Dioleoyl phosphatidylserine (DOPS, ≥99%) was obtained from Avanti Polar Lipids Inc. (U.S.) and soybean phosphatidylserines (SPS) of varying purities (Table 1) were provided by the Lipoid GmbH (Germany). TRIZMA preset-crystals pH 7.4, calcium chloride dihydrate (Ph.Eur.), sodium azide (≥99.5%), EDTA tetrasodium dihydrate (>99%) and chloroform (≥99.0%, Ph.Eur., stabilized with ∼1% ethanol) were obtained from Sigma, sodium chloride (Ph.Eur.) from VWR and Laurdan (6-dodecanoyl-2-dimethylaminonaphtalene) from Molecular Probes (Thermo-Fisher Scientific). Purified water was obtained from a Milli-Q Advantage A10 system (Millipore).

Liposomes (20 mg/ml lipid) were prepared either by the lipid-film method or by directly dispersing the lipid in buffer under mechanical agitation (Heidolph Multi-reax set to 500 rpm; Heidolph Instr., Germany) overnight. If not stated otherwise, 10 mM Tris buffer pH 7.4 preserved with 0.02% (w/v) sodium azide (Tris buffer) was used. To facilitate dispersion of the SPS lipids containing calcium ions, an adequate amount of EDTA was added to the buffer (Table 2). The crude liposome dispersions were submitted to bath sonication (35 kHz, Bandelin Sonorex Digitech, 2–6 cycles à 15 min, Bandelin electronic GmbH & Co. KG, Germany) to obtain suspensions of small unilamellar vesicles. The liposome suspensions were stored at 4–8°C until use. Specifications of the liposomes used in this study are provided in Table 2.

The different methods used for cochleate preparation in this study are schematically presented in Scheme 2.

Vesicle size of the liposome suspensions was determined by DLS (DelsaMax Pro, Beckman Coulter Life Science, U.S.). The diluted (1:1,000 in Tris buffer) liposomes were measured 6 times over 10 s at 20°C in backscattering mode (163.5°). The hydrodynamic diameter (z-average) and polydispersity index (PdI) were calculated by the instrument's cumulant analysis (DelsaMax version 1.0.1.6. Beckman Coulter). Results given as average and standard deviation of the six acquisitions.

Small-angle x-ray diffraction patterns were recorded at room temperature with a SAXSess mc2 instrument (Anton Paar GmbH, Austria, x-ray wavelength: 0.154 nm, CCD-SCX 4,300 detector) using a flow-through capillary. Each sample was measured 50 times over 30 s (1,500 s in total) and desmearing of the raw data was performed with SAXSquant software. The lamellar repeat distance (d-spacing) was calculated from the first order reflection according to Bragg's equation with λ the wavelength of the x-rays (0.154 nm, Cu Kα) and θ the scattering angle:(1)d=λ2sinθIn the figures, the scattering intensities are plotted against the scattering vector s with s = 2sinθ/λ = 1/d.

10 µl 0.1 mM Laurdan solution in anhydrous ethanol was mixed with 500 µl diluted sample (0.5 mg/ml lipid) and equilibrated under light protection on a shaker (IKA Vibrax IKA GmbH & Co. KG, Germany, 250 rpm). Fluorescence emission spectra were recorded with a Cary Eclipse instrument (Varian, Agilent, U.S.) at room temperature from 380 nm to 600 nm and an excitation wavelength of 370 nm. Generalized polarization (GP) values were calculated taking the fluorescence intensities at 430 nm and 490 nm into account:(2)GP=I430−I490I430+I490

To evaluate the general feasibility to prepare cochleate suspensions from naturally derived lipids with similar well-organized lipid organization as found in DOPS cochleates, the soy-PS with highest head-group purity (≥96%; SPS-10) was applied. Cochleate suspensions were prepared by the standard trapping method (10 mg/ml lipid, lipid/calcium ion molar ratio of 1:1).

