Calcium-Induced Formation of Subdomains in Phosphatidylethanolamine−Phosphatidylglycerol Bilayers: A Combined DSC, 31 P NMR, and AFM Study

Calcium-Induced Formation of Subdomains inPhosphatidylethanolamine-Phosphatidylglycerol Bilayers: A Combined DSC, 31P NMR, andAFM Study

Laura Picas,† M. Teresa Montero,†,‡ Antoni Morros,§,| Miquel E. Cabanas,⊥ Bastien Seantier,#Pierre-Emmanuel Milhiet,#,∇ and Jordi Hernandez-Borrell*,†,‡

Departament de Fisicoquımica, Facultat de Farmacia, UniVersitat de Barcelona (UB), and Institut deNanociencia i Nanotecnologia de la UniVersitat de Barcelona (IN2UB), E-08028 Barcelona, Spain, Unitat deBiofısica, Departament de Bioquımica i Biologia Molecular, Facultat de Medicina, Centre d’Estudis enBiofısica (CEB), and SerVei de Ressonancia Magnetica Nuclear (SeRMN), UAB, E-08193 Bellaterra,Barcelona, Spain, and Inserm, Unite 554, Montpellier, France, and Centre de Biochimie Structurale, UniVersitede Montpellier, CNRS, UMR 5048, Montpellier, France.

ReceiVed: NoVember 21, 2008; ReVised Manuscript ReceiVed: January 26, 2009

We study the effect of Ca2+ on the lateral segregation of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethano-lamine (POPE) and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol (POPG) (3:1, mol/mol). Supportedlipid bilayers (SLBs) were observed by atomic force microscopy (AFM). Since SLBs are formed from liposomesof POPE:POPG, we examined the effect of calcium on these suspensions by differential scanning calorimetry(DSC) and 31P nuclear magnetic resonance spectroscopy (31P NMR). AFM images revealed the existence oftwo separated phases, the higher showing a region with protruding subdomains. Force spectroscopy (FS) wasapplied to clarify the nature of each phase. The values of breakthrough force (Fy), adhesion force (Fadh), andheight extracted from the force curves were assigned to the corresponding gel (L!) and fluid (LR) phase. Theendotherms obtained by DSC suggest that, in the presence of Ca2+, phase separation already exists in thesuspensions of POPE:POPG used to form SLBs. Due to the temperature changes applied during preparationof SLBs a 31P NMR study was performed to assess the lamellar nature of the samples before spreading themonto mica. With in situ AFM experiments we showed that the binding of Ca2+ to POPG-enriched domainsonly induces the formation of subdomains in the L! phase.

1. Introduction

The physiological activity of transmembrane proteins maybe influenced by or depend on the physicochemical propertiesof the neighboring phospholipids. According to the matchingprinciple,1 such phospholipids provide a hydrophobic environ-ment suitable to host the protein in its natural folded state. Thus,lactose permease (LacY) from Escherichia coli (E. coli) requires,at least, phosphatidylethanolamine (PE) for appropriate foldingand activity.2-4 Although LacY has been reconstituted intoliposomes or crystallized in two dimensions with zwitterionicphospholipids,5 it is normally reconstituted in a biomimeticenvironment such as 1-palmitoyl-2-oleoyl-sn-glycero-3-phos-phoethanolamine (POPE) and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol (POPG) or the total lipid extract of E. colimembranes.6

In previous studies, we used atomic force microscopy (AFM)to reveal the nanostructure of the supported lipid bilayers (SLBs)of POPE:POPG (3:1, mol/mol) and we observed laterallysegregated domains.7 These SLBs were obtained by spreadingliposomes on a mica substrate, using a buffer containing 0.19

mM CaCl2. Ca2+ favors not only the extension of SLBs butalso interaction between phospholipid species.8 Although Ca2+

improves the spreading of SLBs on mica,9 the concentrationused was higher than physiological levels.10 This could affectdomain formation and, eventually, protein insertion. Bacterialmembrane domains can be studied by fluorescence11 and AFM.12

The role of domains in the membrane physiology of botheukaryotic and prokaryotic cells has now been recognized.13

However, there is little information on the physicochemicalproperties that induce their formation.

In the present study, before attempting protein insertion, westudied the effect of Ca2+ on the lateral segregation of POPEand POPG by observing the SLBs with AFM and theirnanomechanical properties using force spectroscopy (FS). TheSLBs are formed from liposomes of POPE:POPG (3:1, mol/mol); therefore, we examined the effect of calcium on thesesuspensions by differential scanning calorimetry (DSC) and 31Pnuclear magnetic resonance spectroscopy (31P NMR). Since thelateral separation of phospholipid components in differentdomains may affect the insertion of proteins such as LacY intothe SLBs of POPE:POPG (3:1, mol/mol), it is important to studythe influence of Ca2+ on the origin of these domains.

2. Materials and Methods

1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE),specified as 99% pure, and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol (POPG) were purchased from Avanti PolarLipids (Alabaster, AL). Buffer A consisted of 20 mM Hepes

* To whom correspondence should be addressed. E-mail:[email protected].

† Departament de Fisicoquımica, UB.‡ IN2UB.§ Departament de Bioquımica i Biologia Molecular, UAB.| CEB, UAB.⊥ SeRMN, UAB.# Inserm.∇ Universite de Montpellier, CNRS, UMR 5048.

J. Phys. Chem. B 2009, 113, 4648–46554648

10.1021/jp8102468 CCC: $40.75 2009 American Chemical SocietyPublished on Web 03/10/2009

Dow

nloa

ded

by U

NIV

OF

BARC

ELO

NA

on

Aug

ust 1

0, 2

009

Publ

ished

on

Mar

ch 1

0, 2

009

on h

ttp://

pubs

.acs

.org

| do

i: 10

.102

1/jp

8102

468

(pH 7.40) and 150 mM NaCl; buffer B, of 20 mM Hepes (pH7.40), 150 mM NaCl, and 10 mM CaCl2; buffer C, of 20 mMHepes (pH 7.40), 150 mM NaCl, and 20 mM CaCl2; and bufferD, of 20 mM Hepes (pH 7.40), 150 mM NaCl, and 10 mMethylenediaminetetraacetic acid (EDTA). All buffers wereprepared in ultrapure water (Milli Q reverse osmosis system,18.3 M! · cm resistivity). HPLC-grade chloroform and methanoland n-dodecyl-!-D-maltoside (DDM) were purchased fromSIGMA (St. Louis, MO).

