Transverse lipid organization dictates bending fluctuations in model plasma membranes

Brett W. Rickeard a, Michael H. L. Nguyen a, Mitchell DiPasquale a, Caesar G. Yip a, Hamilton Baker a, Frederick A. Heberle bc, Xiaobing Zuo d, Elizabeth G. Kelley e, Michihiro Nagao ef and Drew Marquardt *ag
aDepartment of Chemistry and Biochemistry, University of Windsor, Windsor, ON, Canada. E-mail:
bUniversity of Texas Health Science Center, Houston, TX, USA
cThe Bredesen Center for Interdisciplinary Research and Graduate Education, University of Tennessee, Knoxville, TN, USA
dX-ray Science Division, Argonne National Laboratory, Lemont, IL, USA
eNIST Center for Neutron Research, National Institute of Standards and Technology, Gaithersburg, MD, USA
fCenter for Exploration of Energy and Matter, Department of Physics, Indiana University, Bloomington, IN, USA
gDepartment of Physics, University of Windsor, Windsor, ON, Canada

Received 16th September 2019 , Accepted 8th November 2019

First published on 12th November 2019

Membrane undulations play a vital role in many biological processes, including the regulation of membrane protein activity. The asymmetric lipid composition of most biological membranes complicates theoretical description of these bending fluctuations, yet experimental data that would inform any such a theory is scarce. Here, we used neutron spin-echo (NSE) spectroscopy to measure the bending fluctuations of large unilamellar vesicles (LUV) having an asymmetric transbilayer distribution of high- and low-melting lipids. The asymmetric vesicles were prepared using cyclodextrin-mediated lipid exchange, and were composed of an outer leaflet enriched in egg sphingomyelin (ESM) and an inner leaflet enriched in 1-palmitoyl-2-oleoyl-phosphoethanolamine (POPE), which have main transition temperatures of 37 °C and 25 °C, respectively. The overall membrane bending rigidity was measured at three temperatures: 15 °C, where both lipids are in a gel state; 45 °C, where both lipids are in a fluid state; and 30 °C, where there is gel-fluid co-existence. Remarkably, the dynamics for the fluid asymmetric LUVs (aLUVs) at 30 °C and 45 °C do not follow trends predicted by their symmetric counterparts. At 30 °C, compositional asymmetry suppressed the bending fluctuations, with the asymmetric bilayer exhibiting a larger bending modulus than that of symmetric bilayers corresponding to either the outer or inner leaflet. We conclude that the compositional asymmetry and leaflet coupling influence the internal dissipation within the bilayer and result in membrane properties that cannot be directly predicted from corresponding symmetric bilayers.

1. Introduction

Eukaryotic plasma membranes (PM) are characterized by an asymmetric distribution of lipids between the exoplasmic and cytosolic leaflets of the bilayer.1–3 This compositional distinction has evolved such that high-melting and uncharged lipids predominantly comprise the outer leaflet, while low-melting and negatively charged lipids form most of the inner leaflet. Specifically, healthy mammalian cells sequester nearly all of the high-melting sphingomyelin (SM) in the outer leaflet of the PM, while phosphatidylethanolamine (PE) and negatively charged lipids are almost completely segregated into the inner leaflet.4,5 Consequently, this lipid composition difference contributes to large differences in the physicochemical properties of the two layers. Despite the ubiquity of membrane asymmetry in living cells, there is still much to be learned about the importance of this lipid organization and how it impacts the structure, dynamics, and ultimately the function of the PM.

Most biologically occurring lipids intrinsically form spherical soft bilayers whereby elastic constants can be assigned and used to characterize the mechanical properties of the membrane. For example, the bending modulus κ quantifies the membrane rigidity, which governs the membrane fluctuations on the nanoscale6,7 and has profound biological relevance. Thermally-induced membrane undulations influence numerous biological processes including the interactions and insertions of membrane proteins,8–11 the size and morphology of lipid rafts,12–14 and transport vesicle formation,15–17 to name but a few.

The near ubiquity of asymmetric membranes in nature implies that symmetric bilayers are not adequate for characterizing PM properties. Still, the vast majority of research into membrane mechanical properties has made use of symmetric model membranes, in part owing to the experimental challenges posed by the preparation of asymmetric vesicles. Moreover, asymmetry contributes additional dynamic processes such as transverse lipid diffusion (flip-flop) that can couple to mechanical properties and thereby complicate the analysis of experimental data.18 Recent theoretical efforts examined the effects of asymmetry on the membrane dynamic properties including area compressibility modulus and membrane viscosity, and found that lipid asymmetry couples to the dynamic modes and relaxation times.19 However, within this theoretical framework, the dispersion relations for an asymmetry in bending modulus were deemed non-trivial and thus not reported. Experimental measurements of bending rigidity in asymmetric bilayers will better inform such theoretical treatments, and may provide information about the strength of interleaflet coupling, i.e. the effects that the two leaflets impose on one another. To this end, we investigated the effects of compositional asymmetry on the bending modulus of model asymmetric plasma membranes using neutron spin-echo (NSE). NSE directly measures the effective bending modulus, which contains the effects of internal bilayer dissipation, and is the only technique capable of resolving these effects. These internal dissipations are likely highly sensitive to the differences in leaflet properties present in asymmetric bilayers, but this correlation still requires elucidation and thus necessitates the present study.

The asymmetric composition of the mammalian PM suggests that its two leaflets will have different mechanical properties. In symmetric model membranes, ordered lipids such as those found in the PM outer leaflet typically form more rigid bilayers, while the more disordered lipids found in the PM inner leaflet tend to make softer bilayers.20 How these differences manifest in an asymmetric bilayer depends on the extent of coupling between the leaflets with respect to their differing physicochemical and mechanical properties (termed interleaflet coupling), and possibly other factors. This area of research is largely unexplored. Our experimental strategy was to prepare model membranes that mimic the ordered/disordered lipid asymmetry of a natural PM and measure their membrane structure and mechanical properties. The strength of interleaflet coupling can then be inferred by comparing these properties with symmetric bilayers of the corresponding inner and outer leaflet compositions.

