Floret-Shaped Solid Domains on Giant Fluid Lipid Vesicles Induced

G. Kevrekidis , Frank Bruchertseifer , Alfred Morgenstern , and Stavroula Sofou ... Carl C. Hayden , Hsing-Lin Wang , Atul N. Parikh , and Andrew ...
0 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/Langmuir

Floret-Shaped Solid Domains on Giant Fluid Lipid Vesicles Induced by pH Amey Bandekar and Stavroula Sofou* Biomedical Engineering and Chemical and Biochemical Engineering, Rutgers University, Piscataway, New Jersey 08854, United States S Supporting Information *

ABSTRACT: Lateral lipid phase separation of titratable PS or PA lipids and their assembly in domains induced by changes in pH are significant in liposome-based drug delivery: environmentally responsive lipid heterogeneities can be tuned to alter collective membrane properties such as permeability (altering drug release) and surface topography (altering drug carrier reactivity) impacting, therefore, the therapeutic outcomes. At the micrometer scale fluorescence microscopy on giant unilamellar fluid vesicles (GUVs) shows that lowering pH (from 7.0 to 5.0) promotes condensation of titratable PS or PA lipids into beautiful floret-shaped domains in which lipids are tightly packed via hydrogen-bonding and van der Waals interactions. The order of lipid packing within domains increases radially toward the domain center. Lowering pH enhances the lipid packing order, and at pH 5.0 domains appear to be entirely in the solid (gel) phase. Domains phenomenologically comprise a circular “core” cap beyond which interfacial instabilities emerge resembling leaf-like stripes. At pH 5.0 stripes are of almost vanishing Gaussian curvature independent of GUVs’ preparation path and in agreement with a general condensation mechanism. Increasing incompressibility of domains is strongly correlated with a larger number of thinner stripes per domain and increasing relative rigidity of domains with smaller core cap areas. Line tension drives domain ripening; however, the final domain shape is a result of enhanced incompressibility and rigidity maximized by domain coupling across the bilayer. Introduction of a transmembrane osmotic gradient (hyperosmotic on the outer lipid leaflet) allows the domain condensation process to reach its maximum extent which, however, is limited by the minimal expansivity of the continuous fluid membrane.

1. INTRODUCTION Heterogeneous lipid bilayers containing liquid-ordered domains1−3 are related to critical biological phenomena including membrane trafficking,4,5 lipid and protein transport,6,7 and viral infection mechanisms.6,8 Studies, however, on heterogeneous lipid membranes with gel−fluid coexisting regions are relatively limited9−14 despite their importance on micropatterning of spherical particles relevant to photonic crystals15,16 and their use in medical applications.17,18 We are particularly interested in the potential role of gel−fluid heterogeneities on altering collective membrane properties such as surface topography and functionality, membrane permeability, and/or fusogenicity.19−25 These lipid membranes in the form of small lipid vesicles can be ultimately applied as drug delivery carriers for cancer therapy.26,27 Therefore, environmentally responsive formation of gel−fluid heterogeneities on model lipid bilayers using pH-dependent processes is of practical significance given the role of pH in human tumors.28 We study giant unilamellar vesicles (GUVs) composed of a fluid (low-melting) nontitratable lipid with unsaturated acyl tails, a gel-phase (high-melting) negatively charged titratable lipid with saturated acyl tails, and cholesterol. This study aims to evaluate the critical forces driving lipid phase separation in © 2012 American Chemical Society

GUVs that results in beautiful floret-like gel-phase domains on the spherical surface of fluid vesicle membranes. GUVs were prepared using the gentle hydration method29 which involves formation of fluid membrane GUVs at a temperature higher than the maximum transition temperature of all lipids and gradual cooling of vesicles to room temperature at which all of our observations were performed. In this study the pH (7.0, 6.0, 5.0), cooling rates, acyl tail length of domain-forming lipids, type of titratable lipid headgroup, and content of cholesterol and of D2O were varied and transmembrane pH and osmotic gradients were introduced. We provide evidence that the resulting domains arise from the interplay among the solid material properties (incompressibility and rigidity of domains), fluid material properties (expansivity limits of the continuous fluid membrane), and geometry of the curved closed surface on which they evolve. The terms “solid” and “gel” for domain phase characterization are used here interchangeably. Received: December 2, 2011 Revised: January 23, 2012 Published: January 25, 2012 4113

dx.doi.org/10.1021/la204765r | Langmuir 2012, 28, 4113−4122

Langmuir

Article

Figure 1. Characteristic fluorescence images of GUVs composed of equimolar ratios (1:1:1) of DOPC, DSPS, and cholesterol at different pH values (7.0, 6.0, 5.0) acquired at room temperature vs time after completion of the cooling process from 80 °C at a constant 20 °C/h cooling rate. Contrast is provided by C18-DiI lipids which preferentially partition in liquid-ordered phases. However, C18-DiI lipids are not distributed uniformly within the domains but exhibit radial distribution increasingly localizing at the interfacial edges of domains, suggesting the presence of a region of intermediate lipid packing order (see text for details). Different vesicles were imaged over the time course of the experiment.

Figure 2. Identical experiments as in Figure 1 with contrast provided by C12-DiI lipids which preferentially partition in fluid phases. Different vesicles were imaged over the time course of the experiment. of a thin dried lipid film that was then placed under house vacuum for 2 h for further drying followed by hydration with wet N2 gas at 80 °C for 45 min. After addition of 2 mL of prewarmed (80 °C) buffer (5.0 mM PIPES, 50 mM KCl, 1.0 mM EDTA, of desired pH, unless other buffer specified), lipids were allowed to undergo hydration at 80 °C for 24 h. The lipid suspension was then cooled at a predetermined constant rate using an EchoTherm Programmable Digital Hot Plate/ Stirrer (San Marcos, CA) to room temperature (25 °C). 2.2. Imaging of Phase-Separating Domains on GUVs Using Fluorescence Microscopy. Fluorescence microscopy images were obtained with an Olympus IX 70 inverted microscope (Olympus America Inc., PA) with a 40×, 0.6 NA dry objective. Vesicle suspensions were sandwiched between a coverslip and a microscope glass slide using vacuum grease as barrier. Both fluorescent probes, C12-DiI and C18-DiI, were imaged using an exciter bandpass filter (540 ± 25 nm) and an emitter bandpass filter (605 ± 55 nm) (Chroma

