Nucleic Acid-Loaded Lipid Nanoparticle Interactions with Model Endosomal Membranes

The lipids 1-hexadecanoyl-2-(9Z-octadecenoyl)-sn-glycero-3-phosphocholine (POPC; >99% purity), 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC; >99% purity), 1,2-dioleyl-sn-glycero-3-phosphoethamolamine (DOPE; 100% purity), sn-(3-oleoyl-2-hydroxy)glycerol-1-phospho-sn-1′-(3′-oleoyl-2′-hydroxy)glycerol (ammonium salt), also known as bis(monoleoylglycero)phosphate (BMP_18:1; >99% purity), and sphingomyelin from porcine brain (SM) were all purchased from Avanti Polar Lipids (Alabaster, AL). 1,2-Dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG-2000) was from NOF America Corp. (New York, NY). Cholesterol (≥99% purity), 2-(N-morpholino)ethanesulfonic acid (MES), poly(adenylic acid) [Poly(A); 100–500 kDa], glycerol, phosphate-buffered saline (PBS), and 2-amino-2-(hydroxymethyl)propane-1,3-diol (TRIS) were purchased from Sigma-Aldrich (Poole, U.K.), as were spectroscopic-grade (AnalaR) chloroform and ethanol. Heptatriaconta-6,9,28,31-tetraen-19-yl 4-(dimethylamino)butanoate (DLin-MC3-DMA or MC3) was obtained from Sapala Organics Private Limited (Hyderabad, India), while Firefly luciferase mRNA (FLuc mRNA; 1929 nucleotides, molecular weight = 618.5 kDa calculated assuming an average molecular weight per nucleotide of 320.5 + 159.0) was from CleanCap (TriLink Biotechnologies, San Diego, CA). Ultrapure water was obtained using a Millipore Milli-Q system (Merck Millipore, Darmstadt, Germany) to a resistivity of 18 MΩ cm.

The compositions of model endosomal monolayers were chosen to mimic endosomal compositions found in vivo using, in each case, four lipid components. The EEM comprised 40 mol % POPC, 20 mol % DOPE, 6 mol % SM, and 34 mol % cholesterol, while the LEM comprised 61 mol % POPC, 16 mol % DOPE, 6 mol % BMP_18:1, and 17 mol % cholesterol.^23,32 For each type of monolayer, the various lipid components were weighed separately and individually dissolved in chloroform at a concentration of 1 mg mL^–1. The lipid solutions were then mixed to prepare a spreading solution containing a total of 10 mg of the required lipid mixture. The resulting lipid mixtures were subsequently divided into 1 mg aliquots, dried by evaporating the chloroform under vacuum, and then stored at −20 °C until needed. Immediately prior to every monolayer experiment, a fresh solution of either EEM or LEM lipids was prepared by redissolving the aliquoted lipid in chloroform to a final lipid concentration of 0.1 mg mL^–1.

The LNPs were prepared at room temperature using a Nanoassemblr microfluidic device (Precision NanoSystems, Vancouver, Canada). Stock solutions were first prepared by dissolving the required amounts of DLin-MC3-DMA, DSPC, cholesterol, and DMG-PEG-2000 in ethanol. These stocks were then mixed in a 50:10:38.5:1.5 molar ratio to give a total lipid concentration of 12.5 mM (1.8 mg mL^–1). FLuc mRNA or Poly(A) was diluted in a 50 mM citrate buffer (pH 3.0) to obtain an mRNA/DNA–lipid weight ratio of 1:20 (cationic ionizable lipid–nucleotide with a 5.7:1 molar ratio). Empty (nucleic acid-free) LNPs were prepared in the same way but using (nucleic acid-free) 50 mM citrate buffer as the aqueous phase. The ethanolic and aqueous solutions were mixed in the Nanoassemblr microfluidic device at a 3:1 volume ratio and a total flow rate of 12 mL min^–1. The resulting LNPs were dialyzed in PBS overnight at 4 °C, using Slide-A-Lyzer G2 dialysis cassettes (10 kDa molecular weight cutoff; Thermo Scientific, Loughborough, U.K.). Glycerol was added as a cryoprotectant, with the final formulation containing 10% (v/v) glycerol. Samples were aliquoted, frozen, and defrosted prior to the surface pressure and ellipsometry measurements being performed.

