The Intermediate State of DMPG Is Stabilized by Enhanced Positive

Mar 5, 2010 - The results support the view that the intermediate state consists of vesicles ... intermediate state of DMPG or some aspects of it. Amon...
0 downloads 0 Views 2MB Size
Download PDF
pubs.acs.org/Langmuir © 2010 American Chemical Society

The Intermediate State of DMPG Is Stabilized by Enhanced Positive Spontaneous Curvature Juha-Matti Alakoskela,* Mikko J. Parry, and Paavo K. J. Kinnunen Helsinki Biophysics and Biomembrane Group, Institute of Biomedicine, Division of Biochemistry, P.O. Box 63, 00014 University of Helsinki, Finland Received June 15, 2009. Revised Manuscript Received February 19, 2010 1,2-Dimyristoyl-sn-glycero-3-phospho-rac-glycerol (DMPG) at low salt concentrations has a complex endotherm with at least four components and extending over the span of 20 degrees. During this ongoing melting, the solution becomes viscous and scatters light poorly. This multipeak endotherm was suggested to result from the effects of curvature on the relative free energies of gel and fluid DMPG bilayers, further relating to the formation of an intermediate sponge phase between the lamellar gel and fluid phases. Although later studies appear to exclude a connected bilayer network, the relation of the endotherm peaks to curvature remains an appealing hypothesis. This was tested by including in the system both water-soluble small molecules (dimethyl sulfoxide, ethanol, and urea) as well as amphiphiles (myristoyl-lyso-PG, cholesterol, cholesterol-3-sulfate, and dimyristoylglycerol) known to alter the spontaneous curvature of bilayers. All compounds increasing the monolayer positive spontaneous curvature (ethanol, urea, myristoyl-lyso-PG, cholesterol-3-sulfate) increased the temperature span of the intermediate state and elevated the temperature of its dissolution, while all compounds increasing the negative spontaneous curvature (dimethyl sulfoxide, cholesterol, dimyristoylglycerol) had the opposite effect, implying that the intermediate state contains a structure with positive curvature. The results support the view that the intermediate state consists of vesicles with a large number of holes. The viscosity increase could be related to vesicle expansion needed to accommodate the numerous holes.

Introduction The thermal behavior of the anionic phospholipid 1,2-dimyristoyl-sn-glycero-3-phospho-rac-glycerol (DMPG) at low salt concentrations is profoundly different as compared to corresponding zwitterionic lipids. In brief, the melting of this lipid is continuous over a temperature range spanning from approximately 17.5 to approximately 37.5 °C, with the endotherm comprising multiple components instead of a typical sharp peak at the transition.1 At the onset of the intermediate state, the increased heat capacity in the endotherm is accompanied by a sharp, large decrease in light scattering, and an increase in the bulk viscosity of the DMPG solution. At the end of the intermediate state, light scattering increases, bulk viscosity decreases, and the elevated heat capacity returns to the baseline.1-3 A number of explanations have been put forward to explain this intermediate state of DMPG or some aspects of it. Among them are the formation of a sponge phase,1,2 secondary changes in the aggregation state as a result of changes in Naþ affinity of DMPG,3 and changes in the protonation state.4,5 We and others have recently shown that both the formation of a continuous *To whom correspondence should be addressed. Telephone: þ358 9 191 25426. Fax þ358 9 191 25444. E-mail: [email protected]. (1) Heimburg, T.; Biltonen, R. L. Biochemistry 1994, 33, 9477–9488. (2) Schneider, M. F.; Marsh, D.; Jahn, W.; Kloesgen, B.; Heimburg, T. Proc. Natl. Acad. Sci. 1999, 96, 14312–14317. (3) Riske, K. A.; Politi, M. J.; Reed, W. F.; Lamy-Freund, M. T. Chem. Phys. Lipids 1997, 89, 31–44. (4) Koshinuma, M.; Tajima, K.; Nakamura, A.; Gershfeld, N. L. Langmuir 1999, 15, 3430–3436. (5) Tajima, K.; Koshinuma, M.; Nakamura, A.; Gershfeld, N. L. Langmuir 2000, 16, 2576–2580. (6) Riske, K. A.; Amaral, L. Q.; Lamy-Freund, M. T. Biochim. Biophys. Acta 2001, 1511, 297–308. (7) Riske, K. A.; D€obereiner, H.-G.; Lamy-Freund, M. T. J. Phys. Chem. B 2002, 106, 239–246.

4892 DOI: 10.1021/la100411p

(sponge phase) bilayer network and the dense DMPG vesicle aggregates are unlikely,6-10 although the possibility of dispersed particles with a sponge-phase-like structure could not be excluded.10 A possible structure is a vesicle containing multiple holes in the bilayer with an associated positively curved defect to cover the free edge of the bilayer.8 The matter is not yet resolved, as recent neutron scattering experiments hint at a possible sponge phase at higher concentrations (see refs 12 and 13). The uncertainty concerning the molecular level structure and mechanisms responsible for this behavior is surprising considering the extensive number of techniques used to address the problem so far: NMR, electron spin resonance (ESR), cryo-transmission electron microscopy (cryo-TEM), various fluorescence techniques, optical microscopy, static and dynamic light scattering, together with X-ray and neutron scattering.2,3,6-12 We decided to modify the spontaneous curvature of the DMPG bilayer leaflets by adding different compounds, and assessed the effects of these added compounds on the phase behavior of DMPG. The rationale for choosing the spontaneous curvature as the property to be modified is that the sponge phase and perforated/holey vesicle models show a difference with respect to the spontaneous monolayer curvature: Both perforated/holey vesicles and sponge (L3) phase have a great difference in total Gaussian curvature compared to intact vesicles. A vesicle with a single hole with a lipid-covered rim (8) Riske, K. A.; Amaral, L. Q.; D€obereiner, H.-G.; Lamy, M. T. Biophys. J. 2004, 86, 3722–3733. (9) Lamy-Freund, M. T.; Riske, K. A. Chem. Phys. Lipids 2003, 122, 19–32. (10) Alakoskela, J.-M. I.; Kinnunen, P. K. J. Langmuir 2007, 23, 4203–4213. (11) Riske, K. A.; Amaral, L. Q.; Lamy, M. T. Langmuir 2009, 25, 10083–10091. (12) Preu, J.; Gutberlet, T.; Heimburg, T. Chem. Phys. Lipids 2007, 149S, S40 (poster abstract). (13) Preu, J.; Gutlerbet, T.; Heimburg, T. Biophys. J. 2009, 96(Suppl. 1), 458a (poster abstract).

