Cell medium-dependent dynamic modulation of size and structural transformations of binary phospholipid/Ï‰-3 fatty acid liquid crystalline nano-self-assemblies: Implications in interpretation of cell uptake studies

a r t i c l e i n f o

Naturally occurring phospholipids are alternative biocompatible amphiphiles that could be used for engineering of LLC nanoparticles [36,37]. Phospholipids, however, typically form lamellar mesophases in excess water, but addition of amphiphiles with high propensities to form inverse non-lamellar liquid crystalline phases (e.g., citrem) or apolar solvent modifiers can promote structural transformations to non-lamellar phases [2,36,38,39]. Thus, here we report on the assembly and biophysical characterization of structurally tunable and pH-responsive lamellar and nonlamellar nano-self-assemblies from binary mixtures of a biocompatible phospholipid (phosphatidylglycerol, DOPG) and different types of omega-3 fatty acids (x-3 PUFAs).
Unlike conventional cubosomes and hexosomes [1,38,40], here we show self-assembly of colloidally stable nanoparticles without utilization of a secondary emulsifier or an organic solvent. The inclusion of x-3 PUFAs is also advantageous due to their reported health-promoting effects such as in the prevention and treatment of inflammatory conditions, neurodegenerative diseases, viral infections, and brain cancer [41][42][43][44], and therefore could lead to the development of effective LLC nanopharmaceuticals with broad therapeutic indications. In line with these possibilities, we further selected a well-characterized LLC formulation for uptake studies by human monocytic THP-1 cells (as model scavengers for assessing nanoparticle susceptibility to immune cell clearance) and a patient-derived xenograft glioblastoma GBM T10 cell line. We report on pH-and cell medium-dependent dynamic fluctuations in nanoparticle size, number, and morphology with differential cell uptake kinetics. We discuss the importance of these findings in relation to interpretation of LLC nanoparticle-cell interaction kinetics. . The human monocytic cell line (THP-1) was purchased from the American Type Culture Collection (Manassas, VA). Both cell lines were certified mycoplasma free. Neurobasal Ò -A medium minus phenol red, RPMI1640 cell medium, fetal bovine serum (FBS), Geltrex, N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid (HEPES buffer), sodium pyruvate, and gentamicine were purchased from Gibco TM (NY, USA). B-27 Ò Supplement, GlutaMAX TM , basic fibroblast growth factor (bFGF), epidermal growth factor (EGF), and penicillin/streptomycin were received from ThermoFisher Scientific (MA, USA). All ingredients were used without further purification.

Preparation of DOPG/x-3 PUFA-based aqueous nanodispersions
Using different x-3 PUFAs (EPA, DHA, and DPA), three sets of DOPG/x-3 PUFA nanodispersions were prepared at a fixed total lipid (binary DOPG/x-3 PUFA mixture) content of 5.0 wt%. Samples were prepared at four different DOPG/x-3 PUFA weight ratios: 4:1, 3:2, 1:1, and 2:3. Briefly, the binary DOPG/x-3 PUFA mixtures were vortexed for 1 min and then emulsified in 10 mM excess phosphate buffered saline (PBS), pH 7.4, by using the ultrasonic processor Qsonica 500 (Qsonica LLC, Newtown, CT, USA) for 5 min in pulse mode (5 s pulses interrupted by 2 s breaks) at 30% of its maximum power. The pH of the continuous aqueous medium was adjusted within the range of 6.0-7.4, depending on experimental conditions. Nanodispersions were flushed with nitrogen gas for 2 min, capped, and covered by aluminum foil, and stored at 25°C.

Preparation of fluorescent-labeled nanodispersions
For cell studies, fluorescently labeled DOPG/DHA nanodispersions at DOPG/DHA weight ratio of 3:2 were prepared following the dry lipid film method and application of ultrasonication as described above. In this method, the required amount of binary DOPG/DHA mixture was dissolved in 1 mL chloroform, followed by addition of Rhod-DOPE at 0.2% of total molar lipid content. The organic solvent was evaporated using a gentle stream of nitrogen for 2 h to obtain a lipid film. The latter was left under vacuum for at least 12 h to remove residual chloroform and then hydrated with 10 mM PBS, pH 7.4. The aforementioned ultrasonication method was then employed for preparing the nanodispersions.