The apparent hydrodynamic diameter of the LNPs was determined by dynamic light scattering (DLS) using a Malvern Zetasizer Nano ZS (Malvern Instruments Ltd., Malvern, U.K.). Prior to measurement by DLS, the LNPs were diluted 70-fold using PBS. The resultant apparent hydrodynamic intensity size and polydispersity index (PDI) at 298 ± 0.1 K were recorded at a backscattering angle of 173 °C. The encapsulation efficiency (expressed as %EE) of nucleic acid in the LNPs was quantified by a Quant-iT Ribogreen RNA assay kit (Invitrogen, Paisley, Scotland) following the manufacturer’s instructions. The results of this characterization are shown in Table S1↗ and discussed under the section entitled Characterization of the LNPs↗.

Monolayer experiments were performed at 293 ± 2 K using a G2 Langmuir trough (Kibron, Helsinki, Finland) with a custom-made low-volume stainless steel insert of dimensions 260 × 32 × 1 mm containing a 45 × 32 × 2 mm depression for a tinted glass deflector used to prevent diffuse laser reflections from the trough reaching the detector for the reflectometry measurements. The trough was first cleaned with copious amounts of water, then ethanol, and finally chloroform. It was then filled with 15 mL of an aqueous buffer solution prepared using ultrapure water. The buffers used for different pH values were the following: 5 mM MES buffer for pH 5.5 and 6.5, PBS buffer for pH 7.0 and 7.4, and 10 mM TRIS buffer for pH 8.5. In order to achieve a starting average area per molecule of 120 Ų, different amounts (namely, 54 and 64 μL) of 0.1 mg mL^–1 (total lipid) EEM and LEM mixtures, respectively, were spread using a Hamilton syringe on the surface and left for 15 min to ensure the evaporation of chloroform. With knowledge of the total amount of lipid, and therefore the number of lipid molecules of each type added to the surface, and of the surface area over which the lipid film was spread, it is possible to calculate an average area per molecule (A) according to the molar ratio of the various lipid components present. The surface pressure (π) is defined as the difference in the surface tension of a sample compared with that of pure water. The values were measured using a metal alloy Wilhelmy plate and recorded using Filmware 4.0 software. Furthermore, the reciprocal of the compression modulus (C_s^–1) is used to evaluate the average phase of the lipid monolayers according to⁴¹

It is considered that C_s^–1 values in the range 12.5–50 mN m^–1 indicate a monolayer in the liquid expanded phase and values in the range 100–250 mN m^–1 indicate a monolayer in the liquid condensed phase.⁴²

The area over which the lipid was spread was compressed using the barriers at a speed of 120 cm² min^–1 to a surface pressure of 25 mN m^–1, where it was held for 2 min before being expanded to 15 m N^1– at a speed of 10 cm² min^–1 and held at constant area for 15 min. This procedure was carried out to ensure that the trough edges and barriers were wetted with lipid, and, hence, the lipid monolayer was stable prior to injections of an aqueous solution of LNPs into the subphase. Once the film was determined to be stable, LNPs were gently injected underneath the lipid monolayer using a syringe with a bent 10 cm needle, and the resulting interaction was constantly monitored for at least 2 h by measuring the variation in the surface pressure. The volume of liquid injected into the subphase in each experiment was 1 mL (an overall volume change of about 6%), namely, a 1 in 16 dilution of the LNPs. All of the experiments were repeated, as indicated in the Supporting Information↗.

Data were recorded at 293 ± 2 K using an Nanofilm EP4 instrument (Accurion, Goettingen, Germany) with a blue diode laser at a wavelength of 489 nm and an angle of incidence of 50°. Upon reflection of polarized light at the air/water interface, the attenuation (Ψ) and phase shift (Δ) were determined by the optical properties of the system. As opposed to measurements at the solid/water interface where both parameters can be modeled to reveal the interfacial excess and thickness,³⁹ only the latter parameter is strongly sensitive to the presence of interfacial material at the air/water interface, the change of which can be related generally to the amount of interfacial material in the thin-film limit of a layer thickness of less than a few tens of nanometers.³⁵

The phase shift prior to the injection of LNPs, i.e., for a bare air/water interface or a spread lipid monolayer, Δ₀, recorded for 10–15 min, was subtracted from the measured values following injection to give the phase shift of the resulting interaction, Δ_int = Δ – Δ₀. This process first compensated for the presence of capillary wave roughness⁴³ as well as small systematic errors in the daily calibration of the experiment attributed to instrumental drift and positioning of the deflector and to the starting surface packing of a lipid in the case of interactions with the model monolayers.

