Jeremias
Sibold
a,
Vera E.
Tewaag
a,
Thomas
Vagedes
a,
Ingo
Mey
a and
Claudia
Steinem
*ab
aInstitute of Organic and Biomolecular Chemistry, University of Göttingen, Tammannstr. 2, 37077 Göttingen, Germany. E-mail: csteine@gwdg.de
bMax Planck Institute for Dynamics and Self-Organization, Am Fassberg 17, 37077 Göttingen, Germany
First published on 8th April 2020
Lipid domains in plasma membranes act as molecular sorting platforms for e.g., signalling processes. In model membranes, such as freestanding or supported bilayers, some lipid domains with defined chemical composition, lipid packing and physical behaviour can be reproduced. However, in vivo, the plasma membrane experiences a proteinaceous scaffold underneath, which can sort, compartmentalize and recruit components within the membrane. The influence of such scaffolds on the phase behaviour of lipid membranes has been barely studied. Here, we investigated the partial attachment of a membrane to a support and its influence on the phase behaviour using pore-spanning membranes (PSMs). PSMs were prepared on SiOx=1–2 functionalized silicon substrates with 1.2 μm-sized pores by spreading giant unilamellar vesicles (GUVs) composed of DOPC/sphingomyelin (1:
1) with different cholesterol concentrations. Using two different fluorophores, PSMs were visualized by fluorescence microscopy allowing us to distinguish between different membrane phases, a gel (lβ), a liquid ordered (lo), and a liquid disordered (ld) phase. At low cholesterol concentrations, coexistence of lβ and ld was found, while at higher cholesterol concentrations, coexistence of lo and ld was predominant. Below the mixing temperature, determined by temperature scans, the more ordered phase was always found in the freestanding PSMs, whereas the ld-phase was present in the supported PSMs. We attribute this lipid sorting to a stronger adhesion of the ld-phase lipids to the underlying scaffold. The difference in adhesion alters the phase behaviour from a nominal DOPC/sphingomyelin (1
:
1) mixture to a DOPC/sphingomyelin (1
:
2–1
:
4) mixture compared to phase diagrams obtained from GUVs highlighting the importance of differential adhesive surfaces on lipid domain formation.
Hence, the majority of studies was performed using artificial membranes such as giant unilamellar vesicles (GUVs), multilamellar vesicles or supported lipid bilayers (SLBs), with lipid compositions resembling the outer leaflet of the plasma membrane. These membranes are composed of ternary mixtures of a high melting lipid such as sphingomyelin (SM) or 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) and cholesterol (Chol) and a low melting lipid such as 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) or 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC).4,8–13 Phase diagrams of such ternary lipid mixtures have been obtained for GUVs by means of fluorescence microscopy,14,15 for multilamellar vesicles using NMR spectroscopy,16 while SLBs were investigated by atomic force microscopy.8 The obtained ternary phase diagrams show a coexistent region of two liquid phases, a liquid-ordered (lo)- and a liquid-disordered (ld)-phase. The lo-phase is enriched in Chol and the high melting temperature lipid, while the ld-phase is enriched in the low melting temperature lipid. As an example, the composition of the lo- and ld-phase for the typical phase-separated DOPC/SM/Chol (2:
2
:
1) mixture with SMegg has been calculated using the ternary phase diagram obtained from GUVs with tie lines17 to be lo(DOPC/SMegg/Chol) = 9.5
:
55
:
35.5 (n/n) and ld(DOPC/SMegg/Chol) = 70.5
:
25
:
4.5 (n/n). The ld-phase has loose lateral lipid packing, acyl chains with gauche kinks, and fast lateral diffusion. In contrast, the lo-phase is characterised by tight lipid packing and a high degree of order, but still significant lateral diffusion.18 Even though lo-domains have similarities to raft domains, their size and physical properties are different from the domains found in the plasma membrane.7,10
One significant difference between the in vivo situation and artificial membranes is the fact that in model membranes, such as GUVs or SLBs, the bilayer is either completely freestanding (GUVs) or entirely attached to a solid substrate (SLBs). However, in vivo the plasma membrane is only partly freestanding. A large part of the membrane is linked to other cell structures such as cell–cell junctions, focal adhesions, and the cytoskeleton.19 These linkages greatly influence the mobility and organisation of the membrane components as shown by Schneider et al.20 They found that the diffusion of phospholipids and sphingolipids is considerably hindered in the plasma membranes of cells, while in giant plasma membrane vesicles (GPMVs) derived from these cells, the diffusion was unhindered due to the missing actin cytoskeleton.