2. MATERIALS AND METHODS All lipids (DSPC, DSPS, DOPC, DOPS, DSPA, DPPS, DMPS) were purchased from Avanti Polar Lipids (Alabaster, AL). Cholesterol, potassium chloride, deuterium oxide, calcein, PIPES, and sucrose were purchased from Sigma Aldrich Chemical Co (Milwaukee, WI). EDTA was purchased from Fisher Scientific (Pittsburgh, PA). The fluorescently labeled lipids dioctadecyl tetramethylindocarbocyanine perchlorate (C18-DiI) and dilinoleyl tetramethylindocarbocyanine perchlorate (C12-Dil) were purchased from Molecular Probes (Eugene, OR). 2.1. Preparation of Giant Unilamellar Vesicles. Giant unilamelar vesicles were prepared by the gentle hydration method.29 Briefly, in a 10 mL glass tube 250 μL of organic phase (chloroform and methanol at 2:1 volume ratio) was added to the lipid mixture of 500 nmol total lipid. The organic phase was then evaporated using a Buchi Rotavapor R-200 (Buchi, Flawil, Switzerland) at 55 °C until formation 4114

dx.doi.org/10.1021/la204765r | Langmuir 2012, 28, 4113−4122

Langmuir

Article

Technology Corp., Bellows Falls, VT). To confirm the unilamellarity of GUVs, images of vesicles were acquired by focusing on their equator. Only unilamellar vesicles were imaged and analyzed using the Basic Metamorph (7.5.5.0) software (Downingtown, PA). 2.3. Introduction of a Transbilayer pH Gradient at Constant Temperature. Transient transmembrane pH gradients were introduced in suspensions of GUVs prepared at pH 7.0 (in 50 mM KCl, 5 mM PIPES, and 1 mM EDTA). At t = 0, the pH of the solution facing the outer lipid leaflet was lowered to the value of 5.0. GUV suspensions were maintained at room temperature, and phase separation was monitored over time for up to 96 h. Since single GUVs immobilized on a micropipet could not be retained intact for longer time periods than a few minutes, whereas the phase-separating process requires several hours to be complete, all time measurements were obtained by imaging at least 10 different freely diffusing GUVs at each time point for at least two different GUV preparations. 2.4. Modification of the Transbilayer Osmotic Gradient at Constant Temperature. GUVs were prepared at pH 7.0 in 5 mM PIPES with 50 mM KCl and 1 mM EDTA. At t = 0 the pH was lowered (or not) to 6.0, and 24 h later (t = 24 h) the suspension was split into different aliquots. In the first aliquot equal volumes of PIPES buffer containing 60 mM (100 or 200 mM KCl) were added. In another aliquot buffered solution of identical osmotic concentration (50 mM KCl) was added, and the last aliquot was monitored without any addition. All vesicle suspensions were imaged over time.

Regarding the state of lipids within phase-separated domains, our observations that are described below suggest (1) formation of domains within which the extent of condensation increases radially and is maximized at the domain center (at the center the state of lipids is more condensed than the liquidordered phase) and (2) formation of condensation gradients of lipids within domains that become increasingly steep with lowering pH. In particular, in Figure 1 the observed contrast is generated by the differential distribution of fluorescent lipids (C18-DiI) that preferentially partition in ordered phases.33 However, in Figure 1 fluorescent lipids are not distributed uniformly within the domains as usually demonstrated by liquid-ordered phases.34 Instead, Figure 1 shows that fluorescent lipids exhibit radial distribution within the domains increasingly localizing at the “interfacial boundaries” of domains with the (dark) continuous fluid phase. These “interfacial boundaries” containing the C18-DiI fluorescent lipids occupy relatively significant areas confirming the presence of a radial lipid condensation gradient within the domains and suggesting, at its outer limits, the presence of a region of intermediate lipid packing order (potentially a liquid-ordered phase). The absolute fluorescence intensities in images in Figures 1 and 2 are not normalized across samples; image intensities and contrast were changed to maximize clarity of presentation. In addition, in Figure 1, with lowering pH, the differential partition within the domains of the C18-DiI fluorescent lipids (i.e., the condensation gradient) becomes more pronounced. In particular, at pH 5.0, the C18-DiI fluorescent lipids appear to be entirely expelled from the floret-shaped phase-separated domains which appear dark (Figure 1, lower panel). At pH 5.0, C18-DiI lipids accumulate again at the “interfacial boundaries” between the separated PS-rich domains and the continuous fluid membrane (which also appears dark) and, in particular, at the clefts (“scars”) formed between floret leaves which have been suggested to contain areas of lower lipid order or else defects35 (see discussion). Therefore, again at pH 5.0, we observe formation of two types of phases: first, a phase of high lipid packing order (within the domains) entirely expelling the C18-DiI lipids, accompanied by boundaries of intermediate lipid packing order in which C18-DiI lipids localize and, second, a continuous fluid phase which also appears dark. The shape of the domains and, therefore, the symmetry of the lipid packing gradient within domains are affected by the value of pH. Evidence on the fluidity of the continuous lipid phase is provided by the uniform distribution of C12-DiI (Figure 2) that partitions in the complementary continuous phase. In addition, the measured translational diffusivities of domains (at pH 7.0) are in agreement with a fluid continuous phase36 (measured values range from 0.001 to 0.300 μm2/s, Table S1 and Figure S1 in the Supporting Information). Figures 1 and 2 show that the observed effects of pH and time on the extent and shape of phase-separated domains are identical and independent of the type of fluorescent lipids used, demonstrating, therefore, that these effects are not artifacts of the fluorescent probes. Quantitation of the size and shape of phase-separated domains was calculated as described below for each type of fluorescent lipid which provided the contrast in brightness between coexisting lipid phases. In Figure 2 the boundaries between the separated domain and continuous fluid phase are sharp and were determined by the limits set by the fluorescence intensity of C12-DiI in the fluid phase. In Figure 1, the C18-DiIrich areas are not treated as gel-phase boundaries since