Modeling of Δ_int to a total amount of interfacial material is highly complex for both multicomponent⁴⁴ and lipid systems:⁴⁵ in the former case, the technique alone cannot distinguish different components, while in the latter case, anisotropy exhibited from lipids present in a condensed phase (or any such lateral domains) strongly influences the data. Further complexities are that lateral heterogeneity in the interfacial material on the micrometer scale, i.e., from different lateral regions of the interface passing in and out of the probed area by the laser (∼1 mm²) with time, are exhibited as temporal fluctuations³⁶ and that regions of the interface that exceed the thin-film limit can have a negative contribution to Δ_int as a result of its periodicity.⁴⁶ As a result of these complexities, in the present work, we retain the phase-shift representation of the data, which we interpret in three contexts. First, the magnitude of Δ_int is taken as an approximate measure of the change in the amount of interfacial material per unit area as a result of the interaction. Second, temporal fluctuations above a clear baseline are taken as an indication of thicker regions of the interfacial material or domains of more condensed lipid chains as a result of lipid insertion from LNPs or induced phase separation. Third, temporal fluctuations that are both positive and negative from the baseline are taken as an indication of extended structures in contact with the monolayer as a result of the binding of LNPs. All of the experiments were repeated, as indicated in the Supporting Information↗.

BAM images were recorded at 293 ± 2 K using the same Nanofilm EP4 instrument (Accurion, Goettingen, Germany) with a blue diode laser at a wavelength of 489 nm but in this case at an angle of incidence matching the Brewster angle of the air/water interface of 53.1°. The instrument was used in a mode with p-polarized light, a 10× magnification objective, a polarizer, an analyzer, and a CCD camera. The technique is commonly used to image the lateral dimensions of both anisotropic liquid-condensed domains of phospholipids⁴⁷ and extended structures,³⁶ in films at the air/water interface, because the higher reflectivity of both types of features results in lighter features in the resulting images against a darker background. Images were taken at different surface pressures during compression of the EEM and LEM monolayers and at different time points following the injection of LNPs into the subphase while keeping the Langmuir trough barriers stationary. Background was subtracted from the images using an automatic feature of the instrument software with image focusing enabled. The same Γ correction was applied to all of the images to highlight the morphological features of the interfacial material without changing their relative brightness (Irfanview, Germany). It should be noted that BAM has a spatial resolution of several micrometers and, as a consequence, is unable to detect the binding of individual LNPs to the monolayer. All of the experiments were repeated, as indicated in the Supporting Information↗.

Statistical analysis of the surface pressure and ellipsometry data was performed using GraphPad Prism, version 9. Details of the two-way ANOVA and Šídák’s or Turkey’s multiple comparisons tests are indicated in the Supporting Information↗. For the surface pressure measurements, the value of the surface pressure at the plateau was averaged across the various repeats and subsequently compared between different experiments (namely, empty vs Poly(A)-loaded LNPs and EEM vs LEM at different subphase pH values). For the ellipsometry measurements, the mean value of Δ_int at the plateau for each sample was averaged across repeats and compared across experiments (namely, empty vs Poly(A)-loaded LNPs and EEM vs LEM at different subphase pH values).

Before discussing the results, we describe some of the biophysical processes that should be considered for the systems under study and explain how their effects are manifested in the surface pressure and ellipsometry data. With respect to the Langmuir monolayer experiments, an increase in the surface pressure (π) is related to a lowering of the interfacial free energy and, in crude terms, is determined by an increase in the density of the lipid chains in the monolayer.⁴⁸ On the other hand, the phase shift (Δ_int) obtained from ellipsometry is related to the total amount of interfacial material per unit area, while the nature of the fluctuations in Δ_int can reveal additional information, as noted below. Consideration of the possible ways in which LNPs or LNPs’ components (Scheme) could interact with the model endosomal membranes led to identification of the three possible processes shown in Scheme, namely, lipid insertion, LNP binding, and ionizable lipid–nucleic acid complex delivery.