2.1. Differential Scanning Calorimetry. Multilamellarvesicles (MLVs) for DSC studies were prepared at 50 °C in 20mM Hepes buffer and 150 mM NaCl, pH 7.4 (buffer A), in theabsence or presence of 10 mM CaCl2 to a final concentrationof 2.74 mM. A MicroCal MC-2 calorimeter was used for DSCanalyses, following procedures described elsewhere.14,15 The datawere analyzed by the original calorimeter software. Thetemperature of maximum excess specific heat (Tm) was measuredto the nearest 0.5 °C. Transition enthalpy was calculated in linewith previously described methods.16 The calorimetry accuracyfor Tm and for enthalpy changes was (0.1 °C and (0.2kcal ·mol-1, respectively. Each sample was scanned in triplicateover the temperature range 5-80 °C at a scan rate of 0.44°C ·min-1.

Tm and the intensities of the overlapping components of theDSC endotherm in the presence of 10 mM Ca2+ were assessedas follows. The second derivative, calculated using the Savitsky-Golay procedure (third-grade polynomial, 27° of smoothing),and Fourier self-deconvolution (" factor, 0.3) were used todetermine the minimum number of components and to locatetheir peak positions by means of the program GRAMS/32(Galactic Industries). The curve was fitted by using the fourparameters of the Weibull curve, a built-in function of the SigmaPlot 7.0 program (SPSS Inc.) used to quantify greatly overlap-ping asymmetrical chromatographic peaks.17,18 The sum of twoWeibull functions was fitted to the profile of the complexendotherm by least-squares iteration.

2.2. 31P NMR Spectroscopy. MLVs for 31P NMR studieswere prepared at 50 °C in 20 mM Hepes buffer, 150 mM NaCl,and 10 mM CaCl2, pH 7.4 (buffer B), to a final concentrationof 2.74 mM. A 25 mL aliquot of the resulting unsonicated MLVsuspension was pelleted by ultracentrifugation at 115000g for1 h at 10 °C. The hydrated pellet was then resuspended in 300µL of supernatant and placed in a conventional 5 mm NMRtube. A capillary tube containing 2H2O was added for field-frequency stabilization and the NMR tube was sealed after apurge with N2.

31P NMR spectra were recorded as previously described19 ona Bruker AvanceIII-400 spectrometer (Bruker Espanola, S.A.,Madrid, Spain) operating at 161.98 MHz using a single 90° pulsesequence, with proton-decoupling during signal sampling bymeans of a Waltz-16 composite pulse sequence.20,21 A singlepulse sequence was used instead of the phase-cycled Hahn echopulse sequence22 to obtain spectra with higher signal-to-noiseratios. Each spectrum was the result of accumulating 3072 scanssampled using 2048 complex data points, with a 90° pulse of 9µs (Beff ) 27.78 kHz), an interpulse delay of 2.1 s, and a spectralwidth of 50 kHz. Exponential multiplication resulting in a linebroadening of 50 Hz was applied before Fourier transformationto improve the signal-to-noise ratio.23 Spectra were processedon a personal computer running the TopSpin v. 2.0 software(Bruker Biospin GmbH, Germany) on Debian GNU/Linux v.3.1. All chemical shift values are quoted in parts per million(ppm) with reference to external 85% phosphoric acid in H2O(0 ppm), with positive values referring to low-field shifts.

2.3. SLB Formation. SLBs were prepared according to amethod described elsewhere.24 Briefly, large unilamellar vesicles(LUVs) were prepared by extrusion of MLVs of POPE:POPGin 20 mM Hepes buffer, 150 mM NaCl, and 10 mM CaCl2, pH7.4 (buffer B), through 100 nm pore filters (Nucleopore). LUVswere deposited onto freshly cleaved mica disks mounted on aTeflon O-ring and incubated at 50 °C for 2 h. The bilayers werealways kept in an aqueous environment and carefully rinsedwith the same buffer before imaging.

2.4. AFM Imaging. AFM experiments were performed witha multimode microscope controlled by Nanoscope IIIa electron-ics (Digital Instruments, Santa Barbara, CA). Images wereacquired in tapping mode (TM-AFM) at minimum vertical force,maximizing the amplitude setpoint value and maintaining thevibration amplitude as low as possible. V-shaped Si3N4 canti-levers (MLCT-AUNM, Veeco) with a nominal spring constantof 0.10 N ·m-1 were used in liquid operation.

2.5. Force Spectroscopy. Force spectroscopy experimentswere carried out in liquid media using V-shaped Si3N4 tips(MLCT-AUNM) with a nominal spring constant of 0.10 N ·m-1.However, individual spring constants were calibrated using thethermal noise method.25

Typically, 300 force plots were performed and recorded, withthe laser spot maintained in the same position on the cantileverin order to keep the corresponding photodetector sensitivityconstant (V ·nm-1).26 All spectroscopy experiments were per-formed at a constant cantilever linear velocity of 0.5 µm · s-1 inorder to avoid any velocity-dependent effect. Applied forces Fare given by F ) kc", where " is the cantilever deflection andkc is the cantilever spring constant.

