Investigation on lipid asymmetry using lipid probes: Comparison between spin-labeled lipids and fluorescent lipids.

Abstract

Synthetic lipids with a nitroxide or a fluorescent probe have been extensively used during the last 30 years to determine the transmembrane diffusion of phospholipids in artificial or biological membranes. However, the relevance of data obtained with these modified lipids has sometimes been questioned. Beside possible artefacts introduced by the reporter probe, synthetic lipids used in cells often contain a short fatty acid chain in the sn-2 position, which gives them higher water solubility than naturally occurring lipids. In the present review, we have attempted to give a critical appraisal. Main strategies are recalled and important discoveries obtained with lipid probes on transmembrane lipid traffic in eukaryotic cells are briefly summarized. Examples of artefacts caused by lipid probes are given. Comparisons between data obtained by different techniques such as ESR and fluorescence allow us to emphasize the complementary character of the two approaches and more generally show the necessity to use several probes before drawing conclusions concerning endogenous lipids. In spite of these pitfalls, overall, lipid probes have provided a wealth of useful information that, to date, cannot be obtained with unlabeled lipids.

Full text

Page 1

Chemistry and Physics of Lipids
116 (2002) 115–134
Investigation on lipid asymmetry using lipid probes
Comparison between spin-labeled lipids and fluorescent
lipids
Philippe F. Devaux *, Pierre Fellmann, Paulette Herve ´
Institut de Biologie Physico-Chimique, UMR CNRS 7099, 13 rue Pierre et Marie Curie, F75005 Paris, France
Abstract
Synthetic lipids with a nitroxide or a fluorescent probe have been extensively used during the last 30 years to
determine the transmembrane diffusion of phospholipids in artificial or biological membranes. However, the relevance
of data obtained with these modified lipids has sometimes been questioned. Beside possible artefacts introduced by
the reporter probe, synthetic lipids used in cells often contain a short fatty acid chain in the sn-2 position, which gives
them higher water solubility than naturally occurring lipids. In the present review, we have attempted to give a critical
appraisal. Main strategies are recalled and important discoveries obtained with lipid probes on transmembrane lipid
traffic in eukaryotic cells are briefly summarized. Examples of artefacts caused by lipid probes are given. Comparisons
between data obtained by different techniques such as ESR and fluorescence allow us to emphasize the complemen-
tary character of the two approaches and more generally show the necessity to use several probes before drawing
conclusions concerning endogenous lipids. In spite of these pitfalls, overall, lipid probes have provided a wealth of
useful information that, to date, cannot be obtained with unlabeled lipids. © 2002 Elsevier Science Ireland Ltd. All
rights reserved.
Keywords: Spin-labeled lipids; Fluorescent lipids; ESR; Transmembrane diffusion; Aminophospholipid translocase
www.elsevier.com/locate/chemphyslip
1. Introduction
The transmembrane diffusion of lipids in bio-
logical membranes (also called ‘lipid flip-flop’) has
been measured essentially using lipid probes, that
is with lipids which have been chemically modified
in order to facilitate their transmembrane local-
ization. To this end, various spin-labeled analogs
of naturally occurring lipids were synthesized with
a nitroxide radical either on the polar head-group
or on one of the acyl chains. Fluorescent lipids
were also successfully employed for such investi-
gation and more generally for the investigation of
lipid traffic within eukaryotic cells. Other lipid
probes comprise lipids labeled with13C or2H for
NMR experiments and of course radioactive
lipids labeled with14C or3H atoms that have been
used already a long time ago for the determina-
tion of lipid biosynthetic pathways. This review
presents the main results obtained with the lipid
* Corresponding author. Tel.: +33-1-5841-5105; fax: +33-
1-5841-5024.
E-mail address: philippe.devaux@ibpc.fr (P.F. Devaux).
0009-3084/02/$ - see front matter © 2002 Elsevier Science Ireland Ltd. All rights reserved.
PII: S0009-3084(02)00023-3