To aid in the interpretation of data from the interaction of LNPs with the EEM and LEM, it was necessary first to characterize the endosomal monolayers in the absence of LNPs at a pH of either 5.5 (the pH of the late endosome) or 7.4 (the pH of the cytosol). The variation in the surface pressure (π) with the average area per molecule (A) of the two endosomal monolayers was determined up to 30 mN m^–1 (FigureA), as was the reciprocal of the compression modulus (FigureB). Although the shapes of the isotherms obtained for the two monolayer types are comparable at the same pH, the measured π at the same A is larger for the LEM. Regardless of the pH and endosomal type, at π up to 15 mN m^–1 and with C_s^–1 of less than 50 mN m^–1, the shape of the isotherm indicates that the molecules are in a liquid expanded state, which is widely considered to be analogous to the liquid-crystalline state in a lipid bilayer.⁴⁹ As π increases further and C_s^–1 approaches 100 mN m^–1, there is a transition of the molecules toward the liquid condensed phase, which is analogous to the gel state in a lipid bilayer.⁴⁹ The differences observed between the two monolayers are attributed to the presence, in the LEM, of 6 mol % anionic BMP_18:1. BMP_18:1, a strongly anionic lipid, is only found in the late endosome and lysosome⁵⁰ and will, because of its highly charged nature at both pH values studied coupled with the fact that each of its two alkyl chains contain an unsaturated site, occupy a large area per molecule in the LEM monolayer.

For each monolayer type over the measured range of π, the isotherms obtained at pH 5.5 are more compact in that A is generally lower and C_s^–1 higher at the same π. This effect is likely to be due to the presence of DOPE, because POPC and SM are zwitterionic at both pH values studied, while cholesterol is nonionic in nature and therefore unaffected by the pH. Like BMP_18:1, DOPE contains two unsaturated alkyl chains and a headgroup whose charge depends upon the pH in that it contains both a phosphate moiety (pK_a = 1.7) and an amino group (pK_a ∼ 9.6). Consequently, the amino group becomes increasingly basic as the pH decreases,⁵¹ while the strongly acidic phosphate group is ionized and therefore negatively charged at all pH values examined here. Consequently, as the pH decreases, the headgroup becomes zwitterionic in nature, occupying a smaller headgroup area.

BAM images of the EEM and LEM monolayers were acquired at both pH values at π values of 5, 15, and 25 mN m^–1 (FigureC). These BAM images serve as a reference for the observed exposure to LNPs. Regardless of the monolayer type or pH, bright spots of a few micrometers in size became increasingly abundant with increasing π. The presence of these domains in the EEM are consistent with domains recorded for lipid mixtures of similar composition.⁵² The EEM contains cholesterol (34 mol %), which is known to have a condensing effect in plasma membranes (and regulate lipid segregation), as does SM (6 mol %). Indeed, domains formed in a four-component monolayer of composition similar to that of the EEM were considered to be a result of the mixing of SM (a high chain-melting lipid) with cholesterol and the demixing of POPC (a low chain-melting lipid) with cholesterol.⁵³ On the plasma membrane of animal cells, SM is known to form microdomains with cholesterol and other glycosphingolipids, yielding so-called lipid rafts.⁵⁴ The domains of the LEM, on the other hand, look different in that at surface pressures above 5 mN m^–1 very small and bright spots appear and subsequently remain at all of the higher pressures tested. These domains are smaller, yet more condensed, compared to those in the EEM. Together, these observations are consistent with the values of C_s^–1 and lack of a pronounced phase transition in the monolayers under study.

The interaction of LNPs [with Poly(A)] with a bare air/water interface at pH values of 5.5, 7.4, and 8.5 was examined for up to 5 h after injection into the subphase (Figure) by monitoring π (FigureA) and Δ_int (FigureB) and taking BAM images (FigureC). LNPs were also prepared with no nucleic acid cargo to help understand the role of the nucleic acid cargo in the interaction with model endosomal monolayers. It is worth noting that, while the surface pressure reached at the plateau is extremely consistent between replicates, there was more variability in the lag phase, i.e., time to liftoff as well as the rate of increase in surface pressure, and as a consequence, therefore only trends are reported here.