3. Results and Discussion

DivalentcationsarenormallyusedtoenhanceSLBformation.9,27

Thus, both adsorption and rupture of liposomes on a micasupport are favored by the presence of Ca2+, and they are bothspecific requirements for screening the charge of mica for thosesamples containing negatively charged phospholipids.28 BothCa2+ and Mg2+ neutralize the charge of the mica surface andmay form bridges with anionic phospholipids,28 inducing majorreorganization in charged and zwitterionic phospholipids.29 Suchinteractions can modify the behavior of phospholipids at thesurface, leading to phase separation,30 or they can increase thenanomechanical resistance of the membrane.9

Figure 1 illustrates the effect of Ca2+ on the formation ofSLBs of POPE:POPG (3:1, mol/mol). Concentrations of CaCl2

ranging from 0 to 20 mM were used. As can be seen, no planarstructures are formed either in the absence (Figure 1A) or inthe presence of 5 mM (Figure 1B) CaCl2, and only phospholipidaggregates and nonfused liposomes are found. As reportedelsewhere,9 vesicles may remain intact on the substrate, yieldingsupported vesicular layers. Thus, once a critical vesicularcoverage is reached, vesicle rupture and the formation of bilayerpatches are favored, promoting interaction with adjacent lipidmaterial and leading to the formation of extended bilayerpatches. This can be seen from Figure 1A,B, where the areaformed by lipid patches in the absence of calcium is almostnegligible, but it increases substantially with 5 mM of CaCl2.At 10 (Figure 1C) and 20 (Figure 1D) mM CaCl2 a completecoverage of the substrate is achieved. Judging only fromtopography, these two images could be interpreted as continuousplanar bilayers constituted by two separated phases: the darkerand the brighter regions.

The step height differences between the darker and brighterregions in Figure 1C,D can be established by line profile analysis

Subdomain Formation in POPE-POPG Bilayers J. Phys. Chem. B, Vol. 113, No. 14, 2009 4649

Dow

nloa

ded

by U

NIV

OF

BARC

ELO

NA

on

Aug

ust 1

0, 2

009

Publ

ished

on

Mar

ch 1

0, 2

009

on h

ttp://

pubs

.acs

.org

| do

i: 10

.102

1/jp

8102

468

as 3.5 ( 0.5 nm (n ) 10; Figure 1E) and 3.25 ( 0.25 nm (n )10; Figure 1F), respectively. It is worth mentioning that thebrighter region contains subdomains that protrude 0.50 ( 0.07nm (n ) 10). These height measurements give rise to threehypotheses: (i) SLBs are constituted only by the brighter regionsthat cover a small part of the substrate (dark background); (ii)the images show a continuous bilayer (dark region) with anotherbilayer fragment superimposed (brighter region); or, alterna-tively, (iii) as phospholipids are thermotropic and show well-defined phase transition behavior from gel (L!) to fluid (LR)phases, the images reflect a single continuous bilayer thatdisplays the L! and the LR phases.

To study the nature of the background region and confirm orrefute its lipid nature, a DDM solution was injected into theAFM fluid cell. Figure 2 shows the topographic images before(A) and after (B and C) detergent injection. As can be seen,

while the brighter regions (see white arrows in Figure 2A,B)are clearly disrupted by DDM, the effect on the darker regionis much more discrete. Remarkably, the solubilization effecton the background region becomes more significant with time(Figure 2C). Hence, we conclude that hypothesis (i) is falseand that the dark background in Figures 1C,D and 2 is indeedlipid material, most probably an SLB that covers the substrate.

However, the heights in Figure 1E,F for the brighter regionscompare well with the size obtained by molecular dynamicscalculations31 and may suggest that we have in fact obtained acontinuous bilayer (darker region) with a bilayer structuresuperimposed32 (brighter region). Nevertheless, under ourexperimental conditions, the step height values between theregions may be overestimated as a result of the repulsionbetween the tip and the negative charge conferred by POPG onthe bilayer. This effect has previously been observed in othersystems33 and would tentatively lead us to exclude hypothesis(ii) in favor of hypothesis (iii). Thus, parts C and D of Figure1 would reflect a single continuous bilayer that displays the L!and LR phases at a certain stage of phase transition. To furtherstudy the nature of these two lipid regions and to unambiguouslyassign the corresponding phase to the brighter or darker regions,force curve analysis was performed.

FS has previously been applied to the characterization ofphase separation of phospholipid monolayers and bilayers,34-36

and it can provide quantitative information on the nanostructureand nanomechanics of two separated lipid phases.37 Figure 3Ashows a continuous SLB of the lipid mixture at 10 mM calciumconcentration, and parts B and C of Figure 3 are two examplesof the vertical force versus piezo displacement curve (F#-D)from the highlighted darker and brighter regions, respectively.

The first information we obtain comes from the approachingpart of the FS curves (see magnifications in Figure 3D,E). Thus,when the gradient of the van der Waals attractive forces exceedsthe gradient of the tip spring constant and the repulsive forces,the tip jumps onto the surface.38 In our FS experiments,while this jump to contact can be observed when the tipapproaches the L! phase (arrow in Figure 3D), it is no longerseen when it approaches the LR phase (arrow in Figure 3E).Therefore, we can assume that a certain electrostatic repulsionoccurs when the tip approaches the L! phase, which suggeststhat this region is actually enriched in the negatively chargedphospholipid, POPG. This provides an explanation of theoverestimation of the height of the brighter regions reportedabove and supports hypothesis iii. As reported elsewhere,33 andin further support of this interpretation, deviations in heightdetermination should be attributed to differences in surfacecharge densities that would affect the electrostatic double-layer

Figure 1. TM-AFM images acquired after spreading POPE:POPG (3:1, mol/mol) liposomes in 20 mM Hepes, 150 mM NaCl (pH 7.40) buffersupplemented with the following: 0 (A), 5 (B), 10 (C), and 20 mM (D) CaCl 2. The z color scale is 20 nm, and the scale bar is 2 µm. A cross-sectiontaken along the white line in images C and D is presented beneath in the topographic images (E and F, respectively).

Figure 2. TM-AFM images showing the in situ effect of adding DDMat 2(CMC) to an SLB of POPE:POPG (3:1, mol/mol) in the presenceof 10 mM CaCl2 (A). After adding the detergent, successive images(B and C) were obtained. The z color scale is 20 nm, and the scale baris 1 µm.