2. Results

Asymmetric large unilamellar vesicles (aLUVs) ≈ 200 nm in diameter were prepared by introducing ESM into the outer leaflet of POPE vesicles using cyclodextrin-mediated lipid exchange.21–24 As ESM, and not POPE, possesses a choline group, these aLUVs can be interrogated with NMR to determine the transbilayer lipid distribution, as described below. Diminished bilayer integrity promotes lipid flip-flop,25 which can reduce the asymmetry and potentially modify membrane bending dynamics. To ensure vesicle integrity we rigorously monitored leaflet composition at different stages of sample preparation and data collection. Small-angle neutron scattering (SANS) measurements showed a vesicle form factor at low scattering wave vector transfer, Q, values for both the acceptor POPE vesicles and final aLUVs (Fig. S2), indicating that neither sample preparation conditions nor trace amounts of residual cyclodextrin altered the vesicle structure. To augment the SANS measurements, we used dynamic light scattering (DLS) to periodically check the vesicle size, and found no significant changes during aLUV preparation. All symmetric LUV controls were subjected to the same wash steps and D2O exchanges as the aLUV.

Asymmetric lipid distributions in aLUVs were quantified from solution 1H-NMR spectra measured in the presence of Pr3+, an extravesicular paramagnetic lanthanide ion. Externally added Pr3+ does not permeate into the vesicle interior on the timescale of the NMR measurement21,26 and therefore interacts only with outer leaflet lipid headgroups to induce a downfield shift of the ESM choline resonance. We note that the ethanolamine of POPE does not contribute to the 1H-NMR signal.24 The observed choline signal is a superposition of shifted and unshifted resonances whose relative areas are proportional to the amount of ESM in the outer and inner leaflets, respectively (Fig. 1).26 Immediately prior to SANS and NSE measurements, aLUVs whose outer leaflets were enriched in ESM (i.e., POPEin/ESMout), exhibited an unequal area ratio (Fig. 1A). Further, the lipid distribution in symmetric vesicles was assessed using 1H-NMR to demonstrate that the spontaneous molecular curvature difference between ESM and POPE does not generate spontaneous compositional asymmetry.

image file: c9nr07977g-f1.tif
Fig. 1 (A) 1H-NMR of aLUV in the presence of the shift reagent Pr3+. The red peak represents the protected choline groups on the inner leaflet, the green peak is the outer leaflet choline groups, and the grey peaks arise from other groups on the lipid and residual CD peaks (3.6 and 3.8 ppm). (B) 31P-NMR of the aLUV to determine the ratio of ESM (orange) and POPE (blue) present. (C) Cartoon representation of the aLUV generated. The blue presents POPE lipids, and orange represents ESM, highlighting their distribution. (D) SAXS (main) and SANS (inset) data of the aLUV at 30 °C are represented by circles, and the jointly optimized model fits are represented by solid lines. (E) The resultant electron density (ED) profile calculated from the optimized models from D. The black curve is the total ED which is the sum of the terminal methyl groups of the hydrocarbon chains (CH3), the bulk of the hydrocarbon (CH + CH2), the headgroup (HG) and the water.

Fig. 1B shows a 31P-NMR spectrum which corresponds to the overall distribution of ESM and POPE in the aLUVs. The mol fractions of ESM and POPE were determined to be 0.25 and 0.75, respectively. Combining results from the 1H-NMR and 31P-NMR spectra resulted in the overall lipid distribution of the vesicular bilayers. The aLUV outer leaflet contained an ESM mole fraction of 0.44 (χESM = 0.44) indicating partial replacement of outer leaflet POPE, with a small amount of ESM (χESM = 0.06) translocating to the inner leaflet during the exchange step (Table S1). The cartoon illustration of the aLUV composition is shown in Fig. 1C. This composition remained stable for the duration of the NSE measurements as determined by 1H-NMR, suggesting little to no flip-flop over the course of 3 days. The observed temporal stability of ESM asymmetry is consistent with a recent report of slow flip-flop in both isotopically25 and chemically24,27 asymmetric vesicles.

The asymmetric bilayer structure was determined as a function of temperature by modeling small-angle neutron and X-ray scattering data.23,24 A typical example of the data, fit and X-ray electron density profile are shown in Fig. 1D and E. Scattering profiles were well fit by a model in which ESM is enriched in the outer leaflet and POPE in the inner leaflet and did not change over the course of 24 hours, providing additional evidence for aLUV stability. Bilayer structural parameters, for both LUVs and aLUVs, are summarized in Tables S2 and S3. We note that structural parameters of the symmetric LUVs follow the expected temperature and composition dependence. For example, the bilayer thickness (DB) decreases with increasing temperatures and DB increases with ESM content.

NSE was used to measure the bending dynamics of aLUVs as well as a set of symmetric binary ESM/POPE mixtures at various temperatures, allowing us to observe changes in the membrane stiffness with respect to the bilayer composition and temperature. NSE is well suited for measuring collective dynamic properties of lipid bilayers compared to other techniques (Fig. 2 top). The normalized intermediate scattering function I(Q,t)/I(Q,0) vs. Fourier time for aLUVs at 30 °C is shown in Fig. 2. The measured dynamics for the aLUVs follow the same scaling seen in the symmetric vesicles in this work and those reported in literature,7,28–30 supporting that the same bending fluctuations are being measured in the aLUVs as those in symmetric models. The data follow a stretched exponential as predicted by Zilman and Granek for membrane bending fluctuations based on Helfrich's model that treats the membrane as a thin elastic sheet,

image file: c9nr07977g-t1.tif(1)
with the fits shown as the solid lines.31 The corresponding decay rates, ΓZG, follow the expected Q3-dependence shown in the inset in Fig. 2 with a slope that is inversely related to the effective bending modulus [small kappa, Greek, tilde] by,
image file: c9nr07977g-t2.tif(2)
where η is the solvent viscosity, T is the absolute temperature, and kB is the Boltzmann constant.

image file: c9nr07977g-f2.tif
Fig. 2 Top: Accessible length and time scales and corresponding energy and momentum transfer (Q), for some spectroscopic techniques covering nanoscopic to macroscopic dynamics, covering a number of membrane dynamics. Figure adapted from ref. 34. Bottom: Normalized intermediate scattering function I(Q,t)/I(Q,0) measured by NSE for the aLUV in D2O at 30 °C. The inset shows the linear dependence of the relaxation rate (ΓZG) with respect to Q3. Error bars represent one standard deviation here and throughout the manuscript.

Watson and Brown extended the theory used to treat NSE data to account for the fact that the membrane is not a thin structureless sheet and instead has a finite thickness by incorporating the effects of internal dissipation within the bilayer.32,33 They showed that the effective bending modulus measured by NSE is related to the intrinsic bending modulus, κ, through

[small kappa, Greek, tilde] = κ + 2d2km,(3)
where d is the height of the neutral surface from the bilayer midplane and km is the monolayer compressibility modulus. Accordingly, a stiffer membrane will have a smaller decay constant as the membrane fluctuations are slower.