3. RESULTS 3.1. Formation of Condensed (Gel) Phase Domains Rich in PS-lipids. Although the exact lipid compositions of phase-separated domains, shown in Figures 1 and 2, are not directly measured, our findings collectively suggest that they should be rich in protonated phosphatidyl serine (DSPS) lipids. This is also supported by previous DSC studies on small unilamellar vesicles composed of phosphatidyl serine (PS) lipids and phosphatidylcholine (PC) lipids demonstrating increasing formation of phase-separated domains rich in protonated (at the carboxyl group) PS-lipids with lowering pH.23,30,31 Attraction among protonated PS-lipids, leading to their phase separation, is attributed to hydrogen bonding that is intrinsically designed to form among PS-lipids but has been suggested to be hindered by the explicit negative charge of the carboxyl group.30 The highest reported apparent pKa value of the carboxyl group in PS-lipids when in the form of lipid bilayers is 5.5.32 This value was extracted from studies determining the gel-to-fluid transition temperature of membranes at salt concentrations similar to the values used in this study. This pKa value could be underestimated in the sense that it represents the mean value of the observed transition temperatures of the heating and cooling curves. In charged membranes, however, the direction of the thermal transition between an ordered phase and a fluid phase affects the effective surface charge density since the area per headgroup is different for the gel and the fluid phase. Ordered lipid phases due to smaller areas per headgroup have higher effective surface charge densities attracting higher concentrations of protons at the membrane surface governed by Boltzmann’s law. This decreases the extent of dissociation of the carboxyl group at a given bulk pH and, therefore, increases the value of the apparent pKa. Apparent pKa values of the PS carboxyl group even greater than this value by almost one logarithmic unit have been previously suggested in membranes containing cholesterol based on the observed onset of membrane activity directly related to the extent of PS protonation leading to phase separation.23 4115

dx.doi.org/10.1021/la204765r | Langmuir 2012, 28, 4113−4122

Langmuir

Article

Figure 3. (A) Fraction of the projected total area of phase-separated domains per vesicle vs time. (B) Number of phase-separated domains per GUV vs time at pH 7.0 (filled symbols), 6.0 (shaded gray symbols), and 5.0 (open symbols). GUVs were composed of equimolar ratios of DOPC, DSPS, and cholesterol and were cooled to room temperature at 20 °C/h. Lines are guides to the eye. Error bars correspond to standard deviations of at least 2 vesicle preparations and at least 10 different GUVs per time point per preparation.

Table 1. Quantitative Morphological Characterization of Phase Separated Domainsa

a Unilamellar GUVs were prepared at different pH values and were imaged at room temperature at a late time point (t = 55 hours) after completion of the cooling process from 80 °C at a constant 5 °C/h cooling rate. The error bars correspond to standard deviations of repeated measurements. At least two different vesicle suspensions were prepared and at least ten different GUVs per time point per preparation were imaged. At least two different vesicle suspensions were prepared and ten different GUVs per preparation were imaged. Figure S18 summarizes the first six entries of the above findings. Contrast provided by C12-DiI*, or C18-DiI† lipids. ‡Imaged eleven hours upon initiation of cooling.

from each type of fluorescent image agree with less than 8% deviation in reported values. Figures 1 and 2 exhibit domain ripening over time that is accompanied by growth of the total domain area. At any given time point a decrease in pH is associated with increased instabilities at the interfacial boundaries of domains. Domain boundaries exhibit interfaces that range from circular (at pH

comparison of Figures 1 and 2 shows that C18-DiI and C12-DiI colocalize in these areas. The limits of domain boundaries in the presence of C18-DiI were determined as the coordinates at which the slope of the pixel intensities of the C18-DiI-rich phase intercepts the low pixel intensity valley within the domain; this point is evaluated at 17% of the normalized maximum intensity across all samples. Accordingly, measurements of domain size 4116

dx.doi.org/10.1021/la204765r | Langmuir 2012, 28, 4113−4122

Langmuir

Article

Figure 4. Effect of transient transbilayer pH gradient on the condensation process at constant temperature. Transmembrane pH gradient was introduced at time zero between pH 7.0 (inner leaflet) and pH 5.0 (outer leaflet) at room temperature. GUVs are labeled with C18-DiI (top) or C12-DiI (bottom). Different vesicles were imaged over the time course of the experiment.

time point where one gel-phase domain per vesicle is observed, decreasing pH correlates with a longer interfacial perimeter (Table 1) reflecting the increasing instabilities of the domain interfaces shown in Figures 1 and 2. 3.4. Effect of pH, Cooling Rate, and Time on the Shape of Condensed Domains. Domain area (AD) and perimeter (PD) are used as metrics characterizing the shape of domains. The ratios AD/PD over time (Figure S6, Supporting Information) are independent of cooling rate and converge to values that increase with lowering pH (Table 1). Phase-separated domains comprise a circular “core” cap of radius RC (defined in Table 1) beyond which growth of instabilities may occur resembling leaf-like stripes of length LL (Figure S7, Supporting Information). The average number of leaf-like stripes per domain (Table 1, entries 1−3) is 4.4 ± 0.6 at pH 6.0 (65% of domains contain 4 stripes, 30% contain 5 stripes, and 5% contain 6 stripes) and 6.8 ± 0.6 at pH 5.0 (25% of domains contain 6 stripes, 65% contain 7 stripes, and 10% contain 8 stripes) and is essentially independent of the content of PS-lipids. The ratio LL/RC of the length of leaf-like stripes to the radius of the circular core cap of domains is mostly a function of pH and converges to 7.8 ± 1.2, 0.6 ± 0.2, and 0.0 for pH 5.0, 6.0, and 7.0, respectively. The final LL/RC ratios are unaffected by the cooling rate and content of PS-lipids (Table 1 and Figure S9, Supporting Information). 3.5. Effect of van der Waals and H Bonding on the Shape of Condensed Domains. Since the observed phaseseparated domains appear to be in the gel phase, the mechanical properties of the domains, namely, bending rigidity and incompressibility, were varied in order to measure their effect, if any, on the morphology of domains. Assuming that the major attractive forces between PS-lipids, leading to their phase separation from the continuous membrane, are van der Waals (VdWs) interactions among their acyl chains and intermolecular hydrogen bonding between protonated PS headgroups, we attempted to vary them independently to evaluate their effect on phase separation. To decrease the magnitude of VdWs attraction among the domain forming PS-lipids, shorter acyl tail lengths (16 saturated carbons) of PS-lipids were chosen in an equimolar mixture of DOPC, DPPS, and cholesterol in 100% H2O-based buffered solution (entry 8, Table 1). In parallel

