2.1. Sample preparation

The HSPC (L-α-phosphatidylcholine) liposome powder used consists of phospho­lipids of 1,2-dipalmitoyl-sn-glycero-3-phospho­choline (DPPC) and 1,2-distearoyl-sn-glycero-3-phospho­choline (DSPC) in the molar ratio 1:8. The powder was mixed with cholesterol, sucrose and PEGylated lipid 1,2-distearoyl-sn-glycero-3-phospho­ethano­lamine-N-[meth­oxy­(poly­ethyl­ene glycol)-2000] (mPEG2000-DSPE) with molar ratios HSPC:mPEG2000-DSPE:cholesterol = 9:1:x (with x > 4). The mixed sample powder was used in liposome solution preparation of 10 mM HSPC for AF4/MALS/DLS/dRI and SAXS–WAXS measurements (Dominik et al., 2020). Sample solutions of 10 mM HPSC, without the addition of cholesterol, were also prepared with a similar molar composition of 1 mM DPPC:8.0 mM DSPC, and 1 mM mPEG2000-DSPE, without cholesterol, in co-extrusion processing as previously reported (Yang et al., 2019; Mineart et al., 2017).

2.2. AF4–MALS measurements

A Wyatt Eclipse DualTec system for AF4 was connected to a Wyatt-DAWN MALS spectrometer (with 18-angle light scattering detectors) and an Optilab dRI detector for determination of mass and radius of gyration R_g; one of the 18 MALS detectors was replaced by a DLS device to determine the hydro­dynamic radius R_h of the liposomes (Fig. 1). Sample solutions of 2–30 µl were injected into the AF4 system and measured using a trapezoidal 265 mm-long channel of an RC 10 kDa cut-off membrane and a spacer for a channel height of 350 µm at 293 K. The AF4 parameters used are summarized in Table S1 of the supporting information. The refractive-index increment dn/dc = 0.146 ml g⁻¹, used for deducing the mass of the PEGylated HSPC liposome, was determined from a separate measurement, with the integrated area of the elution profile of the sample solution with the AF4 channel path bypassed (to avoid loss of sample). Assuming 100% mass recovery, the ASTRA program (WYATT Technology) was employed to calculate the dn/dc value from the integrated elution peak area and prescribed sample weight. Details of AF4 analysis were reported previously for liposome characterization (Écija-Arenas et al., 2021).

Figure 1 (a) Integrated AF4-MALS system, comprising an autosampler in the beginning and the AF4 of Wyatt Eclipse DualTec, followed by UV–vis absorption, MALS, DLS and dRI spectrometers, and terminated with a fraction collector. (b) Programmed elution-rate profile over the AF4-MALS elution (∼70 min) of the liposome sample solutions. The AF4 flow parameters are summarized in Table S1.

2.3. Small- and wide-angle X-ray scattering

SAXS and WAXS (SWAXS) measurements were performed at the 13 A BioSWAXS beamline of the Taiwan Photon Source at the National Synchrotron Radiation Research Center. The SWAXS data were collected with an X-ray beam energy of 15.0 keV (or wavelength λ = 0.8266 Å) using the two synchronized in-vacuum detectors Eiger X 9M (SAXS) and 1M (WAXS) of the beamline positioned at sample-to-detector distances of 2500 and 180 mm, respectively. The scattering vector magnitude q = 4πλ⁻¹sinθ (with the scattering angle 2θ) and the projection angles of the WAXS detector plane were calibrated using a mixed powder of silver behenate and lanthanum hexaboride (LaB₆). The absolute intensity (in cm⁻¹) was calibrated using water scattering intensity (Shih et al., 2022). The sample solutions were sealed in thermostated quartz capillaries (2 mm diameter and 20 µm wall thickness) and measured at 25, 40, 50 and 70°C. SAXS data were analyzed using the five-layer model of sharp scattering-length-density (SLD) interfaces, known as the core–multishell model, available in the SASView software platform (https://www.sasview.org/). The X+ software with available Gaussian electron density profiles was also used in SAXS data analysis (Ben-Nun et al., 2010).