Experimentally measured decay rates, normalized by Q3, for symmetric and asymmetric membranes are plotted in Fig. 3A and show a clear temperature dependence, with the membranes becoming more dynamic as the temperature increases. The most dramatic change occurs at the transition from the gel to fluid phase from 15 °C to 45 °C, with 〈ΓZG/Q3〉 increasing by an order of magnitude in the fluid phase, where 〈…〉 denotes averaging over all Q values accessed. This significant increase in dynamics is consistent with the increased hydrocarbon chain disorder in the fluid phase (wherein the chains sample a larger conformation space), as well as an increase in lipid lateral and transverse diffusion, and the loss of long range positional order within the bilayer.25,35–39 The order of magnitude increase in dynamics also agrees with the softening of the bilayers in the fluid phase and the order of magnitude decrease in bending modulus seen upon lipid melting.40

image file: c9nr07977g-f3.tif
Fig. 3 Decay rate ΓZG normalized by Q3 averaged for all Q for both aLUVs and LUVs as a function of (A) ESM mole fraction (χESM) at 15 °C (image file: c9nr07977g-u3.tif), 30 °C (●) and 45 °C (image file: c9nr07977g-u4.tif). aLUV samples are indicated by half-filled points. (B) Bending moduli (κ) for the fluid aLUVs and symmetric vesicles with compositions corresponding to the inner leaflet, outer leaflet and overall aLUV composition measured with NSE (complete temperature series found in Fig. S4. C) Cartoon visualization of the observed bending rigidity for an asymmetric organization of ESM and POPE (top) and a symmetric mixture of ESM and POPE (bottom) at 30 °C.

Fig. 3A also shows that the dynamics are relatively insensitive to composition at temperatures completely below (image file: c9nr07977g-u1.tif) or above (image file: c9nr07977g-u2.tif) the chain melting transitions of both POPE and ESM. The dynamics decrease modestly with increasing amounts of ESM at 45 °C. In contrast, there is a strong composition dependence at the intermediate temperature of 30 °C (●), with the average decay constant decreasing by nearly a factor of 10 with increasing ESM concentration. The sharp change in dynamics at high ESM concentrations indicates that these mixtures are no longer fluid and instead have either phase separated or condensed into an ordered (gel-like) phase, which is further supported by the comparatively high melting temperature of ESM at 37 °C. These observations are consistent with NSE analysis of other binary lipid mixtures that showed a sharp increase in bending modulus upon lipid phase separation.41,42 Meanwhile, the symmetric vesicles with low ESM concentrations (χESM < 0.5) and the aLUVs remain dynamic at the intermediate temperature of 30 °C.

Focusing on the biologically relevant fluid phase membranes reveals unique and unexpected behaviors for asymmetric membranes: the asymmetric vesicle dynamics bear no simple relationship to that of the symmetric membranes corresponding to the inner and outer leaflet compositions. Instead, the dynamics appear to be an emergent property of the asymmetric transbilayer distribution of lipids that themselves have distinct melting transitions and order. The deviations in behavior are evident in the measured relaxation rates, as well as in the corresponding κ values. Values for κ can be calculated by substituting reported values for d and km in eqn (2) and (3) to give7

image file: c9nr07977g-t3.tif(4)

Corresponding values of κ calculated using eqn (4) are presented in Table 1 and plotted in Fig. 3B. The κ for the aLUVs at both temperatures are closer to the stiff outer leaflet, but are not a simple function of the individual leaflet properties (i.e., neither an arithmetic, geometric, or harmonic43 average). Moreover, the anomalies are temperature-dependent: κ for the aLUV falls between the leaflet values at 45 °C while at 30 °C the aLUVs are stiffer than both the outer and inner leaflets. The κ measured for aLUVs at 30 °C is even more surprising given that the symmetric LUV with the outer leaflet composition has overall 75% more high-melting ESM, revealing that the asymmetry increases κ and suppresses the bending fluctuations as illustrated in Fig. 3C.

Table 1 Average composition (χESM) and bending modulus κ of asymmetric and symmetric bilayers composed of ESM and POPE
    κ (kBT)
χ ESM sym/asym 15 °C 30 °C 45 °C
a POPE acceptor, ≈ aLUV inner leaflet composition. b aLUV average composition. c aLUV outer leaflet composition.
0.0a sym 433 ± 51 16.3 ± 0.5 10.4 ± 0.2
0.25b sym 22.2 ± 0.7 15.2 ± 0.2
0.25b asym 360 ± 36 61 ± 2 26.6 ± 0.7
0.44c sym 470 ± 51 48 ± 2 28.6 ± 0.7
0.85 sym 638 ± 77 326 ± 24 32.9 ± 0.9
1.0 sym 705 ± 93 492 ± 45 33 ± 1

3. Discussion

3.1. Influence of lipid spontaneous curvature on bending rigidity

It was recently reported that asymmetric giant unilamellar vesicles (GUVs) prepared by the phase transfer technique and composed of DOPC and POPC in the fluid phase also showed enhanced membrane rigidity as compared to their symmetric counterparts,18 a result that was attributed to the different spontaneous curvatures (J0) of these lipids.44 While the results for ESM/POPE aLUVs at 30 °C are superficially consistent with this explanation, the 45 °C data do not show the same dramatic difference in κ between symmetric and asymmetric LUVs that was seen in the GUV study. It has been reported that the spontaneous curvatures of ESM and POPE exhibit a qualitatively different temperature dependence, wherein J0 of ESM increases, and J0 of POPE decreases, with an increase in temperature.44 As the spontaneous curvature difference between ESM and POPE monolayers increase as the temperature is raised, the effects of the spontaneous curvature should thus be more evident at the higher temperature. It is therefore unlikely that differences in the spontaneous curvature of ESM and POPE can account for the dampened bending fluctuations we observe in aLUVs at 30 °C, but not at 45 °C. Differences in lipid spontaneous curvature have also been addressed in previous structural measurements of aLUVs composed of POPE/POPC.24 In that study, the difference in spontaneous curvature between POPE and POPC could not explain the lack of interleaflet coupling, with respect to the area per lipid, in fluid-phase POPE/POPC aLUVs. Taken together, a structural description of interleaflet coupling is insufficient in explaining the findings in this work. In general, lipid dynamics may be more sensitive to changes in composition and perhaps other system conditions than structural features, a conclusion in line with a previous finding where a dramatic change in lipid flip-flop and intervesicular exchange rates in the presence of methanol was observed before any structural deviations were detected by SANS, SAXS or DLS.45

Another important distinction of this work is that the dynamics were measured on the nanoscale with neutron spin echo and its results are uniquely sensitive to the internal dissipation dynamics within the membrane. Techniques such as flicker spectroscopy and micropipette aspiration probe the long wavelength dynamics that are slow compared to the motions of the lipids and have well-understood relationships to the intrinsic bending modulus. Consider the footprint of a membrane-bound protein compared to the overall size of the membrane itself; it is liken to the Burj Khalifa skyscraper being influenced by the curvature of the Earth. Thus, the relevance of microscopic bending fluctuation to a protein's structure–function is debatable. As mentioned above, NSE directly measures the effective bending modulus, that is the intrinsic bending modulus plus the effects of internal dissipation as suggested by eqn (3).