7.0) to leaf-like stripes (at pH 5.0). The above observations are quantified below. 3.2. Effect of pH, Cooling Rate, and Time on the Extent of Phase Separation. Figure 3A shows that the fraction of the total projected area occupied by condensed (phase-separated) domains increases over time and reaches a maximum value for all cases studied independent of the cooling rate used (Figure 3A is for 20 °C/h, see Figure S2A, Supporting Information, for different cooling rates). At the limit of a high content of PS-lipids (33 mol % shown in Figure 3A) and independent of pH, the total projected area of gel-phase domains does not exceed 18% of the vesicle’s projected surface. In membranes with a lower content of PS-lipid (9 mol %) the total projected area occupied by condensed domains increases with lowering pH and converges toward a value which is less than 18% of the vesicle’s projected surface. (entries 4−6 in Table 1 and Figure S3, Supporting Information). The rate of growth of the total gel-phase domain area (Figure 3A) increases with lowering pH from 7.0 to 6.0 and is affected by the cooling rate only at the early observation time points and only at the highest value, of pH studied, 7.0 (see Figure S2B, Supporting Information). The separation process at pH 5.0 appears to be faster than our observation time frame. 3.3. Effect of pH, Cooling Rate, and Time on Domain Ripening. Figure 3B shows that the number of condensed domains per vesicle decreases over time for pH 7.0 and 6.0, and domain ripening leads to a minimum value of one gel-phase domain per vesicle at longer time points independent of cooling rate (Figure 3B for 20 °C/h, Figure S4A, Supporting Information, for different cooling rate). At the lowest pH of 5.0, GUVs exhibit one condensed domain per vesicle at all time points of observation. At early time points, the number of gelphase domains per vesicle decreases with lowering pH suggesting a faster separation process as pH is lowered. Cooling rates affect the kinetics of domain ripening during the early observation period but only at pH 7.0 and 6.0 (Figures S4B and S4C, Supporting Information). These kinetic profiles are independent of cooling rate at pH 5.0. In agreement with the observed domain ripening is the sum of interfacial domain perimeters PD per vesicle that decreases over time for pH 7.0 and 6.0 (Figure S5, Supporting Information, for 20 °C/h cooling rate). At the long observation 4117

dx.doi.org/10.1021/la204765r | Langmuir 2012, 28, 4113−4122

Langmuir

Article

Figure 5. Effect of temperature variation on the condensation process at constant pH. GUV was composed of DOPC:DMPS:cholesterol (1:1.8:1) at pH 5.0 and labeled with C12-DiI. DMPS-lipid was chosen due to its lower transition temperature relative to DSPS-lipid at pH 5.0 (47 vs 75 °C). Frame 1: GUV before annealing. Frame 2: GUV a few seconds upon removal of heat source. Same vesicle was imaged over the time course of the experiment.

Figure 6. Characteristic images of GUVs at pH 6.0 before (second panel) and after addition of higher salt concentration (40 mOsm difference shown on left column) 24 and 48 h later (third and fifth panels, respectively). (Middle column) Addition of buffered solution of identical osmotic concentration (50 mM KCl). (Right column) No addition. GUVs were composed of equimolar ratios of DOPC, DSPS, and cholesterol at pH 7.0. Upon cooling to room temperature the suspension pH was adjusted to 6.0. Salt addition took place 24 h later to ensure equilibration of pH across the bilayer. GUVs on the left image of each pair of images contains C18-DiI, and the right image contains C12-DiI. Different vesicles were imaged over the time course of the experiment.

entry 8 in Table 1) and do not affect the number of scars per domain. The presence of D2O (resulting in lower intermolecular hydrogen bonding among protonated PS-lipid headgroups) results in domains with larger core cap radii (RC = 3.1 ± 0.1 μm vs RC = 1.1 ± 0.3 μm of the original composition, entries 7 and 3 in Table 1, respectively) and a significant decrease in the number of scars or leaf-like stripes (4.4 ± 0.7) per domain compared to the original composition (6.8 ± 0.6). 3.6. Effect of Transbilayer pH Gradients on the Condensation Process at Constant Temperature. Figure 4 shows characteristic images of GUVs during evolution of the phase separation process in the presence of a transmembrane pH gradient that is introduced at time zero between pH 7.0 (inner leaflet) and pH 5.0 (outer leaflet) at room temperature. Phase separation on the inner lipid leaflet induced by decreasing pH is limited by the diffusion rate of protons across the lipid bilayer of GUVs (Figure S10, Supporting

studies, to decrease the magnitude of interlipid hydrogen bonding among domain-forming PS-lipids, GUVs of the (original) equimolar lipid composition (of DOPC, DSPS, and cholesterol) were prepared in buffered solutions with 10% v/v deuterated water (entry 7, Table 1). In these studies, the effect of D2O on enhancing the hydrophobic effect37 is assumed less significant compared to its role in competing with the hydrogen-bond-donating NH4+ group of PS-lipids to form hydrogen (or deuterium) bonds (solvent isotope effect).38 For both cases, GUVs were prepared at pH 5.0 and the resulting structures were compared to the original composition at pH 5.0 in 100% H2O-based buffered solution (entry 3, Table 1). The fraction of projected area in the gel phase is similar for all three cases (entries 3, 7, and 8, Table 1) at the limit of a high content, 33 mol %, of PS-lipid. Shorter PS-lipid acyl tails (corresponding to lower VdWs attraction among PS-lipids) result in domains with larger core cap radii (RC = 2.4 ± 0.1 μm, 4118

dx.doi.org/10.1021/la204765r | Langmuir 2012, 28, 4113−4122

Langmuir

Article

Addition of sucrose instead of salt results in identical morphologies (Figure S11B, Supporting Information), excluding the possibility of interactions between PS headgroups and ionic species as the source of the observed morphological changes.32,39