3.1. AF4-MALS results

The AF4/MALS/DLS/dRI results for the HSPC liposomes with cholesterol are shown in Fig. 2(a), revealing the number-average particle mass M_n of 118 kDa, with M_w/M_n = 1.0 (with weight-averaged mass M_w). Also shown in Fig. 2(a) are the deduced hydro­dynamic radius R_h and radius of gyration R_g from the DLS and MALS–dRI data, respectively. Dividing the elution mass concentration by the particle mass deduced (Fig. 2) leads to the number density of the liposomes [Fig. 2(b)] as a function of R_h; the result reveals a distribution peak at R_h = 52.4 nm and a polydispersity of ca 10% (Parot et al., 2020). Fig. 2(c) presents the R_g versus R_h plot to illustrate the Burchard–Stockmayer shape factor S_f = R_g/R_h (Mukherjee & Hackley, 2018) for the liposome. This falls close to the line of S_f = 1, corresponding to an ideal thin spherical shell structure. Nevertheless, the average S_f value (1.05) deduced is slightly above unity, which can be attributed to possible deformations of the liposome shape from a thin spherical shell under the asymmetric flow field of AF4. We note that from the known equation R_g² = (3/5)(R₁⁵ − R₂⁵)/(R₁³ − R₂³) for core–shell spheres (Feigin & Svergun, 1987), of core and shell radii R₁ and R₂, it can be deduced that R_g reduces to (3/5)^1/2R_h for solid spheres (i.e. R₂ = 0) and R_g ≃ R_h for thin spherical shells with R₂ ≃ R₁; namely, the shape factor R_g/R_h = (3/5)^1/2 of solid spheres is smaller than that (≃ 1) of thin spherical shells. Further, it can be deduced that the values R_g² = 1/5(a² + 2b²) of ellipsoids (with the semi-major and semi-minor axes a and b) of a common volume have a minimum with a = b for spheroids. Therefore, the measured shape S_f = 1.05 for the liposomes suggests possible deformation of the liposomes from the ideal spherical shape (S_f = 1).

Figure 2 (a) Evolution of the liposome concentration, particle mass, Rg and Rh measured over the AF4 flow of the sample solution of PEGylated HSPC liposome with cholesterol, with Rg and Rh deduced from the MALS and DLS data, respectively. The mass was deduced from the combined analysis of MALS and dRI data. (b) Derived number-density distribution of the liposome as a function of Rh, fitted with a Gaussian profile (solid curve). (c) Rg versus Rh presentation. Data are fitted with a solid line (slope Rg/Rh = 1.05). Also shown is a red dashed line (slope = 1.0) representing the shape factor of ideal shells with Rg/Rh = 1.

3.2. Liposome membrane bilayer structures

Shown in Fig. 3(a) are the integrated SAXS–WAXS data of the PEGylated HSPC liposome with cholesterol added, revealing a characteristic broad hump centered around q ≃ 0.12 Å⁻¹ from the typical vesical bilayers of ca 5 nm thickness. Also observed is an additional [compared with the SAXS data for the neat liposome without cholesterol; Fig. 3(b)] peak at q ≃ 0.05 Å⁻¹ associated with the addition of cholesterol. Correspondingly, the broad hump centered around q ≃ 1.5 Å^−­1 in the high-q region [Fig. 3(a)] indicates a significantly relaxed alkyl chain packing due to the intervening cholesterol. In contrast, the neat HSPC without cholesterol exhibits a relatively sharp peak at a similar q position, revealing a 2D hexagonal-like packing of the phospho­lipids with a Bragg d spacing of 4.2 Å (Geisler et al., 2020; Sreij et al., 2019).

