Published on Web 06/29/2010

Crystalline Domain Structure and Cholesterol Crystal Nucleation

in Single Hydrated DPPC:Cholesterol:POPC Bilayers
Roy Ziblat,† Leslie Leiserowitz,*,‡ and Lia Addadi*,†
Department of Structural Biology and Department of Materials and Interfaces, Weizmann
Institute of Science, 76100 RehoVot, Israel
Received May 10, 2010; E-mail: leslie.leiserowitz@weizmann.ac.il; lia.addadi@weizmann.ac.il

Abstract: Grazing incidence X-ray diffraction measurements were performed on single hydrated bilayers
and monolayers of DPPC:Cholesterol:POPC at varying concentrations. There are substantial differences
in the phase and structure behavior of the crystalline domains formed within the bilayers relative to the
corresponding monolayers, due to interactions between the opposing leaflets. Depending on the lipid
composition, these interactions led to phase separation, changes in molecular tilt angle, or formation of
cholesterol crystals. In monolayers, DPPC and cholesterol form a single crystalline phase at all compositions
studied. In bilayers, a second crystalline phase appears when cholesterol levels are increased: domains of
cholesterol and DPPC form monolayer thick crystals where each of the lipid leaflets diffracts independently,
whereas excess cholesterol forms cholesterol bilayer thick crystals at a DPPC:Chol ratio < 46:54 ( 2 mol
%. The nucleation of the cholesterol crystals occurs at concentrations relevant to the actual cell plasma
membrane composition.

Introduction parisons between the lipid composition of the original membrane

and its DRMs showed elevation in cholesterol (Chol) and
Cell membranes possess lipid domains with different com-
sphingomyelin (SM) concentrations in the latter. This observa-
positionsandstructures,whichareimportanttocellfunctionality.1-4
tion was explained by considering that the sphingolipids in the
Lipid domains were studied extensively during the past two
cell membrane have a larger portion of saturated alkyl chains
decades. Several characteristics of the lipid molecules govern
relative to the glycerolipids; however, the DRMs also contain
their phase and packing structure behavior in the lipid domains,
a significant amount of saturated glycerolipids.12 Glycerophos-
including their chemical structure, molecular charge, and
phocholines are the most abundant lipid molecules in mam-
stereochemistry. There is specifically a growing interest in the
malian cells.14 It is thus important to study domain mixtures of
interactions between the opposing leaflets forming the lipid
cholesterol and glycerolipids, as well as cholesterol and sph-
bilayer.5-10 This study provides insights into such interactions
ingolipids. Several studies have shown that DPPC, a saturated
in symmetrical artificial bilayers composed of cholesterol and
glycerolipid with a phosphocholine headgroup, forms mixed
1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC), specifi-
phases and ordered domains with cholesterol.15-21
cally on the crystalline domains that exist in the lipid bilayer.
Cell membranes contain a large variety of lipid domains, some The phases formed by lipid assemblies are generally sorted
of which are referred to as “lipid rafts”, because of their relative into three main groups,6,22 depending on the melting tempera-
rigidity and stability. These were first studied through the ture; Liquid disordered (ld) phases are characterized by high
detergent resistant membrane (DRM) phenomenon.11-13 Com- diffusion rates. Solid ordered (So) or gel phases have a very
low diffusion rate and are generally composed of lipids with
†
Department of Structural Biology.
‡
Department of Materials and Interfaces. (12) Fiedler, K.; Kobayashi, T.; Kurzchalia, T. V.; Simons, K. Biochemistry
(1) Simons, K.; Van Meer, G. Biochemistry 1988, 27 (17), 6197–6202. 1993, 32 (25), 6365–6373.
(2) Simons, K.; Ikonen, E. Nature 1997, 387 (6633), 569–572. (13) Arnulphi, C.; Sot, J.; Garcia-Pacios, M.; Arrondo, J. L. R.; Alonso,
(3) Simons, K.; Toomre, D. Nat. ReV. Mol. Cell Biol. 2000, 1 (1), 31–39. A.; Goni, F. M. Biophys. J. 2007, 93 (10), 3504–3514.
(4) Hanzal-Bayer, M. F.; Hancock, J. F. FEBS Lett. 2007, 581 (11), 2098– (14) van Meer, G.; Voelker, D. R.; Feigenson, G. W. Nat. ReV. Mol. Cell
104. Biol. 2008, 9 (2), 112–124.
(5) Collins, M. D. Biophys. J. 2008, 94 (5), L32–L34. (15) Estep, T. N.; Mountcastle, D. B.; Barenholz, Y.; Biltonen, R. L.;
(6) Goni, F. M.; Alonso, A.; Bagatolli, L. A.; Brown, R. E.; Marsh, D.; Thompson, T. E. Biochemistry 1979, 18 (10), 2112–2117.
Prieto, M.; Thewalt, J. L. Biochim. Biophys. Acta 2008, 1781 (11- (16) Ratajczak, M. K.; Ege, C.; Ko, Y. T. C.; Majewski, J.; Kjaer, K.;
12), 665–684. Lee, K. Y. C. Biophys. J. 2005, 88 (1), 73a–73a.
(7) Wan, C.; Kiessling, V.; Tamm, L. K. Biochemistry 2008, 47 (7), 2190– (17) Karmakar, S.; Raghunathan, V. A.; Mayor, S. J. Phys.: Condens.
2198. Matter 2005, 17 (14), S1177–S1182.
(8) Bennett, W. F. D.; MacCallum, J. L.; Hinner, M. J.; Marrink, S. J.; (18) Stottrup, B. L.; Keller, S. L. Biophys. J. 2006, 90 (9), 3176–3183.
Tieleman, D. P. J. Am. Chem. Soc. 2009, 131 (35), 12714–12720. (19) Marsh, D. Biochim. Biophys. Acta 2009, 1788 (10), 2114–2123.
(9) Collins, M. D.; Keller, S. L. Biophys. J. 2007, 15a–15a. (20) Bach, D.; Wachtel, E. Biochim. Biophys. Acta 2003, 1610 (2), 187–
(10) Collins, M. D.; Keller, S. L. Proc. Natl. Acad. Sci. U.S.A. 2008, 105 197.
(1), 124–128. (21) Ivankin, A.; Kuzmenko, I.; Gidalevitz, D. Phys. ReV. Lett. 2010, 104
(11) Brown, D. A.; London, E. Biochem. Biophys. Res. Commun. 1997, (10), 108101.
240 (1), 1–7. (22) Feigenson, G. W. Biochim. Biophys. Acta 2009, 1788 (1), 47–52.