Information). The separation process seems to include a “coarse” phase of domain formation followed by a more “refined” rearrangement within the domains. The images in Figure 4 show rearrangement and gradual aggregation of phase-separating domains to a single domain per vesicle. The images do not correspond to the same GUV but to characteristic structures obtained at each time point due to the instability of pipet-immobilized GUVs for long time periods. At early time points (0 ≤ t ≤ 7 h) images show (1) formation of additional separated ordered phases from newly titrated PSlipids (this is demonstrated by the dark thin elongated domains (river-like structures) on the C12-DiI images that stain the fluid phase (Figure 4, bottom panel, arrow)) and (2) lipid rearrangement within pre-existing domains (since the lower pH promotes tighter lipid packing) resulting in expulsion of “impurities”. This is demonstrated by the additional diffusive interfaces formed by the C18-DiI lipid (Figure 4, top, arrow). The images on the bottom of GUVs labeled with C12-DiI suggest that some form of domain coupling takes place throughout the separation process. However, the extent of molecular packing among domain lipids on each opposing leaflet should not be the same at early time points since each leaflet is exposed to a different pH environment. At later time points (9 ≤ t ≤ 24 h), however, when equilibration of the proton concentration across the bilayer should be reached and each domain leaflet should contain more well-packed lipids, further “refined” rearrangement of lipid packing within the domains takes place that results in even more solid-like domains with greater numbers of leaf-like stripes. This is clearly demonstrated by the expulsion of “impurities” observed in the form of the fluorescent C18-DiI from condensing domains (top). Images of GUVs taken 24 h after introduction of a pH gradient exhibit domains indistinguishable from domains formed in GUVs originally in uniform pH 5.0 (Figure 1 and 2). 3.7. Effect of Temperature Variation on the Condensation Process at Constant pH. Figure 5 shows the process of domain reformation on the same GUV at pH 5.0 following thermal annealing. The initially identified floret-like domain (frame 1) on the GUV of interest is dispersed with thermal annealing. Upon removal of the heating source (frame 2) nucleation of new phase-separating domains is observed (frame 3), forming circular structures. Later, the growth pattern of domains switches to floret-like structures before completion of domain ripening to a single domain per vesicle (frame 4) and while the phase-separated area is simultaneously increasing (frame 5). The fluorescent lipid C12-DiI does not partition in the domains shown in Figure 5, suggesting registration/ coupling of the separating lipid phases across the bilayer. 3.8. Effect of Transbilayer Osmotic Gradient on the Condensation Process at Constant Temperature and pH. Upon introduction of additional salt on the side of the outer lipid leaflet of GUVs containing a high content of PSlipids (33 mol %) at pH 6.0 (Figure 6), domain morphologies are observed that are characteristic of phase-separated lipids at conditions of greater molecular packing within domains (i.e., resembling structures formed in pH 5.0 due to more extensive H bonding among protonated PS-lipids). Similar evolution of domain morphologies is observed with addition of greater salt concentrations (Figure S11A, Supporting Information), and in all cases the fraction of projected area occupied by domains does not exceed the cut off value of 18%. Addition at pH 5.0 of even higher salt concentrations in the solution facing the outer lipid leaflet does not result in further growth of domains.

4. DISCUSSION We use pH to induce lateral condensation of titratable lipids into solid domains on the curved surface of fluid vesicle membranes. Domains behave as solid-like materials with their bending rigidity (bending stiffness) and incompressibility (lipid packing density) affecting their morphology. The domains’ interfacial line tension appears to be spatially heterogeneous, and the immediate environment of domains, i.e., the continuous fluid membrane of vesicles, affects phase separation in interesting ways. In particular, in the lipid system studied, lowering pH from 7.0 to 5.0 decreases the electrostatic repulsion among titratable PS-lipids (by protonation of the carboxyl group with apparent pKa ≈ 5.532) and allows mostly the attractive intermolecular H-bonding and, to a lesser extent, VdWs interactions among PS-lipids to cause their phase separation.30,31 All shapes of observed solid domains appear to consist of a circular “core” cap beyond which, with lowering pH, formation of interfacial instabilities occurs resulting in growths that resemble leaf-like stripes. At pH 5.0 (the lowest pH value studied), stripes appear to bend only in one direction exhibiting almost vanishing Gaussian curvature (1/κ1 × 1/κ2, κ1 and κ2 being curvatures along each direction, Figure S12, Supporting Information). Stripes are formed within adjacent grain boundaries (“scars”) which contain dislocations in lipid packing within the domain. Our preliminary studies looking at fluorescence using linearly polarized excitation light show domains with antidiametrical stripes (leaves) exhibiting the same molecular tilt (Figure S13, Supporting Information). The origin of scars potentially identified at the center of the core cap40,41 or at the circular perimeter of the cap or maybe a combination of the above is uncertain. In this study, we demonstrate the following correlations between the phenomenological morphology (core area and stripes) and the material properties of the condensed domain. With lowering pH, an increase of the incompressibility of the phase-separated domain is mostly correlated with a larger number of thinner leaf-like stripes (or scars) per domain and an increase of the domains’ rigidity relative to the rigidity of the continuous membrane is mostly correlated with smaller core caps (beyond which stripes emerge). We demonstrate that the rigidity and compressibility of domain lipids, which control the domain shape, are maximized by coupling of domains across the bilayer. Interestingly, line tension, except for clearly driving domain ripening, does not seem to be critical on the final domain shape. In addition, our studies show that at high contents of domain-forming PS-lipids (33 mol %) the fraction of the projected membrane area in the separated phase reaches a maximum value independent of pH. This upper limit in phase-separated area seems to be controlled by the limited extent of allowable stretching on the continuous fluid membrane (vide infra). Cooling rates to room temperature, during preparation of GUVs, affect the kinetics of domain formation and ripening, but ultimately at long time points the size and shape of the single domain per vesicle is unaffected. In particular, for pH 5.0, where we observe the most intriguing domain shapes, the obtained floret-like domains do not depend on the path of GUV preparation (Figure S14, Supporting 4119