Published on Web 03/05/2010

Langmuir 2010, 26(7), 4892–4900

Alakoskela et al.

makes the monolayers continuous and a transforms the two unconnected spherical monolayer surfaces (inner and outer leaflet) to be topologically (homeomorphically) equivalent to a single monolayer sphere (or a plate, with the hydrophobic core forming the inside instead of any aqueous inner volume), and two holes transform the vesicle to be topologically equivalent to a torus (a circular, cylindrical micelle, or a doughnut), and so forth, decreasing the Euler characteristic of the surface and leading to strongly negative Gaussian curvatures, as described by GaussBonnet theorem. Likewise, the sponge phase is a thermally excited cubic phase, and different cubic phases have strongly negative Gaussian curvatures.14 Accordingly, although Gaussian curvature is undoubtedly important in the energetics of the shape change, the effect is in the same direction for both the perforated vesicles and the sponge phase. The difference, however, comes in the mean curvature: the perforated vesicles are expected to contain defects with positive curvature (the rims of the holes), whereas different cubic phases,15 including the sponge (L3) phase,16,17 become favorable compared to lamellar phase with increasing lipid volume fraction and with slightly negative spontaneous monolayer curvature, because the bilayer Gaussian curvature modulus is dependent on monolayer spontaneous curvature. Therefore, a shift in spontaneous curvature to the positive direction should stabilize the perforated/holey vesicles and destabilize the sponge phase, whereas a shift in spontaneous curvature to the negative direction should have an opposite effect. One well characterized transition clearly affected by monolayer spontaneous curvature is the lamellar to inverse hexagonal transition whose temperature THII is strongly affected by monolayer spontaneous curvature.18 We therefore chose compounds with known effects on THII for the modulation of spontaneous monolayer curvature. Any additive, nevertheless, always changes other properties of the system as well, and even new transitions may emerge because of phase separation into additive-rich and additive-poor domains in bilayers. In an attempt to circumvent this problem, we chose several compounds that are otherwise as different as possible, as long as they have the desired effect on spontaneous curvature. If the measured changes then agree with those expected based on spontaneous curvature, it is unlikely that modification of any other property would cause the effects. To decrease spontaneous monolayer curvature, cholesterol (CHOL), 1,2-dimyristoyl-sn-glycerol (DMG), or dimethyl sulfoxide (DMSO) was added into the mixture, and to increase spontaneous curvature cholesterol-3-sulfate (C3S), 1-myristoyl-2-hydroxy-sn-glycero-3-phosphatidyl-rac-glycerol (lyso-PG), ethanol (EtOH), or urea was added into the mixture. Based on endotherms and the light scattering versus temperature heating scans, CHOL, DMG, and DMSO destabilized the intermediate state, whereas C3S, lyso-PG, EtOH, and urea all stabilized the intermediate state. This strongly implies that increasing positive curvature stabilizes the intermediate state, and further suggests that the intermediate state is likely holey vesicles with positive curvature defects as suggested previously.8,11 The additives also caused a set of complex changes in the relative peak amplitudes and enthalpies, though these changes did not (14) Schwarz, U. S.; Gompper, G. Phys. Rev. E 1999, 59, 5528–5541. (15) Schwarz, U. S.; Gompper, G. Phys. Rev. Lett. 2000, 85, 1472–1475. (16) Wennerstr€om, H; Daicic, J.; Olsson, U.; Jerke, G.; Schurtenberger, P. J. Mol. Liq. 1997, 72, 15–30. (17) Le, T. D.; Olsson, U.; Wennerstr€om, H.; Schurtenberger, P. Phys. Rev. E 1999, 60, 4300–4309. (18) Marsh, D. Biophys. J. 1996, 70, 2248–2255.

Langmuir 2010, 26(7), 4892–4900

Article

appear to be related to changes in spontaneous curvature, at least not in a simple manner.