4650 J. Phys. Chem. B, Vol. 113, No. 14, 2009 Picas et al.

Dow

nloa

ded

by U

NIV

OF

BARC

ELO

NA

on

Aug

ust 1

0, 2

009

Publ

ished

on

Mar

ch 1

0, 2

009

on h

ttp://

pubs

.acs

.org

| do

i: 10

.102

1/jp

8102

468

repulsion between the tip and the sample. In fact, more hydratedcations, such Ca2+ or Mg2+, could exert repulsive forces whenadsorbed to the surfaces.38 In fact, Ca2+ bonds specifically toPOPG,39 promoting the formation of solid “clusters” that coexistwithin the bulk fluid lipid phase. Hence, the repulsion betweenthe tip and the L! phase could result from these hydration forces.

F#-D curves provide the threshold force (Fy), that is, theforce required to indentate a bilayer down to the substrate.37

The values of Fy obtained from 250 F#-D plots are shown inhistogram A in Figure 4. As can be seen, two mean Fy valuesare found at 0.243 ( 0.070 and 0.915 ( 0.190 nN (Table 1),which can be attributed to the LR and L! phases, respectively.Since two subdomains are present in the L! phase, a widerdistribution of the Fy values is observed. The mean Fy valuesfound for both the LR and L! phases are slightly smaller thanthose reported for other SLBs;8,40-42 however, they indicate thatthe two lipid phases exhibit different nanomechanical behavior.

The adhesive force (Fadh) can be determined from themagnitude of the pull of jump in the retraction portion of theF#-D curves34-36 (see arrows in Figure 3B,C). The magnitudeof Fadh for the LR and L! phases and their frequencies aredisplayed in histogram B in Figure 4. The mean value of Fadh

was 1.580 ( 0.125 and 0.243 ( 0.060 nm for the LR and L!

phases, respectively (Table 1). These results are in agreementwith the observation that the fluidity of a bilayer is proportionalto the adhesion force observed.43 Furthermore, the value of Fadh

for the LR phase is similar to the value reported for POPS,44 aheteroacid phospholipid that, like POPG, binds Ca2+ and whichis in fluid phase at room temperature. Finally, by representingFy versus the tip-sample distance, we can obtain the depth ofpenetration and, hence, estimate the bilayer height.37,45 Theheight values obtained by this method are plotted in histogramC in Figure 4. The mean height values were 3.300 ( 0.184 and4.400 ( 0.440 nm for the LR and L! phases, respectively (Table1). Thus, the step height difference between the phases thatresults from FS curves is lower than that obtained from lineprofile analysis (Figure 1E,F) and is in agreement with the stepheight differences between fluid and solid phases found inmixtures of other phospholipids. This result provides anadditional explanation of the overestimation of the height ofthe brighter regions reported above and reinforces the interpreta-tion of the images given in hypothesis iii. The whole of the FS

Figure 3. TM-AFM image of an SLB of POPE:POPG (3:1, mol/mol) acquired in the presence of 10 mM CaCl2 (A). The z color scale is 15 nm,and the scale bar is 2 µm. The pattern of the F#-D curves in the highlighted areas is depicted in B and C, corresponding to the liquid (darkerregion) and gel (brighter region) phases. The approach region of the curve is depicted in blue, whereas the retracting of the cantilever is presentedin red. A magnification of the jump-to-contact region of approach F#-D curves in images B and C is shown in D and E, respectively.

Figure 4. Histograms corresponding to the experimental Fy (A), Fadh (B), and bilayer height (C) values obtained for the two lipid phases studied:LR in gray and L! in black.

TABLE 1: Mean Values of Fy, Fadh, and Bilayer HeightObtained from Experimental F!-D Curves for the Lr(Fluid) and L" (Gel) Phases of a Continuous SLB ofPOPE:POPG (3:1, mol/mol)

lipid phase Fy (nN) Fadh (nN) bilayer height (nm)

LR 0.243 ( 0.070 1.580 ( 0.125 3.300 ( 0.184L! 0.915 ( 0.190 0.243 ( 0.060 4.440 ( 0.440

Subdomain Formation in POPE-POPG Bilayers J. Phys. Chem. B, Vol. 113, No. 14, 2009 4651

Dow

nloa

ded

by U

NIV

OF

BARC

ELO

NA

on

Aug

ust 1

0, 2

009

Publ

ished

on

Mar

ch 1

0, 2

009

on h

ttp://

pubs

.acs

.org

| do

i: 10

.102

1/jp

8102

468

results support the idea that the SLBs shown are a single bilayerdisplaying the LR and L! phases.

To study whether the nanostructure of the POPE:POPG SPBsoriginates on the mica surface or whether it already exists inthe liposomes used to prepare the SLBs, we performed DSCand a 31P NMR study of the POPE:POPG (3:1, mol/mol)multilamellar liposomes.

Figure 5A shows the normalized DSC endotherms for thisphospholipid mixture in the absence (lower trace) and in thepresence (upper trace) of 10 mM Ca2+. Table 2 shows valuesof several thermodynamic parameters of the observed phasetransitions: the temperature of maximal excess heat capacity(Tm, °C), the peak width at half-height ("T, °C) and thetransition enthalpy change ("H, kJ ·mol-1). In the absence ofCa2+, only one phase transition appeared in the temperature

range studied, corresponding to the L!-to-LR main phasetransition; its Tm occurred at 20.0 °C. The heat capacity curveis asymmetrical, skewed toward the low-temperature side.However, in the presence of 10 mM Ca2+ this main transitionoccurred at 23.5 °C and the peak was significantly narrower,whereas the enthalpy change was only slightly smaller than itwas in the absence of the divalent cation (Table 2). We mayconclude that the binding of Ca2+ to POPG induces a reductionin its surface charge and area and causes tighter packing of thelipid lattice, thus increasing the phase transition temperature.In addition to the major transition mentioned, in the presenceof 10 mM Ca2+ a very low enthalpy second endotherm alsoappeared with a peak at 61.1 °C.