dx.doi.org/10.1021/la204765r | Langmuir 2012, 28, 4113−4122

Langmuir

Article

the H-bonding strength is generally 1 order of magnitude greater than VdWs attraction. Therefore, lowering the domain’s bending rigidity without significantly altering its compressibility, by forming domains with condensed C16-PS-lipids, results in domains with 5-fold larger core cap areas relative to the original composition (comprising C18-PS-lipids) and with the same number (6.8 ± 0.5) of stripes per domain compared to 6.8 ± 0.6 of the original composition (entry 3, Table 1). Conversely, addition of D2O is observed to result in domain shapes both with lower numbers of stripes (∼4.4) and with almost 9-fold larger core cap areas (entry 7, Table 1). Since lipid condensation within the domains is primarily driven by H bonding, addition of D2O is expected to decrease both the domain incompressibility (lower lipid compression/packing) and the bending rigidity of phase-separated domains. Similar domain shapes are obtained at pH 6.0 of the original composition in the presence of only H2O (entry 2, Table 1). These domains which should be characterized by less tight lipid packing (higher incompressibility) exhibit similarly low numbers of stripes (∼4.4) and core cap areas similar in size to the above case when D2O is present. Regarding the role of cholesterol in the studied membranes, especially given its high content that is usually observed to oppose phase separation in other lipid systems,1 previous studies suggest differential partition of cholesterol in each of the observed phases.47,48 The effect of pH on the partition distribution of cholesterol in titratable lipid membrane compositions (such as phosphatidyl serine or phosphatidic acid) may, potentially, follow a trend similar to the one imposed by the phase changes of lipid bilayers (liquid crystalline vs liquid ordered vs gel). Given the reported role of cholesterol on decreasing the interfacial line tension,49 complete elimination of cholesterol in GUVs composed of equimolar ratios of DOPC and DSPS results in similar floret-like structures at all three pH values (7.0, 6.0, and 5.0) with a large number (∼6.8) of leaf-like stripes per domain (entries 9−11, Table 1, Figure S15, Supporting Information) and core areas similar to the smallest observed of the cholesterol-containing composition at pH 5.0 (entry 3, Table 1). These findings suggest that (a) cholesterol partitions within the PS-rich domain47,48,50 and acts against tight molecular packing among domain forming PS-lipids and (b) line tension has a secondary role relative to the domain’s incompressibility and rigidity. The suggested correlation of increasing number of scars (or stripes) with increasing domain incompressibility is in agreement with calculations of the free energy of solid (crystalline) domains in fluid lipid vesicles at the limit of high elastic or stretching energy.51 These calculations show that if the domain contains defects of finite length grain boundaries then the stretching energy grows only linearly with the characteristic length of the crystalline domain and not as the area of the domain. Second, coupling of domains between opposing lipid monolayers is suggested to be related to increased lateral ordering between separated phases52 and tighter molecular packing among domain-forming lipids.53 Our studies on GUVs labeled with C12-DiI (Figures 4 and 5) suggest that coupling of domains between the two lipid leaflets is always present. We suggest that coupling maximizes packing (compression) of domain lipids, increasing the rigidity of the domains which then form dislocations (scars) and stripes in order to decrease the overall stretching energy. Single lipid layers of identical

Information). The shapes of domains (LL/RC and the number of stripes per domain) are found to depend on the domain’s incompressibility and rigidity (both tuned by pH) and not on the relative area covered by the domains, i.e. 9 mol % vs 33 mol % PS-lipid content (entries 1−3, and 4−6 in Table 1 and Figure S3, Supporting Information).42 Some discussion points are as follows: (1) at the continuum level, a potential explanation for the floret-like shape of phaseseparated domains, (2) at the molecular level, the potential role of registration/coupling of separating lipid phases across the bilayer in addition to H-bonding and VdWs attraction and the role of coupling determining a general suggested mechanism for the observed kinetics of phase separation, and (3) an explanation of the maximum fraction of projected membrane area covered by domains at all three pH values studied at the limit of high content of domain-forming PS-lipids. First, the observed floret-like domain structures are essentially gel (solid) phase domains evolved on and constrained by the curved surface of spherical fluid vesicle membranes that in one way or another essentially result in a decrease of the vesicle’s encapsulated volume. In general, for solid domains on fluid membranes, domain formation may be affected by43 (1) the line tension at the interface, clearly demonstrated in our studies by domain ripening over time, (2) the domain’s area compression that is generally significant for gel-phase domains and possibly even more important for PSforming phases given their reported smaller headgroup areas,44 and (3) the domain’s bending rigidity that should also play a role here given the solid-like nature of the separated phases and the observation that locally the curvatures of phase-separated domains appear different than the curvature of the continuous fluid vesicle membrane (Figure S12, Supporting Information). Two limiting cases have been discussed by Lipowsky et al.43 for solid domains on fluid vesicles, and our observations seem to correspond to domain properties that fall between these two limits. First, the limit of small domain area compressibility, in which separated domains are suggested to form narrow stripes exhibiting vanishing Gaussian curvature to satisfy the physical constraint of “fixed connectivity”.45 Second, the other limit involves large bending rigidities resulting in faceted domains which are expected to decrease the vesicle’s encapsulated volume. In our studies, stripe-like domains arise with almost vanishing Gaussian curvatures, and in some cases the core caps of domains exhibit curvatures smaller than the curvature of the vesicle’s fluid membrane (Figure S12, Supporting Information). On the basis of our observations, we suggest the possibility of significant accompanying shrinkage of the overall membrane area potentially leading to vesicle volume decrease, balanced by buildup of intravesicular osmotic pressure and increased stretching in the continuous membrane. In an attempt to differentiate between the effects of the domain’s bending rigidity and domain’s compressibility on phase separation we experimentally designed conditions that could somewhat distinguish between these properties. In particular, we evaluated shorter (by two carbon atoms) acyl tail lengths of domain-forming PS-lipids at pH 5.0 (entry 8, Table 1). Shorter domain lipid acyl tails are expected to decrease the bending rigidity of domains (since the bilayer at the region of the phase-separated domain is overall thinner by four carbon atoms)46 but may possibly not significantly alter the membrane compression within the domain. This rationale assumes the attractive H bonding among PS headgroups to be a major contributor to the condensation of PS-lipids given that 4120