Synchrotron small-angle X-ray scattering (SAXS)
SAXS experiments were performed at the Austrian SAXS beamline (synchrotron light source ELETTRA, Trieste, Italy) at X-ray energy of 8 keV, a wavelength of 1.54 Å, and a sample to detector distance of 1314 mm. The detector covered the q-range (q = 4p sinh/k, where k is the wavelength and 2h is the scattering angle) of interest from 0.07 to 5.50 nm À1 . Silver behenate (CH 3 A(CH 2 ) 20 -ACOOAg) with a d-spacing value of 58.38 Å was used as a standard to calibrate the angular scale of the measured intensity. The twodimensional (2D) SAXS patterns of the nanodispersions were acquired using Pilatus3 1 M CCD camera (Dectris Ltd, Baden, Switzerland; active area 169 Â 179 mm 2 with a pixel size of 172 mm), and integrated into one-dimensional (1-D) scattering functions I (q) using Fit2D. They were then analyzed with IGOR pro (Wavemetrics, Inc., Lake Oswego, OR). All measurements were performed with an exposure time of 20 s per frame with 2 s delay between 3 frames at 37°C (±0.1°C) in custom-made glass capillaries with the aid of a Peltier element. In addition to the above, a DOPG/DHA nanodispersion (weight ratio of 3:2) was incubated in two different cell media at a volume ratio (v/v) of 1:1 (nanodispersion/cell medium) at different time points prior to SAXS studies. An optimized nanodispersion:cell medium volume ratio of 1:1 was necessary for obtaining a high signal-to-noise ratio. The scattering from PBS was used as a background and subtracted from SAXS data before further analysis. Lorentzian fitting was used to determine the q-values of the detected Bragg reflections. The lattice parameters, a, and d-spacings of inverse hexagonal (H 2 ), inverse discontinuous cubic (Fd3m), and lamellar (L a ) liquid crystalline phases, respectively, were gleaned from SAXS reflections by calculating the characteristic distance (d = 2p/q) for every reflection, and applying a standard procedure for lattice parameter calculations.

Cryo-transmission electron microscopy (Cryo-TEM)
The morphological characterization of representative DOPG/x-3 PUFA-based nanodispersions was investigated under their frozen-hydrated states as recently described [2]. Briefly, 3-4 mL of an aqueous DOPG/x-3 PUFA nanodispersion or its mixture with cell medium (nanodispersion:cell medium, 1:1 v/v) was applied on Lacey carbon 300 mesh copper grid (Ted Pella Inc., CA, USA) and blotted with filter paper (FEI Vitrobot, Holland) at a blotting time of 5 s, blotting force 0, and 37°C. After vitrification by a rapid plunging into liquid-nitrogen cooled ethane (-180°C), the grids were first hydrophilized by glow-discharge (Leica Inc. EM ACE 200, Germany) to enhance spreading of the samples. A Gatan 626 cryoholder (Gatan, UK) was used to observe the samples in Tecnai G2 20 transmission electron microscope (FEI, Holland) at a voltage of 200 kV under a low-dose rate (~5 e/Å 2 s). The images were recorded using FEI Eagle camera 4 k Â 4 k at a nominal magnification of 69,000 Â resulting in a final image sampling of 0.22 nm/ pixel. Nanoparticle size analysis on randomly selected images was performed with ImageJ (U. S. National Institutes of Health, MD, USA). In addition, Fast Fourier transformation (FFT) analysis using the same software was conducted for characterization of internal structures of selected nanoparticles.

Nanoparticle tracking analysis (NTA)
Nanoparticle size distribution was determined with a Nano-Sight NS300 mounted with a 405 nm laser and microscope (Malvern Panalytical Ltd., Worcestershire, UK), as recently described [2]. All measurements were performed in triplicate at room temperature. Prior to measurements, the nanodispersions were appropriately diluted in filtered 10 mM PBS (pH 7.4) to reach nanoparticle concentration between 10 6 -10 8 nanoparticles/mL. In selected cases, NTA measurements were performed on nanodispersions incubated in different cell media. For all measurements, instrumental settings were identical and the recorded videos were analyzed using Malvern Software (NTA 3.2 Dev Build 3.2.16, Malvern Panalytical Ltd., Worcestershire, UK).

Zeta potential
The zeta potential values were calculated from electrophoretic mobility measurements using Zetasizer Nano-ZS (Malvern Instruments Ltd. Worcestershire, UK) at 21°C, equipped with a 633 nm laser and 173°detection optics. Prior to measurements, samples were dispersed in (x10) diluted PBS, and the pH was adjusted to 6.0, 6.6, and 7.0. Zetasizer Software v.7.11 (Malvern Instruments Ltd.) was used for data acquisition and analysis. For the viscosity and refractive index, the values of pure water were used. Each experiment was repeated three times and the calculated zeta potential values are presented as mean ± SD.

Incubation in cell media
Nanodispersions (hexosomes) at DOPG/DHA weight ratio of 3:2 were incubated in fresh Neurobasal Ò -A and RPMI1640 cell media, which are used in GBM T10 and THP-1 cell culture studies, respectively. Neurobasal Ò -A medium contains 2% B-27 Ò supplement, GlutaMAX TM , bFGF, EGF, and penicillin/streptomycin; whereas RPMI1640 medium is composed of 10% FBS, gentamicine, HEPES, and sodium pyruvate. The nanodispersions and cell media were mixed at a volume ratio (v/v) of 1:1. The incubation studies were carried out at 37°C in 2.0 mL protein LoBind Eppendorf tubes (Hamburg, Germany) using a thermoshaker (MS100, Hangzhou Allsheng Instrument Co. LTD, China). At different time points, the biophysical properties of the samples were investigated using synchrotron SAXS, cryo-TEM, and NTA.