Experimental Section 1,2-Dimyristoyl-sn-glycero-3-phospho-rac-glycerol (DMPG, sodium salt) and 1-myristoyl-sn-glycero-3-phospho-rac-glycerol (lyso-PG) were from Avanti Polar Lipids (Alabaster, AL). Cholest-5-en-3β-ol (cholesterol, CHOL), cholest-5-en-3β-ol-3-sulfate, sodium salt (cholesterol-sulfate, C3S), and urea were from Sigma (St. Louis, MO). The concentrations of DMPG, CHOL, lyso-PG, and C3S stock solutions were determined gravimetrically with a high precision Cahn 2000 microbalance. Dimethyl sulfoxide, HPLC grade, was from Lab-Scan (Dublin, Ireland). Ethanol (99.5%, spectroscopical grade) was from Altia (Helsinki, Finland). All samples were prepared hydrated with 1 mM EDTA and 10 mM Hepes adjusted to pH 7.4 with NaOH, yielding a Naþ concentration of 6.2 mM. For DMSO- and ethanol-containing samples, a more concentrated buffer was used, diluted with water and DMSO/EtOH to give 1 mM EDTA, 10 mM Hepes, as well as the desired DMSO/water or EtOH/water volume ratio. Unless otherwise indicated, all samples contained 1 mM DMPG. Accordingly, to give, for example, a cholesterol (or lyso-PG) mole fraction of 0.20, samples contained 1 mM DMPG and 0.25 mM cholesterol (or lyso-PG). As the work was done over a long period of time requiring the use of several different batches of DMPG, we remeasured the pure DMPG for each new stock solution (and series) to ensure that there were no differences between batches (e.g., due to different amount of residual protonated DMPG) or the prepared stock solutions (e.g., due to contamination or erraneous concentration). For this reason, the zero sample values are not exactly the same for the different series. For cholesterolcontaining samples, triplicates were initially measured at the same time, but since there seemed to be more deviation between the same sample at different measurement dates, an additional three samples were measured for most cholesterol mole fractions, and for all the other additives the samples were prepared and measured in interleaved manner. Differential scanning calorimetry (DSC) measurements were done using a highly sensitive microcalorimeter (VP-DSC, Microcal Inc.) at a scan rate of 10 °C/hour. Static light scattering for DMPG was measured at an angle of 90° using a Cary Eclipse spectrofluorometer. The incident wavelength was set as 550 nm, and the intensity of scattered light was measured at 551 nm. For most samples, both light scattering and DSC were measured from ∼3 to ∼60 °C; for some samples, the light scattering measurements were extended for temperatures up to approximately 85 °C. The transition temperatures from the light scattering scans were extracted as follows. First the numerical differentiation of the (relative) scattering intensity versus T was performed, and then the spline interpolation in Matlab was used to create continuous curves for the evaluation of the maxima and minima of the numerical derivative. The quality of the interpolated data and the extracted maxima and minima were always checked against the original numerical derivative (see Supporting Information Figures 1 and 2). Four temperatures can typically be extracted from the light scattering scans by the above procedure: the pretransition temperature (Tp), the temperature for the appearance of the intermediate state (Ton), the temperature for the disappearance of the intermediate state (Toff), and the higher temperature transition (Tpost), which is seen in light scattering scans but not in the endotherms. Light scattering data are particularly useful to confirm which DSC peaks correspond to the onset and disappearance of the (low scattering) intermediate state. The scattering intensities are presented either relative to average scattering intensities of pure DMPG samples (SI, main text) or relative to scattering intensity of each curve at the lowest temperature (RSI, Supporting Information). DOI: 10.1021/la100411p

4893

Article

Alakoskela et al.

Figure 1. Components of the main transition of 1 mM DMPG. The components include a second order polynomial (green) to account for the baseline and four Gaussian components centered at 17.8 (blue), 19.0 (gray), 22.2 (violet), and 36.7 °C (cyan), shown with dash-dotted lines. The sum of the components is the red dashed line, and the recorded endotherm is shown with a continuous black line. ζ-Potential measurements for pure 1 mM DMPG in the 10 mM Hepes, 1 mM EDTA, pH 7.4 buffer were done using a Malvern Zetasizer Nano instrument and were analyzed using the Smoluchowski model with constant 3/2 in the dedicated software by the instrument manufacturer. The conductivity of the solution was also measured, by using both the Malvern Zetasizer Nano instrument and a hand-held electrode (Mettler Toledo, InLab 731).

Figure 2. Heating scans of DMPG/CHOL mixtures. Two scans for each concentration are shown. The mole fraction of cholesterol is indicated above the curve. 1, 2, and 4 indicate the scaling of the y-axis relative to pure DMPG. The bar on the left shows marks 20 kJ/mol/K for pure DMPG, or 5 kJ/mol/K for XCHOL = 0.200 sample.

Results We characterized the effects of the spontaneous-curvatureincreasing compounds EtOH, urea, C3S, and lyso-PG and the spontaneous-curvature-decreasing compounds CHOL, DMG, and DMSO, on the thermal phase behavior of 1.0 mM DMPG vesicle preparations by recording the endotherms and 90° light scattering intensities. The light scattering of the DMPG preparations detects at high temperatures (Tpost) a clear, additional shift that is not observed in the endotherms. Additionally, the fourth, wide, shallow endotherm which in the presence of the additives merges into the heat capacity baselines is associated with a large, sharp, and easily detectable increase in light scattering intensity. In the endotherms of pure DMPG, the melting transition peak has at least four components: three peaks closely spaced, with the first sharp peak at approximately 17.8 °C, followed by two wide peaks, with the first centered at approximately 19.0-19.5 °C and the second at 22-23 °C. A fourth component is at a higher temperature of approximately 37 °C (Figure 1). Notably, the fitting by Gaussians has little significance, since the peak shape may not be Gaussian and the normal sharp main transition peak of DPPC, for instance, cannot be fitted satisfactorily by a single Gaussian. Ton and Toff are clearly separate peaks, however, and different modulation of T1 and T2 by cholesterol (Figure 2), for instance, suggests that they, too, are separate entities. For brevity, we will use the symbols Ton, T1, T2, and Toff for the peak (or shoulder) temperatures.11 The first peak at Ton corresponds to the appearance of the intermediate state, and the last peak at Toff corresponds to its disappearance. Spontaneous-Curvature-Decreasing Compounds. Cholesterol, DMG, and DMSO tend to shift a bilayer leaflet toward negative spontaneous curvature. If spontaneous curvature has an impact on endotherm shape, then cholesterol, DMG, and DMSO should cause similar chages in the endotherms. In the heating 4894 DOI: 10.1021/la100411p

Figure 3. Heating scans of 1 mM DMPG dissolved in 10 mM Hepes, 1 mM EDTA, pH 7.4 in H2O/DMSO systems. The numbers indicate the volume fractions of DMSO.

scans of cholesterol-containing samples, the temperature of intermediate state appearance, Ton, decreases gradually with increasing mole fraction of cholesterol (Figure 2). Yet, cholesterol slighly decreases the temperature span of the intermediate state as the temperature for the intermediate state disappearance, Toff, decreases slightly more. Cholesterol strongly alters the relative enthalpies of the peaks, however, greatly increasing the relative enthalphy of the peak at T1 already at the cholesterol mole fraction of as low as 0.015; this peak also becomes narrower in the presence of cholesterol (Figure 2). The 90° light scattering scans in the presence of cholesterol likewise demonstrate a small decrease in the temperature span of the intermediate state but otherwise are little affected (see Supporting Information Figures 3 and 4). Langmuir 2010, 26(7), 4892–4900

Alakoskela et al.