9920 9 J. AM. CHEM. SOC. 2010, 132, 9920–9927 10.1021/ja103975g  2010 American Chemical Society
Single Hydrated DPPC:Cholesterol:POPC Bilayers ARTICLES

saturated alkyl chains. Liquid ordered (lo) phases are character- by cells in excess, cholesterol was found to participate in several
ized by an intermediate diffusion coefficient and may contain pathological phenomena37,38 such as cataract39 and atheroscle-
high levels of cholesterol. The nature of the domains formed in rosis.40 Large amounts of cholesterol monohydrate crystals are
the phases described above are defined by their lipid composi- found in gall stones41 and foam cells42 at the site of athero-
tion, and their sizes can grow to micrometer scale under sclerotic plaques. It seems that these crystals are an early cause
equilibrium conditions.23 The lo and So phases include crystalline rather than a late outcome of atherosclerosis inflammation.43
domains detectable by X-ray diffraction.16,24-29 Their coherence The nucleation mechanism by which cholesterol crystals are
lengths are typically nanometers in the lo phase and tens of formed in and around cells is not well understood. It was
nanometers in the So phase. recently shown that cholesterol enriched macrophage cells form
Grazing incidence X-ray diffraction (GIXD) is a method cholesterol crystals near the cytoplasm membrane,44 and it was
developed to probe the structure of two-dimensional crystals suggested that these crystals nucleate at the lipid membrane.20,40,45
and can yield direct information on the organization of laterally We hereby demonstrate that the nucleation of cholesterol crystals
ordered lipid sheets.30,31 To enable the measurement of the thin can occur within DPPC:Chol:POPC bilayers and discuss the
flat samples, the incidence X-ray beam is set at a grazing angle, conditions for nucleation.
to enhance the diffraction signal from the sample relative to
scattering from its supporting substrate. These studies are Experimental Section
generally performed on artificial lipid monolayers at the air/
Bilayer Deposition. Silicon wafers (SESO, France) were treated
water interface. Monolayers differ, however, from the single with n-hexane, acetone, ethanol; left in Piranha for 20 min; and
lipid bilayer constituents of the cell membrane, especially washed. Deposition of a first polymer cushion is by immersion of
because the hydrophobic parts of the molecules are exposed to wafers in 1000 ppm polyethyleneimine (PEI) overnight. The
air and therefore do not interact with an opposing lipid leaflet. minimum thickness of this layer is estimated to be 5 nm.46 The
In bilayers, it is thus of fundamental interest to study the desired proportion of 1,2-dipalmitoyl-sn-glycero-3-phosphocholine
interactions between opposing leaflets, which were experimen- (DPPC), 1-palmotoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC)
tally shown to influence membrane domain formation in model (Avanti Polar Lipids), and cholesterol (>99% Sigma-Aldrich) was
membranes.5,7 Interdigitation of alkyl chains in the bilayer core, dissolved in chloroform and spread (∼50 µL) on a preheated
i.e. at the interface between the two lipid leaflets, was also Langmuir trough. Prior to spreading, cholesterol was purified by
crystallization.47 The temperature of the Langmuir trough was set
suggested to occur.32
to be higher than the melting temperature of the lipid composition,
GIXD measurements on single hydrated lipid bilayers were when possible, typically above 40-50 °C. After 10 min (chloroform
only recently conducted to successfully reveal crystalline evaporation time) the lipid monolayer at the air-water interface is
order.33-35 In our work34 GIXD was performed on a single compressed to 30 mN/m. (Figure 1) The water is then drained out
hydrated bilayer composed of SM:Chol:POPC, while maintain- of the trough, lowering the lipid monolayer on top of a silicon/
ing the sample close to the dew point, such that the bilayer is quartz wafer covered with a polymer cushion which lies at the