Quantum yield measurements
The absorption and emission spectra of the fluorescently labeled DOPG/DHA (3:2) nanodispersions were collected using Jasco V-650 spectrophotometer (JASCO, Easton, U.S.A.) and Fluotime 300 instrument (Picoquant, Berlin, Germany), respectively. To investigate the impact of pH and temperature on the photophysical stability, measurements at both temperatures of 25 and 37°C were conducted, and absorption and emission spectra were collected from nanodispersions prepared at three different pH values (5.0, 6.5, and 7.4). Time-resolved fluorescence anisotropy decays were acquired on Fluotime 300 instrument (Picoquant, Berlin, Germany) using picosecond laser diodes at k = 507 ± 3 nm as an excitation source (see Supplementary Information for further details).

Cell cultures
Patient-derived xenograft GBM T10 and human monocytic THP-1 cells were cultured in Neurobasal Ò -A and RPMI1640 cell media, respectively, at 37°C in a humidified atmosphere in the presence of 5% CO 2 and harvested before they reached 80-90% confluence.

Cell viability assay
Cell viability was determined using WST-1 reagent ready-touse colorimetric cell proliferation assay (Roche Diagnostics GmbH, Mannheim, Germany). Briefly, GBM T10 and THP-1 cells were seeded in 96-well plate at a cell density of 2 Â 10 3 cells/well and 5 Â 10 4 cells/well, respectively, and incubated overnight at 37 C and 5% CO 2 in a humidified atmosphere. Cells were then exposed for 24 h to DOPG/DHA (3:2) nanodispersion at a concentration in the range of 20-250 mg/mL. Untreated cells represented the negative control. Afterwards, the extracellular media were removed and replaced with fresh media, then WST-1 reagent (10 mL/well, 1:10 final dilution) was added, and the obtained mixture was incubated for 1 h at 37°C in the presence of 5% CO 2 . The absorbance was then measured at 450 nm by using Multiskan TM FC Microplate Photometer (Massachusetts, U.S.). The metabolic activity at standard growth conditions was considered as 100%. Results are expressed as relative percentages of viable cells compared with untreated cells. All incubations were in triplicate and the data are expressed as mean ± SD.

Cellular uptake studies
GBM T10 and THP-1 cells were seeded in 6-well plates at 25 Â 10 4 cells/well, and 5 Â 10 5 cells/well, respectively, and incubated overnight at 37°C and 5% CO 2 . These cells were then exposed to DOPG/DHA (3:2) nanodispersion at different concentrations (5-60 mg/mL). At 24 h post-incubation, the medium was removed, and the cells were then washed twice with PBS and kept in fresh cell medium. Cellular uptake of fluorescently labeled DOPG/DHA (3:2) nanodispersion was quantified by flow cytometry using BD Accuri TM C6 Plus Flow Cytometer (New Jersey, U.S.). The procedure was repeated in three independent measurements with at least 20.000 cells analyzed in each measurement, and data were presented as mean ± SD.

Statistical analysis
Statistical significance was assessed by One-way ANOVA using GraphPad Prism version 8.0.0 for Windows (GraphPad Software, San Diego, USA), and with Dunnett's multiple comparison tests.
Collectively, the aforementioned analysis suggests that differences in acyl chain length and unsaturation degree of the investigated x-3 PUFAs do not induce notable alterations in the structural features of DOPG/x-3 PUFA nano-self-assemblies (Table 1). For the identified H 2 and Fd3m phases, our results indicate that an increase in acyl chain length at the same unsaturation degree [i.e., DPA (22:5, n-3) compared with EPA (20:5, n-3)] has no major effect on the internal structural features. Both phases exhibited comparable lattice parameters in DPA-and EPA-containing DOPG nanodispersions. This is consistent with our recent report on x-3 PUFA monoglycerides [MAG-DPA and MAG-EPA], which also displayed H 2 phase in excess water with comparable lattice parameters to each other [48]. Compared with DPA-and EPAcontaining nanoparticles, a slight increase in the lattice parameters of both coexisting H 2 and Fd3m phases was observed on loading DHA (22:6, n-3) to DOPG nanodispersions ( Table 1). Assuming that these fatty acids are in a similar ionization state at pH 6.0, this effect could be attributed to alterations in the molecular organiza-  tion of DHA at DOPG-water interface and its mode of interaction with DOPG as compared with DPA and EPA, which have a lower unsaturation degree than DHA.
In relation to the promotion effect of x-3 PUFAs on the formation of non-lamellar liquid crystalline phases, it is important to take into account a possible impact of the protonation state of the embedded x-3 PUFA molecules at DOPG-water interface, and hence the role of electrostatic forces in modulating the structural features of these nanodispersions. Thus, for gaining more insight into possible alterations in the protonation state of embedded x-3 PUFA molecules on varying pH, a set of additional SAXS experiments was performed at pH 6.6 and 7.0 with binary DOPG/DHA mixtures at different ratios (4:1, 3:2, 1:1, and 2:3). As shown in Fig. 2A, the structural features of the nano-self-assemblies were pH-sensitive, causing significant structural alterations or phase transitions. For instance, increasing pH from 6.0 to 6.6 and 7.0 was associated with a slight decrease in the lattice parameter of the H 2 phase at DOPG/DHA weight ratio of 3:2 from 7.87 ± 0.02 n m to 7.59 ± 0.02 nm and 7.49 ± 0.01 nm, respectively. However, the SAXS patterns became more diffusive with increasing pH, particularly at pH 7.0, presumably indicating an increase in the fraction of coexisting vesicles with hexosomes. A more dramatic effect was observed at a higher DHA content (DOPG/DHA weight ratios of 1:1 and 2:3), including a transition from nanoparticles with a biphasic H 2 /Fd3m feature to hexosomes (with an internal H 2 phase having a lattice parameter of 7.18 ± 0.03 nm) and vesicles, respectively, on increasing the pH from 6.0 to 7.0. In addition to the formation of vesicles, as indicated with the diffuse-dominating scattering pattern at DOPG/DHA weight ratio of 2:3, a very weak peak at q %1.06 nm À1 (marked with an asterisk, Fig. 2A) was present. This might indicate the presence of cubic Fd3m traces. These observations suggest a modulatory role for electrostatic forces in pH-triggered nonlamellar-lamellar structural transitions, and consistent with previous studies on nanodispersions containing longchain fatty acids [59][60][61][62]. Here, increasing pH from 6.0 to 7.0 is most likely associated with enhanced electrostatic repulsions among a greater fraction of deprotonating carboxyl moieties of x-3 PUFAs that are incorporated into DOPG-water interface, which consequently expand areas occupied by the hydrophilic headgroup of DOPG and x-3 PUFA molecules (an effect that is simultaneously associated with an increase in the water uptake by the selfassembled interiors). Eventually, this leads to destabilization of non-lamellar liquid phases and the formation of flattened lipid bilayers in form of vesicles. Finally, the estimated pK a of incorporated unsaturated fatty acids into lipid-water interfaces is approximately 7.0 [63] compared with~4.8 for the carboxylic acid moiety of the corresponding fatty acids in solution [64]. It is, therefore, expected that increasing pH from 6.0 to 7.0 would increase the negative surface-charge density at DOPG-water interface, due to an enhanced DHA ionization (an increase in the concentration of ionized DHA carboxylic groups), thereby leading to the aforementioned nonlamellar-lamellar phase transitions [63,64].