Figure 3 (a) SWAXS data of the PEGylated HSPC liposomes with cholesterol added. (b) Corresponding neat HPSC liposomes without cholesterol. The SAXS data are fitted using a five-layer core–multishell model (dash–dotted curves) with core radii of (c) 530 Å and (d) 428 Å and a multilayer model with five Gaussian electron density profiles (solid curves). In (c) and (d), the units of relative electron density Δρ (with respect to the water solvent) are used in the Gaussian interface model, with zero representing the absolute electron density of water (0.334 e− Å−3). (e) Cartoon of the local structure of the PEGylated HSPC liposome, with z = 0 for the center of the bilayer. PtP represents the lipid head-to-head distance of the bilayer. The thin arrows in the WAXS region of (a) and (b) indicate the characteristic 2D hexagonal packing of the bilayer lipids of the liposomes.

To reveal the detailed bilayer structure of the liposomes, we fitted the SAXS data with a core–multishell model, having a five-layer SLD profile with sharp interfaces (Yang et al., 2019; Mineart et al., 2017); a multilayer model comprising five Gaussian electron density profiles for smooth density transitions across the sublayer interfaces (Schilt et al., 2016; Ben-Nun et al., 2010) is also used to fit the same sets of data. The five-layer core–multishell model comprises the central alkyl-dominated zone sandwiched by the head-group sublayers of the phospho­lipids, which are further sandwiched by two outer PEGylated layers, as illustrated in Fig. 3(e). As shown in Figs. 3(a) and 3(b), the SAXS data are better fitted in the higher-q region (>0.2 Å⁻¹) using the asymmetric Gaussian electron density profiles compared with the core–multishell SLD profiles [Figs. 3(c) and 3(d)]; nevertheless, both models could fit the lower-q data equally well down to ∼0.01 Å⁻¹, with qualitatively consistent electron density profiles. We note that the asymmetry in the electron density profile revealed consistently from both models is crucial in the data fitting. We also attempted a seven-layer core–multishell model fitting by adding an additional thin layer to the center of the lipid tail region; the fitting result, however, reduces to that of the five-layer model.

Figs. 3(c) and 3(d) illustrate the best-fitted asymmetric Gaussian electron density profiles for the HSPC liposome bilayer (Su et al., 2013, 2018), which is sandwiched by two PEGylated layers each of ca 45 Å thickness. We attribute the higher electron density sublayers dominated by the phospho­lipid heads and the mPEG lipids in Figs. 3(c) and 3(d) to the inner leaflet of the bilayers. Presumably, the inner leaflet of the liposome bilayer, owing to its smaller shell radius (hence smaller shell area), might have tighter packing of the phospho­lipids and mPEG chains, resulting in sublayers of higher electron density. In contrast, the outer leaflet of a larger shell radius and facing open solvent tends to have more broadened peaks of lower electron density. On cholesterol intercalation, all the characteristic density peaks of the inner and outer leaflets of the bilayer [Fig. 3(c)] are broadened from that of the neat HSPC liposomes [Fig. 3(d)], leading to an enlarged bilayer thickness and peak-to-peak (PtP) distance (between the two phospho­lipid head sublayers of the inner and outer leaflets). Consistently, the thicker cholesterol-intercalated bilayer, with presumably larger bending modulus, is found to have larger liposome sizes as shown in Fig. S2 of the supporting information. These results suggest a significant association of the cholesterol with the alkyl chain zone. We note that the cholesterol–lipid interactions affect the global and local ordering of the bilayer concomitantly, as revealed from the nearly collapsed scattering hump at q ≃ 0.4 Å⁻¹ (Su et al., 2013) and the much broadened hump at q ≃ 1.5 Å⁻¹ (the characteristic peak that represents the 2D hexagonal packing of the gel phase) from that observed for the neat HSPC liposome (Geisler et al., 2020; Sreij et al., 2019). Similar deteriorations in the scattering features are also consistently observed with the temperature-dependent SWAXS data of the pure PEGylated HSPC liposomes (Fig. S1), when the sample temperature increased from 25°C (gel phase) to 70°C (fluid phase).

