Sorting of lipidated peptides in fluid bilayers: A molecular-level investigation

Article abstract of DOI:10.1021/ja3074825

Nearest-neighbor recognition (NNR) measurements have been made for two lipidated forms of GlyCys, interacting with analogues of cholesterol and 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) in the liquid-ordered (l _o) and liquid-disordered (l_d) phases. Interaction free energies that have been determined from these measurements have been used in Monte Carlo simulations to quantify the distribution of the peptides between liquid-ordered and liquid-disordered regions. These simulations have shown that significant differences in the lipid chains have a very weak influence on the partitioning of the peptide between these two phases. They have also revealed an insensitivity of the peptide partition coefficient, K_p, to the size of the l_o and l_d domains that are present. In a broader context, these findings strongly suggest that the sorting of peripheral proteins in cellular membranes via differential lipidation may be more subtle than previously thought.

Full text of DOI:10.1021/ja3074825

Article
pubs.acs.org/JACS
Sorting of Lipidated Peptides in Fluid Bilayers: A Molecular-Level
Investigation
Trevor A. Daly,† Paulo F. Almeida,*,‡ and Steven L. Regen*,†
†Department of Chemistry, Lehigh University, Bethlehem, Pennsylvania 18105, United States
‡Department of Chemistry and Biochemistry, University of North Carolina Wilmington, Wilmington, North Carolina 28403, United
States
S
* Supporting Information
ABSTRACT: Nearest-neighbor recognition (NNR) measurements have
been made for two lipidated forms of GlyCys, interacting with analogues
of cholesterol and 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC)
in the liquid-ordered (lo) and liquid-disordered (ld) phases. Interaction
free energies that have been determined from these measurements have
been used in Monte Carlo simulations to quantify the distribution of the
peptides between liquid-ordered and liquid-disordered regions. These
simulations have shown that significant differences in the lipid chains have
a very weak influence on the partitioning of the peptide between these two
phases. They have also revealed an insensitivity of the peptide partition
coefficient, Kp, to the size of the lo and ld domains that are present. In a broader context, these findings strongly suggest that the
sorting of peripheral proteins in cellular membranes via differential lipidation may be more subtle than previously thought.
We have begun a program that is aimed at gaining deeper
insight into lipid sorting by placing it on a more quantitative
basis. Our approach uses experimentally determined values of
nearest-neighbor interaction free energies, ωAB, in combination
with Monte Carlo computer simulations.14−16 Of particular
significance is the fact that this approach is applicable to domains
of any size and that it can provide a detailed molecular-level view of
membrane organization.
Here we show how this strategy yields detailed molecular-
level insight into the mixing behavior of two lipidated peptides
(1 and 2, Chart 1) with analogues of cholesterol and DPPC (3
and 4, Chart 1) in host membranes of DPPC and cholesterol in
the lo/ld coexistence region. This binary system was of special
interest to us because its phase behavior has been well studied
and because discrete microdomains are thought to be present in
this coexistence region but have never actually been visually
observed, a situation that is analogous to putative micro-
domains in cellular membranes.17−19 The two lipidated
peptides designed for this investigation are analogs of
[(myristoyl)GlyCys(palmitoyl)-]. It has been suggested that
this moiety represents a minimal sequence that is necessary for
promoting efficient association with lipid rafts; for example,
with the lck-EGFP chimera protein in COS-7 cells.9 In one of
our analogues, a permanent kink was introduced in its C14
chain to assess the consequences of coiling the lipid anchor on
the mixing properties of the lipidated peptide.
INTRODUCTION
■
One of the greatest challenges presently facing chemists,
biochemists and biophysicists is to define the two-dimensional
structure of cell membranes. In particular, the time-averaged
lateral distribution of the lipids and proteins that make up these
natural enclosures remains to be established. Over the past
decade, a popular model for mammalian membranes has
emerged that is based on the “lipid raft hypothesis.”1−7
Specifically, it has been proposed that cholesterol and high-
melting sphingolipids form fluctuating nanoscale assemblies
(lipid rafts) that float in a “sea” of low-melting lipids. It has also
been postulated that lipid rafts serve as organizing media for
peripheral proteins and that the nature of the lipids used to
anchor these proteins to membranes controls their partitioning
between raft and nonraft regions, a process that has been
referred to as lipid sorting.1 Thus, saturated hydrocarbon chains
and sterols are thought to favor the association with lipid rafts
while short, branched, and unsaturated hydrocarbon chains
having one or more cis-double bonds (chains having permanent
“kinks”) are believed to favor partitioning into more fluid
regions.1,8
Fluorescence spectroscopy and microscopy, atomic force
microscopy, and surface plasmon resonance have been used to
investigate sorting of lipidated proteins in ternary model
membranes, such as mixtures of 1,2-dipalmitoyl-sn-glycero-3-
phosphocholine (DPPC), 1,2-dioleoyl-sn-glycero-3-phospho-
choline (DOPC), and cholesterol.9−13 Ternary systems have
proven especially popular because they produce macroscopic
(μm-size) liquid-ordered (lo) and liquid-disordered (ld)
domains that coexist and can be readily visualized.