Article

Figure 4. Effect of DMG on the endotherms (A) and scattering intensity of 1 mM DMPG (B). (A) Numbers on the right indicate the DMG mole fraction in the bilayer and the scaling for the y-axis of the curve compared to the endotherm in the absence of DMG. (B) Light scattering intensity (SI) relative to average scattering intensity of pure DMPG samples at the lowest temperature. The black squares, red circles, green up triangles, blue down triangles, and magenta tilted squares represent data for XDMG = 0, 0.005, 0.015, 0.048, and 0.091, respectively. The scans are normalized relative to scattering intensity at the first point of each curve. For scans normalized to the first point of each scan (and thus with smaller error) see Supporting Information Figure 7.

Dehydration of lipid headgroups has been suggested to be the basis for the stabilization of the HII phase by DMSO,19,20 and this dehydration has been observed in simulations21 and experiments.22,23 So, the effect of DMSO to shift the monolayer spontaneous curvature to the negative direction likely relates to solvation of lipid headgroups. In the DSC heating scans of the samples dissolved in DMSO-containing buffer, Ton increases and Toff decreases with increasing DMSO volume fraction, indicating that DMSO destabilizes the intermediate state (Figure 3). Like in the case of cholesterol, also DMSO changes the relative enthalpies and peak widths. The enthalpy of the peak at Toff increases with increasing DMSO fraction, with this peak also narrowing. The peak at Ton splits into two distinct peaks at the DMSO volume fraction of approximately 0.10-0.15. Surprisingly, in contrast to the case of cholesterol, even 25 vol % DMSO in the buffer did not significantly decrease the peak amplitudes, while the peak at Toff became narrower and increased in amplitude. Light scattering data for DMSO agreed with the endotherms (see Supporting Information Figures 5 and 6) The third compound shifting spontaneous curvature toward negative, DMG, decreased Toff and the temperature span of the intermediate state at low mole fractions of 0.005 and 0.015, similarly to the cases of CHOL and DMSO (Figure 4A). At higher DMG mole fractions (0.048 and 0.091), new high temperature peaks appeared. Scattering heating scans at 90° revealed that these peaks are not connected to the intermediate state (Figure 4B) but instead represent DMG-induced phase separation. Similar high temperature phase separation induced by DMG has been reported for a DMPC matrix.24 (19) Yu, Z.-W.; Williams, W. P.; Quinn, P. J. Arch. Biochem. Biophys. 1996, 332, 187–195. (20) Kinoshita, K.; Li, S. L.; Yamazaki, M. Eur. Biophys. J. 2001, 30, 207–220. (21) Sum, A. K.; de Pablo, J. J. Biophys. J. 2003, 85, 3636–3645. (22) Krasteva, N.; Vollhardt, D.; Brezesinski, G.; M€ohwald, H. J. Phys. Chem. B 2001, 105, 1185–1190. (23) Krasteva, N.; Krustev, R.; M€uller, H. J.; Vollhardt, D.; M€ohwald, H. Langmuir 2001, 17, 1209–1214. (24) Heimburg, T.; W€urz, U.; Marsh, D. Biophys. J. 1992, 63, 1369–1378.

Langmuir 2010, 26(7), 4892–4900

Figure 5. Effect of C3S on the endotherms of 1 mM DMPG. The numbers on the right indicate the C3S mole fraction in the bilayer.

Spontaneous-Curvature-Increasing Compounds. C3S, lyso-PG, ethanol, and urea tend to shift a bilayer leaflet toward positive spontaneous curvature. If spontaneous curvature has a major impact on endotherm shape or transition temperatures, then C3S, lyso-PG, ethanol, and urea should not only affect some features of the endotherms similarly, but these effects should also be opposite to those of cholesterol, DMSO, and DMG. C3S is structurally similar to cholesterol except for the large, anionic sulfate group attached to the hydroxyl group of cholesterol. The sulfate headgroup gives cholesterol-3-sulfate a shape that prefers positive spontaneous curvature, unlike cholesterol, whose small hydroxyl headgroup gives it a slight preference for negative spontaneous curvature. For C3S, the effects on Toff and the temperature span of the intermediate state were opposite to DOI: 10.1021/la100411p

4895

Article

Alakoskela et al.

Figure 6. Effect of lyso-PG on the endotherms of 1 mM DMPG (A). Numbers on the right indicate the lyso-PG mole fraction in the bilayer. Scattering intensities (B). The black squares are the data in the absence of lyso-PG, and the red circles, green up triangles, blue down triangles, and orange stars represent data for Xlyso-PG = 0.015, 0.048, 0.091, and 0.130, respectively. The scans are normalized relative to scattering intensity at the first point of each curve. Note that the actual scattering intensities for Xlyso-PG = 0.130 are low at all temperatures. For scans normalized to the lowest temperature scattering intensity of each scan, see Supporting Information Figure 9.

Figure 7. Effect of urea on the endotherms of 1 mM DMPG. The concentration of urea in buffer for the scans is shown on right, as well as the scaling compared to scans without urea.

those of cholesterol; both Toff and the T span increased (Figure 5). The effect of C3S on the peak at T1 was similar to that of cholesterol: the amplitude increased and the peak became narrower. This is likely related to the effects of the sterol ring structure on the bilayer. The light scattering data agreed with the DSC results (see Supporting Information Figure 8). Lyso-PG also increased Toff and the temperature span of the intermediate phase (Figure 6A). The peak at Toff also decreased in amplitude and widened with increasing lyso-PG mole fraction, making it difficult to determine Toff from the data, particularly because small baseline errors may have a strong effect on the peak value. For Xlyso-PG > 0.048, the peak at Toff could no longer be observed in the endotherms, whereas the transition remained salient in the light scattering data up to Xlyso-PG = 0.091, but 4896 DOI: 10.1021/la100411p