3.3. Cholesterol effect on liposome chain packing revealed by WAXS analysis

Neat HSPC liposomes were reported to have pre- and main gel-ordered-to-fluid-disordered phase transition temperatures at T_pre = 47.8°C and T_m = 53.6°C, respectively (Kitayama et al., 2014). Shown in Fig. 4 are the WAXS data measured at 25, 40 and 70°C for the HSPC liposomes, with and without cholesterol. The neat PEGylated HSPC liposome manifests a primary sharp peak of the 2D hexagonal packing centered at q₂ = 1.517 Å⁻¹ at 25 and 40°C, which is significantly reduced at 70°C [Fig. 4(b)]. The corresponding coherent length L_c ≃ 2π/Δq, deduced from the q₂ peak width Δq, increases from 117 Å at 25°C to 163 Å at 40°C, showing a pre- to main ordering behavior similar to that mentioned previously. At 70°C, the q₂ peak deteriorates significantly for a much reduced L_c = 36 Å, illustrating the gel-to-fluid phase transition. From the q₂ peak position, the deduced lipid–lipid d spacing D = 2π/q₂ of the neat HSPC liposome changes from 4.14 to 4.17 to 4.34 Å, for 25, 40 and 70°C; the corresponding area per lipid A_L estimated from the 2D hexagonal packing with 16π²/( 3^1/2q₂²) (Geisler et al., 2020) increases from 39.6 to 40.2 to 43.5 Ų. The deduced feature sizes are summarized in Table 1.

Figure 4 (a) Temperature-dependent WAXS data of the PEGylated HSPC liposome with the incorporation of cholesterol. (b) Parallel WAXS data for the PEGylated HPSC liposome without cholesterol. (c) Deconvoluted scattering humps from that shown in (a), with a common background subtracted. (d) Evolution of the three peak positions shown in (c), as indicated.

In contrast, the WAXS data measured at 25°C for the HPSC liposome with cholesterol exhibit a convoluted broad hump that can be decomposed into three broad humps centered at q₁ = 1.234 Å⁻¹, q₂ = 1.497 Å⁻¹ and q₃ = 1.788 Å⁻¹; all three humps have similar small L_c values of 16–17 Å. We assign the q₂ peak observed for the PEGylated HSPC liposomes with cholesterol to a deteriorated 2D hexagonal packing of the lipid–cholesterol complex. We notice that the q₂ hump position of the liposome with cholesterol shifts to lower q values as the temperature increases from 25 to 40 to 50 to 70°C. The values deduced for A_L with cholesterol A_L-chol are also larger than those of the corresponding liposome without cholesterol, as shown in Table 1, especially in the fluid phase. The result suggests that cholesterol can interact substantially with the lipid chains, especially in the fluid phase of reduced lipid self-interactions.

We notice that q₁ and q₃ show little or no peak position shifting following the temperature changes, and the q₁ peak even disappears at 50°C [Fig. 4(a)]. It is likely that these two peaks are associated with the packing of excess cholesterol phase segregated from the 2D hexagonal domains of the lipid–cholesterol complex, or an additional orthorhombic-like packing as suggested in previous reports (Sreij et al., 2019; Geisler et al., 2020). To further clarify the origin of these two peaks, we measured cholesterol-concentration-dependent SWAXS (Fig. S3). The results indeed show that the q₁ and q₃ peaks emerge with higher cholesterol content roughly above the molar ratio HSPC:mPEG:cholesterol = 9:1:2 (i.e. 20% cholesterol). The q₁ peak of lower thermal stability may be associated with the 2D monolayer packing of cholesterol (Rapaport et al., 2001); the less temperature-dependent q₃ peak, however, may be of different origin.