Received: July 30, 2012
© XXXX American Chemical Society
A
dx.doi.org/10.1021/ja3074825 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
are exchanged by randomly selecting partners on the lattice, and a
Glauber step,27 in which the lipid is switched between gel, ld and lo
states. Cholesterol is considered to have only one state. The choice
between attempted moves is aleatory. Acceptance is based on the
Metropolis criterion28 with a move probability that depends
exponentially on the free energy change,22−24,28 using a random
number for the decision.29 The simulations were performed in 100 ×
100 lattices, but it was previously shown that simulations in lattices of
200 × 200 and 300 × 300 sites yield equivalent results in this type of
system.22,30 A Monte Carlo cycle is defined as a number of attempted
moves identical to the number of lattice sites. The calculations
included a pre-equilibration period of 5 × 104 Monte Carlo cycles
followed by a period of 106 acquisition cycles, which were more than
sufficient to obtain equilibrium properties. One lipid−lipid interaction
parameter (ωAB) is used for each pair of possible states (gel, lo and ld)
and lipid species (Chart 1, 1, 2, 3, and 4) present in the system, which
is defined by16
EXPERIMENTAL METHODS
■
Nearest-Neighbor Recognition Measurements. Thin films of
lipid were prepared by evaporating a chloroform solution containing
11.7 μmol total lipid (2.5 mol % of each exchangeable lipid, and 95
mol % DPPC/1,2-dipalmitoyl-sn-glycero-3-phosphoglycerol (DPPG)/
Cholesterol). For the exact composition of the lipid mixtures used for
all NNR reactions, see the Supporting Information. After drying the
thin film overnight under reduced pressure (0.4 mmHg) at room
temperature, 2.0 mL of a 10 mM Tris-HCl buffer (10 mM Tris, 150
mM NaCl, 2 mM NaN3, 1 mM EDTA, pH = 7.4) was added to the
dried film. The lipids were then dispersed by vortex mixing for 30 s,
followed by incubation for 5 min at 60 °C, vortex mixing for an
additional 30 s, and incubation for an additional 30 min at 60 °C with
intermittent vortexing. The resulting dispersion was then subjected to
six freeze/thaw cycles (liquid nitrogen/60 °C water bath) and
extruded 20 times through a 200 nm pore diameter polycarbonate
filter (Nuclepore, Whatman, Inc.) using argon at a pressure of ∼100
psi. An 60 μL aliquot of 1.68 μM monensin in Tris-HCl buffer was
then added to aid in pH equilibration across the membrane during
NNR reactions.
εAA + εBB
ω
= εAB −
AB
(1)
2
Here, εAB represents the contact (nearest-neighbor) interaction
between lipids A and B, which can be any combination of species and
states. In addition, the simulations use the experimental values of the
enthalpy (ΔH) and the transition temperatures (Tm) of the main
phase transition of the phospholipids and lipidated peptides to
calculate the probabilities of changing the lipid state. The model of
Almeida22 was used for DPPC and its analogue, 4. In this model, the
phospholipid accesses essentially only the gel and the ld states in the
absence of cholesterol, but has one more accessible thermodynamic
state, lo, which is intermediate in enthalpy and entropy, in the presence
of cholesterol. The enthalpy of the lo state is assumed to lie at 40% of
the way between those of the gel and ld.22 Because lipidated peptides 1
and 2 resemble phospholipids, having a polar headgroup and two
hydrocarbon chains, we have assumed that they behave similarly to
DPPC regarding gel-to-fluid phase transitions. The important feature
of the model for its present use is to treat the existence of gel, lo and ld
states for the phospholipids and for the presumed gel, lo, and ld states
of the lipidated peptides in a simple way. The model parameters
pertaining to 1 and 2, however, have little effect on the phase behavior
of DPPC/cholesterol, especially because the lipidated peptides only
occur in small amounts in the mixtures studied. Except in the cases
determined here experimentally by nearest-neighbor recognition
measurements (Table 1), the ωAB interaction parameters were the
same as those previously used for DPPC/cholesterol,20 Namely, for
gel-lo and ld−lo interactions ωAB = +330 cal/mol, and for gel-ld
interactions ωAB = +360 cal/mol, where A and B are any phospholipids
(DPPC or 4) or lipidated peptides (1 or 2). The complete set of
parameters is listed in Table SI4 (Supporting Information). The Tm
and ΔH values of the chain-melting transition for 1 and 2 were
estimated from differential scanning calorimetry (DSC) experiments.
The vesicle dispersion (1600 μL) was heated to 45 °C and oxygen
was removed by purging with argon. A thiolate-disulfide interchange
reaction was then initiated by adding threo-dithiothreitol (12.0 μL of a
25.1 mM solution in pH 7.4 Tris buffer, 1.0 equiv with respect to
disulfide content) and sufficient amount of 0.1 M NaOH to bring the
pH to 7.4 at 45 °C. Aliquots (250 μL) were withdrawn as a function of
time, and the exchange reaction quenched by adding 25 μL of 8.3 M
acetic acid to the test tube containing these aliquots, along with vortex
mixing 10 s and quickly freezing the dispersion using liquid nitrogen.
Aliquots were stored at −20 °C until analysis by HPLC was carried
out. To each thawed aliquot was then added 1000 μL of CHCl3/
MeOH (2/1, v/v) and aldrithiol-2 (2,2′-dipyridyldisulfide, 30 μL of a
10 mM solution in CHCl3), and the tube vortex mixed, centrifuged,
and the aqueous phases removed using a Pasteur pipet. The organic
phase was then concentrated under reduced pressure using a Savant
SVC-100 SpeedVac concentrator equipped with a cold trap and
vacuum pump (∼1 h at ∼0.4 Torr). The lipids were then dissolved in
20 μL of CHCl3 and 80 μL of the HPLC mobile phase. These samples
were then analyzed by HPLC using a C18 reversed phase column and
a flow rate of 0.9 mL/min. The mobile phases were composed of 10
mM n-Bu4NOAc in ethanol/water/hexane 76/13/10 (v/v/v) (mobile
phase A) or 77.5/11/11 (v/v/v) (mobile phase B). (See the
Supporting Information for the gradients used for each analysis.)
The column was maintained at 31 °C and the components were
monitored at 203 nm. Values of K (K = [AB]2/([AA] × [BB])) were
calculated from peak areas obtained from the HPLC chromatograms
using appropriate calibration curves; values that are reported are mean
values after dimer equilibrium was reached, typically, within 12 h.
Differential Scanning Calorimetry. Suspensions of lipidated
peptide samples (heterodimers {1−4} and {2−4} (Chart 1), and 2′,
the methyl sulfide derivative of 2) were prepared by hydrating the lipid
film at 85−95 °C in buffer, pH 7.5 (10 mM potassium phosphate or
20 mM MOPS, 0.1 mM EGTA, 0.02% NaN3, and 100 mM KCl).