could no longer be observed for Xlyso-PG = 0.130 (Figure 8). The effects of lyso-PG on the shapes of the other endotherms are not pronounced, mainly increasing the width and decreasing the amplitude of the peak at Ton (Figure 6A). The two soluble compounds shifting monolayer spontaneous curvature to the positive direction, urea and EtOH, also produced similar results, shifting Toff to higher temperaratures and increasing the temperature span of the intermediate state. For urea, the effect on the peak shapes was mostly limited to decreasing the amplitude of the peak at Ton (Figure 7). The light scattering scans showed the same increase in Toff, though the quality of the numerical derivatives for unaveraged scans did not allow for the determination of Toff in the presence of 1.5 or 2.0 M urea (see Supporting Information Figure 10-11). With increasing ethanol volume fraction, in the heating scans, Ton remains nearly constant, but Toff strongly increases. The peak at Toff also becomes broader and its enthalpy decreases, becoming indistinguishable from the fluctuations of baseline at EtOH volume fractions > 0.01 (Figure 8A), and Toff could not be found from the scattering intensity (SI) versus T heating scans above 1% (v/v) EtOH (Figure 10). The disappearance of the peak at Toff in the endotherms coincides with the disappearance of the peak at T1. Unlike the other compounds, EtOH had a peculiar effect of strongly increasing the amplitude of the peak at Ton and strongly decreasing its half-width at half-height (Figure 8A). In the presence of higher concentrations of EtOH, the SI versus T scans show another decrease at high temperatures, with the temperature of this transition decreasing with increasing EtOH (Figure 8B). A similar transition can be seen for Xlyso-PG = 0.091 and 0.130 (Figure 6B, Supporting Information Figure 9). Four temperatures can be evaluated from DSC scans: Tp, Ton, T1, and Toff. The temperature of the third, wide peak of the first peak complex cannot be attained reliably in the presence of the added compounds, and even in the absence of the additives it can only be evaluated by fitting the endotherm. For a summary of the effects of the additives on the different transition temperatures, see Supporting Information Figures 13 and 14. Langmuir 2010, 26(7), 4892–4900

Alakoskela et al.

Article

Figure 8. Endotherms of 1 mM DMPG in the presence of EtOH (A) and scattering intensities (B). (A) DSC scans for each ethanol concentration are shown. The numbers on right show the ethanol volume fraction in buffer for the scan. (B) The 90° light scattering heating scans in the presence of EtOH. The black squares, red circles, blue up triangles, dark cyan down triangles, and magenta tilted squares correspond to ethanol vol % of 0, 0.5, 1.0, 2.5, and 5, respectively. The scans are normalized relative to the average scattering intensity of pure DMPG samples at the first point of each curve. For scans normalized to their first temperature, see Supporting Information Figure 12.

Figure 10. Comparison of dToff/dx values from DSC data (black squares) and SI data (red circles) to dTHII/dx values from literature for lipidic additives (CHOL, C3S, DMG, lyso-PG). The dotted lines represent least-squares linear fits weighed with errors. Figure 9. ζ-Potential and conductivity measurements. (A) ζ-Po-

tential of DMPG vesicles (squares) and the wall ζ-potential (circles). (B) Conductivity of the 1 mM DMPG solution as a function of temperature, measured with the conductivity meter of Malvern Zetasizer (up triangles) and with a hand-held conductivity electrode (circles).

ζ-Potential Data. One possible trigger to change the spontaneous curvature at the transition has been suggested to be the charging of the DMPG headgroups by increased dissociation of Naþ ions from DMPG headgroups.10,25 The data on this rely on the use of partioning of a cationic spin label between lipid and water, where increased partioning is observed at Ton.25 However, in general, the partioning of most compounds into lipid bilayers increases in the fluid phase, as even compounds associating to the (25) Riske, K. A.; Nascimento, O. R.; Peric, M.; Bales, B. L.; Lamy-Freund, M. T. Biochim. Biophys. Acta 1999, 1418, 133–146.

Langmuir 2010, 26(7), 4892–4900

headgroup region tend to perturb the gel phase lattice. To check for changes in electrostatics, we measured the ζ-potential of 1 mM DMPG vesicles, the wall ζ-potential that relates to vesicle adsorption on measurement cuvette walls,26 and the conductivity of the solution as a function of temperature (Figure 9). There appears to be only a light change in the slope of the ζ-potential versus temperature curve at Ton, another slight change at about 30 °C corresponding to the temperature at which the T2-centered component returns to baseline in endotherms, and finally, after Tpost, a complete neutralization of the vesicle charge (Figure 9A). Any abrupt change at Ton or Toff is missing also from the conductivity measured both with the conductivity meter of Malvern Zetasizer (triangles in Figure 9B) and with a hand-held electrode, though the absolute values are slightly different. Interestingly, wall ζ-potential implies that there is likely a significant decrease in vesicle adsorption on the measurement cell walls at (26) Reboiras, M. D.; Kaszuba, M.; Connah, M. T.; Jones, M. N. Langmuir 2001, 17, 5314–5318.

DOI: 10.1021/la100411p

4897

Article

Alakoskela et al. Table 1. Change in Toff with Additive Mole Fraction, vol %, and Concentration