completely hydrated. We study here, using the same method, bottom of the trough. Some water remains sandwiched between
the structure of lipid bilayers of DPPC:Chol:POPC, and we the wafer and the lipid monolayer (Figure 1). This layer, though
analyze the interactions between the opposing leaflets. covered with a lipid monolayer, undergoes a slow evaporation. Once
the water has almost completely evaporated, the wafer is lifted and
The cholesterol molecule participates in membrane structure
placed upside down on top of a water-filled Teflon plate, placed in
and determines its mechanical and dynamical properties. advance at the bottom of the trough, and now covered with a lipid
Cholesterol is believed to be an essential component in the monolayer as well. Once placed on the Teflon plate, the wafer is
formation of functional ordered domains.2 Its concentration was gradually cooled down to 4 °C and kept so until AFM or GIXD
found to be as high as 20-45% of the phospholipids in the experiments are performed.
plasma membrane in mammalian cells.14,36 When accumulated Prior to the GIXD experiment a second PEI cushion layer is
deposited on top of the bilayer by immersion in 1000 ppm PEI for
(23) Veatch, S. L.; Keller, S. L. Biophys. J. 2003, 85 (5), 3074–3083. 10 min and washed. Water comprises 85% of the PEI cushion
(24) Vaknin, D.; Kelley, M. S. Biophys. J. 2000, 79 (5), 2616–2623. volume.46 The bilayer sample is constantly kept under either water
(25) Majewski, J.; Kuhl, T. L.; Kjaer, K.; Smith, G. S. Biophys. J. 2001, or high humidity. After deposition of the PEI layer, the contact
81 (5), 2707–2715. angle of the bilayer is 0°, while in the absence of PEI the contact
(26) Tristram-Nagle, S.; Liu, Y. F.; Legleiter, J.; Nagle, J. F. Biophys. J. angle is hydrophilic but higher than 30°. Samples which contain
2002, 83 (6), 3324–3335.
(27) Lee, K. Y. C. In CNLS conference, Santa Fe, New Mexico, May 21,
2007; Santa Fe, New Mexico, 2007. (37) Ikonen, E. Physiol. ReV. 2006, 86 (4), 1237–1261.
(28) Mills, T. T.; Huang, J. Y.; Feigenson, G. W.; Nagle, J. F. Gen. Physiol. (38) Maxfield, F. R.; Tabas, I. Nature 2005, 438 (7068), 612–621.
Biophys. 2009, 28 (2), 126–139. (39) Cenedella, R. J. JAMA, J. Am. Med. Assoc. 1987, 257 (12), 1602–
(29) Ratajczak, M. K.; Chi, E. Y.; Frey, S. L.; Cao, K. D.; Luther, L. M.; 1602.
Lee, K. Y. C.; Majewski, J.; Kjaer, K. Phys. ReV. Lett. 2009, 103 (2), (40) Mason, R. P.; Tulenko, T. N.; Jacob, R. F. Biochim. Biophys. Acta
028103. 2003, 1610 (2), 198–207.
(30) Als-Nielsen, J.; Jacquemain, D.; Kjaer, K.; Leveiller, F.; Lahav, M.; (41) Weihs, D.; Schmidt, J.; Goldiner, I.; Danino, D.; Rubin, M.; Talmon,
Leiserowitz, L. Phys. Rep. 1994, 246 (5), 252–313. Y.; Konikoff, F. M. J. Lipid Res. 2005, 46 (5), 942–948.
(31) Jensen, T. R.; Kjaer, K. In NoVel methods to study interfacial layers; (42) Tulenko, T. N.; Chen, M.; Mason, P. E.; Mason, R. P. J. Lipid Res.
Moebius, D., Miller, R., Eds.; Elsevier: Amsterdam, 2001; pp 205- 1998, 39 (5), 947–956.
254. (43) Duewell, P.; et al. Nature 2010, 464 (7293), 1357–1361.
(32) Pinto, S. N.; Silva, L. C.; de Almeida, R. F. M.; Prieto, M. Biophys. (44) Ong, D. S.; Anzinger, J. J.; Leyva, F. J.; Rubin, N.; Addadi, L.; Kruth,
J. 2008, 95 (6), 2867–2879. H. S. J. Lipid Res. 2010, in press.
(33) Miller, C. E.; Majewski, J.; Watkins, E. B.; Mulder, D. J.; Gog, T.; (45) Rapaport, H.; Kuzmenko, I.; Lafont, S.; Kjaer, K.; Howes, P. B.; Als-
Kuhl, T. L. Phys. ReV. Lett. 2008, 1 (5), 058103. Nielsen, J.; Lahav, M.; Leiserowitz, L. Biophys. J. 2001, 81 (5), 2729–
(34) Ziblat, R.; Kjaer, K.; Leiserowitz, L.; Addadi, L. Angew. Chem., Int. 2736.
Ed. 2009, 48 (47), 8958–8961. (46) Wong, J. Y.; Majewski, J.; Seitz, M.; Park, C. K.; Israelachvili, J. N.;
(35) Watkins, E. B.; Miller, C. E.; Mulder, D. J.; Kuhl, T. L.; Majewski, Smith, G. S. Biophys. J. 1999, 77 (3), 1445–1457.
J. Phys. ReV. Lett. 2009, 102 (23), 238101. (47) Perl-Treves, D.; Kessler, N.; Izhaky, D.; Addadi, L. Chem. Biol. 1996,
(36) Ikonen, E. Nat. ReV. Mol. Cell Biol. 2008, 9 (2), 125–138. 3 (7), 567–77.
J. AM. CHEM. SOC. 9 VOL. 132, NO. 28, 2010 9921
ARTICLES Ziblat et al.