Most noticeably, at the lowest pH of 5.5, the addition of both nucleic acid-containing and nucleic acid-free LNPs resulted in large increases in Δ_int and π, although the maximum changes in the value were only reached 4 h after LNP injection. In contrast, at the higher pH values of 7.4 and 8.5, the changes in Δ_int and π recorded were smaller but quicker, with maximum values being reached within 40 min of LNP injection (Statistical Analysis and Figures S7 and S8↗). At pH 7.4, the fluctuations in Δ_int after interaction, at the air/water interface, with nucleic acid-free LNPs were greater than those after interaction with Poly(A)-loaded LNPs, while both interfaces exhibit comparable fluctuations at pH 8.5. These fluctuations start about 80 min after LNP injection, with Δ_int values varying about a mean, as demonstrated by the bimodal distribution of the values (inset in FigureB). Such a bimodal distribution of the ellipsometry values indicates that, over time, the interfacial material adopts a morphology with regions that have considerably different thicknesses within the plane of the interface. This is the case of laterally segregated domains, which move in and out of the laser beam with time, possibly as a result of Marangoni flow from spreading events.⁵⁵ The distinct values of the bimodal distribution suggest that, over the first hour of the interaction, the lateral size of the domains increases to the millimeter length scale, matching the size of the interfacial area probed by the laser.

Together these observations suggest that, after LNP injection, the lipids comprising LNP transfer to the air/water interface cause increases in π and Δ_int. This transfer is clearly influenced by the pH, occurring more slowly, but to a greater extent, at pH 5.5. At this pH, Poly(A)-loaded LNPs resulted in higher values of π and Δ_int compared to nucleic acid-free LNPs, while no such difference was observed at the higher pH values of 7.4 and 8.5. At a pH of 5.5, the MC3 lipid becomes predominately ionized (cationic), possibly causing the remaining LNPs to become unstable,²⁰ releasing their component lipids and, when present, forming water-insoluble complexes of MC3 and nucleic acid.⁷

At pH 7.4 and 8.5, although the PEGylated lipids will be shed upon injection, MC3 is predominately un-ionized. Therefore, less disintegration of the LNPs is expected at these pH values and, when Poly(A) is present, less formation and lower delivery of ionizable lipid–nucleic acid complexes. This hypothesis can explain the lower increase in π and the nature of the fluctuations observed in Δ_int and the similarity in behavior of both the Poly(A)-loaded and nucleic acid-free LNPs at these pH values.⁷

This hypothesis is further supported by the BAM experiments, where images taken at the late time points at pH 7.4 and 8.5 showed the presence of much larger domains than those seen at pH 5.5, indicating that the reduced LNP disintegration at higher pH values results in the formation of large thick domains at the air/water interface. Regardless of the pH and the absence/presence of Poly(A), a few minutes after injection of the LNPs, the BAM images show the appearance of domains of a few microns in size, which considerably increase in size over the next couple of hours. The one exception to this is the Poly(A)-loaded LNPs at pH 5.5 at 210 min after injection, where a distinct extended network structure is observed. Such a feature is similar to those seen in mixed films containing macromolecules,^56,57 and we infer that the domains may be formed due to the presence of water-insoluble ionizable lipid–nucleic acid complexes at the air/water interface.

In summary, the following observations can be made:

(i) Higher π and Δ_int were reached at pH 5.5 compared to pH 7.4 and 8.5, as a result of MC3 protonation, LNP disruption, and lipid insertion and ionizable lipid–nucleic acid delivery [Poly(A)-loaded LNPs only] to the surface.

(ii) Poly(A)-loaded LNPs reached higher π and Δ_int compared to nucleic acid-free LNPs but only at pH 5.5, thereby demonstrating the impact of the lipid–nucleic acid complex formation and delivery.

(iii) The fluctuations in Δ_int recorded at pH 7.4 and 8.5 are the result of domain formation, as observed with BAM, due to whole LNP translocation to the surface.