Addition of calcium ions to the liposome suspensions resulted in immediate flocculation for both lipids. In the DOPS suspension, the typical cochleate structures with high aspect ratio were observed in SEM in addition to more spherical particles (Figures 1A–C). In good agreement with literature (3, 13), some of the cochleate cylinders had an inner water channel (marked with arrows in Figure 1C). In the SPS-10 suspensions, the particles were considerably smaller, but the elongated particles indicate the formation of cochleate cylinders (Figures 1D–F). As in the DOPS suspension, the cochleate cylinders were coexisting with more spherical particles. Due to the solid nature of the particles, the formulations could directly be visualized in TEM (e.g., direct analysis of the dried sample) where they had a rather similar appearance as observed in cryo-TEM (Figure 2). The highly organized lamellar internal structure of the cochleate particles could be visualized for both DOPS (Figures 2A–C) and SPS-10 (Figures 2D–F) cochleates. The lamellar repeat distances of the lipid lamellae were 5.1 ± 0.1 nm (n = 11) and 4.8 ± 0.1 nm (n = 13) for DOPS and SPS-10 cochleates, respectively. Importantly, the tightly packed lamellar structure could also be visualized in the more spherically shaped particles (marked in Figures 2C,F). To get more quantitative information about the lipid dehydration, Laurdan GP analysis was carried out. As expected, the Laurdan emission spectra of DOPS and SPS-10 liposomes were bimodal (Figure 3A, dotted lines) resulting in slightly negative GP values between −0.1 and −0.2, as typically for membranes in the fluid state. Binding of calcium ions results in dehydration of the phospholipid headgroups and formation of an anhydrous multilamellar structure. Accordingly, a distinct change in the Laurdan emission spectra with an emission maximum around 430 nm was observed (Figure 3A, closed lines). Similar GP values around 0.5 were measured for both DOPS and SPS-10 cochleates indicating a similar degree of dehydration of the lipid bilayers. SAXD results confirmed the well-organized lamellar structure of the lipid particles (Figure 3B, closed lines) by the presence of a very sharp 1st order reflection. Even the much weaker reflections of 2nd and 3rd order (marked with arrows in Figure 3B) could be detected. Based on the position of the reflections, the thickness of the lipid lamella (d-spacing) was determined to be of 5.1 and 4.9 nm for the DOPS and SPS-10 cochleates, respectively. In contrast, only weak and broad reflections were measured for the corresponding liposome suspensions (Figure 3B, dotted lines). Interestingly, the reflection was more distinct for SPS-10 than DOPS liposomes which may be explained by the presence of residual calcium ions in the starting lipid. As the calcium salt of PS cannot be dispersed in an aqueous medium, some crystalline lipid material will be present in the liposome suspension. This is likely also the reason for the larger diameter and PdI in DLS measurements of these liposomes (Table 2). This explanation is supported by the observation that addition of small amounts of EDTA upon liposome preparation resulted in a distinctly smaller size and similar PdI values as measured for the other vesicle suspensions (Table 2).

Despite the differences in particle shape, the results clearly indicate that cochleates with similar structural lipid organization as in DOPS cochleates can be prepared from naturally derived phosphatidylserines with high headgroup purity. Similarly as DOPS cochleates, the SPS-10 cochleate suspensions were physically stable (particle structure and redispersibility) for at least half a year.

DOPS and SPS-10 cochleate suspension were also prepared by controlled mixing of equal volumes of liposomes and calcium chloride solution at a mixing speed of 80 µl/s where suspensions with similar properties were obtained (Supplementary Figure S3). All further samples were thus prepared by controlled mixing at 80 µl/s if not stated otherwise.

In the next step, a range of naturally derived phosphatidylserines of varying purities (PS between 53 and 96%) and salt forms (sodium or calcium salt or a mixture of both, Table 1) were screened for their ability to form cochleates. To facilitate dispersion of the lipid, EDTA was added in an adequate amount (Table 2) and the concentration of calcium chloride solution used for cochleate preparation was adjusted accordingly to reach lipid/calcium ion molar ratios of about 1:1 or 1:2 for the sodium (SPS-10, SPS-13, SPS-14) and calcium salts (SPS-15, SPS-19, SPS-34), respectively (Table 2).