Concentrations were estimated by weight and (for phosphate-
containing dimers, {1−4} and {2−4}) by a modified Bartlett
phosphate method.20 The heat capacity of the aqueous suspensions
(degassed under vacuum of 500 mmHg for 10 min) was measured
using a high sensitivity Nano DSC (TA Instruments, New Castle, DE),
equipped with 300 μL twin gold capillary cells, under a slight pressure
(set once to 3 atm). The scan rate was 0.1 °C/min. The DSC curves
were corrected by baseline subtraction as previously described.21
Monte Carlo Simulations. Monte Carlo simulations were
performed using the model and methods recently described for
DPPC/cholesterol,22 with standard Monte Carlo procedures.23−25 The
lipid membrane was represented by a triangular lattice, with skew-
periodic boundary conditions, where each site is occupied by a
phospholipid, a lipidated peptide, or a cholesterol molecule.
Equilibrium configurations of the lattice were generated using two
types of steps: a non-nearest-neighbor Kawasaki step,26 in which lipids
RESULTS AND DISCUSSION
■
The Nearest-Neighbor Recognition (NNR) Method. As
discussed elsewhere, the nearest-neighbor recognition (NNR)
method is a chemical technique that probes lipid mixing at the
molecular level.14,31 Thus, NNR measurements take molecular-
level snapshots of bilayer organization by detecting and
quantifying the thermodynamic tendency of exchangeable
monomers to become nearest-neighbors of one another.
Typically, two lipids of interest (A and B) are converted into
exchangeable dimers (homodimers AA and BB, and hetero-
dimer AB) via the introduction of disulfide bonds, which are
then allowed to undergo monomer interchange via thiolate-
disulfide exchange (Figure 1). The resulting equilibrium that is
established is governed by an equilibrium constant, K = [AB]2/
([AA][BB]). When lipid monomers A and B mix ideally, this is
reflected by an equilibrium constant that equals 4.0; when
homoassociations are favored, K < 4.0, and when hetero-
B
dx.doi.org/10.1021/ja3074825 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
to-liquid crystalline phase transition temperatures, which are
41.5 and 39.9 °C, respectively.34 We also note that 4 has an
acidic head group, as DPPG, and because the mixing and
physical properties of DPPC and DPPG are known to be
similar, this exchangeable lipid can serve as a mimic for both
phospholipids.
The synthetic method that was used to prepare the
homodimer of 1 (i.e., {1−1}) is outlined in Scheme 1. Thus,
Scheme 1. Synthesis of {1−1}
alkylation of Boc-protected cysteine with 1-bromohexadecane
to give 5, followed by condensation with cystamine and
deprotection afforded 6. Subsequent acylation with a
myristoylated form of glycine (7) produced the requisite
dimer, {1−1}. The corresponding heterodimer, composed of 1
and 3 (i.e., {1−3}), was prepared using a sequence of reactions
shown in Scheme 2. Thus, activation of cholesterol with N,N′-
Figure 1. A stylized illustration showing the exchangeable homo-
dimers, AA and BB and the corresponding heterodimer, AB, as well as
the equations that describe the dimer equilibrium and the relationship
between the equilibrium constant, K, and the corresponding nearest-
neighbor interaction free energy, ωAB, between A and B.
associations are favored K > 4.0. Taking statistical consid-
erations into account, nearest-neighbor interaction free energies
between A and B are then given by ωAB= −1/2RT ln(K/4).16
Values of ωAB are the primary information that is sought in all
NNR measurements.
Scheme 2. Synthesis of {1−3}
The Exchangeable Lipidated Peptides. The exchange-
able lipids that were designed for this study are shown in Chart
1. Lipidated peptide 1 is a mimic of [(myristoyl)GlyCys-
Chart 1
disuccinimidyl carbonate and condensation with cystamine
afforded 8, followed by condensation with 5, deprotection and
coupling to 7 afforded {1,3}. Heterodimer {1−4} was prepared
by forming an activated derivative of 1,2-dipalmitoyl-sn-glycero-
3-phosphoethanolamine (DPPE), 10, followed by reaction with
the thiol monomer of 1 (i.e., 1-SH) (Scheme 3). The latter was
obtained by reduction of {1−1} with tris(2-carboxyethyl)-
phosphine (TCEP) (not shown). Related methods were used
for the synthesis of homodimer {2−2} and the corresponding
heterodimers, {2−3} and {2−4} (not shown), where myristic
acid was replaced by the cylopropyl adduct of myristoleic acid
(Supporting Informaton).
(palmitoyl)-], having the thioester carbonyl group replaced by a
thioether linkage for enhanced stability. Lipidated peptide 2 is
similar to 1, except for a permanent kink in its C14 chain.
Because double bonds undergo cis-/trans-isomerization under
NNR conditions, a cis-cyclopropyl moiety was used to “lock in”
a kink in 2.32 Exchangeable lipids 3 and 4 have previously been
shown to be excellent mimics for cholesterol and DPPC,
respectively, as judged by their membrane physical properties
and their mutual mixing behavior.33,34 Additionally, DPPC and
4′, the methyl sulfide derivative of 4, exhibit nearly identical gel-
Nearest-Neighbor Interactions. For all of the NNR
experiments that are reported herein, an equimolar mixture of
C
dx.doi.org/10.1021/ja3074825 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Scheme 3. Synthesis of {1−4}
the exchangeable lipids (2.5 mol % of each lipid) was included
in host membranes (95 mol %) that were made from a mixture
of DPPC and cholesterol or only DPPC. For reactions using
{1−3} or {2−3}, 2.5 mol % DPPG was included to give all
liposomes used in these studies the same net negative charge.