To investigate the impact of the cytoskeleton on the phase behaviour of GUVs, Liu et al.21 polymerized an F-actin network on the GUV membranes inducing phase separation of initially homogenous vesicles. Honigmann et al.22 pinned an F-actin network to lo/ld-phase-separated SLBs via streptavidin–biotin linkages, which significantly altered the lipid domain structure. Along these lines, several groups investigated the influence of an underlying pre-patterned substrate on the phase behaviour of phase-separated SLBs.23 Patterning was either realized physically using different surface topographies24–26 or chemically with different polymer coatings. By using a patterning of polymers with either a saturated or an unsaturated lipid, the binding of particular lipids to the surface was altered leading to the recruitment of certain membrane components.27,28 Alternatively, Morigaki and coworkers29–31 used polymers that were partially or fully removed in confined areas to control membrane adhesion to the surface. They found that the lo-phase is localized in areas without polymer coating and the ld-phase in areas with partial polymer coating. The results demonstrate that the differential adhesion of the membrane to a support alters its phase behaviour. These findings imply that not only the membrane but its local adhesion to the cytoskeleton considerably impacts the lateral organisation of the plasma membrane in vivo. However, compared to SLBs, the plasma membrane is composed of freestanding bilayer areas with an only partial attachment to the underlying scaffold.
To mimic such situation, a membrane model, where a bilayer only partially experiences a support is highly desirable and can be generated by using pore-spanning membranes (PSMs).32 In this system, a membrane with an underlying nano- to micrometre-sized scaffold providing freestanding (f-PSMs) as well as supported (s-PSMs) bilayer areas is generated in a planar geometry readily accessible by fluorescence microscopy.19,33–42 As yet, PSMs composed of PC/SM/Chol have been formed on gold functionalized pore rims, and hence, owing to the quenching of the fluorescence by the gold layer, s-PSMs could not be fluorescently observed lacking the information about lo/ld-phase separation of the entire PSM.43,44 To overcome this drawback, we here make use of SiOx=1–2 functionalized porous silicon substrates33 in combination with two fluorescent (ld- and lo-) markers allowing the visualization of f-PSMs and s-PSMs. Our results on DOPC/SM/Chol PSMs with different Chol contents highlight the importance of the differences in adhesion of particular membrane lipids considerably altering the phase behaviour.
From the fluorescence micrographs, temperature dependent curves were obtained, from which the mixing temperature TM was extracted. A detailed description of the analysis of the fluorescence micrographs of PSMs as a function of temperature is given in the ESI† (Fig. S1 and S2).
Fig. 1 shows typical fluorescence micrographs of PSMs with BODIPY-Chol (green) and TexRed-DHPE (red) at T = 25 °C and T = 55 °C. In all images the regions of the freestanding PSMs (f-PSMs) can be distinguished from the supported PSMs (s-PSMs). The background intensity IBackground (ESI†) was determined from an uncovered pore, which is indicated with “x” in the images. In Fig. 1A, PSMs with a lipid composition of DOPC/SMporc/Chol (41.5:
41.5
:
17) are depicted. At 25 °C, the PSM is phase-separated into an lo-phase indicated by the green BODIPY-Chol fluorescence, while the ld-phase is visualized by the red TexRed-DHPE dye. From the fluorescence images at 25 °C it can be concluded that the lo-phase is only found in the f-PSMs, while the s-PSMs are only composed of ld-phase. At 55 °C, both fluorophores are co-localized and distributed equally in both parts of the membranes, the s-PSMs and f-PSMs, respectively. Overall, the fluorescence intensity of the s-PSMs is diminished compared to that of the f-PSMs owing to partial fluorescence quenching on the SiOx=1–2 surface, which can be readily seen if homogeneous pure DOPC membranes are formed on the porous substrates (Fig. S4, ESI†) and which has been also shown previously for other non-phase separated lipid compositions.33
For the lipid composition DOPC/SMporc/Chol (50:
50
:
0), a different scenario is observed (Fig. 1B). While the membrane is in a completely mixed state at 55 °C, two phases can be distinguished at 25 °C. Again, the s-PSMs are only composed of ld-phase indicated by the red TexRed-DHPE fluorescence. However, the f-PSMs contain an area, in which both fluorophores are excluded, which can be readily visualized by an overlay of both fluorescence images (Fig. S5, ESI†). Even though these areas appear black, the fluorescence images at 55 °C unambiguously show that f-PSMs are present (Fig. 1B, right). We assign these membrane areas lacking both fluorophores to a gel phase (lβ).47,49 From ternary phase diagrams of DOPC/SM/Chol, it is known that a gel phase occurs at low Chol and high SM concentrations.10,47,50–54
As the three detectable phases were clearly assigned by the two different fluorophores, we next asked the question how the Chol content influences the phase behaviour of the PSMs. We systematically varied the Chol content in the membranes, while keeping the DOPC/SM ratio constant to 1:
1. We determined the transition temperature (TM) of mixing/demixing by monitoring fluorescence images as a function of temperature. The TexRed-DHPE fluorescence intensity was used to extract quantitative information from the images, as TexRed-DHPE is known to partition to almost 100% in the ld-phase.46,47 In a previous study, we moreover analysed whether the chosen ld-phase marker alters the phase behaviour of ternary lipid mixtures, which turned out to be not the case.55 In contrast, labelling the lo-phase of model membranes is much more limited. We have chosen the well-studied BODIPY-Chol, which is known to partition into the lo-phase in model membrane systems with about 50–80%.3,48 Other fluorophores, which have been shown to partition in ordered phases of giant plasma membrane vesicles and plasma membranes partition, however in the ld-phase in model membrane systems.56–58 BODIPY-Chol was used to unambiguously distinguish between the lβ- and lo-phase in the demixed state. However, owing to its only moderate partition in the lo-phase, it was not suited to determine the mixing temperature TM. To determine TM, the ratio of the TexRed-DHPE fluorescence intensities of the f-PSM (If-PSM) and s-PSM (Is-PSM) were monitored. The analysis is described in detail in the ESI† (Fig. S1 and S2). The intensity ratios were plotted vs. temperature. To ensure that the system is in thermodynamic equilibrium, we performed heating–cooling cycles indicating only a small hysteresis (Fig. S6, ESI†).
An example of a phase transition curve is shown in Fig. 2. A PSM composed of DOPC/SMporc/Chol (46:
46
:
8) is heated up from 25 °C to 55 °C. The fluorescence micrographs (Fig. 2A and B) at T = 25 °C clearly indicate that both fluorophores are excluded from the f-PSMs meaning that the f-PSMs are in the lβ-phase (Fig. S7, ESI†). The TexRed-DHPE fluorescence was observed in the s-PSMs representing the ld-phase. At 55 °C (Fig. 2A), a homogeneous distribution of the TexRed-DHPE fluorophores in the f-PSM and in the s-PSM was observed. Within a Chol concentration range of about 0–10 mol%, an lβ-phase was identified in the demixed membrane state. All corresponding If-PSM/Is-PSM (T) curves show a sigmoidal behaviour. A sigmoidal Boltzmann function (eqn (1)):59
![]() | (1) |
![]() | ||
Fig. 2 Fluorescence micrographs of a PSM composed of DOPC/SMporc/Chol (46![]() ![]() ![]() ![]() |
If the Chol concentration is increased above 10 mol%, the f-PSMs are composed of the lo-phase instead of the lβ-phase. A PSM composed of DOPC/SMporc/Chol (42.5:
42.5
:
15) is shown in Fig. 3. At 25 °C, TexRed-DHPE is excluded from the f-PSMs, while BODIPY-Chol is enriched proving the existence of the lo-phase in the f-PSMs (Fig. 3A and B, 25 °C). TexRed-DHPE fluorescence is observed in the s-PSMs labelling the ld-phase. From the corresponding If-PSM/Is-PSM (T) curve (Fig. 3C), TM = 39.1 °C was obtained. Within a Chol concentration range of about 10–50 mol% we always observed an lo- and ld-phase in the demixed state.
![]() | ||
Fig. 3 Fluorescence micrographs of a PSM composed of DOPC/SMporc/Chol (42.5![]() ![]() ![]() ![]() |
Independent of whether the transition from lo/ld to ld or from lβ/ld to ld was observed, both phases, lβ/ld (Fig. 2A, 40 °C) or lo/ld (Fig. 3A, 40 °C) coexist in the f-PSMs during the transition. At this stage, lβ- or lo-domains become discernible, even though the time resolution was not sufficient to resolve the circular appearance of such domains in the f-PSMs that we have visualized previously.44 Circular domains are indicative of a significant line tension, which has been determined by Schwille and coworkers for a similar lipid composition (DOPC/SM C18:0/Chol, 2:
2
:
1) to be 1.2 pN.60 Tsai and Feigenson61 later demonstrated that an increase in the Chol content slightly reduces the line tension in such membrane mixtures. We found the described behaviour of the membranes for Chol concentrations of 0–18 mol%. However, at larger Chol concentrations, the transition regime became substantially broader.
An example is given in Fig. 4 for a Chol concentration of 25 mol%. While the fluorescence micrographs still unambiguously indicate the presence of lo-and ld-phase at 25 °C, no lo/ld coexisting phase is observed in the f-PSMs during the phase transition (Fig. 4A, 45 °C). From the If-PSM/Is-PSM (T) curve (Fig. 4C), a TM = 45.7 °C was determined, even though it cannot be ruled out that the transition is more complex and might contain a second non-resolved transition.
![]() | ||
Fig. 4 Fluorescence micrographs of a PSM composed of DOPC/SMporc/Chol (37.5![]() ![]() ![]() ![]() |
We analysed 22 different Chol concentrations. When the membrane is heated up or cooled down, it passes the transition temperature TM, which results in a coexistence of lo/ld- or lβ/ld-phases within the PSMs below TM independent of the lipid composition. The s-PSMs remain always in the ld-phase independent of temperature. The TM values are plotted as a function of Chol concentration in a partial phase diagram (Fig. 5A). In a Chol range of 0–8 mol%, only a phase separation between lβ- and ld-phase was found below TM. At a Chol concentration ≥11 mol%, only lo/ld-phase separation was detected. At Chol concentrations of 9–10 mol%, both types of phase separation were observed. This uncertainty is probably a result of the compositional heterogeneity of the GUVs of up to 5% if obtained by electroformation.52,62–64 The largest Chol concentration under investigation was 50 mol%, which is still well below the maximum of soluble Chol that was determined to be 66 mol% in DOPC membranes.65,66
![]() | ||
Fig. 5 (A) Partial phase diagram for PSMs composed of DOPC/SMporc/Chol with a constant ratio of DOPC and SMporc (1![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Ternary phase diagrams composed of PC, SM and Chol have been determined for model membranes such as small vesicles and GUVs as well as for SLBs.67 In case of small vesicles and GUVs, the membranes do not experience any adhesion. However, if GUVs adhere with each other, and the system is driven towards the demixing transition, Gordon et al.68 found that phase separation invariably occurs first in the adhered region, which they discuss in terms of suppression of thermal shape fluctuations in these regions. In case of SLBs, the membranes experience a homogeneous adhesion, if the surface itself is not patterned, and membrane fluctuations are fully suppressed. In this study, membranes were prepared with two different micrometre-sized areas, a freestanding membrane part, where the lipids do not experience adhesion to a substrate and a supported part, where the lipids are in contact to the SiOx=1–2 support. These different adhesion areas influence the distribution of the lipids in the entire membrane, i.e., lipid sorting as a function of surface adhesion energy occurs. Membrane fluctuations in the f-PSMs can be neglected as the lateral membrane tension is with about 1 mN m−1 or larger32,69,70 at least one to two orders of magnitude larger than lateral membrane tensions found in GUVs.71 The influence of surface adhesion on the distribution of lipids is most obvious in the example presented in Fig. 1A. Below TM, only ld-phase is observed in the s-PSMs, while there is only lo-phase in the f-PSMs.
Such reorganisation and preferential localization of certain lipid phases as a function of a difference in surface adhesion has been described previously by Sarmento et al.72 using surface adhered GUVs. Lipowsky et al.73,74 treated this phenomenon theoretically relating the phase diagram of a binary DPPC/Chol mixture to the surface adhesion energies (affinity contrast). Our observation clearly demonstrates that the interaction of the ld-phase lipids with the support is stronger than that of the lo-phase lipids and leads to lipid sorting as a function of a partial interaction with the support. Two aspects need to be considered to explain this finding.