scattering vector, where 2θxy is the angle between the incident and
diffracted beam projected onto the horizontal plane. The diffraction
data are represented in two ways: (1) Bragg peaks I(qxy) are the
integral of the GIXD pattern I(qxy,qz) over qz; (2) Bragg rod intensity
profiles I(qz) are the scattered intensity I(qxy,qz) integrated across
the qxy range of each Bragg peak. The qxy positions of the Bragg
peaks yield the lattice repeat distances d ) 2π/qxy, which may be
indexed by the two Miller indices h,k to yield the unit cell. The
full width at half-maximum (fwhm) of the Bragg peaks yields the
lateral 2-D crystalline coherence length Lxy ≈ 0.9(2π)/fwhm(qxy).
The Bragg rod profile along qz similarly gives a measure of the
thickness of the crystalline film. Control studies on monolayers
deposited at the air/water interface were performed at the same
beamline using a liquid surface diffractometer.31,48 All measure-
ments were performed at sample temperatures of ∼6-7 °C.

Results
Bilayer samples were deposited at the HASYLAB synchro-
tron facility using a Langmuir trough (Figure 1). From the
moment of deposition and until the end of the measurements
the sample is kept constantly under water. When placed inside
the humidity chamber for GIXD measurements, the sample is
Figure 1. Schematic representation of the bilayer deposition process. (a) covered with an ∼1 mm thick water layer, which subsequently
Lipid molecules are spread at the air-water interface and compressed to a undergoes evaporation under controlled conditions. Scattering
surface pressure of 30 mN/m. (b) Water is drained, and a lipid monolayer calculations show that when the thickness of the condensed
is deposited on top of both a preplaced wafer and Teflon plate. (c) The
wafer is lifted and placed upside down on top of the Teflon plate to form water layer on top of the lipid bilayer is lower than 1 µm, GIXD
a lipid bilayer. An ∼1 mm thick water layer remains between the bilayer measurements can be successfully performed.
and the Teflon plate. The entire deposition process is performed at a Throughout the diffraction measurements the sample is kept
temperature > 45 °C. under high humidity (>94%). To avoid even local and temporary
drying of the sample, the lipid bilayer is covered with a
cholesterol molecules remain intact for long periods when 10%
POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) is added polyethyleneimine (PEI) layer with a thickness of a few to tens
to the mixture. POPC separates from cholesterol and DPPC6 and of nanometers. The highly hydrophilic PEI layer, deposited prior
does not participate in the crystalline domain structures, because to GIXD measurements, adsorbs water guaranteeing the pres-
of the presence of one kinked acyl chain. ence of a water reservoir that does not evaporate even at
AFM. AFM scans were performed with a BIOSCOPE (Veeco) humidity <100%. Prior to extraction from the humidity chamber,
or a Molecular Imaging PICO (Agilent Technologies) system under a thick water layer is condensed on top of the sample. Samples
wet contact or tapping mode, using Silicon Nitride tips (Digital are then characterized to ensure that no damage occurred to the
Instruments). Scan frequencies were typically 2-3 Hz, and a line bilayer sheet, during the process.
resolution of either 256 or 512 was used.
The conventional method to confirm the thickness of thin
Humidity GIXD Setup. A humidity chamber was designed and
constructed at the Weizmann Institute and adapted to be mounted layer samples that are measured by GIXD is X-ray reflectivity.
on beamline BW1 at the DORIS III Synchrotron DESY (Hamburg, In the case of the hydrated lipid bilayer the sample is composed
Germany). The methodology was recently introduced and de- of a large number of blocks, including silicon, silicon oxide,
scribed.34 Briefly, the sample is completely wet when placed inside polymer cushion, and the lipid bilayer that is composed of at
the humidity chamber. A controlled evaporation process is then least four varying electron density blocks, followed by a second
initiated to enable GIXD measurements. The temperature of the polymer cushion. In addition to this complicated electron density
chamber walls is controlled by a circulating flow from a refrigerated profile, the top layer of water on the sample may vary over
bath, and the sample temperature is adjusted by a Peltier element. time according to relative humidity oscillations, all together
The chamber is filled with humidified helium. Relative humidity, making reflectivity scans an impractical method. In our case,
gas temperature, and sample temperature are monitored and
therefore, AFM scans are deemed more reliable.34 We note that
maintained by PID feedback loops. The measurements are per-
formed under continuous control of temperature and humidity. the lipid bilayer samples were deposited at relatively high
Although this setup was designed for bilayer studies, it can be used temperature (45-50 °C) but were imaged by AFM (Figure 2)
on any two-dimensional crystal which requires hydration. at ambient temperature. The surface pressure/area isotherms of
GIXD Measurements. A monochromatic X-ray beam (λ ) DPPC at these two temperatures are significantly different, the
1.304 Å) was adjusted to strike the sample surface at an incident area per molecule at any given pressure being higher at 45 °C
angle Ri ) 0.85Rc, where Rc is the critical angle for total external than at ambient T. Therefore, bilayer samples, which were
reflection for the He-silicon or the He-water interface for bilayer deposited at high temperature and scanned at ambient T, contain
and monolayer setups respectively. The dimensions of the footprint empty regions at the boundaries of which the height profile of
of the incoming X-ray beam on the sample surface were 1 mm × the deposited layer can be easily evaluated. The PEI cushion,
36 mm for silicon and 1 mm × 50 mm for water. GIXD signals
which consists mostly of water, is soft and is easily displaced
were obtained from 2-D crystallites randomly oriented about the
wafer surface normal in the form of a 2-D “powder”. The scattered by the pressure of the tip; therefore, its effect on the height
intensity was collected by means of a vertical position-sensitive image measured by AFM is negligible. The AFM tip is,
detector, measuring qz ≈ 2π sin(Rf)/λ, which is the out-of-plane
component of the scattering vector. Measurements were performed (48) Als-Nielsen, J.; Jacquemain, D.; Kjaer, K.; Leveiller, F.; Lahav, M.;
by scanning the horizontal component, qxy ≈ 4π sin(2θxy)/λ, of the Leiserowitz, L. Phys. Rep. 1994, 246 (5), 252–313.
9922 J. AM. CHEM. SOC. 9 VOL. 132, NO. 28, 2010
Single Hydrated DPPC:Cholesterol:POPC Bilayers ARTICLES