Cholesterol-rich and cholesterol-poor membranes were pre-
pared with a total sterol content of 40 and 2.5 mol %,
respectively. At the temperature used in this work (45 °C), the
former is in the liquid-ordered (lo) phase, and the latter is in the
liquid-disordered (ld), which are the phases that are commonly
used as models for lipid rafts and more fluid regions of cell
membranes.35
Figure 2. Energy diagram showing the nearest-neighbor interaction
free energies, ωAB, for various pair of lipids and lipidated peptides in
the lo phase (left) and the ld phase (right), separated into peptide-
cholesterol (Pep-Chol), cholesterol-phospholipid (Chol-PL) and
peptide-phospholipid (Pep-PL) interactions.
Using standard conditions that have been developed for
carrying out thiolate-disulfide interchange in liposomal
membranes, NNR measurements were made for the
interactions of (i) 1 with 3 and 4, and (ii) 2 with 3 and 4 in
the lo and ld states. The interaction of 3 with 4 has previously
been reported.35 Our principal results are shown in Table 1.
and 4 mix randomly in the lo phase and show a slight preference
for heteroassociation in the ld phase, 2 and 4 have a strong
preference for heteroassociation in both phases. The fact that
the changes that occur within each trend are similar in
magnitude strongly suggests that they have a common origin.
We posit that these trends are best explained by (i) the well-
known preference of cholesterol to associcate with ordered
rather than “kinked” chains, (ii) a very unfavorable homointer-
action between two molecules of 2 (kinked chains) in bilayers
based on DPPC, and (iii) a tendency of lipid bilayers to
maximize hydrocarbon chain packing. In considering these
trends, it is important to keep in mind that our experimental
values of ωAB do not represent absolute energies for the interactions
between A and B. Rather, they are a measure of the difference in
free energy between the heterointeractions and the average of the
homointeractions (eq 1).
First, let us consider the interactions between each of the
lipidated peptides (1 and 2) and the phospholipid (4). In the
case of 1 interacting with 4, the mixing is, essentially, ideal in
both the ld and lo phases. This finding implies that there are no
special homo- or heterointeractions occurring in either phase.
In contrast, there is a strong preference for heteroassociation
between 2 and 4 in these same two phases. Since the
headgroups of 1 and 2 are identical, this strong heteroassoci-
ation must be due to the “kink” that is present in 2. Intuitively,
one might suppose that some type of complex is being formed
between 2 and 4 based on these large negative values of ω24.
However, the probability of complexation occurring in both the
lo and the ld phases is very low. A much more plausible scenario
is one in which favorable heteroassociation between 2 and 4 is
driven by unfavorable homointeractions between two molecules
of 2. Thus, the poor packing efficiency of 2, resulting from the
presence of a “kink,” leads to a homodimer, {2−2}, that is
much less stable than the corresponding heterodimer, {2−4}.
The heteroassociation between 2 and 4 becomes more
favorable in the ld phase because their packing behavior
becomes better tolerated in a disordered state.16
a
Table 1. Nearest-Neighbor Recognition
lo
ωAB (cal/mol)
ld
lipid pair
K
K
ωAB(cal/mol)
1,3
1,4
2,3
2,4
3,4
5.7 0.4
4.0 0.2
5.0 0.3
8.0 1.3
9.8 0.5
−110 23
0.8 14
2.8 0.4
4.8 0.3
2.7 0.3
9.8 0.6
3.5 0.1
108 41
−55 20
123 31
−74 21
−219 51
−282 15
−283 19
40
7
a
All measurements were made at 45 °C.
In Figure 2 is shown an energy diagram that summarizes the
interaction free energies, ωAB, for each pair of lipids listed in
Table 1. For ease of interpretation, they have been separated
into three general categories: lipidated peptide-cholesterol
(Pep-Chol), cholesterol-phospholipid (Chol-PL) and peptide-
phospholipid (Pep-PL) interactions.
Inspection of this diagram reveals two distinct trends. First,
the association of the cholesterol analogue (3) with both
lipidated peptides (1 and 2) and with the phospholipid
analogue (4) becomes more favorable on going from the ld to
the lo phase. In particular, favored homoassociation in the ld
phase crosses over to favored heteroassociation in the lo phase.
In addition, this change is similar in magnitude in each case.
Second, the association of each of the lipidated peptides (1 and
2) with the phospholipid analogue 4 becomes less favorable as
one goes from the ld to the lo phase. Here also, the magnitude of
these changes is similar in each case. Additionally, the difference
between 1 and 2 associating with 4 is dramatic. Thus, whereas 1
D
dx.doi.org/10.1021/ja3074825 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Now let us consider the lipidated peptide-sterol and
phospholipid-sterol interactions. Within experimental error,
the interactions of the lipidated peptides (1 and 2) with the
sterol (3) are identical in both the lo and ld phases (Table 1).
Also, these heteroassociations are slightly less favorable (by
∼100 cal/mol) than that found for 3 interacting with 4
(cholesterol and DPPC analogues). Whereas the mixing of 3
with 4 is close to ideal (i.e., ω34 ≈ 0) in the ld phase,
homoassociations are favored for 1 and 2 interacting with 3 in
this same phase. The latter is a likely consequence of poor
interactions between the rigid sterol and the disordered
hydrocarbon chains of the lipidated peptides, which are spaced
further apart than in the phospholipid. If the interactions
between two molecules of 2 are, in fact, very unfavorable, then
the association between 2 and 3 is also expected to be weak.