The shape of the main transition peak in the presence of Ca2+

was characteristic of a complex endotherm. This indicates aphase separation in the POPE:POPG mixture. Fitting the resultsto two asymmetrical curves (Weibull functions; see Materialsand Methods) skewed to low temperatures yielded a much betterfit (Figure 5C) than that obtained with symmetrical curves(Gauss functions, results not shown). Conversely, the applicationof the same resolution enhancement techniques to the maintransition peak in the absence of Ca2+ did not reveal overlappingpeaks (Figure 5B,D,F). Thus, we conclude that, on binding toPOPG, Ca2+ induces lipid phase separation. A solid phaseconsisting of a Ca2+-POPG complex is probably formed withinthe fluid phase of POPE.39

Figure 5. Normalized DSC thermograms of POPE:POPG (3:1, mol/mol) multilamellar liposomes in the absence of Ca2+ (lower line) andwith 10 mM CaCl2 (upper line) added to 20 mM Hepes buffer, 150mM NaCl, pH 7.4. Only the first heating scan is shown for each study,as there were no significant differences in the second and third scans.A 10! magnified profile is included as an inset to show the bilayer tononbilayer phase transition in the presence of 10 mM Ca2+. The totalphospholipid concentration was 2.0 mg ·mL-1 and the heating scan ratewas 0.44 °C ·min-1 (A).Resolution enhancement of the main phasetransition profile in the absence of Ca2+ (B, D, and F) and in thepresence of 10 mM Ca2+ (C, E, and G). Parts B and C show theexpanded experimental trace of the main transition (continuous line)and, in the presence of 10 mM Ca2+ (C), the best fit (dotted lines) witha sum of two Weibull asymmetrical peaks (broken lines). Parts D andE show the second derivatives; and F and G show the Fourier self-deconvoluted traces.

TABLE 2: Thermodynamic Parameters of the PhaseTransitions of POPE:POPG (3:1) Mixtures (Mean ( SEM)a

sample n Tm (°C) "T (°C) "H (kJ ·mol-1)

absence of Ca2+ 4 20.0 ( 0.6 4.5 ( 0.2 19.8 ( 1.210 mM Ca2+ 2 23.5 ( 0.1 2.8 ( 0.2 17.5 ( 0.4

2 (61.1 ( 1.1) (5.8 ( 0.5) (1.7 ( 0.2)

a The thermodynamic parameters of the bilayer to non-bilayerphase transition are shown in brackets.

Figure 6. Solid-state 31P NMR spectra of POPE:POPG (3:1, mol/mol)multilamellar liposomes in 10 mM CaCl2 added to 20 mM Hepes buffer,150 mM NaCl, pH 7.4. Spectra are shown at full scale. All chemicalshift values are quoted in parts per million with reference to external85% phosphoric acid in H2O (0 ppm), with positive values referring tolow-field shifts. By comparison with the DSC thermogram (see Figure5), these spectra can be assigned as follows: 5 and 10 °C, gel phase;23 °C, main phase transition; 30 °C, liquid-crystal phase; 62 and 65°C, very low enthalpy second endotherm.

4652 J. Phys. Chem. B, Vol. 113, No. 14, 2009 Picas et al.

Dow

nloa

ded

by U

NIV

OF

BARC

ELO

NA

on

Aug

ust 1

0, 2

009

Publ

ished

on

Mar

ch 1

0, 2

009

on h

ttp://

pubs

.acs

.org

| do

i: 10

.102

1/jp

8102

468

In the process of SLB formation (see Materials and Methods,subsection 2.3), the sample was incubated at 50 °C for 2 h andobservation was carried out at room temperature. This meansthat the sample has a thermal history and that phases other thanthe lamellar phase may appear (see inset Figure 5A). So weperformed a 31P NMR study in the range of temperatures usedin the formation of the SLBs. Thus, Figure 6 shows thecorresponding 31P NMR powder pattern spectra of POPE:POPG(3:1, mol/mol) liposomes in the presence of 10 mM CaCl2 atsix different temperatures. By comparison with the DSCthermogram in Figure 5, the spectra at 0 and 10 °C can beassigned to the L! phase, for which 23 °C corresponds to themain phase transition temperature and the 30 °C spectrum,therefore, corresponds to the LR phase. The shapes of all thesespectra are characteristic of lamellar organization,46 whichindicates that the presence of non-bilayer phases can be ruledout in this temperature range. However, the spectrum at 62 °Ccorresponds to the temperature of the very low enthalpy secondendotherm (inset in Figure 5A) and shows the coexistence ofthe hexagonal phase (HII) and an isotropic phase, probably acubic phase. The spectrum at 50 °C shows a minor shoulderbetween 5 and 10 ppm, which could be attributed to a smallamount of hexagonal and/or isotropic phases. This shoulder

would explain the observed increase at half-height from 10.4ppm at 30 °C to 15.7 ppm at 50 °C (see Figure 6).

Transition temperatures in liposomes and planar-supportedbilayers do not coincide due to the effect of the underlyingsubstrate surface on this property. Thus, because of theuncoupling of the two leaflets of the bilayer, the range oftemperatures that encompasses the transition is wider insupported bilayers than in liposomes. Therefore, when preparingthe SLBs, at least during incubation at 50 °C, the occurrenceof non-bilayer structures (HII phases) can be expected. However,as recently demonstrated,42 when the sample is cooled to roomtemperature for AFM observations, only lamellar structuresremain.

According to the information obtained from DSC, phaseseparation in SLBs (Figure 5B,C) can be expected. However,the existence of subdomains within one of the phases is stillunexplained. To examine how Ca2+ affects the organization ofPOPE:POPG (3:1, mol/mol) at the surface, we performed insitu AFM experiments. Thus, Figure 7A shows an SLB ofPOPE:POPG (3:1, mol/mol) formed in the presence of 10 mMCaCl2 (buffer B; see Material and Methods), where the L! phase(brighter regions) with protruding subdomains and the LR phase(darker regions) are observed. Once the image was obtained,

Figure 7. Sequence of TM-AFM images of SLBs of POPE:POPG (3:1, mol/mol) in 20 mM Hepes buffer, 150 mM NaCl (pH 7.40), showing theeffect of Ca2+: in the presence of 10 mM CaCl2 (A), after calcium removal with CaCl2-free buffer (B), after adding 20 mM CaCl2 buffer (C), andafter removing Ca2+ with a calcium-free buffer (D). The z color scale is 20 nm, and the scale bar is 2 µm.

Subdomain Formation in POPE-POPG Bilayers J. Phys. Chem. B, Vol. 113, No. 14, 2009 4653

Dow

nloa

ded

by U

NIV

OF

BARC

ELO

NA

on

Aug

ust 1

0, 2

009

Publ

ished

on

Mar

ch 1

0, 2

009

on h

ttp://

pubs

.acs

.org

| do

i: 10

.102

1/jp

8102

468

the tip was lifted up a few nanometers and calcium was removedby exchanging the liquid with a calcium-free buffer (buffer A).As can be seen, after 10 min for tip stabilization, there is noevidence of any segregated subdomain on top of the L! phase(Figure 7B). These observations demonstrate that only thesubdomains are calcium-dependent. Remarkably, when bufferA was exchanged for a buffer supplemented with 20 mMcalcium (buffer C), the subdomains reappear in the L! phase(Figure 7C). In corroboration, when buffer C is exchanged forbuffer A (calcium-free), the subdomains disappear again andflat featureless SLBs are obtained (Figure 7D).

In conclusion, while DSC experiments indicated that thereis lateral segregation of POPE:POPG liposomes in the presenceof Ca2+, the direct observation of SLBs suggests a more complexsituation. Thus, according to DSC and AFM, we conclude thatthere is a lateral phase separation between the L! and LR phasesin the POPE:POPG binary mixture. However, AFM reveals Ca2+

subdomains in the L! phase. In addition, we have evidence ofthe reversibility of subdomain formation. Quantitative informa-tion obtained from the height profile analysis shown in Figure7A,B is given in Table 3.

The most striking result is that the occurrence and reversibilityof subdomains were only observed in the L! phase and not inthe LR phase. Since the phospholipids that constitute the LR phasedid not bind Ca2+, the distal leaflet in the SLBs should be mainlyconstituted of POPE and almost depleted of POPG. In fact, ithas been shown elsewhere28,47 that when liposomes composedof anionic phospholipids are deposited on a charged substrate,there is an asymmetric distribution of phospholipids. The POPE:POPG (3:1, mol/mol) mixture may behave in the same way.Hence, in the fluid phase, POPG may preferentially be adsorbedonto the underlying mica, mainly driven by the electrostaticbridging effect of Ca2+. Consequently, the leaflet in immediatecontact with the mica becomes enriched in POPG, while thedistal leaflet becomes enriched in POPE (the darker backgrounddomain in Figures 1 and 7). Conversely, in the L! phase, POPEand POPG in the presence of Ca2+ are immiscible, leading tothe formation of subdomains (bright regions in Figures 1 and7). These subdomains reflect the fact that the L! phase is muchricher in POPG than the LR phase is, which is the reason whyonly this phase responds to Ca2+. Once the SLBs are washedout with Ca2+-free buffer, both phospholipid components mixideally and the subdomains are no longer observed (Figure7B,D).

The capacity of phospholipids to segregate into domains isof physiological interest, since it points to microenvironmentsin the vicinity of membrane proteins.48 Although the Ca2+

concentration used in this study is above physiological levels,concentrations could be higher in some pathological conditions,such as in the formation of kidney stones, where the presenceof negatively charged phospholipids enhances the formation ofcalcium oxalate crystals.49

Acknowledgment. L.P. is the recipient of an R&D fellowshipfrom the University of Barcelona. Special thanks to Dr. GerardOncins for helpful discussions. This study is supported byHispano-French bilateral action (Grant HF 2007-0028) and theGrant CTQ-2008-03922/BQU from the Spanish Government.

References and Notes

(1) Jensen, M. O.; Mouritsen, O. G. Lipids Do Influence ProteinFunction-The Hydrophobic Matching Hypothesis Revisited. Biochim. Bio-phys. Acta 2004, 1666, 205–226.

(2) Bogdanov, M.; Dowhan, W. Phosphatidylethanolamine is Requiredfor in Vivo Function of the Membrane-Associated Lactose Permease ofEscherichia coli. J. Biol. Chem. 1995, 270, 732–739.

(3) Bogdanov, M.; Dowhan, W. Phospholipid-Assisted Protein Folding:Phosphatidylethanolamine Is Required at a Late Step of the ConformationalMaturation of the Polytopic Membrane Protein Lactose Permease. EMBOJ. 1998, 17, 5255–5264.

(4) Bogdanov, M.; Xie, J.; Heacock, P.; Dowhan, W. To Flip Or Notto Flip: Lipid-Protein Charge Interactions Are a Determinant of FinalMembrane Protein Topology. J. Cell Biol. 2008, 182, 925–935.

(5) Zhuang, J.; Prive, G. G.; Werner, G. E.; Ringler, P.; Kaback, H. R.;Engel, A. Two-Dimensional Crystallization of Escherichia coli LactosePermease. J. Struct. Biol. 1999, 125, 63–75.

(6) Vazquez-Ibar, J. L.; Weinglass, A. B.; Kaback, H. R. Engineeringa Terbium-Binding Site into an Integral Membrane Protein for LuminescenceEnergy Transfer. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 3487–3492.

(7) Merino, S.; Domenech, O.; Vinas, M.; Montero, M. T.; Hernandez-Borrell, J. Effects of Lactose Permease on the Phospholipid Environmentin Which It Is Reconstituted: A Fluorescence and Atomic Force MicroscopyStudy. Langmuir 2005, 21, 4642–4647.

(8) Garcia-Manyes, S.; Oncins, G.; Sanz, F. Effect of pH and IonicStrength on Phospholipid Nanomechanics and on Deposition Process ontoHydrophilic Surfaces Measured by AFM. Electrochim. Acta 2006, 51, 5029–5036.

(9) Richter, R. P.; Berat, R.; Brisson, A. R. Formation of Solid-Supported Lipid Bilayers: An Integrated View. Langmuir 2006, 22, 3497–3505.

(10) Naseem, R.; Holland, I. B.; Jacq, A.; Wann, K. T.; Campbell, A. K.pH and Monovalent Cations Regulate Cytosolic Free Ca(2+) in E. coli.Biochim. Biophys. Acta 2008, 1778, 1415–1422.

(11) Vanounou, S.; Parola, A. H.; Fishov, I. Phosphatidylethanolamineand Phosphatidylglycerol Are Segregated into Different Domains inBacterial Membrane. A Study with Pyrene-Labelled Phospholipids. Mol.Microbiol. 2003, 49, 1067–1079.

(12) Zerrouk, Z.; Alexandre, S.; Lafontaine, C.; Norris, V.; Valleton,J. M. Inner Membrane Lipids of Escherichia coli Form Domains. ColloidsSurf., B 2008, 63, 306–310.

(13) Milhiet, P. E.; Domec, C.; Giocondi, M. C.; Van Mau, N.; Heitz,F.; Le Grimellec, C. Domain Formation in Models of the Renal BrushBorder Membrane Outer Leaflet. Biophys. J. 2001, 81, 547–555.

(14) Villaverde, J.; Cladera, J.; Padros, E.; Rigaud, J. L.; Dunach, M.Effect of Nucleotides on the Thermal Stability and on the DeuterationKinetics of the Thermophilic F0F1 ATP Synthase. Eur. J. Biochem. 1997,244, 441–448.

(15) Villaverde, J.; Cladera, J.; Hartog, A.; Berden, J.; Padros, E.;Dunach, M. Nucleotide and Mg2+ Dependency of the Thermal Denaturationof Mitochondrial F1-ATPase. Biophys. J. 1998, 75, 1980–1988.

(16) Tilcock, C. P.; Fisher, D. Interaction of Phospholipid Membraneswith Poly(ethylene glycol)s. Biochim. Biophys. Acta 1979, 557, 53–61.

(17) Hervas, C.; Martinez, A. C.; Silva, M.; Serrano, J. M. Improvingthe Quantification of Highly Overlapping Chromatographic Peaks by UsingProduct Unit Neural Networks Modeled by an Evolutionary Algorithm.J. Chem. Inf. Model. 2005, 45, 894–903.

(18) Grimalt, J.; Iturriaga, H.; Tomas, X. The Resolution of Chromato-grams with Overlapping Peaks by Means of Different Statistical Functions.Anal. Chim. Acta 1982, 139, 155–166.

(19) Grancelli, A.; Morros, A.; Cabanas, M. E.; Domenech, O.; Merino,S.; Vazquez, J. L.; Montero, M. T.; Vinas, M.; Hernandez-Borrell, J.Interaction of 6-Fluoroquinolones with DipalmitoylphosphatidylcholineMonolayers and Liposomes. Langmuir 2002, 18, 9177–9182.

(20) Killian, J. A.; Fabrie, C. H.; Baart, W.; Morein, S.; de Kruijff, B.Effects of Temperature Variation and Phenethyl Alcohol Addition on AcylChain Order and Lipid Organization in Escherichia coli Derived MembraneSystems. A 2H- and 31P-NMR Study. Biochim. Biophys. Acta 1992, 1105,253–262.

(21) Gomez-Fernandez, J. C.; Llamas, M. A.; Aranda, F. J. TheInteraction of Coenzyme Q with Phosphatidylethanolamine Membranes.Eur. J. Biochem. 1999, 259, 739–746.

(22) Rance, M.; Byrd, R. A. Obtaining High-Fidelity Spin-1. J. Magn.Reson. 1983, 52, 221–240.

TABLE 3: Height (h) Values Obtained from theCross-Section Profile Analysis of SLBs of POPE:POPG (3:1,mol/mol) in the Absence of Calcium (Figure 7B) and with 10mM Calcium (Figure 7A)a

h (nm)

n calcium-free buffer 10 mM calcium buffer

Darker domain 10 3.75 3.67Brighter domain 10 4.32

a The h (nm) values are the mean of 10 different sections (n )10) and were obtained in tapping mode.

4654 J. Phys. Chem. B, Vol. 113, No. 14, 2009 Picas et al.

Dow

nloa

ded

by U

NIV

OF

BARC

ELO

NA

on

Aug

ust 1

0, 2

009

Publ

ished

on

Mar

ch 1

0, 2

009

on h

ttp://

pubs

.acs

.org

| do

i: 10

.102

1/jp

8102

468

(23) Lewis, R. N.; Sykes, B. D.; McElhaney, R. N. Thermotropic PhaseBehavior of Model Membranes Composed of Phosphatidylcholines Contain-ing Cis-Monounsaturated Acyl Chain Homologues of Oleic Acid: Dif-ferential Scanning Calorimetric and 31P NMR Spectroscopic Studies.Biochemistry 1988, 27, 880–887.

(24) Milhiet, P. E.; Vie, V.; Giocondi, M. C.; Le Grimellec, C. AFMCharacterization of Model Rafts in Supported Bilayers. Single Mol. 2001,2, 109–112.

(25) Florin, E. L.; Rief, M.; Lehmann, H.; Ludwig, M.; Dornmair, C.;Moy, V. T.; Gaub, H. E. Sensing Specific Molecular Interactions with theAtomic Force Microscope. Biosens. Bioelectron. 1995, 10, 895–901.

(26) Proksch, R.; Schaffer, T. E.; Cleveland, J. P.; Callahan, R. C.; Viani,M. B. Finite Optical Spot Size and Position Corrections in Thermal SpringConstant Calibration. Nanotechnology 2004, 15, 1344–1350.

(27) Seantier, B.; Breffa, C.; Felix, O.; Decher, G. Dissipation-EnhancedQuartz Crystal Microbalance Studies on the Experimental ParametersControlling the Formation of Supported Lipid Bilayers. J. Phys. Chem. B2005, 109, 21755–21765.