Figure 3. Bragg peak data from (a) bilayers and (b) monolayers at several
DPPC:Chol:POPC compositions. Representative schemes of the samples
Figure 2. AFM bioscope height images, measured in wet contact mode. are shown on top. The Bragg peaks were fitted with Voight functions and
Each scan is 5 µm × 5 µm, and the z color range is between 0 and 25 nm. are shown after deconvolution: (c) bilayers; (d) monolayers. Peaks were
Images were linearly flattened. The AFM scans are performed after GIXD colored according to their phases: DPPC phase (black), mixed DPPC:Chol
experiments. Each sample is composed of DPPC:Chol:POPC composition phase (blue), and Chol monohydrate bilayer phase (red). The intensities of
according to (a) 100:0:0, (b) 72:18:10, (c) 54:36:10, and (d) 36:54:10. Black all Bragg peaks were factorized for easier comparison. Most DPPC Bragg
lines on the AFM images represent the locations of height profile cuts which peaks obey the multiplicity rule where {(11),(11j)} and (02) have a ratio of
are given at the bottom of each scan image. These profiles, 5-6 nm thick, integrated intensities of ∼1. The 100:0:0 bilayer has a 10° undulation which
correspond to a bilayer. widens the {(11),(11j)} peak altering the ratio.

however, forced through the polymer, thus introducing some The lipid samples form several phases, all of which were
noise to the scan images. The height profiles of the images identified previously by GIXD on a Langmuir trough,16,21,49 i.e.
obtained from DPPC samples after GIXD measurements at the air/water interface. For convenience, the separated Bragg
peaks are colored according to their phases: pure DPPC (black),
consistently show a sample thickness of 5-6 nm, corresponding
mixed DPPC:Chol (blue), and pure Chol (red). For the 100:0:0
to a single bilayer (Figure 2). Each sample was scanned at 3-5
and 72:18:10 compositions both the monolayer and the bilayer
arbitrary locations, millimeters apart. Unlike bilayers, the height
samples show one crystalline phase. At 72:18:10 the GIXD
profiles of supported wet monolayers display a thickness of
patterns of the bilayer and the monolayer are similar, and thus
∼1-1.5 nm and a pronounced roughness (Figure S1). AFM also are their molecular arrangements, yet the molecular tilt,
studies show that, in lipid bilayers which were not always fully deduced from the fwhm of their Bragg rods (Figure 4a,b), is
hydrated, the bilayer structure is irreversibly damaged, and this distinctively lower in the bilayer system. The molecular tilt in
is reflected in the AFM scans. the monolayer is 26° whereas in the bilayer it is 5° (Table 1).
GIXD measurements were performed on both bilayer and Increasing the concentration of cholesterol to 36% yields in the
monolayer systems with the following DPPC:Chol:POPC monolayer a new wide peak which belongs to the DPPC:Chol
compositions: 100:0:0, 72:18:10, 54:36:10, and 36:54:10. POPC mixed phase, positioned at qxy ) 1.39 Å-1. In the bilayer,
is introduced only to extend the bilayer stability in time but however, even though the composition is identical to that of
does not mix with cholesterol and DPPC. Figure 3 shows the the monolayer, two crystalline phases coexist. The position of
Bragg peaks of bilayers and monolayers, extracted from the the mixed phase peak at qxy ) 1.33 Å-1 is distinctly lower than
diffraction data. The intensities of the Bragg peaks were that in the monolayer sample. This shift to lower qxy is indicative
independently normalized for convenience (the un-normalized of a lower DPPC:Chol ratio in the mixed phase,21,29,50 which
data is shown in Figure S2). Each of the GIXD scans is
composed of several diffraction peaks. Data were fitted using (49) Solomonov, I.; Weygand, M. J.; Kjaer, K.; Rapaport, H.; Leiserowitz,
L. Biophys. J. 2005, 88 (3), 1809–1817.
Voight functions, and the separated peaks are shown in Figure (50) Ege, C.; Ratajczak, M. K.; Majewski, J.; Kjaer, K.; Lee, K. Y. C.
3c, d for bilayer and monolayer accordingly. Biophys. J. 2006, 91 (1), L1–L3.
J. AM. CHEM. SOC. 9 VOL. 132, NO. 28, 2010 9923
ARTICLES Ziblat et al.

monolayer and the bilayer. The monolayer yields one broad

peak, which belongs to a mixed phase. Its position at qxy )
1.25 Å-1 indicates that the monolayer has a higher cholesterol
concentration than the monolayer composed of 36% cholesterol.
On the other hand, the position of the broad mixed phase peak
in the bilayer is at qxy ) 1.31 Å-1, indicating that the domains
formed by the mixed phase in the bilayer have a higher DPPC:
Chol ratio than those observed in the monolayer. The remaining
cholesterol gives rise to the two sharp Bragg peaks located at
qxy ) 1.04, 1.23 Å-1, which belong to a crystalline cholesterol
phase with a high coherence length of 60 nm.
As a bilayer, the exocyclic chains of the cholesterol molecules
are partially interdigitated, leading to a well packed structure
with a high coherence length and a rectangular 10 × 7.5 Å2
unit cell.49 This unit cell structure corresponds to the bilayer of
the macroscopically metastable phase of Chol · H2O.49 The Bragg
rods of the GIXD peaks (Figure 4g), which are fingerprint
evidence of the bilayer motif, yield a thickness of 35 ( 1.3 Å,
deduced from their fwhm. In contrast, the Bragg rods from the
mixed DPPC:Chol and pure DPPC structural films, shown in
Figure 4a-f, belong to crystalline phases with thicknesses of
less than 20 Å corresponding to monolayers. The position along
qz of the Bragg rods of all the mixed phases show that the
molecules are aligned essentially perpendicular to the membrane
plane. The tilt angle of the alkyl chains in the pure DPPC phase
varies significantly in the bilayer, depending on the cholesterol
concentration; in the monolayer structures the tilt is distinctly
higher, varying little as a function of cholesterol concentration
(Table 1).
The position of the Bragg peak of the mixed DPPC:Chol
phase changes linearly with the composition ratio.21 We can
then deduce from the position of the mixed phase peak in the
bilayer at 40% cholesterol total composition that the DPPC:
Chol ratio in the mixed domains is 51:49 mol %. The position
of the Bragg peak along qxy of the mixed DPPC:Chol domain
barely shifts when increasing the cholesterol concentration from
40 to 60 mol %, yielding a composition of 48:52 mol % for the
mixed phase. Therefore this is likely to be the nucleation ratio
for formation of cholesterol bilayer crystals.
An estimate of the threshold DPPC:Chol ratio required for
nucleation of cholesterol may also be made utilizing the Bragg
peak intensities. In the previous study that we conducted on
SM:Chol mixtures in bilayers, X-ray diffraction from the
cholesterol bilayer was observed at both 40:60 and 60:40 SM:
Chol ratios.34 The composition where cholesterol nucleation
occurs was determined then by extrapolating to zero the intensity
of the cholesterol Bragg peaks. In the DPPC:Chol system,
cholesterol bilayer diffraction was observed only for the 40:60
ratio; therefore no intensity extrapolation can be made. Never-
theless, by assuming a similar dependence of the cholesterol
Figure 4. Bragg rods data from (a, c, e, g) bilayers and (b, d, f) monolayers diffraction intensity on composition as in the SM:Chol system,
at several DPPC:Chol:POPC compositions; (a-d) DPPC phase; (e-f) mixed the nucleation ratio is evaluated to be 44:56 mol %.
DPPC:Chol phase; (g) cholesterol bilayer phase. The Bragg rods are colored Both methods mentioned above, although based on assump-
according to their Bragg peaks in Figure 3. All Bragg rods show monolayer
thick crystals. The Bragg rod of the 100:0:0 bilayer sample is wider relative
tions, result in similar values of 52 and 56 mol %. We thus
to the 100:0:0 monolayer, due to an ∼10° undulation of the bilayer sheet. estimate that cholesterol nucleation in DPPC:cholesterol bilayer
The cholesterol Bragg rod shows a 34 Å thick crystal corresponding to a mixtures occurs at 54 ( 2 mol % cholesterol at 7 °C.
bilayer.
Discussion
is in agreement with the observed formation of a second
crystalline phase composed mainly of DPPC, identified by two We have compared the structures of crystalline domains and
Bragg peaks located at qxy ) 1.41, 1.51 Å-1. the phases formed in monolayers and bilayers composed of
At the increased cholesterol concentration of 54% the GIXD DPPC:cholesterol:POPC mixtures at several cholesterol con-
pattern also shows a different phase behavior between the centrations. As POPC does not mix well with DPPC and
9924 J. AM. CHEM. SOC. 9 VOL. 132, NO. 28, 2010
Single Hydrated DPPC:Cholesterol:POPC Bilayers ARTICLES

Table 1. Crystallographic Data on the Lipid Monolayer and Bilayer Samplesa

a
Unit cell dimensions (UC), Area per molecule (A/mol), Projected area (PA), Coherence length (CL). The {11} represents both (11) and (11j).

cholesterol at 7 °C, we shall refer below, for simplicity, to

DPPC:cholesterol proportions within the mixed domains only.
The differences in phase behaviors between the monolayer
and bilayer are schematically illustrated in Figure 5. At 20 mol
% cholesterol, the molecular tilt was found to be lower in the
bilayer system relative to the monolayer, but the same DPPC
phase is formed in both systems. At 40% mole cholesterol, the
monolayer is composed of one mixed DPPC:Chol phase,
whereas in the bilayer phase separation occurs into a mixed
DPPC:Chol phase and a DPPC phase. At 60 mol % cholesterol,
again phase separation occurs in the bilayer, but this time
between a mixed DPPC:Chol phase and a cholesterol bilayer
crystal. This indicates that the DPPC:Chol miscibility range in
symmetrical bilayers is less than 20%. Figure 5. Scheme of (a) bilayer (left) and monolayer (right) of DPPC:
Interestingly, the DPPC crystalline phase in the bilayer Chol (40:60) and (b) bilayer (left) and monolayer (right) of DPPC:Chol
(60:40). POPC was omitted from the scheme. The phase behavior at both
containing 60% DPPC has sharper peaks relative to the DPPC compositions in the bilayer differs from the corresponding monolayers. One
phases in the bilayer samples with lower cholesterol concentra- crystalline mixed phase exists in each monolayer, colored in light blue for
tions; in other words a dramatic increase in the coherence length DPPC:Chol (60:40) and dark blue for DPPC:Chol (40:60). There is a phase
from 40 Å to 240 Å in the direction of d*{11} and from 110 Å separation in their bilayers. Both bilayers have a similar mixed phase
composition which is estimated to be DPPC:Chol (∼50:50). The additional
to 220 Å in the d*{02} direction (see Table 1) is observed, as phases are in (a) crystalline cholesterol bilayer (red) and (c) crystalline DPPC
deduced from the fwhm values (Figure 3c). The 60% DPPC phase (gray).
bilayer has two coexisting phases, which introduces a line
tension between their domains. This line tension is likely to To understand the reason for the observed phase separation
drive the system to form larger crystalline domains, by minimiz- of cholesterol in the bilayer system, let us consider first the
ing their boundaries, therefore increasing the coherence lengths. crystalline behavior of sphingolipids; When spread at the
J. AM. CHEM. SOC. 9 VOL. 132, NO. 28, 2010 9925
ARTICLES Ziblat et al.

Chart 1. Chemical Formulas 10-13 Å thick (see Table 1), which corresponds to the length
of the steroid moiety only, indicating that the alkyl chains of
DPPC are likely to be packed well with the steroid moiety but
not with the exocyclic chains.
The crystalline domains in the DPPC monolayer have a chain
tilt of 27°, which remains the same when the cholesterol levels
are increased to 20%. In the bilayer the tilt is 21° at 100% DPPC
and is decreased further to 5° when increasing the cholesterol
levels to 20%. We deduce from these data that the interaction
of an opposing leaflet in the bilayer is accompanied by
interactions between the molecule extremities, presumably by
filling the voids between molecules. The thicknesses of all the
mixed phase crystalline domains, both monolayer and bilayer,
are in the range of 10.4-13.4 Å. This thickness indicates that
the domains in the opposing leaflets diffract independently and
no repetitive structural correlations exist there. Interactions
between the leaflets made by partial interdigitation in the
crystalline domains are not trivially formed since this would
require appropriate alkyl chain orientation and packing registry
between the opposing leaflets, demanding, at least, 2-fold
symmetry operation. Whether or not the crystalline domains in
opposing leaflets are independently arranged or coupled by any
short-range correlation between leaflets remains unclear.
In contrast to the mixed phase, partial interdigitation is
air-water interface, ceramide C16 (Chart 1) spontaneously observed in the cholesterol bilayer crystals. These are formed
forms crystalline domains even at negligible surface pressure.24,51 by virtue of interplay between the lateral packing requirements
Sphingomyelin, however, which consists of a ceramide back- of the rigid steroid and flexible exocyclic moieties, and the
bone esterified with a phosphocholine headgroup, does not form interlayer contact between exocyclic moieties at the interface
crystalline monolayer domains, unless it is subjected to high of the two leaflets, yielding 35 Å thick crystals.
compression.16,29 This lack of self-assembly was attributed to The threshold value at which cholesterol starts forming a
the large hydrophilic headgroup of SM, which disturbs the crystalline bilayer in DPPC:Chol bilayers is 54 mol %, which
molecular packing, leading to the “umbrella effect” hypothesis.52 is significantly higher than that in SM:Chol bilayers where
A corollary of this concept is that molecules with small head cholesterol crystal nucleation occurs at 38 mol %, measured
groups such as cholesterol or ceramide readily form mixed under the same conditions. The sphingoid backbone of SM has
phases with SM, because they dilute the interactions between two hydrogen donor groups, hydroxyl and amide, which are
the SM bulky head groups. DPPC is similar to SM in the head not present in the glycerol backbone of DPPC. It is conceivable
group and the alkyl chain length but differs in the backbone that the cholesterol hydroxyl group may interact with these
chemistry. Unlike SM, DPPC does, however, spontaneously groups, thus affecting the positioning of the cholesterol per-
form crystalline domains at the air-water interface at negligible pendicular to the bilayer surface.
surface pressure. Using molecular dynamics simulations, T. Rog et al. showed
The driving force of DPPC to form stable crystalline domains that the average distribution of the cholesterol location in one
suggests that DPPC should tend to phase separate from leaflet of the DPPC:Chol bilayer system is 1.6 Å further away
cholesterol to crystallize at a lower concentration than that of from its opposing cholesterol molecule in the other leaflet than
SM, which cannot self-assemble into crystals. This suggestion in the SM:Chol system.53 They also suggested that the DPPC:
may explain why, in the bilayers with an overall composition Chol bilayer is less dynamic in terms of the cholesterol
of DPPC:Chol ) 60:40, DPPC separates from the DPPC:Chol oscillations along the membrane perpendicular axis. The
mixture and forms a pure phase, whereas, in the corresponding consequence of these differences between the SM:Chol and the
bilayers with composition SM:Chol ) 60:40, no pure SM phase DPPC:Chol bilayers is that cholesterol molecules in the former
separation is observed. The logic of this argument would suggest bilayer will more frequently make contact and interdigitate,
similar phase separation behaviors in monolayers composed of nucleate, and subsequently form cholesterol bilayer crystals.
mixed DPPC:Chol and SM:Chol. However, as monolayers, both The position of the Bragg peak of the mixed DPPC:Chol
systems have only one crystalline phase at all ratios. The phase phase changes linearly with the composition ratio.29 We can
separation observed in the bilayers thus must be rather due to then deduce from the position of the mixed phase peak in the
interactions between the opposing leaflets. bilayer at 60% cholesterol total composition that the DPPC:
The cross section of the DPPC molecule is composed of two Chol ratio in the mixed domains is 46:54 mol %. Assuming
alkyl chains with an area of ∼20 Å2 each. The widest average that the system is isotropic and homogeneous, and that the
cross section of the rigid moiety of the steroid in the cholesterol coherence length equals the crystalline domain size, we can
molecule is ∼38 Å2, whereas its branched exocyclic chain has calculate by knowing the total lipid composition that the average
a cross-sectional area of ∼28 Å2. The crystalline domains distance between cholesterol bilayer crystals is 500 ( 250 nm.
formed in the mixed phase in both monolayers and bilayers are The bilayer cholesterol crystals are not expected to dissolve
at body temperatures, although their nucleation threshold in
(51) Scheffer, L.; Solomonov, I.; Weygand, M. J.; Kjaer, K.; Leiserowitz,
L.; Addadi, L. Biophys. J. 2005, 88 (5), 3381–3391. (53) Rog, T.; Pasenkiewicz-Gierula, M. Biophys. J. 2006, 91 (10), 3756–
(52) Huang, J. Y.; Feigenson, G. W. Biophys. J. 1999, 76 (4), 2142–2157. 3767.
9926 J. AM. CHEM. SOC. 9 VOL. 132, NO. 28, 2010
Single Hydrated DPPC:Cholesterol:POPC Bilayers ARTICLES

terms of cholesterol composition is likely to be higher than that interlayer interactions, stabilized by partial interdigitation of the
at 7 °C. When these bilayers become stacked via interleaving exocyclic chains, leads to bilayer crystallinity. Structural inter-
H-bonded water molecules the resulting monoclinic cholesterol digitation was not observed for other phases. Similar phase
monohydrate structure spontaneously transforms into the ther- separation of cholesterol into bilayer crystals was observed also
modynamically stable triclinic monohydrate phase,45 which in cholesterol:sphingomyelin bilayer domains at cholesterol
incorporates the 12.4 × 12.4 Å2 bilayer motif, and exists as a concentrations above 38%. We argue that in both systems the
three-dimensional crystal. Thus, it is conceivable that once nucleation of the cholesterol crystals occurs at concentrations
formed within the biomembrane the two-dimensional cholesterol relevant to the actual cell plasma membrane composition. The
crystals, incorporating the 10 × 7.5 Å2 bilayer motif, serve as formation of bilayer thick crystalline domains in the membrane
nucleating domains, which eventually lead to the formation of may lead to nucleation of 3D cholesterol crystals under
three-dimensional cholesterol crystals. Cholesterol is found in appropriate conditions.
large amounts at the cytoplasm membrane and can reach levels
Acknowledgment. We thank HASYLAB for synchrotron
as high as 45 mol % of the total lipid composition. The
beamtime and ELISA: EU financial support of access to synchrotrons/
cholesterol is unlikely to be spread homogeneously, especially
FELs in Europe. This work was supported by the Israel Science
because the membrane is largely composed of unsaturated lipids
Foundation, and by the Clore Center for Biological Physics. We
that tend to phase separate from it. Therefore, locally, cholesterol
are grateful to the support of the Helen and Martin Kimmel Center
concentrations can be presumably much higher.
for Nanoscale Science. We thank Dr. Kristian Kjaer for his
Concluding Remarks tremendous help in the development and implementation of the
GIXD humidity chamber in beamline BW1.
We have monitored here the consequences of interactions
between opposing leaflets in a lipid bilayer composed of Supporting Information Available: Complete ref 43. AFM
cholesterol and DPPC separated by immiscible POPC disordered scan of supported lipid monolayer (Figure S1) and un-normal-
areas. Depending on the lipid composition, the interactions ized Bragg peaks of both mono- and bilayer samples (Figure
between leaflets led to phase separation, changes in molecular S2). This material is available free of charge via the Internet at
tilt angle, or formation of cholesterol bilayer thick crystals. The
http://pubs.acs.org.
cholesterol bilayer crystals begin to form with a DPPC:Chol
ratio of 46:54 ( 2 mol %, and the interplay between intra- and JA103975G

J. AM. CHEM. SOC. 9 VOL. 132, NO. 28, 2010 9927