This can account for the positive ω23. It should be noted, in this
regard, that unfavorable interactions between disordered
(kinked) phospholipid chains and cholesterol are quite
common.16
Chart 2
In the lo phase, the net heteroassociation between 3 and 1, 2,
and 4 (i.e., ω13, ω23 and ω34, respectively) is enhanced. At
present, we believe this is mainly due to more favorable
interactions between the sterol and the hydrocarbon chains that
are now more ordered. Although unfavorable sterol−sterol
interactions (ε) in the lo phase can also account for more
favorable heteroassociations (eq 1), we think that such
contributions are likely to be of minor importance since only
1, 2 and 4, are capable of undergoing significant conformational
changes between the ld to the lo phase; the conformation of the
rigid sterol is, essentially, constant. The similarity in behavior
among 1, 2 and 4 that we have observed in these studies
strongly suggests that these lipidated peptides access conforma-
tional states that are similar to those of DPPC, for which 4 is an
analogue. However, the smaller negative values of ω13 and ω23,
relative to ω34, indicate that the association between these
lipidated peptides and the sterol are not quite as favorable as
phospholipid−sterol association. In the Supporting Information
section, we present a more quantitative analysis of these energy
differences, which supports the conclusion that nearest-neighbor
interactions between two molecules of 2, and also between a
molecule of 2 interacting with a molecule of 3, are especially weak.
Chain Melting of the Lipidated Peptides. Both of the
exchangeable lipidated peptides, 1 and 2, bear a resemblance to
common phospholipids by having one polar headgroup and
two hydrocarbon chains. One might expect, therefore, that
these peptides would exhibit similar gel to liquid-crystalline
phase transition behavior. In an effort to explore this possibility,
we first synthesized nonexchangeable derivatives of 1 and 2 (1′
and 2′, respectively), where the thiol moiety was “capped” with
a methyl group (not shown). Attempted dispersal of 1′ in
buffer proved impossible, even at temperatures as high as 80
°C. In contrast, 2′ could be dispersed at 80 °C. Examination of
the latter by high-sensitivity DSC showed a well-defined
endotherm with an apparent Tm of 61 °C. Because of the
difficulty in quantifying the amount of 2′ present in this
dispersion, a reliable enthalpy value could not be obtained.
In an alternative approach for gaining insight into the melting
behavior of 1 and 2, we examined the thermal properties of
dispersions made from heterodimers {1−4} and {2−4}.
Previously, we showed that the melting of exchangeable
homodimers formed from 4, as well as analogues bearing
myristoyl and stearoyl chains (4a and 4b, Chart 2) occurs at Tm
that are nearly identical to those of phosphocholines with the
same acyl chains.34 We have also shown that the corresponding
dimers exhibit enthalpies (ΔH) that are essentially additive of
those of the monomers, and that the Tm values tend to be
weighted more heavily in favor of the lower-melting lipid.34
This bias toward the lower-melting lipid is analogous to what is
known for phospholipids bearing two different acyl chains. For
example, the Tm values for 1-palmitoyl-2-oleyol-sn-glycero-3-
phosphocholine (POPC), 1,2-dioleyol-sn-glycero-3-phospho-
choline (DOPC), and DPPC are −3, −18, and 41.5 °C,
respectively.36 This behavior is probably a consequence of
disordering of the longer hydrocarbon chains by the lower-
melting partner in the bilayer phase. On the basis of the data in
Table 2, we can estimate ΔH and Tm for the chain melting of 1
and 2 from phospholipid-based heterodimers; that is, by
assuming that ΔH for {1−4} and {2−4} are simply the sum of
the corresponding monomers and, to a first approximation, that
high
Tm = (0.65 × Tmlow) + (0.35 × Tmhigh), where Tmlow and Tm
Table 2. Melting Behavior of Dimers and Monomers
Tm (°C)
monomer
ΔH (kcal/mol)
dimer monomer
a
a
lipid
dimer
DMPC
{4a−4a}
DPPC
23.8
6.1
b
22.7
41.9
14.7
18.7
41.5
39.9
54.8
8.5
9.3
b
c
c
{4−4}
DSPC
10.9
b
{4b−4b}
{4a−4}
{4a−4b}
{1−4}
55.4
−31.2
33.9
55
21.7
16.7
18.7
d
d
d
23.8, 41.5
∼7 , 10
∼7 , 11
∼10 , 10
∼7 , 10
b
d
23.8, 54.8
d
d
d
∼80 , 41.5
∼24
∼17
e
d
d
{2−4}
45
61 , 41.5
a
The values given are for phosphocholines, 1,2-myristoyl-sn-glycero-3-
phosphocholine (DMPC), DPPC, and 1,2-stearoyl-sn-glycero-3-
phosphocholine (DSPC), and are from a combination of references
b
c
d
(i.e., 36−44). Taken from reference 34. Value of 4′. Estimated
e
value. Value of 2′.
E
dx.doi.org/10.1021/ja3074825 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Figure 3. Snapshots of Monte Carlo simulations of mixtures of DPPC and cholesterol containing 2.5 mol % of 1: (A) ld phase, DPPC/cholesterol/1,
95/2.5/2.5 (mol/mol/mol); (B) lo phase, DPPC/cholesterol/1, 57.5/40/2.5 (mol/mol/mol); (C) ld/lo coexistence region, DPPC/cholesterol/1,
77.5/20/2.5 (mol/mol/mol); (D) same as panel C, except that ω34 (nearest-neighbor interaction free energy between 3 and 4 in the ld phase) has
been artificially set at 400 cal/mol instead of the observed value of 40 cal/mol. DPPC is shown in black (gel), white (ld), or blue (lo); cholesterol is
shown in red; and 1 is shown in green. No distinction is made between DPPC and 4, and between cholesterol and 3..
are the Tm values for the low and high-melting monomers (see
Supporting Information for details).
represents a phospholipid, a lipidated peptide, or a sterol
molecule. The phospholipids can exist in three states: gel, lo,
and ld state. The lo state is intermediate to the gel and ld states
in terms of its enthalpy, entropy, and chain order. Simulations
were performed at the same temperature (45 °C) using NNR
values that were now experimentally determined in each phase,
in addition to those previously used for DPPC/cholesterol.22
Snapshots of the simulations are shown in Figure 3 for the lo, ld,
and lo/ld (20 mol % cholesterol) coexistence phases. DPPC is
shown in black (gel), white (ld), or blue (lo); cholesterol is
shown in red, and 1 is shown in green. No distinction is made
between DPPC and 4, or between cholesterol and 3.