3.2. Influence of bilayer phase state and lateral organization on bending rigidity

The inability to recover asymmetric κ values through a combination of the symmetric systems is not completely surprising. Examining the κ values for the symmetric LUVs there is a clear χESM dependence at 30 °C and 45 °C (at low χESM); however, the κ values for the symmetric LUVs are not clear weighted averages of the pure lipid systems either. We have ruled out lateral phase-separation of the outer leaflet as the driver for these observations because of the large increase in AL for aLUVs at 30 °C (54.2 Å2) from 15 °C (45.7 Å2), which corresponds closely to AL of the fluid-phase LUVs (Tables S2 and S3). From this, it is clear that the aLUVs are also fluid phase. Further, the presence of the faster dynamics indicates that these membranes are fully fluid at 30 °C. This observation is supported by the DSC thermogram of ESM/POPE aLUVs (Fig. S1) and agrees with previous NMR studies by Soni et al. that showed symmetric binary mixtures with a similar lipid compositions were disordered and fluid at 30 °C.46,47 Further, wide angle X-ray scattering (WAXS) data demonstrate that mixtures composed of χESM = 0, 0.25 and 0.44 are fluid at 30 °C. The lack of a Bragg peak in these WAXS curves is indicative of no phase coexistence as well (Fig. S2).

The simplest, yet still plausible, explanation for the differences in the POPE/ESM mixed membranes dynamics (both LUVs and aLUVs) is the non-ideal mixing of POPE and ESM, even when both lipids are fluid. Past works have reported that PE and PC lipids de-mix based on differences in hydrogen bonding capabilities and hydration properties.48 Given ESM possesses a sphingosine backbone and a choline headgroup, it seems reasonable to conclude the hydration and hydrogen bond differences are present in the ESM/POPE system. Furthermore, it has been demonstrated that low-melting PE lipids have minimal miscibility with high-melting PC lipids.49 Local inhomogeneities in the membrane composition due to non-ideal lipid mixing could indeed influence the dynamics at the length scale of the bilayer measured with NSE.42,43

3.3. Vesicle and bilayer structure

Although effective, the use of the phase transfer method to generate the asymmetric bilayers is a cause for concern. When working in an oil-rich environment, similar to those in the phase transfer set-up, there exists potential for oil entrapment within the vesicle bilayer, which can cause membrane defects and influence relevant bilayer properties, such as membrane thickness. Therefore, asymmetric bilayers constructed in the absence of hydrocarbon solvents, and where the bilayer structure has been intimately characterized, may constitute a more biologically faithful platform for these mechanical measurements.

Asymmetric vesicles generated by CD-mediated exchange were meticulously assessed for artifacts resulting from the preparation conditions. The analysis of SANS and SAXS data from isotopically asymmetric aLUVs showed that POPC and DPPC vesicles prepared by CD-mediated exchange had bilayer structures that were identical to those prepared by conventional vesicle preparation techniques.21,25 X-ray and neutron based techniques are particularly sensitive to the lipid interface and can reveal information at the bilayer surface, such as CD interactions.50,51 In the experiments presented here, both the initial POPE acceptor LUVs and the final aLUVs exhibit a spherical shell scattering form factor consistent with a bilayered vesicle, a clear indication that sample preparation conditions did not compromise the vesicle integrity. Moreover, the diameter of the aLUVs used in NSE experiments was ≈200 nm, rendering the possibility of bilayer curvature or lipid packing effects unlikely.5 Further, we showed that neither the bilayer curvature or the molecular curvature of the lipids cause spontaneous asymmetry.

Potential changes in the membrane structure and order do not explain the disparate temperature behavior of the aLUVs. The increase in κ upon lowering the temperature from 45 °C to 30 °C in the symmetric ESM and ESM/POPE (χESM = 0.44) bilayers is consistent with the measured changes in the bilayer thickness (Table S3) according to the widely-used polymer brush model.52 Structural measurements of the aLUVs indicate that the aLUV bilayer thickness changes only by ≈2% on cooling from 45 °C to 30 °C, which cannot account for the nearly 200% change in κ, ruling out that the aLUV dynamics are explained solely by changes in the bilayer thickness.

3.4. Mechanical properties of asymmetric vesicles

The emergent behavior seen in the aLUV dynamics cannot be explained by mechanisms commonly seen in symmetric bilayers such as changes in membrane phase state, lateral organization, thickness or spontaneous curvature discussed above. Instead, we speculate that combining leaflets with unique chemical and physical properties into an asymmetric membrane influences the bending fluctuations measured on the nanoscale with NSE. The fluctuations at these short length and time scales are strongly influenced by not only the bending modulus, but also the interleaflet friction.

The extended theory by Watson and Brown that relates the effective bending modulus [small kappa, Greek, tilde] measured with NSE to the true bending modulus κ incorporates Seifert–Langer theory to account for the leaflet density and interleaflet friction.32,33 In other words, the lipids do not redistribute between the leaflets quickly enough, leading to an effectively larger bending rigidity at the nanoscale. The effects of such leaflet density fluctuations have been confirmed in various symmetric lipid membranes,7,28,30,53 but the asymmetric membranes studied here show dynamics that cannot be accounted for by this simple explanation. One likely possibility is that the asymmetric leaflet compositions and densities are influencing the internal dissipation within the bilayer. Alternatively, eqn (3) that links the true bending modulus κ to the effective bending modulus [small kappa, Greek, tilde], inherently assumes that the membrane is symmetric and depends on the definition of the neutral surface and the monolayer compressibility modulus.