Potential mechanism of lamellar-nonlamellar phase transitions by x-3 PUFAs
We attribute the observed concentration-dependent effect of x-3 PUFAs on induction of lamellar-nonlamellar liquid crystalline phase transitions to the classical fusion pathway as illustrated in Fig. 3A. Here, inclusion of x-3 PUFAs (DHA, DPA, or EPA) at a relatively low concentration is associated with the symmetricasymmetric transition of DOPG bilayers, presumably due to an even distribution of the fatty acid molecules into the bilayer leaflets, and further is mediated by electrostatic and van der Waals forces [2,47].
Through the point-defect route [47,65], a further increase in x-3 PUFA concentration is associated with an enhanced lateral diffusion of embedded DOPG and x-3 PUFA molecules and increased tensions between the bilayer leaflets. This leads to evolvement of intermembrane attachment sites with stalk-like topologies that are then converted to extended intermembrane contact zones (otherwise known as transmonolayer contacts or hemifusions [65]) (Fig. 3A). The eventual breakdown of these zones promotes the formation of numerous pores within the interconnected vesicular structures that render a phase transition to H 2 phase. At a relatively high x-3 PUFA concentration, the transition from a neat H 2 phase to a biphasic H 2 /cubic Fd3m feature could be attributed to an increase in the negative spontaneous curvature owing to localization of the fatty acid molecules in the hydrophobic domains of the self-assembled nanostructure as discussed above.
In addition to the proposed classical fusion pathway, the observed direct L a -H 2 phase transition on loading x-3 PUFAs to DOPG nanodispersion might also proceed through an alternative process involving the line-defect route (Fig. 3B). It is well known that the inclusion of oils, including fatty acids, is associated with a dehydration effect (a decrease in the amount of solubilized water in nanoparticles' self-interiors) [47,53,59,[66][67][68]. As previously suggested, such a direct transition may occur through the linedefect route that takes into account possible shortage in water content during the transition [65,69,70]. Thus, we propose that line defects may spontaneously be formed between membranes of opposed bilayers, leading eventually to the formation of tubular lipid structures that initiate a transition to H 2 phase (Fig. 3B).

Morphological characteristics of DOPG/x-3 PUFA nanodispersions
Taking into account the tendency of non-lamellar liquid crystalline nanoparticles to coexist with other colloidal nanoobjects (typically vesicular structures), cryo-TEM was used to gain insights into the morphological features and population heterogeneity of DOPG/x-3 PUFA nano-self-assemblies dispersed at pH 6.0. We focused on DOPG/DPA and DOPG/DHA nano-self-assemblies with different weight ratios (4:1, 3:2, and 2:3), and their corresponding cryo-TEM images are presented in Fig. 4A-D and Fig. 5A,B. At a relatively low DPA concentration (DOPG/DPA 4:1), the cryo-TEM images (Fig. 4A, and S1) show the presence of small spherical uni-and oligo-lamellar vesicles (ULVs and OLVs) of 5-100 nm in diameter, which is in accordance with SAXS results above. The cryo-TEM images were also in agreement with SAXS analysis on colloidal transformation from vesicles to hexosomes (a curved striation typically observed in hexosomes is marked with a yellow star in Fig. 4B) on increasing DPA concentration (at DOPG/DPA weight ratio of 3:2). Curved striations are typically observed in hexosomes based on unsaturated monoglycerides or PHYT and stabilized with Pluronic F127 [12,21,48,49,51,57,71,72]. The cryo-TEM micrographs also show the presence of additional nanoparticles (marked with black dashed arrows) with surface vesicular structures (denoted with blue arrows, Fig. 4B), and sizes in the range of 110-225 nm. These nanoparticles might resemble hexosomes; however, the occurrence of disordered internal nanostructures did not allow unambiguous determination of their structural features by cryo-TEM. Furthermore, relatively small ULVs and OLVs with sizes of 10-50 nm are also present (marked with white asterisks, Fig. 4B). The formation of such coexisting vesicular structures is consistent with previous studies on the heterogeneous nature of cubosome and hexosome preparations [10,12,71], where nonlamellar liquid crystalline nanoparticles are typically coexist with vesicles. Hexosomes with typical curved striations (marked with yellow stars in Fig. 5A,B) are clearly detected in DOPG/DHA (3:2) nanodispersion as well as a population of relatively small coexisting vesicular structures including ULVs (marked with white asterisks in Fig. 5A,B), and adsorbed vesicles on hexosomes (marked with a white arrow in Fig. 5A). At a higher DPA concentration (DOPG/DPA of 2:3), a colloidal transformation to micellar cubosomes (nanoparticles with an internal cubic Fd3m phase, marked with a red star in Fig. 4C) was indicated and supported by FFT analysis. Similar morphological features have also been reported in Pluronic F127-stabilized micellar cubosomes [18,59]. In contrast to SAXS findings ( Fig. 2A and Table 1), there was no indication from cryo-TEM images on formation of nanoparticles with a biphasic H 2 /Fd3m feature or a coexistence of hexosomes with micellar cubosomes, as evident by SAXS analysis. It is most likely that the aforementioned coexisting nanoparticles present in relatively very small fractions in these nanodispersions [10,12,13,73], and future examination of numerous cryo-TEM images are necessary to detect and visualize them. Finally, the electron micrograph images also show the presence of relatively smaller non-lamellar liquid crystalline nanoparticles with sizes in the range of 30-60 nm (marked with red arrows in Fig. 4C,D) compared with those of the most investigated Pluronic F127-stabilized cubosomes and hexosomes (typically 100-200 nm) [1,4,48,54]. Due to current limitations with cryo-TEM image resolution, we have not been able to visually identify these species, but they might represent relatively small micellar cubosomes (Fig. 4D).

Nanoparticle size analysis and zeta-potential determination
NTA was used for determination of the nanoparticle hydrodynamic size distributions suspended in 10 mM PBS at pH 7.4, and the results are summarized in Table 2. The results show the influ-ence of the fatty acid type and concentration on nanoparticle mean sizes as compared with the control sample (DOPG vesicles). Size increases on fatty acid loading is consistent with the structural alterations and transitions observed by SAXS analysis and cryo-TEM. While increasing concentration of either DPA or EPA in the binary fatty acid/DOPG mixtures gradually increased nanoparticle sizes, increasing DHA concentration had no dramatic effect on the mean nanoparticle size (ranging from 188.4 ± 66.3 nm to 191.7 ± 73.2 nm), and the sizes remained comparable regardless of DHA concentration (Table 2).
Next, zeta potentials of all DOPG/x-3 PUFA nanoparticles were initially determined in diluted PBS at pH 6.0 and the results are summarized in Table 2. All nano-self-assemblies exhibited a high negative zeta potential confirming colloidal stability through electrostatic repulsion forces. On increasing the concentration of loaded x-3 PUFAs, the zeta potential values became slightly more negative. This was most likely attributed to a partial ionization of fatty acids embedded at DOPG-water interface, and further supported in experiments when the pH was increased from 6.0 to 7.0 ( Table 2).

The effect of cell incubation media on nanoparticle structural characteristics and size distribution
Dispersion media can dramatically affect LLC nanoparticle swelling and modulate structural transformation from one population to another [2,14,51,74,75]. Thus, prior to studies on the interaction of nanodispersions with cells, it is imperative to determine the effect of the cell incubation medium on the biophysical characteristics of nanodispersions. Accordingly, we selected DOPG/DHA (3:2) nanodispersions for such studies in RPMI1640 (containing 10% FBS) and Neurobasal Ò -A (a serum-free medium) in relation to THP-1 and GBM T10 cells, respectively. The results in Fig. 5D,E compares time-dependent changes in nanodispersion SAXS patterns in Neurobasal Ò -A and RPMI1640 media (with initial pH of 7.4). In both media, all three characteristics Bragg peaks of internal H 2 phase were detectable at 60 min post-exposure, but they slightly shifted to higher q values leading to a decrease in the lattice parameter of the H 2 phase from 7.87 ± 0. 02 nm to 7.53 ± 0.01 and 7.52 ± 0.01 nm in Neurobasal Ò -A and RPMI1640 media, respectively. The loss of the H 2 phase was faster in RPMI1640 than Neurobasal Ò -A, and presumably a reflection of pH responsiveness of nanodispersions (arising from partial ionization of fatty acids embedded at DOPG-aqueous interface, with a final pH elevation from 6.0 to 6.6 in both media) and their swelling rates. However, the role of different media components on nanodispersion swelling was not studied. In addition to the slight decrease in the lattice parameter of the H 2 phase, the SAXS patterns became mainly dominated diffusive scattering with increasing incubation times, leading to a gradual transition from hexosomes to vesicles. By 120 min, vesicles were predominantly present as evident from the mainly dominated diffusive scattering patterns. These represented most likely uni-and oligo-lamellar vesicles, due to the presence of positionally uncorrelated bilayers [47]. A weak single peak at q % 0.97 nm À1 (marked with a purple asterisk, Fig. 5D) was also observed, indicating most likely the presence of traces of the H 2 phase in Neurobasal Ò -A medium. At different time points, the SAXS patterns showed an approximate q -3 dependence in the low q regime, which indicated the formation of vesicles (Fig. 5F,G). This is consistent with previous reports on the formation of vesicles with this characteristic dependence in the low q regime [76,77]. An additional diffuse Bragg peak started also to appear at q % 0.5 nm À1 after 30 min of incubation in both media (marked with black asterisks in Fig. 5D,E). This presumably indicate the formation of coexisting sponge (L 3 ) phase in the nanodispersion.
Cryo-TEM was then used for determination of the morphological characteristics of nanoparticles. Since SAXS experiments indicated similar cell medium-dependent trends and behaviors in nanoparticle characteristics on exposure to both cell media, we limited cryo-TEM studies to the Neurobasal Ò -A medium. In parallel with SAXS, cryo-TEM studies also confirmed the transformation of DOPG/DHA (3:2) hexosomes to vesicles in coexistence with spongosomes (nanoparticles with L 3 phase) in Neurobasal Ò -A medium ( Fig. 5A-C). At 120 min post-incubation, the cryo-TEM micrographs show the presence of a heterogeneous population of nanoparticles with varying sizes in the broad range of 30-215 nm. In addition to ULVs and OLVs (marked with white asterisks), spongosomes with highly disordered interiors (denoted by blue stars), adsorbed vesicles (or sponge phases) on hexosomes, and less developed spongosomes with adsorbed vesicles (marked as white arrows) are also evident (Fig. 5C). The electron micrographs also depict interlamellar attachments (ILAs) between bilayers (denoted with dashed black arrows) and membrane-linking pores (circular features denoted with dashed white arrows) (Fig. 5C). Such ILAs and pore features were previously observed in spongosomes [65,[78][79][80][81][82]. Thus, these observations further support the suggestion that the point-defect route (Fig. 3A) might be responsible for cell medium-mediated transition from the H 2 phase to a biphasic L a /L 3 feature.
Next, we studied the effect of cell media on hydrodynamic size distribution of DOPG/DHA (3:2) nanoparticles by NTA. The data in Table 3 and Fig. 6 show nanoparticle size distribution changes as a function of nanoparticle concentration. The corresponding scatterplots of relative light scattering intensity versus the estimate of the nanoparticle size are also shown (Fig. 6). A number of striking observations are apparent. Thus, within 30 min of incubation in either cell media the total concentration of nanoparticles dramatically increased compared with control incubations in PBS (control sample: nanodispersion produced on dispersing binary DOPG/DHA mixture in excess PBS). Afterwards, nanoparticle concentration declines, but by 120 min the total nanoparticle concentration was still much higher than the control sample (DOPG/DHA mixture incubated in excess PBS) ( Table 3). These fluctuations were more prominent with nanoparticles within the 130-160 nm size range (Fig. 6). However, nanoparticle concentration fluctuations were more notable in RPMI1640 than in Neurobasal Ò -A medium, and presumably a reflection of the presence of relatively larger vesicular structures in RPMI1640 medium. Incubation in RPMI1640 medium also led to an increase in the population of nanoparticles with sizes ! 200 nm compared with both control sample and Neurobasal Ò -A incubations, and further reflected in the corresponding scatterplots (Fig. 6). Finally, the aforementioned fluctuations in nanoparticle population sizes were also in line with changes in median and mode size values (Table 3). In conclusion, the reported fluctuations in nanoparticle population sizes and concentrations are in agreement with SAXS and cryo-TEM findings, which reflect pH-and cell medium-dependent dynamic transformation from hexosomes to vesicles and spongosomes with variable swelling rates. It is also tempting to speculate that interactions of DOPG/ DHA nano-self-assemblies with some components of cell media could trigger DOPG or DOPG and DHA release from hexosomes, thereby leading to spontaneous formation of further coexisting vesicles. Finally, since RPMI1640 medium contains FBS, adsorption of serum proteins to some nanoparticle populations may further explain the size increase on prolonged incubation periods.

Cell-interaction studies
The abovementioned findings indicate that the combination of cell medium composition and nanodispersion pH sensitivity within the pH 6.0-7.4 range dramatically modulate the nanoparticle structural features and morphology. These in turn could affect nanoparticle interaction and uptake differently by different cells. Indeed, the mode of nanoparticle interaction with cells and internalization pathways are not universal, but are dependent on both nanoparticle characteristics as well as the cell type and culture conditions [83][84][85]. For example, nanoparticle type, size, shape, and surface characteristics could determine through which of endocytic pathways nanoparticles can enter a cell, but often multiple pathways simultaneously operate or inhibition of one pathway could modulate another. There is a general consensus that clathrin-mediated endocytosis, clathrin-independent/dynamin-in dependent endocytosis, and fast endophilin-mediated endocytosis could all participate in internalization of sub-200 nm nanoparticles [86], whereas larger particles can be internalized via macropincocytosis and phagocytosis [87]. On the other hand, some endocytic  Table 3 SAXS-derived structural parameters and hydrodynamic size characteristics of DOPG/DHA (3:2) nanodispersions a before and after incubation in Neurobasal Ò -A and RPMI1640 cell media.
Incubation time Nanodispersion Space Group a b (nm) Size (nm) ± SD pathways are not only limited to certain cells, but nanoparticle internalization can be a constitutive or an induced process [85,88]. An example of the latter is the induction of macropinocytosis in dendritic cells in response to lipopolysaachride [89]. Thus, by considering Neurobasal Ò -A-and RPMI1640-induced dynamic changes in nanoparticle size, number, and morphology, internalization processes will be most likely stochastic and competitive. Furthermore, apart from endocytosis, uptake might also involve vesicular fusion with the plasma membrane, since the proportion of vesicles seems to increase on prolonged incubation periods in either cell media (Fig. 5). Accordingly, we restricted nanoparticlecell interaction studies to the DOPG/DHA (3:2) formulation, which was characterized in detail, by two different cell lines (GBM T10 and THP-1) as a function of the initial nanoparticle concentration.
For following nanoparticle-cell interactions, DOPG/DHA (3:2) nano-self-assemblies were labeled with the fluorescent marker Rhod-DOPE. Labeling neither perturbed internal H 2 nanostructure of DOPG/DHA hexosomes ( Figure S2) nor modulated their size bio-  (Table S2). Next, we evaluated nanoparticle quantum yield as a function of pH (5.0, 6.5, and 7.4) and temperature (25 and 37°C). The corresponding UV-Vis absorption and emission spectra are depicted in Figure S3, showing peaks at k abs = 520 nm and k em = 544 nm, and the calculated fluorescence quantum yields are shown in Table S1. The results confirm pH and temperature stability of labeled DOPG/DHA nanoparticles as well as their high luminescence efficiency, thereby suggesting their suitability for biological evaluation.
Prior to gaining insight on nanoparticle-cell interaction kinetics, we evaluated cellular safety as a function of DOPG/DHA (3:2) nanoparticle concentration in comparison to DOPG vesicles. Exposure of both GBM T10 and THP-1 cells to DOPG vesicles in the concentration range of 25-200 mg/mL had no dramatic effect on cell viability compared with untreated cells (Fig. 7A,B). This was expected since DOPG is a major anionic lipid constituent of mammalian cell membranes [90]. However, the two cell lines differently responded to DOPG/DHA (3:2) nanoparticles in a dose-dependent manner. With GBM T10 cells, cell viability was not significantly compromised when challenged with DOPG/DHA (3:2) nanoparticles at concentrations up to 100 mg/mL; however, the cell viability drastically decreased at 200 mg/mL (Fig. 7A). THP-1 cells were more sensitive to DOPG/DHA (3:2) nanoparticles. Cell viability was marginally affected at low nanoparticle concentrations (25 mg/mL) compared with untreated cells but significantly reduced on exposure to nanoparticles up to 100 mg/mL (cell viability > 65%) (Fig. 7B). The observed different sensitivity of GBM T10 and THP-1 cells to DOPG/DHA (3:2) nanoparticles is not clear, but may be attributed to differences in uptake kinetics and intracellular responses such as the extent of DHA accumulation in intracellular lipid vesicles, and initiation of lipid-mediated autophagy and other cell death pathways [91][92][93][94].
On the basis of cell viability experiments, we next studied DOPG/DHA (3:2) nanoparticle uptake kinetics in the range of 5-60 mg/mL at 24 h post-incubation. The results in Fig. 7C,D show a concentration-dependent nanoparticle uptake by both cell lines. GBM T10 cells were more responsive to these nanoparticles. Even at low nanoparticle concentrations (5 mg/mL) approximately 80% of cells participated in nanoparticle uptake (Fig. 7E). Despite their inherent scavenging properties, THP-1 cells were less responsive to DOPG/DHA nanoparticles. Here, the cell-associated fluorescence intensity increased at nanoparticle concentrations above 30 mg/mL (Fig. 7D); however, compared with GBM T10 cells, an initial concentration of 60 mg/mL nanoparticles was necessary to engage 80% of cells in nanoparticle uptake (Fig. 7F). Nevertheless, nanoparticle uptake in both cell lines was qualitatively confirmed by fluorescence microscopy (Fig. 7G). As discussed above, the observed difference in nanoparticle uptake by GBM T10 and THP-1 cells might be due to operation of different cell-dependent endocytic mechanisms. Such mechanisms are further modulated by dynamic transformations in nanoparticle morphology, number, and size in the extracellular media (as well as cell-derived factors). For instance, THP-1 cells might only respond to nanoparticle aggregates or species with > 200 nm size and ingest them by phagocytosis and/or macropinocytosis. Furthermore, poor nanoparticle opsonization by the third complement protein (C3) in FBS might have contributed to inefficient nanoparticle binding and uptake by THP-1 cells through complement receptors. Indeed, C3 functional activity is rather undetectable in most fetal bovine sera [95], and FBS further contains <3% of conglutinin, C1, and C6, thereby limiting the activity of the calcium-sensitive pathways of the complement system. Accordingly, uptake mechanisms by THP-1 cells under prescribed conditions could be limited to adsorptive endocytosis or other receptor-mediated processes such as nanoparticle recognition by some classes of scavenger receptors. Non-specific serum protein adsorption to nanoparticles might also exert dysopsonic activity and reduce nanoparticle binding to THP-1 cells, particularly at low lipid concentrations.
In addition to the above, it is also likely that extracellular modifications and transformations in nanoparticle morphology and size could inadvertently affect the distribution of the fluorescent lipid marker. Thus, some populations of nanoparticles and vesicles could be free from the lipid marker, yet competing with nanoparticles bearing the marker on cell binding and internalization. This might project estimation inaccuracies in nanoparticle uptake measurement with fluorescent markers. Considering the detection sensitivity limitation with fluorescent tracers in cell uptake studies, particularly at early time points (and at low lipid concentrations) where uptake levels are very low, our cell experiments were limited to analysis of nanoparticle uptake signal at 24 h postincubation. Accordingly, future experiments should introduce simultaneous nanoparticle labeling with radioactive aqueous and lipid phase markers for better evaluation of nanoparticle stability in biological media and for overcoming the aforementioned limitations in nanoparticle uptake kinetic studies. Such approaches might allow for better mapping of cellular responses (nanoparticle binding and uptake kinetics) in relation to early dynamic changes in nanoparticle biophysical characteristics. Notwithstanding, experiments here have demonstrated complexities associated in monitoring cellular uptake of cell medium-dependent transformative ISAsomes.

Conclusions
Previous efforts have extensively studied the structural features of Pluronic F127-stabilized cubosomes or hexosomes based on unsaturated monoglycerides or phytantriol [59,68]. Contrary to these efforts, here we introduced a simple-by-design approach for engineering of stabilizer-free fatty acid-based non-lamellar liquid crystalline nanoparticles. This was achieved through the formation of a diverse library of colloidally stable lamellar and non-lamellar liquid crystalline nanoparticles from binary mixtures of DOPG and three different types of x-3 PUFAs (DHA, DPA, or EPA) in the absence of a secondary emulsifier and organic solvents. Through extensive biophysical characterization, we demonstrated structural diversity among the nano-selfassemblies that spanned into three distinct phases (L a , H 2 , and Fd3m). SAXS and cryo-TEM studies confirmed that x-3 PUFAs modulate the internal architecture of LLC nanoparticles in concentration-and pH-dependent manners. Our results have also shown dynamic and progressive changes in LLC nanoparticle size distribution, number, and morphological characteristics in cell culture medium. Considering that dynamic changes in physicochemical properties of nanoparticles can affect the mode of nanoparticle interaction with cells, our work cautions on over interpretation of LLC nanoparticle-cell interaction kinetics and internalization mechanisms per se. The therapeutic efficacy of a particulate delivery system primarily depends on internalization and intracellular trafficking. Thus, future cell experiments must consider which populations of LLC nanoparticles are therapeutically relevant. This in turn requires improved techniques for nanoparticle labeling and tracing as well as nanoparticle fractionation (based on morphology), and approaches that normalize nanoparticle binding to cells, assess uptake in knockout models of key components in the internalization pathways, and most importantly determine whether these pathways operate in validated in vivo knockout models. Such integrated approaches could lead to the development of multifunctional LLC nanocarriers for site-specific targeting and effective therapeutic interventions through single and combination drugs delivery.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements
Financial support by the Danish Council for Independent Research | Technology and Production Sciences, reference DFF-7017-00065 (to AY & SMM) is gratefully acknowledged. AY further acknowledges financial support from the Danish Natural Sciences Research Council (DanScatt) for SAXS experiments. The authors are grateful to the beamline scientist Dr. Heinz Amenitsch (Institute of Inorganic Chemistry, Graz University of Technology) for the technical support at the Austrian SAXS beamline (ELETTRA, Trieste, Italy). They thank also Tillman Pape (Core Facility for Integrated Microscopy, University of Copenhagen) for the technical assistance with cryo-TEM imaging, and Dr. Tom André Jos Vosh and Cecilia Cerretani (Department of Chemistry, University of Copenhagen) for their support and technical assistance with quantum yield measurements. The authors acknowledge the CERIC-ERIC Consortium for the access to experimental facilities and financial support.