A partition coefficient for each lipidated peptide, 1 or 2, was
defined by their distribution between lo and ld regions of the
membrane according to eq 2. Here, [1 or 2]o and [1 or 2]d
represent the number of lipidated peptides in the lo and ld
regions, and [lo] and [ld] are the number of phospholipid sites
belonging to each region.
Examination of {1−4} by DSC revealed a main transition at
55 °C with an apparent ΔH ≈ 24 4 kcal/mol. Given the
monomer chains involved (similar to DPPC), this seems a little
too large, and a value close to the lower bound, ∼20 kcal/mol,
seems more reasonable. On the basis of these considerations,
ΔH for 1 is estimated to be ∼10 kcal/mol. Further, using the
approximation above, with Tm = 55 °C and Tmlow = 41.5 °C, we
obtain Tmhigh∼80 °C for 1. Note that this high Tm value readily
explains our inability to disperse 1′. In contrast, {2−4} gave a
broad and complex transition that was centered around 45 °C,
with a significant dependence on the hydration temperature
and thermal history (Supporting Information). Our best
estimate for its overall enthalpy change is 17
4 kcal/mol,
which leads to an estimated ΔH of ∼7 kcal/mol for 2, assuming
that there is again a ∼10 kcal/mol contribution from the 4
moiety. It should be noted that the level of uncertainty affecting
ΔH has a negligible effect on the Monte Carlo simulation
results.30 We consider the Tm of 61 °C, which we measured
directly for 2′, to be a more reliable estimate for the Tm of 2
than what could be calculated from the transition of the
heterodimer {2−4}. We did not investigate the melting
behavior of {2−4} further. Judging from the DSC, the phase
transitions of all the heterodimers containing lipidated peptides
are certainly more complex than a gel to fluid transition.
However, all that is required for this study is an estimate of the
tendency of monomers 1 and 2 to be in an ordered or
disordered state, which is provided by Tm and, to a lesser
extent, by ΔH. This is all the information we extract from the
DSC results.
[1 or 2]o /[lo]
Kp =
[1 or 2]d /[ld]
(2)
For these calculations, a lipidated peptide molecule was
considered to be located in an lo or ld region according to the
majority of nearest neighbor lipids surrounding it. In defining
Kp, only phospholipid molecules were counted; the sites
occupied by sterol were not counted for either phase. The
results obtained for 1 and 2 in the ld/lo coexistence region (20
mol % cholesterol) are summarized in Table 3.
Because of the presence of long saturated hydrocarbon
chains in 1, we expected that this lipidated peptide would favor
the lo phase. This, however, did not prove to be the case.
Rather, the peptide was evenly distributed between both phases
(Kp = 1.0). In contrast, the presence of a permanent kink in 2
Monte Carlo Simulations. Lipid membranes were
simulated as 100 × 100 triangular lattices, where each site
F
dx.doi.org/10.1021/ja3074825 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
a
values along with the interaction free energy values for ω13, ω14,
ω23, and ω24 in the two phases (determined by NNR
measurements) in eqs 3 and 4, the mole fractions of cholesterol
Table 3. Partition Coefficients for 1 and 2
b
ω34
Kp
lipid
phase behavior
(cal/mol)
(lo/ld)
in the lo and ld regions are calculated to be fco = 0.36 and fcd
=
1
1
2
2
small domains
phase separation
small domains
phase separation
40
400
40
1.0 0.15
1.2 0.2
0.08. In other words, the cholesterol content in the lo and ld
phases are estimated to be 36 and 8 mol %, respectively, which
are very reasonable values given an overall cholesterol content
of 20 mol % in this mixture and the phase diagram for DPPC/
cholesterol.22
It should also be noted that this treatment then allows one to
calculate Kp directly from the interaction parameters, ωAB, if the
cholesterol content in the two phases is known even if a true
phase coexistence does not actually exist in the lo/ld region, but
only small domains are present. This further illustrates the
power of NNR measurements for probing lipid sorting at the
molecular level. In summary, the values of Kp that we have
calculated from these Monte Carlo simulations can be
understood very simply and directly in terms of the
experimental values of ωAB. If they appear small, it is because
the preference of the lipidated peptides between the ld and lo
regions is, indeed, very weak.
0.46 0.07
0.46 0.07
400
a
Determined from Monte Carlo simulations in membranes of DPPC/
cholesterol/(1 or 2) 77.5/20/2.5 (mol/mol/mol). No distinction is
made between DPPC and 4, and between cholesterol and 3. The
nearest-neighbor interaction free energy between 3 and 4 in the ld
phase was set either at the experimentally determined value of 40 cal/
mol or at a hypothetical value of 400 cal/mol to strongly favor phase
separation.
b
led to an expected preference for the ld phase, albeit only
modest (Kp ≈ 0.5). To test if Kp actually depended on domain
size, we carried out additional Monte Carlo simulations in
which the nearest-neighbor interaction free energy between
DPPC and cholesterol (ω34) in the ld phase was changed from
the experimental value of 40 cal/mol to an artificial value of 400
cal/mol. As expected, this led to a system approaching true
phase separation (Figure 3D). However, the calculated Kp in
this artificial situation was essentially unchanged (Table 3).
Thus, the lack of a significant preference for these lipidated peptides
to reside in either phase appears to be a robust property.
Partitioning of other lipidated forms of GlyCys between a lo
and ld phase have previously been investigated by Silvius and
co-workers in host membranes made from DPPC, 1,2-dioleoyl-
sn-glycero-3-phosphocholine and cholesterol using a fluores-
cence quenching assay.9 In that study, it was shown that the
introduction of a single cis-double bond in a myristoyl chain
was sufficient to increase the peptide preference for the ld
phase. Qualitatively, our results are in complete agreement with
those earlier findings.