The interaction of nucleic acid-free LNPs with the two model endosomal monolayers was examined after their injection into a subphase at either pH 5.5 or 7.4 (Figure). A starting π of 15 mN m^–1 was selected as a compromise between a higher, more biologically relevant π of around 30–35 mN m^–158 and a lower value that results in a more significant change in π.^59,60 In general, π was monitored for at least 2 h, and in some cases up to 5 h, after injection. It should be noted that the injection of a pure buffer under the EEM and LEM at pH 5.5 was investigated as a control to ensure that the injection caused no mechanical disturbance to the monolayer and introduced no artifacts into the data (Figure S1↗).

In the presence of model endosomal membranes at pH 5.5, a lag phase was observed before any interaction of the nucleic acid-free LNPs was seen, and although this interaction was slow, it was strong (FigureA). In the case of the EEM, a lag of about 45 min was seen before π and Δ_int increased, whereupon the interaction continued for a total of 75 min until maximum values of Δ_int and π (40 mN m^–1) were recorded. With the LEM, comparable maximum values of Δ_int and π (38 mN m^–1) were observed about 90 min after LNP injection.

When LNPs were injected under an EEM at a subphase of pH 7.4, a lag of about 45 min was seen before any change in π occurred, whereupon π increased to 18 mN m^–1, further increasing to 25 mN m^–1 after 3 h. Δ_int showed a slight increase but only at later time points, at which time some temporal fluctuations were observed (FigureB). By way of comparison, the lag phase and fluctuations in Δ_int seen for a LEM were minimal, with π increasing very slowly up to about 20 mN m^–1 at 3 h after LNP injection, with only a very small change in Δ_int being observed. These observations suggest that at pH 5.5 material from LNPs was incorporated into the model membranes, but only over an extended time period. This time lag was comparable to the data obtained for the interaction of nucleic acid-free LNPs at the bare air/water interface at pH 5.5 (FigureB) when a π of 15 mN m^–1 was reached 2 h after LNP injection, with a maximum value of around 35 mN m^–1 being observed at 4 h.

In contrast, at pH 7.4, the Δ_int fluctuations observed with the EEM suggest the formation of domains as seen for the LNPs alone. Significantly, at both pH values, the interaction of nucleic acid-free LNPs with either the EEM or LEM monolayers is qualitatively similar to no obvious differences being recorded for the two membrane types.

The BAM images (FigureC) highlight that, prior to LNP injection at pH 5.5, micrometer-sized domains were present in the monolayers However, upon injection, there was a rapid formation of larger domains, which remained only for a few minutes, providing further evidence for the proposed lipid exchange. When a plateau in π of about 40 mN m^–1 was reached, the domains were less visible, probably as result of the insertion of lipids from the LNPs into the monolayer. Interestingly, this phenomenon was not observed at the higher pH of 7.4 because the domains remained at the later time points, with some domains appearing occasionally in the field of view of the camera (Figure S2↗); these domains are attributed to the fluctuations in Δ_int observed for the EEM. For the LEM at both pH values, bright spots are the most evident feature even after LNP injection, although they tend to become more sparse over time as π increased.

In summary, the following observations can be made:

(i) Nucleic acid-free LNPs injected underneath the EEM or LEM at pH 5.5 resulted in much higher π and Δ_int values compared to the corresponding system at pH 7.4 as a result of MC3 protonation, LNP disruption, and lipid components from LNP insertion into the monolayer (SchemeA).

(ii) At pH 7.4, the injection of nucleic acid-free LNPs below the EEM showed some Δ_int fluctuations, indicating the formation of domains, as confirmed by the BAM data, due to whole LNP binding (SchemeB).

A range of subphase pH values (5.5, 6.5, 7.0, 7.4, and 8.5) were selected to determine in more detail the effect of the pH on the interaction of LNPs with the two model endosomal membranes. Because different buffers were used to obtain these different pH values, it was necessary to ensure the absence of any effects due to the varying nature of the buffer. As a control measurement, therefore, a comparison was made of the results obtained using 5 mM MES and PBS buffers at pH 5.5 (Figure S3↗) to exclude any buffer related effects. Because the effects of LNPs on the behavior of the EEM and LEM were similar, both are discussed together.