The numerical pre-factor in eqn (4) used to calculate κ assumes that the neutral surface, d, of each leaflet is at the interface between the lipid headgroup and hydrophobic tail. This assumption has worked well for other lipid systems,7,44 but the location of d within the bilayer remains a topic of discussion even for symmetric bilayers, and it is not obvious how d would be defined in an asymmetric bilayer. Moreover, it is likely that the two leaflets of an asymmetric bilayer have different compressibility moduli which is not explicitly accounted for in the current model. It is also possible that the membrane asymmetry introduces other dissipation mechanisms or dynamic modes not yet considered theoretically. For example, asymmetry is predicted to induce a coupling between membrane thickness fluctuations and the movement of the membrane internal surface, which would lead to different relaxation behavior in an asymmetric and symmetric membranes.19 These results indicate that the bending modulus of an asymmetric bilayer is not simply the sum of each monolayer bending modulus. This was also observed previously in studies of asymmetric GUVs.18,54 Therefore, the bending mechanisms of the leaflets could be different between symmetric and asymmetric membranes, suggesting differences in interleaflet coupling and dissipation on a nanoscale.

4. Conclusion

To summarize, the data presented reveal that bending fluctuations in asymmetric membranes cannot be simply predicted from the properties of corresponding symmetric bilayers. We stress that the observations of anomalous dynamics in the aLUV are a true representation of the behavior, despite the current theoretical frameworks being unable to fully describe the mechanisms. Our data suggest that the stiffness of an asymmetric membrane is dictated by the more rigid leaflet, and that coupling of two leaflets with independent physicochemical properties introduces new contributions to the dissipation within the bilayer. The asymmetry-induced changes in the dynamics on the length-scale of the membrane itself are especially relevant to protein conformation changes and binding as well as intermembrane interactions that have been linked to the bending fluctuations.55–57 These changes in rigidity undoubtedly have consequences for integral membrane proteins such as mechanosensitive pores,58,59 and likely impact processes such as endocytosis, exocytosis, and membrane trafficking.60 Moreover, the strong interleaflet coupling observed here is entirely driven by lipids; indeed, lipids may be the major driving force for coupling in natural membranes, as peripheral proteins have not demonstrated such an ability.61 As membrane viscosity directly influences protein diffusion, and thus their ability to find interaction partners, these lipid-driven fluctuations may even be required for the proper function of some membrane proteins.

5. Methods

5.1. Materials

N-Acyl-4-sphingenyl-1-O-phosphorylcholine (egg sphingomyelin, egg SM), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (16:0/18:1 PE, POPE) 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1′-rac-glycerol) (sodium salt) [16:0/18:1 PG, POPG] were purchased from Avanti Polar Lipids (Alabaster, AL) and used as received (Fig. S5). Lipid stock solutions were prepared by dissolving the dry lipid powder in either HPLC-grade chloroform (ACP Chemicals Inc., Saint Lonard, QC) or in a combination of HPLC-grade chloroform and HPLC-grade methanol (Sigma-Aldrich, Oakville, ON) in a 2[thin space (1/6-em)]:[thin space (1/6-em)]1 ratio respectively. Methyl-beta-cyclodextrin (mβCD) was purchased from Sigma-Aldrich (Oakville, ON) and prepared as a 35 mmol L−1 (mM) stock solution in ultrapure H2O. The sucrose was purchased from Sigma-Aldrich (Oakville, ON) and prepared as a 0.632 M stock solution in ultrapure H2O. Sodium chloride was purchased from Sigma-Aldrich (Oakville, ON) and prepared as a 20 mM solution in D2O. Praseodymium(III) nitrate hexahydrate (Pr(NO3)3·6H2O) [Pr3+] was purchased from Sigma-Aldrich (Oakville, ON) and prepared as 40 mM stock solution in D2O. The deuterated chloroform (CDCl3) was purchased from Cambridge Isotopes (Andover, MA). Ultrapure H2O was obtained from a Milli-Q purification system (Millipore-Sigma, Oakville, ON) and 99.9% D2O was purchased from Cambridge Isotopes (Andover, MA).

5.2. Preparation of asymmetric vesicles

POPE (acceptor) and ESM (donor) lipid films were prepared by transferring the desired volumes of stock lipid solutions to separate glass scintillation vials. Bulk organic solvent was removed under gentle house vacuum and the resulting films were dried overnight (≈12 h) under vacuum at 50 °C. POPE films were doped with a mole fraction of 5% POPG to help facilitate unilamellarity. Donor multilamellar vesicles (MLVs) were prepared by hydrating the ESM film with 0.632 M sucrose solution to a lipid concentration of 20 mg mL−1. The film was preheated to 50 °C before hydration, followed by incubation at 50 °C with intermittent vortexing to fully disperse the lipids. The donor MLVs were then subjected to 3 freeze/thaw cycles at −80 °C and 50 °C with intermittent vortexing. The donor MLVs were diluted 20-fold with H2O, immediately followed by centrifugation at 20[thin space (1/6-em)]000g for 30 minutes. The resulting supernatant was discarded, and the pellet was re-suspended with the 35 mM mβCD solution to an 8[thin space (1/6-em)]:[thin space (1/6-em)]1 mβCD to lipid ratio. The donor lipid/mβCD mixture was then incubated at room temperature for 2 h with mild stirring (4 Hz). Acceptor large unilamellar vesicles (LUVs) were prepared by hydrating the POPE film with the 20 mM NaCl solution to a lipid concentration of 10 mg mL−1. The subsequent MLVs were then subjected to 5 freeze/thaw cycles at −80 °C and 50 °C including vortexing after each thawing step. The acceptor LUVs were prepared by a Nanosizer AUTO Liposome Extruder (T & T Scientific, Knoxville, TN) equipped with 100 nm polycarbonate filter (T & T Scientific, Knoxville, TN). The lipid suspension was passed through the filter 31 times at 45 °C. Asymmetric large unilamellar vesicles (aLUVs) were prepared by combining the acceptor LUVs with the donor/mβCD mixture at a ESM[thin space (1/6-em)]:[thin space (1/6-em)]POPE molar ratio of 2[thin space (1/6-em)]:[thin space (1/6-em)]1. Prior to bringing the donors and acceptors together, the donor/mβCD slurry temperature was raised to 35 °C. The donor/mβCD/acceptor mixture was then stirred gently (at 35 °C) for 30 minutes, followed by an 8-fold dilution with H2O and centrifugation at 20[thin space (1/6-em)]000g for 45 minutes. The supernatant was carefully removed as not to disturb the pellet. The supernatant was concentrated to 10 mL using an Amicon Stirred Cells (Millipore-Sigma, Oakville, ON) with a pre-washed 100 kDa Ultrafiltration Disc (Millipore-Sigma, Oakville, ON). The unwanted sucrose and mβCD was removed from the asymmetric LUVs via consecutive dilution/concentration cycles with D2O to a total H2O/D2O dilution of 1000×. On the aLUVs wash/concentration the volume was brought ≈6 mL of concentrated aLUVs.