dx.doi.org/10.1021/la204765r | Langmuir 2012, 28, 4113−4122

Langmuir

Article

the corresponding PC lipid (from 72.5 to 48.1 Å2). The pronounced decrease in the headgroup area of the PS-lipid may also induce reduction of the encapsulated volume from the vesicle to release some tension from stretching. In addition to the shrinking membrane area, the locally observed (almost) faceted domains (Figure S12, Supporting Information) would also induce reduction on the vesicle’s encapsulated volume.43 Volume decrease, however, should be stopped by the buildup of a transmembrane osmotic gradient since lipid bilayers are semipermeable to water and all our studies are performed in the presence of salt. We demonstrate that inversion (or minimization) of the osmotic gradient allows phase separation to proceed toward more rigid (smaller core cap areas) and more compressed domains (larger numbers of thinner stripes). Nevertheless, in all cases the fraction of projected phase-separated areas does not exceed 18%, which suggests a major role for the limited (fluid) membrane expansivity. Replacement of PS-lipids with titratable PA lipids of lower apparent pKa30 results in similar phase separation behavior with lowering pH just shifted toward lower pH values (Figure S16, Supporting Information). Finally, membranes containing DSPC- and DSPS-lipids demonstrate increasing formation of phase-separated domains containing interfacial instabilities with lowering pH, suggesting a critical role for H bonding in the observed separation relative to VdWs interactions (Figure S17, Supporting Information). Our studies suggest the significance in formation of floretlike domains of the coupling of domains across both bilayer leaflets. Such coupling is observed in symmetrical vesicles, but it is highly unlikely to occur in most natural membranes given the asymmetrically distributed PS.55 Symmetrically designed lipid bilayer vesicles, however, used as drug carriers may benefit from such extensive phase separation that is expected to alter collective membrane properties such as permeability of contents and effective functionality. These properties may have an impact in applications such as drug delivery.

composition forming microbubbles do not exhibit florets (data not shown). In particular, monitoring of the phase-separation process following different paths, i.e., by changing the order of lowering pH (from 7.0 to 5.0) and temperature (from 80 °C to room temperature), supports a proposed general mechanism as suggested above with a critical role for domain incompressibility and rigidity controlled/maximized by coupling across the bilayer. In the first condensation route (constant temperature, transient transmembrane pH gradient, Figure 4) we observe a two-state separation process: initially a “coarse” separation of lipids into coupled domain leaflets of different lipid order (since each leaflet experiences different pH) and of relatively low lipid order (since at least one of the leaflets experiences pH > 5.0) leading to minimal interfacial line boundaries as expected for a relatively significant contribution from line tension. Upon exposure of both domain leaflets to the same high lipid ordering (pH 5.0) the separation process evolves into a “refined” rearrangement of lipids of higher order within domains and formation of floret-like structures. In the second route (constant pH, transient thermal annealing, followed by fast quenching, Figure 5) we observe registration of separating phases that are of the same (high) lipid order (same pH 5.0 and same decreasing temperature). The lipid order is expected to increase rapidly with time as GUVs are cooled to room temperature almost immediately upon removal of the heating source. This separation process appears to follow a path closer to the “refined” mechanism that could be attributed to more ordered lipids within both domain leaflets at all time points of observation. The kinetics of this separation process appear faster than the process of domain ripening, and phaseseparating domains evolve into stripe-like structures before complete domain ripening contrary to the first separation path. During this “refined” condensation process it is possible that the trajectory of outflowing “impurities” coincides with the grain boundaries (scars) of domains based on reports on diffusivities of “impurities” being faster within scars.35 This, therefore, implies that the order of molecular packing within these scars is lower than the packing order within the domain, and thus, the inflowing PS-lipids may preferentially partition at the interfaces exhibiting higher order, i.e., at the moving front boundary of the solid domains. Such differential molecular packing could maintain the heterogeneous interfacial tension along the domain’s boundaries. Third, lipid membranes are essentially inextensible (2−5% extension before rupture).54 Therefore, the limited stretching capacity of the vesicle’s continuous fluid lipid membrane may partially explain the observed maximum area fraction occupied by the separated phase at the limit of a high content of domainforming PS-lipids. We suggest that due to significant compression of lipids within the domains the vesicle’s overall membrane shrinks since it consists of a finite number of lipids. Studies on DOPS and DOPC lipids demonstrate a condensing effect among PS-lipids relative to PC in both the fluid and the gel state.44 Therefore, by analogy, for PS-lipids undergoing a fluid-to-gel transition, i.e. undergoing a transition from partitioning in the continuous fluid phase of the GUV membrane to the phase-separated gel domain, the condensing effect in their cross-sectional area could range from 37.5% (assuming 65.3 Å2 in the fluid phase with DOPC to 40.8 Å2 in the solid domain) to as much as 43.7% (assuming 72.5 Å2 in the fluid phase with DOPC to 40.8 Å2 in the solid phase). This condensation is greater than the value of 33.6% calculated for

■

ASSOCIATED CONTENT

S Supporting Information *

Translational diffusion coefficients of domains, detailed effect of pH and cooling rates on domains’ size, number, shape etc., effect of DSPS content on domain shape and size, monitoring of internal pH in large vesicles in the presence of a transmembrane gradient, effect of osmotic stress on the extent of phase separation, curvature of domains, effect of cholesterol, PA titratable lipids, and lack of difference in lipid acyl−tail lengths on domain shape. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Phone: 732-445-4500 ext 6219. Fax: 732-445-3155. E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors thank Dr. Gerald W. Feigenson at Cornell University for teaching them the GUV preparation method and Dr. Nada Boustany at Rutgers University for help with the polarized fluorescence measurements. This work was supported 4121

dx.doi.org/10.1021/la204765r | Langmuir 2012, 28, 4113−4122

Langmuir

Article

(36) Cicuta, P.; Keller, S. L.; Veatch, S. L. J. Phys. Chem. B 2007, 111 (13), 3328−3331. (37) Nostro, P. L.; Stubicar, N.; Chen, S.-H. Langmuir 1994, 10 (4), 1040−1043. (38) Sheu, S.-Y.; Schlag, E. W.; Selzle, H. L.; Yang, D.-Y. J. Phys. Chem. A 2008, 112 (5), 797−802. (39) Huang, S.-L.; MacDonald, R. C. Biochim. Biophys. Acta 2004, 1665, 134−141. (40) Bernchou, U.; Brewer, J.; Midtiby, H. S.; Ipsen, J. H.; Bagatolli, L. A.; Simonsen, A. C. J. Am. Chem. Soc. 2009, 131 (40), 14130− 14131. (41) Watkins, E. B.; Miller, C. E.; Majewski, J.; Kuhl, T. L. Proc. Natl. Acad. Sci. U.S.A. 2011, 108 (17), 6975−6980. (42) Schneider, S.; Gompper, G. Europhys. Lett. 2005, 70, 136−142. (43) Lipowsky, R.; Dimova, R. J. Phys. Condens. Matter 2003, 15, S31−S45. (44) Petrache, H. I.; Tristram-Nagle, S.; Gawrisch, K.; Harries, D.; Parsegian, V. A.; Nagle, J. F. Biophys. J. 2004, 86 (3), 1574−1586. (45) Komura, S.; Lipowsky, R. J. Phys. II Fr. 1992, 2, 1563−1575. (46) Evans, E. A. Biophys. J. 1974, 14 (12), 923−931. (47) Barnes, J. P.; Freed, J. H. Biophys. J. 1998, 75 (5), 2532−2546. (48) McMullen, T. P. W; Lewis, R. N. A. H.; McElhaney, R. N. Biophys. J. 2000, 79 (4), 2056−2065. (49) Lin, W.-C.; Blanchette, C. D.; Ratto, T. V.; Longo, M. L. Biophys. J. 2006, 90 (1), 228−237. (50) Van Dijck, P. W. M. Biochim. Biophys. Acta 1979, 555, 89−101. (51) Chushak, Y.; Travesset, A. Europhys. Lett. 2005, 72, 767−773. (52) Collins, M. D.; Keller, S. L. Proc. Natl. Acad. Sci. U.S.A. 2008, 105 (1), 124−128. (53) Allender, D. W.; Schick, M. Biophys. J. 2006, 91, 2928−2935. (54) Needham, D.; Nunn, R. S. Biophys. J. 1990, 58 (4), 997−1009. (55) Devaux, P. F. Ann. Rev. Biophys. Biomol. Struct. 1992, 21 (1), 417−439.

primarily by the MRSEC Program of the National Science Foundation under Award Number DMR-0820341 and the Career Catalyst Award from the Susan G. Komen Foundation.

■

REFERENCES

(1) Feigenson, G. W. Nat. Chem. Biol. 2006, 2 (11), 560−563. (2) Feigenson, G. W. Biochim. Biophys. Acta 2009, 1788 (1), 47−52. (3) Veatch, S. L.; Keller, S. L. Biochim. Biophys. Acta 2005, 1746 (3), 172−185. (4) Hanzal-Bayer, M. F.; Hancock, J. F. FEBS Lett. 2007, 581, 2098− 2104. (5) Parthasarathy, R.; Groves, J. T. Soft Matter 2007, 3, 24−33. (6) Heinrich, M.; Tian, A.; Esposito, C.; Baumgart, T. Proc. Natl. Acad. Sci. U.S.A. 2010, 107 (16), 7208−7213. (7) Machta, B. B.; Papanikolaou, S.; Sethna, J. P.; Veatch, S. L. Biophys. J. 2011, 100 (7), 1668−1677. (8) Edidin, M. Curr. Opin. Struct. Biol 1997, 7 (4), 528−532. (9) Bagatolli, L. A.; Gratton, E. Biophys. J. 2000, 79 (1), 434−447. (10) Gordon, V. D.; et al. J Phys: Condens. Matter 2006, 18 (32), L415. (11) Korlach, J.; Schwille, P.; Webb, W. W.; Feigenson, G. W. Proc. Natl. Acad. Sci. U.S.A. 1999, 96 (15), 8461−8466. (12) Scherfeld, D.; Kahya, N.; Schwille, P. Biophys. J. 2003, 85 (6), 3758−3768. (13) Li; Cheng, J.-X. Biochemistry 2006, 45 (39), 11819−11826. (14) Szmodis, A. W.; Blanchette, C. D.; Levchenko, A. A.; Navrotsky, A.; Longo, M. L.; Orme, C. A.; Parikh, A. N. Soft Matter 2008, 4 (6), 1161−1164. (15) Manoharan, V. N.; Elsesser, M. T.; Pine, D. J. Science 2003, 301, 483−487. (16) Masuda, Y.; Itoh, T.; Koumoto, K. Adv. Mater. 2005, 17 (7), 841−845. (17) Needham, D.; Dewhirst, M. W. Adv. Drug Delivery Rev. 2001, 53 (3), 285−305. (18) Alaouie, A.; Sofou, S. J. Biomed. Nanotechnol. 2008, 4, 234−244. (19) Carruthers, A.; Melchior, D. L. Biochemistry 1983, 22, 5797− 5807. (20) Clerc, S. G.; Thompson, T. E. Biophys. J. 1995, 68 (6), 2333− 2341. (21) Mouritsen, O. G.; Zucherman, M. J. Phys. Rev. Lett. 1987, 58, 389−392. (22) Papahadjopoulos, D.; Jacobson, K.; Nir, S.; Isac, I. Biochim. Biophys. Acta 1973, 311 (3), 330−348. (23) Bajagur Kempegowda, G.; Karve, S.; Bandekar, A.; Adhikari, A.; Khaimchayev, T.; Sofou, S. Langmuir 2009, 25, 8144−8151. (24) Bhagat, M.; Sofou, S. Langmuir 2010, 26, 1666−1673. (25) Karve, S.; Bajagur Kempegowda, G.; Sofou, S. Langmuir 2008, 24, 5679−5688. (26) Karve, S.; Alaouie, A.; Zhou, Y.; Rotolo, J.; Sofou, S. Biomaterials 2009, 30, 6055−6064. (27) Karve, S.; Bandekar, A.; Ali, M. R.; Sofou, S. Biomaterials 2010, 31, 4409−4416. (28) Tannock, I.; Rotin, D. Cancer Res. 1989, 49, 4373−4384. (29) Akashi, K.; Miyata, H.; Itoh, H.; Kinosita, K. Jr. Biophys. J. 1996, 71 (6), 3242−3250. (30) Boggs, J. M. Biochim. Biophys. Acta 1987, 906, 353−404. (31) Tokutomi, S.; Ohki, K.; Ohnishi, S. I. Biochim. Biophys. Acta 1980, 596, 192−200. (32) Cevc, G.; Watts, A.; Marsh, D. Biochemistry 1981, 20, 4955− 4965. (33) Spink, C. H.; Yeager, M. D.; Feigenson, G. W. Biochim. Biophys. Acta 1990, 1023 (1), 25−33. (34) Zhao, J.; Wu, J.; Heberle, F. A.; Mills, T. T.; Klawitter, P.; Huang, G.; Costanza, G.; Feigenson, G. W. Biochim. Biophys. Acta 2007, 1768 (11), 2764−2776. (35) Schneider, M. B.; Chan, W. K.; Webb, W. W. Biophys. J. 1983, 43 (2), 157−165. 4122

dx.doi.org/10.1021/la204765r | Langmuir 2012, 28, 4113−4122