additive

dTHII/dx, °C/mol %a

LRfHΙΙ matrix

dToff/dx, °C/mol %a

dToff/dx °C/mol %a

lipidic DMG CHOL C3S lyso-PG

ref 28b -2.45c -0.6 1.7 1.8

DEPEb DEPE DEPE DEPE

from DSC -2.4 ( 0.3 -0.53 ( 0.06 0.8 ( 0.3 1.4 ( 0.2

from SI -2.12 ( 0.12 -0.38 ( 0.09 0.58 ( 0.19 1.90 ( 0.10

additive

dTHII/dx, °C/mol %a

LRfHΙΙ matrix

dToff/dx, °C/mol %a

dToff/dx °C/mol %a

soluble refs 19, 29, and 30d from DSC from SI DOPEc -1.1 ( 0.1d -0.80 ( 0.04d DMSO -0.80 ( 0.07d,e d, f d EtOH 5.4 ( 0.5 DOPE 9.5 ( 1.7 7.2 ( 1.1d urea 6.15 ( 0.16d, f DOPE 4.7 ( 0.2d 3.8 ( 0.3d a The change in THII with the additive mole fraction (lipidic additives), volume percentage (DMSO and EtOH), or concentration (urea), dTHII/dx, obtained from literature, and the corresponding values for Toff (dToff/dx) estimated from the (initial) linear parts of the additive effects on Toff. b The values represent the shift in the temperature of lamellar-to-inverse-hexagonal phase transition of dielaidoylphosphatidylethanolamine (DEPE), given as °C/mol % of lipidic additive.28 c Evaluated by linear interpolation based on the values for dilauroylglycerol (-2.2 °C/mol %) and distearoylglycerol (-2.7 °C/mol %).28 For soluble additives, values for dTHII/dx in a dielaidoylphosphatidylethanolamine (DEPE) matrix could not be found. Values in dioleoylphosphatidylethanolamine (DOPE) matrix could be found for each compound. The values are given as °C/vol % for DMSO and EtOH, and as °C/M for urea. d Evaluated from Yu et al.19 e Evaluated from Kinoshita and Yamazaki.29 The value should be viewed with some caution, as it is only based on the concentration regime 10-14 vol % ethanol, and comparison to the 0% additive value from refs 19 and 30 implies that nonlinearities are likely present. f Evaluated from Feng et al.30

about 30 °C after the completion of the peak at T2 in the endotherms (Figure 9A). The near complete protonation of DMPG at high temperatures has been suggested earlier,4,5 but it is nevertheless quite surprising, particularly since the increased protonation cannot be attributed to increased surface proton concentration, but instead must derive from changes in the effective pKa value of DMPG in these conditions, if the apparent neutrality is not a result of some unknown artifact (see Discussion). There are possible confounding factors, however, for example, because of an electrical force related to the relaxation of the ionic atmosphere, the absolute value of ζ-potential at low electrolyte concentrations appears smaller for small particles, as observed by Eisenberg et al. who compared their small particles of approximately 1-5 μm and large particles of approximately 13-20 μm diameter.27 In our case, the particles have a diameter in the order of 0.15 μm.10 Importantly, the ζ-potential data show that there are no abrupt, large changes related to the transitions except the neutralization seen when T > Tpost.

Discussion The Effects of Spontaneous Curvature Modification on Transition Temperatures. The effects of additives on the transition temperatures were consistent with their effects on spontaneous curvature only for Toff and Tpost; both temperatures were strongly increased by the additives increasing spontaneous curvature and decreased by additives decreasing spontaneous curvature (see Supporting Information Figures 13 and 15). This is expected if the phase at the low temperature side of the transition has larger curvature than on the high temperature side. This is in agreement with the holey/perforated vesicle model8,11 but not with the sponge phase model. Seeking further confirmation that the detected effect on Toff derives from the effect of the additives on spontaneous curvature, we compared the effects of the additives on Toff to their effects on THII. For a bilayer-tocylindrical transition such as the LRfHII (or bilayer-to-rim of a very large hole) whose temperature is modified by an additive (27) Eisenberg, M.; Gresalfi, T.; Riccio, T.; McLaughlin, S. Biochemistry 1979, 18, 5213–5223. (28) Janes, N. Chem. Phys. Lipids 1996, 81, 133–150. (29) Kinoshita, K.; Yamazaki, M. Biochim. Biophys. Acta 1997, 1330, 199–206. (30) Feng, Y.; Yu, Z.-W.; Quinn, P. J. Chem. Phys. Lipids 2002, 114, 149–157.

4898 DOI: 10.1021/la100411p

affecting the spontaneous monolayer curvature, a linear shift in transition temperature at very low additive mole fractions is expected, with the extent of the shift being dependent on the molecular shape of the additive, the shape of the parent matrix molecule, and the enthalpy of the transition.18 Unfortunately, we could not find reference data for the effects of all the additives on THII in the same lipid matrix, but we managed to find reference data for all the lipidic additives in dielaidoylphosphatidylethanolamine (DEPE; see Table 1) and for all the water-soluble additives in dioleoylphosphatidylethanolamine (DOPE; see Table 1). (Note that the effect of a compound on THII is expected to be different in DEPE and DOPE matrices, so these reference values cannot be compared directly.) Since the parent matrices and the change of the radius of curvature in the transitions (LRfHII and bilayerfrim) are not exactly the same, the absolute effects of the additives on THII and Toff are not expected to be equal, but the relative effects should be similar, in agreement with the results for lipidic additives (Table 1 and Figure 10), with the exception that C3S has a smaller effect on Toff than expected. The water-soluble additives urea, DMSO, and EtOH do not show as good agreement, which may derive from two factors: (i) the reference data for EtOH were only available in the concentration range 1014 vol %, and the dTHII/d[EtOH] values may be larger at small EtOH volume percentrages, and (ii) there may be significant differences in the lipid/aqueous partition coefficients of the additives for DMPG and DOPE. Nevertheless, the effect of additives on Ton did not show the expected opposite effect compared to the effect on Toff by all the additives (i.e., consistent decrease by additives increasing spontaneous curvature and increase by additives decreasing spontaneous curvature). We originally expected the compounds to have many confounding effects unrelated to their effect on spontaneous curvature, and we observed such effects, for instance, their effects on endotherm shapes. At Toff, the bilayer is practically completely fluid (as can be seen from endotherms); in other words, at Toff, there is effectively a transition between two fluid phases, whereas below Ton DMPG is in the gel phase. It seems likely that somehow the complex interactions of different additives with the gel phase structure are responsible for the discrepancy. Holey Vesicle Phase. It thus appears that in the intermediate state DMPG exists as vesicles with extensive numbers of holes as Langmuir 2010, 26(7), 4892–4900

Alakoskela et al.

suggested previously.8,11 Such holey vesicles are sometimes called stomatosomes and are observed, for instance, in the phase diagrams of aqueous solutions of mixtures of bilayer- and micelle-forming amphiphiles.31,32 In the mixture of bilayer-forming DMPC and micelle-forming dihexanoyl-PC, for instance, with increasing fraction of dihexanoyl-PC, the intact DMPC vesicles first change into holey vesicles, then into branched cylindrical micelles, into nonbranched cylindrical micelles, into small bicelles, and finally, with high enough fraction of dihexanoyl-PC, into spherical micelles.33 In simulations of amphiphilic block copolymers, the evolution of vesicles into cagelike structures formed by a mesh of branched cylindrical micelles has also been observed,34 and it seems possible that such a state exists between the holey vesicles and the branched cylindrical micelles. In the holey vesicles, the micelle-forming amphiphile is most likely enriched in the rims of the holes. Accordingly, the holey vesicle for DMPG in the coexistence region fits the known phase diagrams of other systems. Yet, the Gaussian curvature free energy cost of forming a hole is quite considerable, and the positive spontaneous curvature required so that the free energy from the relaxation of curvature stress can compensate for this is quite high, basically requiring that DMPG molecules in the rims have spontaneous curvature similar to that of micelle-forming surfactants. While holey vesicles have been observed in the mixtures of bilayerforming and micelle-forming amphiphiles,33 where the micelleforming surfactant provides the highly positive spontaneous curvature component needed for the formation of holes, in the case of DMPG, the origin of the shift toward positive curvature at the onset of the intermediate state is less clear. Since there appears to be no major change in the ionization state of DMPG (see Figure 9), the positive spontaneous curvature of DMPG molecules probably must result from the loss of attractive interactions between the headgroups. Though hydrogen-bonding for DMPG in these conditions has not been studied as far as we know, in several kinds of systems, extensive inter-PG hydrogen-bonding has been described.35,36 We suggest that in the gel-like phase extensive inter-DMPG hydrogen bonding balances the electrostatic repulsion between the anionic headgroups, but that this interaction becomes lost upon the onset of bilayer melting when the gel phase lattice breaks and the area per DMPG molecule increases. Because of the loss of hydrogen-bonding, the electrostatic repulsion becomes dominant and the effective spontaneous curvature changes to highly positive, and to accommodate the increased spontaneous curvature of the non-hydrogen-bonded DMPG into the rims of the holes the vesicles become holey. With increasing temperature in the intermediate state, the chain disorder gradually increases, the bilayer becomes thinner, and as a result the spontaneous curvature decreases, and DMPG molecules can again be accommodated into a hole-free vesicle. The biggest drawback of the holey vesicle model is that, unlike the sponge phase hypothesis, the holey vesicle model does not readily offer explanations for the increased viscosity in the holey vesicle phase or for the origin of the multipeaked endotherm. Possible Origin of High Viscosity in the Holey Vesicle Phase. Since the holey vesicle phase does not form a continuous (31) Almgren, M. J. Dispersion Sci. Technol. 2007, 28, 43–54. (32) Kakehashi, R.; Karlsson, G.; Almgren, M. J. Colloid Interface Sci. 2009, 331, 484–493. (33) van Dam, L.; Karlsson, G.; Edwards, K. Biochim. Biophys. Acta 2004, 1664, 241–256. (34) He, X.; Schmid, F. Phys. Rev. Lett. 2008, 100, 137802-1–137802-4. (35) Zhang, Y.-P.; Lewis, R. N. A. H.; McElhaney, R. N. Biophys. J. 1997, 72, 779–793. (36) Epand, R. M.; Gabel, B.; Epand, R. F.; Sen, A.; Hui, S. W.; Muga, A.; Surewicz, W. Biophys. J. 1992, 63, 327–332.

Langmuir 2010, 26(7), 4892–4900

Article

network, an alternative explanation for viscosity is required. It has been suggested that the increased viscosity originates from increased electrostatic repulsion between DMPG vesicles within the intermediate state.10,25 Based on the lipid/aqueous partitioning of a cationic spin label, it was suggested that the absolute value of negative surface potential of DMPG increases strongly upon entering the intermediate state.25 Yet, our ζ-potential measurement shows no significant change at Ton (Figure 9), in contrast to the cationic spin label partitioning.25 In the earlier work, it was assumed, however, that the change in partitioning derives from the change in surface potential, whereas in the light of our ζ-potential measurements the change in the partitioning more likely results from the higher intrinsic partitioning coefficient of the spin label into the fluid compared to the gel phase or into the positive curvature defects compared to gel phase. When the melting begins and the positive curvature defects form, the binding of the cationic spin label increases drastically even without a change in the surface potential. Notably, there is a caveat with respect to our ζ-potential measurements, as we did not measure the viscosity of the samples and used aqueous solution viscosity in measurement. At relatively dilute 1 mM DMPG, the effect of the intermediate state on viscosity is small, nevertheless, and since the vesicle surface potential should be directly proportional to viscosity (i.e., if vesicle surface potential = 0, viscosity = 0) for the changes in ζ-potential to be masked by changes in surface-potential-induced changes in viscosity, it seems unlikely that the lack of any change in ζ-potential is an artifact resulting from using aqueous viscosity values in the calculations. A lack of significant change in the ionization degree in this dilute salt solution (7.2 mM Naþ, 1 mM DMPG-, 1 mM EDTA2-, 4.2 mM Hepes-) is further corroborated by the lack of abrupt changes in the conductivities measured with the Malvern Zetasizer Nano instrument using a high frequency waveform. The holes themselves may, however, indirectly explain the increased viscosity of the solution. Based on the previously published data, the radius of the vesicles below the intermediate state is on the order of 70 nm.10 In X-ray scattering measurements, the repeat distance related to the supposed holes was 40 nm.8 This sets the upper limit for the size of the holes as we define the hole size, that is, from the edge of the normal bilayer, with the curved rim covering the edge of the normal bilayer considered to be within the hole. When the hole diameter is significantly larger than the thickness of the bilayer, the volume of the lipid bilayer that would be required to fill the hole is larger than the volume of the lipid rim that covers the edge of the bilayer facing the hole. Accordingly, in order to accommodate a hole or holes, the bilayer outside the hole needs to become rippled or the vesicle needs to slightly expand. Assuming the latter to be the dominant mechanism, we calculated the increase in the apparent radius of the vesicle upon introduction of increasing number of holes, as well as the fraction of the volume of the rims of the total lipid volume, that is, the fraction of lipids in the rims of the holes, and the effective relative volume fraction within the outer surface of the vesicles (see Supporting Information section 2 for details on the calculations). The results show that the apparent volume fraction within the holey vesicles increases strongly with increasing numbers of holes (Figure 11). The increased apparent volume fraction, a major determinant of viscosity,37 could explain the observed viscosity increase.2 Multipeaked Endotherm. The only explanation forwarded for the multipeaked endotherm suggests that the free energy levels for flat and curved bilayers cross at the beginning and at the end of (37) Douglas, J. F.; Garboczi, E. J. Adv. Chem. Phys. 1995, 91, 85–153.

DOI: 10.1021/la100411p

4899

Article

Figure 11. Apparent holey vesicle radius R (continuous line), volume fraction of the rims of the total lipid volume Vrims/V0 (dashed line), and relative volume fraction occupied by the particles (dotted line) as a function of the number of 20 nm (red lines) or 40 nm (black lines) diameter holes.

the intermediate state, leading to transitions from flat, gel-like bilayers to the curved intermediate state, melting transition during the intermediate state, and the transition from the fluid intermediate state to a fluid, flat bilayer.2 This behavior (if complete) would thus result in three different components in the endotherms. The endotherms, however, appear to have four components, so this explanation would be incomplete. In all recent studies1-13 on the intermediate state, a mixture of two different chemical compounds has been used. 1,2-Dimyristoyl-sn-glycero3-phospho-rac-glycerol is a mixture of two diastereoisomers: 1,2dimyristoyl-sn-glycero-3-phospho-sn-10 -glycerol and 1,2-dimyristoyl-sn-glycero-3-phospho-sn-30 -glycerol. In contrast to enantiomers, diastereoisomers typically have somewhat different chemical and physical properties, and the two DMPG diastereoisomers are known to have different interactions with Naþ.38 The endotherms of pure diastereoisomers show the presence of peaks at Ton and Toff, but peaks at T1 and T2 appear to be missing in the samples scanned immediately after hydration,39 suggesting that the peaks at T1 and T2 could be related to either the mixing of the diastereoisomers or to slightly different transition temperatures for the two diastereoisomers, though perhaps the different buffer and hydration protocol used in ref 39 could also cause the difference observed. In this case, the reorganization responsible for these two additional peaks remains unknown, though they could reflect continuing melting in the rims/or bilayer. Tpost and Apparent Neutralization of Vesicles Surfaces after Tpost. It is difficult to explain with this hole model why the disappearance of holes should come about in two stages, first at Toff with associated enthalpy and second at Tpost without noticeable enthalpy in DSC. It also appears that above Tpost DMPG (38) Lotta, T. I.; Salonen, I. S.; Virtanen, J. A.; Eklund, K. K.; Kinnunen, P. K. J. Biochemistry 1988, 27, 8158–8169. (39) Salonen, I. S.; Eklund, K. K.; Virtanen, J. A.; Kinnunen, P. K. J. Biochim. Biophys. Acta 1989, 982, 205–215. (40) Egorova, E. M. Colloids Surf., A 1998, 131, 7–18.

4900 DOI: 10.1021/la100411p

Alakoskela et al.

becomes nonionic (Figure 9), unless this is an artifact deriving from unknown causes. Gershfeld and co-workers reported that DMPG above the intermediate state (pH was 6.2 in their measurements) becomes almost completely protonated, resulting in neutral DMPGH.4,5 The observed ζ-potential changes are in agreement with their reports, though it is difficult to understand what would shift the balance by so much, when the surface pKa for the PG headgroup is in the order of 2.8,40 particularly since the surface potential contribution becomes negligible when the surface potential approaches zero. For the same reason, counterion condensation appears unlikely as well. Since a transition related to headgroup conformation of DMPG has been suggested to take place at Tpost,41,42 it is possible that changed headgroup conformation that both decreases local dielectric permittivity and either strongly favors the chelation of Naþ ions or allows for additional hydrogen bonds within a PG headgroup or between PG headgroups for the protonated PG only could explain both our ζ-potential and the results of others.4,5

Conclusion We have showed that the thermal stability of the intermediate state can be changed in a highly predictable way by modifying the spontaneous curvature of the DMPG films by additives, and thus provided evidence that the intermediate state is a structure with positive curvature (defects), presumably a perforated/holey vesicle suggested before.8,11 This structure likely forms because the melting breaks inter-DMPG- hydrogen bonds, resulting in DMPG- with highly positive intrinsic curvature, and disappears when increasing chain disorder again shifts the spontaneous curvature in favor of the vesicles. The expansion of the vesicles to accommodate the holes should increase the effective volume fraction of the vesicles, leading to the observed increased viscosity within the intermediate state. The continuous network at high concentrations12,13 may represent the bulk holey lamellar phase, which appears for holey vesicles at high concentrations, similar to the bulk lamellar phase for normal vesicles.43 Acknowledgment. We thank Kristiina S€oderholm for technical assistance with the experiments. HBBG is supported by grants from the Sigrid Juselius Foundation and the Finnish Academy. J.-M.A. and M.J.P. were partly supported by Research Foundation of Orion Corporation and Finnish Medical Foundation, respectively, during this study. Supporting Information Available: Additional figures; calculation of the area and volume in the rims of the hole. This material is available free of charge via the Internet at http://pubs.acs.org. (41) Tajima, K.; Imai, Y.; Nakamura, A.; Koshinuma, M. Adv. Colloid Interface Sci. 2000, 88, 79–97. (42) Tajima, K.; Tsutsui, T.; Imai, Y.; Nakamura, A.; Koshinuma, M. Chem. Lett. 2002, 50–51. (43) Rasmusson, M.; Olsson, U. Prog. Colloid Polym. Sci. 2002, 120, 74–82.

Langmuir 2010, 26(7), 4892–4900