5.3. Determination of bilayer composition via NMR

The relative lipid composition of the aLUV was determined via31P-NMR. Water was removed from an aLUV aliquot using a CentriVap Benchtop Centrifugal Vacuum Concentrator (Labconco, Kansas City, MO) for 16 h at 50 °C to ensure all the water was removed. The resulting film was dissolved with 1 mL of deuterated chloroform (CDCl3) to a lipid concentration of ≈1 mg mL−1. 700 μL was loaded into an NMR tube for 31P-NMR measurement. 31P-NMR spectra were collected on an Avance III 500 MHz spectrometer (Bruker, Billerica, MA) using Bruker TopSpin acquisition software, and analyzed with TopSpin 3.5. The observed chemical shifts were relative to the external standard for 31P (85% H3PO4). The mole fraction (χ) of the POPE and ESM were determined by the intensity under the respective peaks. The relative distribution ESM between the aLUV leaflets was determined via1H-NMR and the addition of a Pr3+ shift reagent as outlined in Doktorova et al.221H-NMR spectra were collected on an Avance III 600 MHz spectrometer (Bruker, Billerica, MA) using Bruker TopSpin acquisition software, and analyzed with TopSpin 3.5. The overall lipid organization of the aLUVs was calculated from the mole fractions of ESM and POPE determined from the integrated peak intensities of the lipid phosphates as measured by 31P-NMR and the distribution of ESM from 1H-NMR following the procedures for Heberle et al.21 and Doktorova et al.22 The composition of our aLUVs is summarized in Fig. 1C.

5.4. Preparation of symmetric control vesicles

LUVs were prepared by transferring the desired volumes of stock lipid solutions to achieve the desired ratio. Bulk organic solvent was removed under a gentle stream of N2 and the resulting films were dried overnight (≈12 h) under vacuum at 50 °C. All symmetric LUV films were doped with a mole fraction of 5% POPG to help facilitate unilamellarity. Films were hydrated with the 20 mM NaCl solution to a lipid concentration of 20 mg mL−1. One POPE control was hydrated with pure D2O. The hydrated films were subjected to 5 freeze/thaw cycles at −80 °C and 50 °C with intermittent vortexing after each thawing step. LUVs were generated by a Nanosizer AUTO Liposome Extruder (T & T Scientific, Knoxville, TN) equipped with 100 nm polycarbonate filter (T & T Scientific, Knoxville, TN). The vesicle suspensions were passed through the filter 31 times at 45 °C. LUVs were then subjected to the same D2O washing/concentration steps as the aLUVs. The LUV controls were: POPE, POPE (hydrated with pure D2O), ESM[thin space (1/6-em)]:[thin space (1/6-em)]POPE (1[thin space (1/6-em)]:[thin space (1/6-em)]1), ESM[thin space (1/6-em)]:[thin space (1/6-em)]POPE (4[thin space (1/6-em)]:[thin space (1/6-em)]1) and ESM.

5.5. Neutron spin-echo (NSE) spectroscopy

Data were taken on the spectrometer located on the NG-A guide at the NIST Center for Neutron Research (NCNR; Gaithersburg, MD).62 The momentum transfer, Q, ranged from 0.04 Å−1 to 0.11 Å−1. The wavelengths used in this experiment were 8 Å, allowing the higher region of the scattering vector to be reached, and 11 Å, which was chosen to access the lower momentum transfer regions at the expenditure of reduced beam intensity. These experimental settings allowed for Fourier times of up to 100 ns, corresponding to bilayer motions on length scales of ≈0.1 nm to 10 nm and time scales of 0.1 ns to 100 ns. The cells used for the NSE measurements had a path length of 4 mm. Temperature was controlled within 0.5 °C and the samples were allowed to equilibrate the desired temperature for at least 30 minutes prior to starting the measurement. The subsequent NSE data was reduced using Data Analysis and Visualization Environment (DAVE) software package.63

5.6. SANS and SAXS measurements

SANS measurements were conducted on the NGB 30 m SANS instrument located at the National Institute of Standards and Technology Center for Neutron Research (NCNR, Gaithersburg, MD).64 Wavelengths of 6 Å and 8.4 Å were used, as well as three sample-to-detector distances (1.33 m, 4 m, and 13.2 m), to access a scattering vector range of approximately 0.001 Å−1 to 0.5 Å−1. The scattered beam was counted on a 2D 3He detector and radially-averaged to produce 1D scattering curves of total intensity against the scattering vector (I vs. Q). Data were reduced using Igor Pro and NCNR developed reduction scripts.65 Small angle X-ray scattering (SAXS) aLUV data were collected with a Rigaku BioSAXS-2000 small angle instrument outfitted with a Pilatus 100 K detector and a HF007 rotating copper anode (Rigaku Americas, The Woodlands, TX). LUV data were collected at Beamline 12-ID-B of at the Advanced Photon Source, Argonne National Laboratory. aLUV and LUV data were collected at 288 K, 303 K and 318 K. SAXS data were analyzed by the joint refinement of SAXS and SANS data using a symmetric 5-strip model similar to ref. 25 and 27.

5.7. aLUV assessment

Asymmetric lipid distributions in aLUVs were quantified from solution 1H-NMR spectra measured in the presence of extravesicular paramagnetic lanthanide ions Pr3+. Externally added Pr3+ does not permeate into the vesicle interior on the timescale of the NMR measurement,22 and therefore interacts only with outer leaflet lipid headgroups to induce a downfield shift of the ESM choline resonance. We note that the ethanolamine of POPE does not contribute to the 1H-NMR signal. The observed choline signal is a superposition of shifted and unshifted resonances whose relative areas are proportional to the amount of ESM in the outer and inner leaflets, respectively (Fig. 1A).22 Immediately prior to SANS and NSE measurements aLUVs whose outer leaflets were enriched in ESM (i.e., POPEin/ESMout), exhibited an unequal area ratio (Fig. 1, upper panel). The ESM mole fraction in the outer and inner leaflet were 0.44 and 0.06 respectively. This aLUV lipid configuration remained over the course of the NSE measurements, as confirmed by NMR post measurement. The observed temporal stability of asymmetry is consistent with recent flip-flop studies of long chained phospholipids.25

Conflicts of interest

There are no conflicts to declare.