At pH 7.0 and 7.4, the changes in π of both endosomal membranes after injection of the LNPs were modest (FigureA), equilibrating at a value of just 22 mN m^–1, suggesting only a limited insertion of LNP lipids into the monolayers. Similarly, the temporal fluctuations in Δ_int were minimal, indicating either the absence or limited phase separation of the lipid and the formation of limited size domains (FigureB).

At pH 8.5, the data are subtly different. Specifically, while a modest change in π indicates a limited insertion of LNP lipids into each of the monolayers, the positive and negative temporal fluctuations of Δ_int observed for the EEM indicate that the interfacial film contains material that possesses a thickness greater than the thin-film limit of a few tens of nanometers. This finding is in line with the situation observed with LNPs in the absence of an endosomal monolayer at pH 8.5 (FigureB).

At pH 5.5 and 6.5, the observed differences are even more pronounced. Here, after a lag phase, there is a sharp increase in π to around 40 mN m^–1, which is mirrored by an increase in the baseline value of Δ_int and which is consistent with an increase in the lipid density per unit area (Statistical Analysis and Figures S5 and S6↗).

BAM images of the interactions were acquired at three pH values, representative of the different types of behavior discussed above (FigureC). The images highlighted that the characteristic small domains of the EEM remained visible, even after LNP injection. At pH 7.4, significant changes over time were not observed, consistent with the minimal effects of interaction inferred above. However, at pH 8.5, large domains were observed at later time points, confirming the phase separation inferred from ellipsometry. The very bright spots typical of LEM monolayers were also observed after LNP injection at every pH value measured. The distinct laterally extended network structures seen for Poly(A)-loaded LNPs at the bare air/water interface (FigureC) were also seen for both types of endosomal membranes at pH 5.5 after an increase in π due to the presence of Poly(A)-loaded LNPs, which plateaued at about 140 min (FigureC).

In summary, the following observations can be made:

(i) At pH <7, significant increases in π and in the baseline of Δ_int were recorded after the injection of Poly(A)-loaded LNPs underneath the EEM and LEM monolayers as a result of MC3 protonation, LNP disruption, lipid insertion, and ionizable lipid–nucleic acid delivery (SchemeA,C)

(ii) At pH 5.5, pronounced fluctuations in Δ_int were observed, consistent with the formation of laterally extended network structures at late time points, as observed with BAM.

(iii) At pH >6.5, only minimal changes in π and Δ_int were recorded, except at pH 8.5 with the EEM, where fluctuations in Δ_int and larger domains with BAM were observed.

FigureA gives a comparison of the changes in π and Δ_int of mRNA-loaded LNPs and Poly(A)-loaded and nucleic acid-free LNPs at pH 5.5. The interaction of mRNA-loaded LNPs with the two model monolayers was comparable to that of the Poly(A)-loaded LNPs in terms of π. In fact, their π values increased more quickly compared to those of the nucleic acid-free LNPs, with lag phase times of 30 min for EEM and 30–40 min for LEM, reaching 40 and 35 mN m^–1, respectively, about 1 h after LNP injection. Furthermore, the duration of the lag phase observed with ellipsometry (FigureB) was comparable. As observed above, Poly(A)-loaded LNPs showed pronounced fluctuations in Δ_int, and, interestingly, the baseline of Δ_int of mRNA-loaded LNPs exhibited a greater increase than the Poly(A)-loaded and nucleic acid-free LNPs. This is interpreted as a greater extent of delivery of the MC3 lipid–nucleic acid complex, implying that the effect of mRNA on the interface is greater, meriting further work on the subject. On the other hand, no significant differences were observed between the three types of LNPs with either endosomal membrane at pH 7.4 (Figure S4↗), thus proving the crucial role played by the pH-sensitive ionizable lipid MC3 in the interaction with model endosomal monolayers.

In summary, the following observation can be made:

(i) Although the same π was reached, higher Δ_int values were exhibited at pH 5.5 with mRNA-loaded LNPs compared to nucleic acid-free and Poly(A)-loaded LNPs, indicating that larger nucleic acid molecules may be able to interact more strongly with ionizable lipid molecules and, consequently, interact more with model endosomal membranes.