All samples showed immediate flocculation upon mixing with calcium ions, but the characteristic cochleate structures (e.g., cochleate cylinders) could not be seen in SEM (Supplementary Figure S4). To get more information about the inner structure of the particles, selected suspensions (SPS-10, SPS-13, SPS-15 and SPS-19) were stained with osmium tetroxide, embedded in epoxy resin and thin sections were then viewed in the TEM (Figure 4). The overall appearance (images on the left) was similar as observed for DOPS cochleates in a previous study (3). Well-organized lamellar structures could clearly be visualized at higher magnification allowing an estimation of the lamellar repeat distance (see also Supplementary Figure S5). The average lamellar repeat distances were between 4.2 and 4.4 nm (n = 4), e.g., somewhat smaller than those determined by TEM original samples and SAXD (4.9 nm). In good agreement with this results, considerably high Laurdan GP values were determined (Figure 5A) with a trend of decreasing GP with decreasing amount of negatively charged lipids. This trend was very clear for the suspensions prepared from the sodium salts (100%, 88%, 85% and 75% of total negatively charged lipids for SPS-10, SPS-13, SPS-16 and SPS-14, Figure 5A). High GP values were also obtained for the SPS-19 and SPS-34 suspensions, which, however, contained a higher amount of calcium ions (lipid/calcium ion molar ratio about 1:2). Moreover, these lipids likely will also contain other (not specified) negatively charged lipids, which may contribute to the high degree of lipid dehydration and organization. In all suspensions, the characteristic SAXD reflection was detected, and the lamellar repeat distance (d = 4.9 nm) was similar for all suspensions. The reflections were, however, broader and less intensive as shown for selected formulations in Figure 5B.

From a practical and economic point of view, direct processing of the water-insoluble calcium salt of PS to prepare the cochleate suspensions is of interest and has been evaluated in this study. Cochleate suspensions were prepared from the most promising SPS calcium salt from the screening experiments (SPS-19). To disperse the water-insoluble calcium PS in the aqueous buffer for liposome preparation, EDTA was added to the buffer (Table 2). Cochleate formulations with different lipid/calcium ion molar ratios (1:1, 1:2, 1:5 and 1:10) were then prepared to determine the optimal composition with respect to homogeneity (aggregate size) and degree of dehydration (high GP values). In addition, a suspension was also prepared by dialyzing liposomes directly against a calcium chloride solution to remove the EDTA, which initially was added for liposome preparation. All cochleate suspensions were prepared in triplicate.

Independently on the lipid/calcium ion ratio, flocculated suspensions were obtained and the characteristic SAXD reflection (d-spacing 4.9 nm) was detected in all suspensions shortly after preparation (not shown). As expected, the aggregation tendency (aggregate size) increased with increasing amounts of added calcium chloride. There was also an increase in GP with increasing lipid/calcium molar ratio (Figure 6A) up to a ratio of 1:5 reaching then GP values between 0.4 and 0.5 (Figure 6A). However, GP values declined distinctly during storage in all samples prepared by direct mixing (Figure 6A). It can be speculated that the EDTA in the suspension interferes with the particle structure over time. To circumvent this problem, cochleate suspensions were prepared by dialysis. The process of lipid dehydration during dialysis could clearly be followed by the increasing Laurdan GP values and a plateau was reached after about 1 h (Figure 6B). Most importantly, the GP values did not decrease distinctly upon storage (Figure 6A) indicating improved stability compared to the samples prepared by direct mixing. Remarkably, the SPS-19 cochleate suspensions prepared with lipid/calcium chloride molar ratio of 1:5 and by dialysis had a rather similar morphology as those prepared from the purer SPS-10 (Figure 7). Importantly, some cochleates with the typical cylindrical particle shape could be detected (marked with arrows in Figures 7B,D).