The authors thank Rob Brinson for NMR support. This work acknowledges support from: the Natural Sciences and Engineering Research Council (NSERC) of Canada [RGPIN-2018-04841] (D. M.); the Ontario Graduate Scholarship (OGS) program (M. H. L. N. and M. D.). M. N. acknowledges funding support of cooperative agreement No. 70NANB15H259 from the National Institute of Standards and Technology, U.S. Department of Commerce. F. A. H. acknowledges support from National Science Foundation (NSF) grant No. MCB-1817929. This material is based upon activities supported by the NSF under Agreement No. DMR-9986442. Access to the Neutron Spin Echo Spectrometer and NGB30 SANS Instrument was provided by the Center for High Resolution Neutron Scattering, a partnership between NIST and the NSF under Agreement No. DMR-1508249. Identification of any commercial product or trade names does not imply recommendation or endorsement by NIST. This research used resources of the Advanced Photon Source, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357.


  1. R. M. Bell, L. M. Ballas and R. A. Coleman, J. Lipid Res., 1981, 22, 391–403 CAS.
  2. T. Pomorski, S. Hrafnsdóttir, P. F. Devaux and G. van Meer, Semin. Cell Dev. Biol., 2001, 12, 139–148 CrossRef CAS.
  3. M. A. Kol, A. I. de Kroon, J. A. Killian and B. de Kruijff, Biochemistry, 2004, 43, 2673–2681 CrossRef CAS.
  4. P. F. Devaux and R. Morris, Traffic, 2004, 5, 241–246 CrossRef CAS.
  5. D. Marquardt, B. Geier and G. Pabst, Membranes, 2015, 5, 180–196 CrossRef CAS.
  6. A. C. Woodka, P. D. Butler, L. Porcar, B. Farago and M. Nagao, Phys. Rev. Lett., 2012, 109, 058102 CrossRef.
  7. M. Nagao, E. G. Kelley, R. Ashkar, R. Bradbury and P. D. Butler, J. Phys. Chem. Lett., 2017, 8, 4679–4684 CrossRef CAS.
  8. D. Marsh, Chem. Phys. Lipids, 2006, 144, 146–159 CrossRef CAS.
  9. M. Simunovic and G. A. Voth, Nat. Commun., 2015, 6, 7219 CrossRef CAS.
  10. T. Chou, K. S. Kim and G. Oster, Biophys. J., 2001, 80, 1075–1087 CrossRef CAS.
  11. J. A. Lundbæk, P. Birn, S. E. Tape, G. E. S. Toombes, R. Søgaard, R. E. Koeppe, S. M. Gruner, A. J. Hansen and O. S. Andersen, Mol. Pharmacol., 2005, 68, 680–689 CrossRef.
  12. P. I. Kuzmin, S. A. Akimov, Y. A. Chizmadzhev, J. Zimmerberg and F. S. Cohen, Biophys. J., 2005, 88, 1120–1133 CrossRef CAS.
  13. T. Baumgart, S. T. Hess and W. W. Webb, Nature, 2003, 425, 821–824 CrossRef CAS.
  14. R. Lipowsky, J. Biol. Phys., 2002, 28, 195–210 CrossRef CAS.
  15. F. Jülicher and R. Lipowsky, Phys. Rev. Lett., 1993, 70, 2964–2967 CrossRef.
  16. J. Liu, M. Kaksonen, D. G. Drubin and G. Oster, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 10277–10282 CrossRef CAS.
  17. E. Sackmann, FEBS Lett., 1994, 346, 3–16 CrossRef CAS.
  18. Y. Elani, S. Purushothaman, P. J. Booth, J. M. Seddon, N. J. Brooks, R. V. Law and O. Ces, Chem. Commun., 2015, 51, 6976–6979 RSC.
  19. R. J. Bingham, S. W. Smye and P. D. Olmsted, EPL, 2015, 111, 18004 CrossRef.
  20. G. van Meer, D. R. Voelker and G. W. Feigenson, Nat. Rev. Mol. Cell Biol., 2008, 9, 112–124 CrossRef CAS.
  21. F. A. Heberle, D. Marquardt, M. Doktorova, B. Geier, R. F. Standaert, P. Heftberger, B. Kollmitzer, J. D. Nickels, R. A. Dick, G. W. Feigenson, J. Katsaras, E. London and G. Pabst, Langmuir, 2016, 32, 5195–5200 CrossRef CAS PubMed.
  22. M. Doktorova, F. A. Heberle, B. Eicher, R. F. Standaert, J. Katsaras, E. London, G. Pabst and D. Marquardt, Nat. Protoc., 2018, 13, 2086–2101 CrossRef CAS PubMed.
  23. B. Eicher, F. A. Heberle, D. Marquardt, G. N. Rechberger, J. Katsaras and G. Pabst, J. Appl. Crystallogr., 2017, 50, 419–429 CrossRef CAS.
  24. B. Eicher, D. Marquardt, F. A. Heberle, I. Letofsky-Papst, G. N. Rechberger, M.-S. Appavou, J. Katsaras and G. Pabst, Biophys. J., 2018, 114, 146–157 CrossRef CAS.
  25. D. Marquardt, F. A. Heberle, T. Miti, B. Eicher, E. London, J. Katsaras and G. Pabst, Langmuir, 2017, 33, 3731–3741 CrossRef CAS PubMed.
  26. B. Perly, I. C. P. Smith, L. Hughes, G. W. Burton and K. U. Ingold, Biochim. Biophys. Acta, 1985, 819, 131–135 CrossRef CAS.
  27. M. Doktorova, F. A. Heberle, D. Marquardt, R. Rusinova, R. L. Sanford, T. A. Peyear, J. Katsaras, G. W. Feigenson, H. Weinstein and O. S. Andersen, Biophys. J., 2019, 116, 860–873 CrossRef CAS.
  28. J.-H. Lee, S.-M. Choi, C. Doe, A. Faraone, P. A. Pincus and S. R. Kline, Phys. Rev. Lett., 2010, 105, 038101 CrossRef PubMed.