(28) Rossetti, F. F.; Textor, M.; Reviakine, I. Asymmetric Distributionof Phosphatidyl Serine in Supported Phospholipid Bilayers on TitaniumDioxide. Langmuir 2006, 22, 3467–3473.

(29) Cha, T.; Guo, A.; Zhu, X. Y. Formation of Supported PhospholipidBilayers on Molecular Surfaces: Role of Surface Charge Density andElectrostatic Interaction. Biophys. J. 2006, 90, 1270–1274.

(30) Binder, H.; Zschornig, O. The Effect of Metal Cations on the PhaseBehavior and Hydration Characteristics of Phospholipid Membranes. Chem.Phys. Lipids 2002, 115, 39–61.

(31) Murzyn, K.; Rog, T.; Pasenkiewicz-Gierula, M. Phosphatidyletha-nolamine-Phosphatidylglycerol Bilayer as a Model of the Inner BacterialMembrane. Biophys. J. 2005, 88, 1091–1103.

(32) Egawa, H.; Furusawa, K. Liposome Adhesion on Mica SurfaceStudied by Atomic Force Microscopy. Langmuir 1999, 15, 1660–1666.

(33) Muller, D. J.; Engel, A. The Height of Biomolecules Measuredwith the Atomic Force Microscope Depends on Electrostatic Interactions.Biophys. J. 1997, 73, 1633–1644.

(34) Dufrene, Y. F.; Barger, W. R.; Green, J. B. D.; Lee, G. U.Nanometer-Scale Surface Properties of Mixed Phospholipid Monolayersand Bilayers. Langmuir 1997, 13, 4779–4784.

(35) Dufrene, Y. F.; Boland, T.; Schneider, J. W.; Barger, W. R.; Lee,G. U. Characterization of the Physical Properties of Model Biomembranesat the Nanometer Scale with the Atomic Force Microscope. FaradayDiscuss. 1998, 79, 94.

(36) Dufrene, Y. F.; Lee, G. U. Advances in the Characterization ofSupported Lipid Films with the Atomic Force Microscope. Biochim.Biophys. Acta 2000, 1509, 14–41.

(37) Garcia-Manyes, S.; Oncins, G.; Sanz, F. Effect of Ion-Binding andChemical Phospholipid Structure on the Nanomechanics of Lipid BilayersStudied by Force Spectroscopy. Biophys. J. 2005, 89, 1812–1826.

(38) Butt, H. J. Measuring Electrostatic, Van der Waals, and HydrationForces in Electrolyte-Solutions with an Atomic Force Microscope. Biophys.J. 1991, 60, 1438–1444.

(39) Houslay, M. D.; Stanley, K. K. Dynamics of Biological Membranes,influence on synthesis, structure and function; Wiley Interscience: NewYork, 1982.

(40) Garcia-Manyes, S.; Oncins, G.; Sanz, F. Effect of Temperature onthe Nanomechanics of Lipid Bilayers Studied by Force Spectroscopy.Biophys. J. 2005, 89, 4261–4274.

(41) Oncins, G.; Picas, L.; Hernandez-Borrell, J.; Garcia-Manyes, S.;Sanz, F. Thermal Response of Langmuir-Blodgett Films of Dipalmi-toylphosphatidylcholine Studied by Atomic Force Microscopy and ForceSpectroscopy. Biophys. J. 2007, 93, 2713–2725.

(42) Picas, L.; Montero, M. T.; Morros, A.; Oncins, G.; Hernandez-Borrell, J. Phase Changes in Supported Planar Bilayers of 1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine. J. Phys. Chem. B 2008, 112,10181–10187.

(43) Leonenko, Z. V.; Finot, E.; Ma, H.; Dahms, T. E.; Cramb, D. T.Investigation of Temperature-Induced Phase Transitions in DOPC and DPPCPhospholipid Bilayers Using Temperature-Controlled Scanning ForceMicroscopy. Biophys. J. 2004, 86, 3783–3793.

(44) Kunneke, S.; Kruger, D.; Janshoff, A. Scrutiny of the Failure ofLipid Membranes as a Function of Headgroups, Chain Length, andLamellarity Measured by Scanning Force Microscopy. Biophys. J. 2004,86, 1545–1553.

(45) Franz, V.; Loi, S.; Muller, H.; Bamberg, E.; Butt, H. H. TipPenetration through Lipid Bilayers in Atomic Force Microscopy. ColloidsSurf., B 2002, 23, 191–200.

(46) Cullis, R. R.; de Kruijff, B.; Hope, M. J.; Verkleij, A. J.; Nayar,R.; Farren, S. B.; Tilcock, C. P. S.; Madden, T. D.; Bally, M. B. In StructuralProperties of Lipids and Their Functional Roles in Biological Membranes;Aloia, R. C., Ed.; Academic Press: New York, 1983; Vol. 1, pp39-81.

(47) Richter, R. P.; Maury, N.; Brisson, A. R. On the Effect of the SolidSupport on the Interleaflet Distribution of Lipids in Supported Lipid Bilayers.Langmuir 2005, 21, 299–304.

(48) Jacobson, K.; Papahadjopoulos, D. Phase-Transitions and Phase Separa-tions in Phospholipid Membranes Induced by Changes in Temperature, pH, andConcentration of Bivalent-Cations. Biochemistry 1975, 14, 152–161.

(49) Khan, S. R.; Glenton, P. A.; Backov, R.; Talham, D. R. Presenceof Lipids in Urine, Crystals and Stones: Implications for the Formation ofKidney Stones. Kidney Int. 2002, 62, 2062–2072.

JP8102468

Subdomain Formation in POPE-POPG Bilayers J. Phys. Chem. B, Vol. 113, No. 14, 2009 4655

Dow

nloa

ded

by U

NIV

OF

BARC

ELO

NA

on

Aug

ust 1

0, 2

009

Publ

ished

on

Mar

ch 1

0, 2

009

on h

ttp://

pubs

.acs

.org

| do

i: 10

.102

1/jp

8102

468