Interpretation of Partition Coefficients. The partition
coefficients, Kp, which characterize the partitioning of 1 and 2
between liquid-ordered and liquid-disordered regions, can be
understood in a very simple way in terms of our nearest-
neighbor interaction free energies. First, we can relate Kp to the
change in interactions experienced by the lipidated peptides as
they are transferred from the ld to the lo phase. Let us define a
mean-field, Δω, for each lipidated peptide, 1 and 2, which
measures the change in the interactions upon transfer from the
ld to the lo regions by a weighted average of the interactions in
each region. Here, fco and fcd are the mole fractions of
cholesterol in the ordered and disordered regions, respectively,
(the superscripts o and d indicate those two phases in eqs 3 and
4).
CONCLUSIONS
■
In this paper, we have shown that nearest-neighbor recognition
measurements, in combination with Monte Carlo simulations,
afford reasonable estimates of the partition coefficients of
lipidated peptides between liquid-ordered and liquid-disordered
regions of fluid membranes. We have also shown that these
partition coefficients are insensitive to the size of lo and ld
domains. Our results further indicate that the NNR method, in
combination with a phase diagram, should provide a good
estimate of partitioning of a species between two phases.
In a broader context, the very weak directing effect that
lipidation of GlyCys has on its partitioning between lo and ld
regions, observed here, strongly suggests that the sorting of
lipidated peripheral proteins on the basis of their fatty acyl
chain modifications is more subtle than previously realized. Our
results are reminiscent of those reported for the dual-lipidated
peripheral protein N-Ras. With one farnesyl and one hexadecyl
chain, N-Ras partitions to the ld phase or to ld−lo interfaces, but
not to the lo phase.11,12 This behavior is similar to that of our
lipidated peptide 2, which also contains a Cys-linked, ordered,
hexadecyl chain and a kinked chain. Further, our lipidated
peptide 1, bearing one myristoyl and one hexadecyl chain,
shows no preference for the lo phase. In fact, even with two
hexadecyl chains, N-Ras still prefers the ld phase.11 In other
words, this peptide and protein, bearing long hydrocarbon
chains that can exist in an all-anti conformation, show no
special affinity to the lo phase. .
Finally, from a biological standpoint, it should be noted that
our results with these simple model systems are in complete
agreement with a recent study of the dynamic colocalization of
lipid-anchored fluorescent proteins in live COS 7 cells, using
pulsed-interleaved excitation fluorescence cross-correlation
spectroscopy (PIE-FCCS) and fluorescent lifetime analysis.45
Specifically, neither study supports the view that the
organization of lipidated proteins is determined, alone, by the
partitioning of their lipid anchors between two lipid phases.
Δω1 = (ω1o3f o + ω1o4(1 − f o )) − (ω1d3f d + ω1d4(1 − f d ))
(3)
Δω2 = (ω2o3f o + ω2o4(1 − f o )) − (ω2d3f d + ω2d4(1 − f d ))
(4)
c
c
c
c
c
c
c
c
According to the mean-field approximation, the partition
coefficient is then related to Δω by
Kp = e−6Δω/RT
(5)
ASSOCIATED CONTENT
* Supporting Information
Complete experimental procedures for the synthesis of the
exchangeable lipids and for NNR measurements as well as raw
■
S
where there are 6 nearest-neighbors in a triangular lattice. Using
the values of Kp of 1.0 for 1 and 0.46 for 2, we obtain Δω1 = 0
for 1 and Δω2= +80 cal/mol for 2 at 45 °C. Using these two
G
dx.doi.org/10.1021/ja3074825 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(21) Pokorny, A.; Yandek, L. E.; Elegbede, A. I.; Hinderliter, A.;
Almeida, P. F. F. Biophys. J. 2006, 91, 2184−2197.
(22) Almeida, P. F. Biophys. J. 2011, 100, 420−429.
(23) Heimburg, T. Thermal Biophysics of Membranes; Wiley-VCH:
Weinheim, Germany, 2007; pp 123−140.
NNR data, DSC data, and a quantitative interpretation of the
NNR results. This material is available free of charge via the
Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
slr0@lehigh.edu; almeidap@uncw.edu
■
(24) Sugar, I. P.; Biltonen, R. L.; Mitchard, N. Methods Enzymol.
1994, 240, 569−593.
(25) Binder, K.; Heerman, D. W. Monte Carlo Simulations in
Statistical Physics, 3rd ed.; Springer: New York, 1997.
(26) Kawasaki, K. Kinetics of Ising Models. In Phase Transitions and
Critical Phenomena; Domb, C., Green, M. S., Eds.; Academic Press:
New York, 1972; Vol. 2, pp 443−501.
(27) Glauber, R. J. J. Math. Phys. 1963, 4, 294−307.
(28) Metropolis, N.; Rosenbluth, A. W.; Rosenbluth, W. N.; Teller,
A. H.; Teller, E. J. Chem. Phys. 1953, 21, 1087−1092.
(29) Press, W. H.; Teukolsky, S. A.; Vetterling, W. T.; Flannery, B. P.
Numerical Recipes in FORTRAN: The Art of Scientific Computing, 2nd
ed.; Cambridge University Press: Cambridge U.K., 1994.
(30) Svetlovics, J. A.; Wheaten, S. A.; Almeida, P. F. Biophys. J. 2012,
102, 2526−2535.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We are grateful to our colleagues, Drs. Serhan Turkyilmaz,
Ravil Petrov and Vaclav Janout for helpful discussions. This
work was funded by the National Science Foundation (CHE-
1145500), the National Institutes of Health (PHS GM56149)
and North Carolina Biotechnology Center grant 2009-IDG-
1031.
(31) In contrast to other methods that have been used to determine
ωAB values (i.e., differential scanning calorimetry, fluorescence
resonance energy transfer, isothermal titration calorimetry, and
analyses of phase diagrams), the NNR method does not require any
matching of experimental data with theoretical curves.16 It is also
highly sensitive and can detect changes in free energies of interaction
down to tens of calories per mole.
REFERENCES
■
(1) Lingwood, D.; Simons, K. Science 2010, 327, 46−50.
(2) McIntosh, T. J., Ed. Lipid Rafts; Springer Verlag: New York, NY,
2007.