By AFM indentation experiments, we have shown previously that, owing to the fact that PSMs experience two different adhesive surfaces, the adhesion strength of the lipids to the solid substrate (s-PSMs) greatly influences the lateral pre-stress in the f-PSMs.69 To investigate how the adhesion of the two different lipid types, lo-phase lipids and ld-phase lipids, influences the lateral membrane tension, which is the surface free energy per unit area, in the f-PSMs, we performed AFM indentation experiments as described previously.75,76 We indented f-PSMs composed of only DOPC mimicking the ld-phase and f-PSMs composed of SMporc/Chol (60:
40) resembling the lo-phase. For DOPC, a mean membrane tension of 10.5 ± 4.3 mN m−1 was determined, while on SMporc/Chol membranes a slightly lower mean membrane tension of 6.7 ± 4.0 mN m−1 was found (Fig. S8, ESI†). This result implies that the ld-phase lipids adhere more strongly to the SiOx=1–2 surface than the lo-phase lipids consistent with our observation that the ld-phase lipids are preferentially localized in the s-PSMs and the lo-phase lipids in the f-PSMs.
A second aspect that needs to be considered is the surface roughness of the SiOx=1–2 substrate. We determined a mean surface roughness of rms = 0.49 ± 0.03 nm by means of atomic force micrographs.33 Even though the surface roughness is in the sub-nanometre regime, the lipid bilayer has to adopt to the underlying structure. This process is energetically less costly in case of a membrane composed of ld-phase lipids as these membranes have a lower bending stiffness than membranes composed of lo-phase lipids.77–79 On the molecular scale, the more ordered structure of the individual lipid tails in the lo-phase will favour a position of the lo-phase lipids on the smooth water surface (f-PSMs) compared to the rough SiOx=1–2 substrate (s-PSMs).
If we compare the lo fraction in GUVs calculated according to the ternary phase diagram provided by Bezylepkina et al.17 with that found in our PSMs (Fig. S9, ESI†), which is basically determined by the surface porosity (Fig. S3, ESI†), we clearly see that the lo fraction in GUVs increases with increasing Chol content, while it remains constant in PSMs. However, we cannot rule out that nanometre-sized lo domains are present in the s-PSMs, which are not resolvable by optical microscopy. We analysed the maximum fluorescence change (Irel,T=55°C − Irel,T=25°C) for lo/ld-coexisting phases in PSMs and found that it decreases in the range of a Chol content = 10–50% (Fig. S9, ESI†), which might indicate that nanometre-sized lo-phase domains with excluded TexRed-DHPE dye are located in the s-PSMs diminishing the relative fluorescence intensity at T = 25 °C.
Furthermore, we compare the observed partial phase diagram (Fig. 5A) obtained for a nominal DOPC/SMporc (1:
1) with those obtained by Veatch and Keller50 (Fig. 5B–D) derived from GUVs, which are freestanding membranes. They used a mixture composed of DOPC/SM with a synthetic SM C16:0, while we used SMporc, which is mainly (70%) composed of saturated SM with SM C18:0 as the major component (49%). 20% of the SM is a monounsaturated SM C24:1 (see ESI†), which is known to prevent domain segregation.80 Even though the SM composition is different from the one used by Veatch and Keller, we clearly see a trend in the phase boundaries as a function of Chol concentration. The phase boundaries for a DOPC/SM ratio of 1
:
1 derived from GUVs do not match our results (Fig. 5B). If SM C24:1 significantly contributed to the phase transition temperatures TM, we would expect to see decreased TM values instead of increased ones. A better match is found for DOPC/SM (1
:
2) (Fig. 5C) and DOPC/SM (1
:
4) (Fig. 5D) ratios. This finding suggests that lipid sorting occurs, resulting in a change in the DOPC/SM ratio. It agrees with the observation that the s-PSMs are enriched in fluid phase DOPC lipids. As the overall lipid composition is constant, the ratio of DOPC/SM is locally shifted to larger SM concentrations in the f-PSMs.53,81
Another observation is that the lipid mixtures with a Chol content of 50 mol% are still phase-separated, which is not observed for freestanding membranes in GUVs. For example, by means of fluorescence resonance energy transfer experiments, Petruzielo et al.82 found phase separation in DOPC/SMbrain multilamellar vesicles with up to 40 mol% of Chol. Carravilla et al.83 visualized phase separation by Laurdan in GUVs also only up to 40 mol% of Chol. The same limit was observed for membranes attached to a solid support. Lipid bilayers composed of DOPC/SMegg/Chol fully adhered to a support showed phase separation up to a Chol content of 40 mol%.8
Our results show that ternary phase diagrams derived from simple lipid compositions only partially reproduce the nanoscale heterogeneity of eukaryotic plasma membranes. It is the cytoskeletal scaffold underneath the membrane providing different adhesive areas that also contributes significantly to the phase behaviour of a lipid bilayer.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0cp00335b |
This journal is © the Owner Societies 2020 |