  29. M. Mell, L. H. Moleiro, Y. Hertle, P. Fouquet, R. Schweins, I. Lopez-Montero, T. Hellweg and F. Monroy, Eur. Phys. J. E: Soft Matter Biol. Phys., 2013, 75, 75–97 CrossRef PubMed.
  30. M. Mell, L. H. Moleiro, Y. Hertle, I. Lopez-Montero, F. J. Cao, P. Fouquet, T. Hellweg and F. Monroy, Chem. Phys. Lipids, 2015, 185, 61–77 CrossRef CAS PubMed.
  31. A. G. Zilman and R. Granek, Phys. Rev. Lett., 1996, 77, 4788–4791 CrossRef CAS.
  32. U. Seifert and S. A. Langer, Europhys. Lett., 1993, 23, 71–76 CrossRef.
  33. M. C. Watson and F. L. H. Brown, Biophys. J., 2010, 98, L09–L11 Search PubMed.
  34. E. G. Kelley, P. D. Butler and M. Nagao, Characterization of Biological Membranes, De Gruyter, Berlin, Boston, 2019, pp. 131–176 Search PubMed.
  35. J. H. Ipsen, G. Karlström, O. G. Mouritsen, H. Wennerström and M. J. Zuckermann, Biochim. Biophys. Acta, 1987, 905, 162–172 CrossRef CAS.
  36. N. Kučerka, M.-P. Nieh and J. Katsaras, Biochim. Biophys. Acta, 2011, 1808, 2761–2771 CrossRef.
  37. J. L. Rubenstein, B. A. Smith and H. M. McConnell, Proc. Natl. Acad. Sci. U. S. A., 1979, 76, 15–18 CrossRef CAS PubMed.
  38. B. Wah, J. M. Breidigan, J. Adams, P. Horbal, S. Garg, L. Porcar and U. Perez-Salas, Langmuir, 2017, 33, 3384–3394 CrossRef CAS.
  39. F. A. Heberle and G. W. Feigenson, Cold Spring Harbor Perspect. Biol., 2011, 3, a004630–a004630 Search PubMed.
  40. R. Dimova, Adv. Colloid Interface Sci., 2014, 208, 225–234 CrossRef CAS.
  41. R. Ashkar, M. Nagao, P. D. Butler, A. C. Woodka, M. K. Sen and T. Koga, Biophys. J., 2015, 109, 106–112 CrossRef CAS.
  42. E. G. Kelley, P. D. Butler and M. Nagao, Soft Matter, 2019, 15, 2762–2767 RSC.
  43. R. R. Netz and P. Pincus, Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip. Top., 1995, 52, 4114–4128 CrossRef CAS.
  44. B. Kollmitzer, P. Heftberger, M. Rappolt and G. Pabst, Soft Matter, 2013, 9, 10877 RSC.
  45. M. Nguyen, M. DiPasquale, B. Rickeard, C. Stanley, E. Kelley and D. Marquardt, Biophys. J., 2019, 116, 755–759 CrossRef CAS.
  46. S. R. Shaikh, A. C. Dumaual, A. Castillo, D. LoCascio, R. A. Siddiqui, W. Stillwell and S. R. Wassall, Biophys. J., 2004, 87, 1752–1766 CrossRef CAS PubMed.
  47. S. P. Soni, D. S. LoCascio, Y. Liu, J. A. Williams, R. Bittman, W. Stillwell and S. R. Wassall, Biophys. J., 2008, 95, 203–214 CrossRef CAS PubMed.
  48. I. V. Polozov and K. Gawrisch, Biophys. J., 2004, 87, 1741–1751 CrossRef CAS PubMed.
  49. J. R. Silvius, Lipid-Protein Interact., 1982, 2, 239–281 CAS.
  50. L. Wu and X. Jiang, Anal. Bioanal. Chem., 2016, 408, 2743–2758 CrossRef CAS.
  51. L. Wang, P. Quan, S. H. Chen, W. Bu, Y.-F. Li, X. Wu, J. Wu, L. Zhang, Y. Zhao, X. Jiang, B. Lin, R. Zhou and C. Chen, ACS Nano, 2019, 13, 8680–8693 CrossRef CAS.
  52. W. Rawicz, K. Olbrich, T. McIntosh, D. Needham and E. Evans, Biophys. J., 2000, 79, 328–339 CrossRef CAS.
  53. L. R. Arriaga, R. Rodríguez-García, I. López-Montero, B. Farago, T. Hellweg and F. Monroy, Eur. Phys. J. E: Soft Matter Biol. Phys., 2010, 31, 105–113 CrossRef CAS.
  54. K. Karamdad, R. V. Law, J. M. Seddon, N. J. Brooks and O. Ces, Chem. Commun., 2016, 52, 5277–5280 RSC.
  55. W. Helfrich, Z. Naturforsch., A, 1978, 33, 304–315 Search PubMed.
  56. N. Dan, P. Pincus and S. A. Safran, Langmuir, 1993, 9, 2768–2771 CrossRef CAS.
  57. S. F. Fenz, T. Bihr, R. Merkel, U. Seifert, K. Sengupta and A.-S. Smith, Nat. Phys., 2017, 13, 906–913 Search PubMed.
  58. G. Chang, R. H. Spencer, A. T. Lee, M. T. Barclay and D. C. Rees, Science, 1998, 282, 2220–2226 CrossRef CAS.
  59. E. Perozo and D. C. Rees, Curr. Opin. Struct. Biol., 2003, 13, 432–442 CrossRef CAS.
  60. S. Dmitrieff and F. Nédélec, PLoS Comput. Biol., 2015, 11, e1004538 CrossRef.
  61. P. W. Fowler, J. Hélie, A. Duncan, M. Chavent, H. Koldsø and M. S. P. Sansom, Soft Matter, 2016, 12, 7792–7803 RSC.
  62. N. Rosov, S. Rathgeber and M. Monkenbusch, Scattering from Polymers, American Chemical Society, 1999, ch. 7, vol. 739, pp. 103–116 Search PubMed.
  63. R. Azuah, L. Kneller, Y. Qiu, P. Tregenna-Piggott, C. Brown, J. Copley and D. RM, J. Res. Natl. Inst. Stand. Technol., 2009, 114, 347–352 CrossRef.
  64. C. J. Glinka, J. G. Barker, B. Hammouda, S. Krueger, J. J. Moyer and W. J. Orts, J. Appl. Crystallogr., 1998, 31, 430–445 CrossRef CAS.
  65. S. R. Kline, J. Appl. Crystallogr., 2006, 39, 895–900 CrossRef CAS.


Electronic supplementary information (ESI) available. See DOI: 10.1039/C9NR07977G
These authors contributed equally to this work.

This journal is © The Royal Society of Chemistry 2020