(3) (a) Lichtenberg, D.; Goni, F. M.; Heerklotz, H. Trends Biochem.
Sci. 2005, 30, 430−436. (b) Heerklotz, H. Biophys. J. 2002, 83, 2693−
2701.
(32) Jing, B.; Tokutake, N.; McCullough, D. H., III; Regen, S. L. J.
Am. Chem. Soc. 2004, 126, 15344−15345.
(4) Edidin, M. Annu. Rev. Biomol. Struct. 2003, 32, 257−283.
(5) Munro, S. Cell 2003, 115, 377−388.
(33) Zhang, J.; Cao, H.; Jing, B.; Almeida, P. F.; Regen, S. L. Biophys.
J. 2006, 91, 1402−1406.
(6) McMullen, T. P. W.; Lewis, R. N. A. H.; McElhaney, R. N. Curr.
Opin. Colloid Interface Sci. 2004, 8, 459−468.
(34) Kristovitch, S. M.; Regen, S. L. J. Am. Chem. Soc. 1992, 114,
9828−9835.
(7) Veatch, S.; Cicuta, P.; Sengupta, P.; Honerkamp-Smith, A;
Holowka, D.; Baird, B. ACS Chem. Biol. 2008, 3, 287−293.
(8) Levental, I.; Grzybek, M.; Simons, K. Biochemistry 2010, 49,
6305−6316.
(35) Daly, T. A.; Wang, M.; Regen, S. L. Langmuir 2011, 27, 2159−
2161.
(36) Phospholipids Handbook; Cevc., G., Ed.; Marcel Dekker: New
York, 2001; pp 939−956.
(37) Hac, A. E.; Seeger, H. M.; Fidorra, M.; Heimburg, T. Biophys. J.
2005, 88, 317−333.
(38) Ivanova, V. P.; Heimburg, T. Phys. Rev. E. 2001, 63, 1914−1925.
(39) Suurkuusk, J.; Lentz, B. R.; Barenholz, Y.; Biltonen, R. L.;
Thompson, T. E. Biochemistry 1976, 15, 1393−1401.
(40) Mabrey, S.; Sturtevant, J. Proc. Natl. Acad. Sci. U.S.A. 1976, 73,
3862−3866.
(41) Estep, T. N.; Mountcastle, D. B.; Biltonen, R. L.; Thompson, T.
E. Biochemistry 1978, 17, 1984−1989.
(42) Blume, A. Biochemistry 1983, 22, 5436−5442.
(43) Marsh, D. CRC Handbook of Phospholipid Bilayers; CRC Press:
Boca Raton, FL, 1990; p 149.
(9) Wang, T.-Y.; Leventis, R.; Silvius, J. R. Biochemistry 2001, 40,
13031−13040.
(10) Janosch, S.; Nicolini, C.; Ludloph, B.; Peters, C.; Voklert, M.;
Hazlet, T. L.; Gratton, E.; Waldmann; Winter, R. J. Am. Chem. Soc.
2004, 126, 7496−7503.
(11) Weise, K.; Triola, G.; Brunsveld, L.; Waldmann, H.; Winter, R. J.
Am. Chem. Soc. 2009, 131, 1557−1564.
(12) Weise, K.; Triola, G.; Janosch, S.; Waldmann, H.; Winter, R.
Biochim. Biophys. Acta 2010, 1798, 1409−1417.
(13) Gohlke, A.; Triola, G.; Waldmann, H.; Winter, R. Biophys. J.
2010, 98, 2226−2235. Weise, K.; Kapoor, S.; Denter, C.; Nikolaus, J.;
Opitz, N.; Koch, S.; Triola, G.; Herrmann, A.; Waldmann, H.; Winter,
R. J. Am. Chem. Soc. 2011, 133, 880−887.
(14) (a) Davidson, S. K. M.; Regen, S. L. Chem. Rev. 1997, 97, 1269−
1279. (b) Regen, S. L. Curr. Opin. Chem. Biol. 2002, 6, 720−735.
(c) Cao, H.; Tokutake, N.; Regen, S. L. J. Am. Chem. Soc. 2003, 125,
16182−16183. (d) Sugahara, M.; Uragami, M.; Yan, X.; Regen, S. L. J.
Am. Chem. Soc. 2001, 123, 7939−7940. (e) Mitomo, H.; Chen, W.-H.;
Regen, S. L. J. Am. Chem. Soc. 2009, 131, 12354−12357. (f) Cao, H.;
Zhang, J.; Jing, B.; Regen, S. L. J. Am. Chem. Soc. 2005, 127, 8813−
8816. (g) Zhang, J.; Jing, B.; Tokutake, N.; Regen, S. L. Biochemistry
2005, 44, 3598−3603.
(44) Koyonova, R.; Caffrey, M. Biochim. Biophys. Acta 1998, 1376,
91−145.
(45) Triffo, S. B.; Huang, H. H.; Smith, A. W.; Chou, E. T.; Groves, J.
T. J. Am. Chem. Soc. 2012, 134, 10833−10842.
(15) Frazier, M. L.; Wright, J. R.; Pokorny, A.; Almeida, P. F. F.
Biophys. J. 2007, 92, 2422−2433.
(16) Almeida, P. F. F. Biochim. Biophys. Acta 2009, 1788, 72−85.
(17) Recktenwald, D. J.; McConnell, H. M. Biochemistry 1981, 20,
4505−4510.
(18) Subramaniam, S.; McConnell, H. M. J. Phys. Chem. 1987, 91,
1715−1718.
(19) Collins, M. D.; Keller, S. L. Proc. Natl. Acad. Sci. U.S.A. 2008,
105, 124−128.
(20) Bartlett, G. R. J. Biol. Chem. 1959, 234, 466−468. Pokorny, A.;
Birkbeck, T. H.; Almeida, P. F. F. Biochemistry 2002, 41, 11044−
11056.
H
dx.doi.org/10.1021/ja3074825 | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX