate at the bottom of the centrifuge tube. Agarose gel electrophoresis further revealed that most of the encapsulated nucleic acid had been released (Figure 5A). In contrast, samples stored at 4 °C (Group A) showed no significant changes in appearance or physicochemical state compared to their condition 7 days earlier, indicating good stability under this storage condition.

To further investigate the impact of freeze–thaw cycling on LNP structural integrity, all groups were analyzed by S-DGC. The results showed that samples from the 4 °C storage group (Group A) were uniformly distributed within the 0–20% density layers. Samples from the −80 °C freeze–thaw group with cryoprotectants (Group C) still displayed partial distribution in the 0–20% density range, suggesting that a portion of the LNPs remained homogeneous and stable. However, samples from the −80 °C freeze–thaw group without cryoprotectants (Group B) exhibited an abnormal distribution pattern: some components accumulated in the upper layers, and particle aggregation was observed around the 20–30% density interface (Figure 5B,C). Characterization of particle size and zeta potential for the post-S-DGC fractions indicated that both the 4 °C storage group (Group A) and the −80 °C freeze–thaw group with cryoprotectants (Group C) maintained uniform size and zeta potential distributions within the 0–20% density region. In contrast, fractions from the −80 °C freeze–thaw group without cryoprotectants (Group B) showed significantly increased particle size and decreased zeta potential after treatment (Figure 5D,E). Results from cell transfection experiments demonstrated that following S-DGC separation, the transfection efficiency of components within the 0–20% density range exhibited no significant difference compared to pre-freezing and pre-separation samples in both the 4 °C storage group (Group A) and the −80 °C freeze–thaw + cryoprotectant group (Group C). These samples maintained high transfection activity at 60–80% (Figure 5F). In contrast, for the −80 °C freeze–thaw group without cryoprotectants (Group B), only samples collected near the 20–30% density interface after freeze–thaw cycling exhibited about 30% transfection efficiency. In summary, combining freeze–thaw testing with S-DGC separation and subsequent characterization revealed that the addition of cryoprotectants effectively prevents structural damage to LNPs during freeze–thaw cycles. In the absence of cryoprotectants, freeze–thaw treatment leads to looser LNP structures, particle aggregation, leakage of initially encapsulated nucleic acid, an increase in particle density, and a consequent significant reduction in transfection function.

The GFP-encoding plasmid was transformed into E. coli DH5α competent cells using a standard heat-shock protocol. Subsequently, the transformed cells were inoculated into liquid LB medium supplemented with 0.1% kanamycin and cultured overnight (12–16 h) at 37 °C with shaking at 250 rpm. Plasmid DNA was then extracted using a commercial endotoxin-free maxiprep kit (TIANGEN, Beijing, China) according to the manufacturer’s instructions.

DNA-loaded LNPs were prepared using a clinically relevant LNP formulation (denoted as the “standard formulation”) composed of the ionizable cationic lipid SM-102, DSPC, cholesterol, and DMG-PEG 2000 at a molar ratio of 50:10:38.5:1.5. The lipid mixture was dissolved in anhydrous ethanol to form the organic phase. The nucleic acid was diluted in 20 mM sodium citrate buffer (pH 4.0) to prepare the aqueous phase. The two phases were mixed at a volume ratio of 3:1 (aqueous: organic) using a microfluidic chip. The resulting nanoparticles were subsequently dialyzed for 2–3 h at 4 °C against a dialysis buffer (pH 7.8) using 10 kDa MWCO dialysis tubing (16 mm width). Particle size and zeta potential of the DNA-LNPs before and after dialysis were measured using a Malvern Zetasizer Pro laser diffraction analyzer (Malvern Panalytical Ltd., Malvern, United Kingdom). Encapsulation efficiency was assessed by agarose gel electrophoresis. For each sample, two aliquots equivalent to 500 ng of theoretical nucleic acid content were prepared. One aliquot was treated with 1% Triton X-100 and incubated at 37 °C for 10 min to disrupt the LNPs and release the nucleic acid, allowing evaluation of nucleic acid integrity within the complexes.

Prepare sucrose solutions of 0%, 10%, 20%, and 30% concentrations using ultrapure water, with densities of 0.998 g/mL, 1.038 g/mL, 1.081 g/mL, and 1.127 g/mL, respectively (sterilized by filtration through a 0.22 μm microporous membrane). Discontinuous sucrose density gradient columns (approximately 0–30% w/v sucrose) were prepared by sequentially layering 2 mL of each sucrose solution into centrifuge tubes in order of decreasing concentration (i.e., 30%, 20%, 10%, and 0% from bottom to top). The prepared gradient columns were stored at −80 °C until use. Prior to separation, the gradient columns were thawed at 4 °C. DNA-LNP samples were carefully loaded onto the top of the gradient. The loaded columns were placed in a rotor and centrifuged at 110,000× g for 16 h at 13–16 °C using an Avanti JXN-30 floor-type ultracentrifuge (BECKMAN COULTER, Shanghai, China). After centrifugation, fractions of approximately 450–500 μL were collected sequentially from the top of the gradient using a micropipette, and each fraction was subsequently characterized.

HeLa cells were cultured in complete medium, consisting of high-glucose DMEM supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin, at 37 °C under 5% CO₂. When cells reached approximately 80% confluence, they were seeded in 24-well plates at a density of 8 × 10⁴ cells per well and allowed to adhere for at least 6 h prior to transfection. Once cell attachment was confirmed and confluence reached 70–90%, the cultures were divided into two groups: one subjected to serum starvation for 2 h (−FBS) and the other maintained in complete medium (Fed). Following the respective treatments, cells were transfected with DNA-LNPs corresponding to 2 μg of GFP plasmid per well and incubated overnight. The medium was then replaced with fresh complete medium, and live cell nuclei were stained with 10 μg/mL Hoechst 33342 (Thermo Fisher Scientific, Meridian Rd. Rockford, IL, USA) for 30 min. Images were acquired using a high-content imaging system, and transfection efficiency was quantified with ImageJ 1.8.0 software.

Statistical analysis was performed using GraphPad Prism 8. All values are presented as mean ± standard deviation (SD) from at least three independent experiments. Comparisons between groups were conducted using two-way analysis of variance (ANOVA). Differences were considered statistically significant at p < 0.05, and highly significant at p < 0.01. In this study, two-way analysis of variance (ANOVA) was primarily employed to compare differences in specific physicochemical or functional parameters between various LNP formulations across different density layers (i.e., distinct collected fractions) following separation via S-DGC. Consequently, the two factors typically incorporated into the analysis were: formulation factor and density layer/separated fraction factor.

The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding author.