Page 2

P.F. De?aux et al. / Chemistry and Physics of Lipids 116 (2002) 115–134116
probes in relation with transmembrane lipid diffu-
sion and lipid asymmetry in eukaryotic cells. Be-
cause the relevance of lipid probes has often been
disputed, emphasis will be put on the significance
of data obtained with chemically modified lipids.
Critical comparison between spin-labeled lipids
and fluorescent lipids will be attempted.
There are a few naturally occurring lipids with
fluorescent properties. For example dehydroergos-
terol is present in some cells where it replaces
cholesterol (Schroeder et al., 1991). There are also
a few fatty acids with conjugated double bonds
provided with fluorescence properties. However,
the transmembrane diffusion and transmembrane
transport of lipids in cell membranes have been
investigated essentially with chemically modified
lipid probes. Undoubtedly, spin-labeled lipids and
lipids bearing a fluorescent probe are different
from naturally occurring ones. For example, one
can legitimately ask to what extent a phos-
phatidylethanolamine analog containing a fluores-
ceine group that has been reacted with the NH
canstillbeconsidered
phatidylethanolamine molecule.
may arise with spin-labeled or fluorescent lipid
analogs containing a rather bulky group attached
to a fatty acid chain. In the latter case, the steric
perturbation is aggravated by the fact that the
polarity of the reporter group is more important
than one would expect for a residue buried in the
hydrophobic core of a membrane. The label by
itself is not always the only perturbation. Indeed,
because long chain phospholipids are insoluble in
water, lipid probes are frequently designed with
one (or two) relatively short chain (C6or C5),
which enhances their water solubility, and allows
an easy incorporation in the plasma membrane as
well as probe removal by back exchange. How-
ever, the slightly detergent character of lipids with
one short chain can modify the physical proper-
ties of a membrane and eventually, if added at a
high concentration, can even disrupt the lipid
bilayer. Note also that lipid peroxidation gener-
ates lipids with one short chain due to the rupture
of the acyl chain at the position of the double
bond. In vivo peroxidized acyl chains are deleted
and the lipid reacylated. We have found that a
short sn-2 chain is rapidly hydrolyzed in vivo,
asa phos-
doubtsOther
indicating that these lipid probes are better sub-
strates than ordinary long chain lipids for the
endogenous phospholipase A2that is present in
the cytosol of most eukaryotic cells. A possible
explanation is that the lipid probes with a short
sn-2 chain resemble peroxidized lipids that are the
natural target of these enzymes.
So lipid probes with one short chain have some
drawbacks. However, when long chain lipid
probes are used, they generally contain no double
bonds because saturated lipids are more stable
and easier to manipulate. Yet in vivo lipids with
one or several double bonds are more widespread
than saturated lipids. In that respect
lipids can sometimes be considered as well as
modified lipids.
14C or
1H
2. Spin-label flip-flop in lipid vesicles and in
erythrocytes
2.1. Transmembrane diffusion in sonicated ?esicles
The historical experiment that showed for the
first time the (very) slow transmembrane diffusion
of lipids in a lipid bilayer was carried out in 1971
by Kornberg and McConnell (Kornberg and Mc-
Connell, 1971). These authors used a spin-labeled
phosphatidylcholine analog carrying a 6 member
nitroxide ring on the phospholipid head-group in
place of one of the methyl of the quaternary
ammonium (molecule I, Fig. 1). The experiment
carried out by Kornberg and McConnell was
performed with sonicated unilamellar vesicles
(SUVs) containing a small proportion of the spin-
labeled lipids (?1%). The distribution of spin-la-
beled lipids was made initially asymmetrical by
chemical reduction by sodium ascorbate of the
probes exposed on the outer monolayers of SUVs.
Fig. 1. Spin-label I: Tempo-phosphatidylcholine. The probe is
exposed to the aqueous phase and can be reduced very rapidly
by a few mM sodium ascorbate.

Page 3

P.F. De?aux et al. / Chemistry and Physics of Lipids 116 (2002) 115–134 117
After filtering out ascorbate with a sepharose
column, SUVs were allowed to incubate and,
afterwards, the fraction of the probe that diffused
to the outer monolayer was determined by adding
fresh ascorbate. A half time of residence of 6.5 h
at 30 °C was determined by this technique in
sonicated vesicles.
This experiment had a crucial historical impor-
tance because there was no knowledge of even the
order of magnitude of the rate of lipid flip-flop in
a lipid bilayer. On the other hand of course, this
experiment dealt with a phospholipid with an
abnormally large head-group flipping in SUVs,
which are membranes highly curved and hence
under tension, unlike most biological membranes.
Nevertheless, by setting the time scale, the spin-
label experiment was an incitement for other lab-
oratories to use techniques that allows one to
measure flip-flop of naturally occurring lipids for
example by enzymatic assays. The phospholipase
A2assay confirmed the slow flip-flop of endoge-
nous phosphatidylcholine (Zwaal et al., 1973). In
1986, Middelkoop and collaborators showed by
exposing red cell membranes to exogenous phos-
pholipase A2that phospholipids with at least one
unsaturated chain experience a more rapid flip-
flop than saturated lipids. Half-times varied from
26.3 to 2.9 h for 1,2-dipalmitoyl-PC and 1-palmi-
toyl 2-lineoyl-PC respectively, thus the length of
the chain is an important factor (Middelkoop et
al., 1986). As one might have anticipated, the
details of the chemical structure of a lipid influ-
ences the flip-flop rate and therefore the presence
of a nitroxide either on the head-group or on one
of the acyl chains can modify the diffusion rate.
However, the main obstacle for transmembrane
diffusion being the charge carried by the lipid
head-group, the nitroxide on the chain, although
polar, is not a determinant factor for lipid flip-
flop. In conclusion, 30 years after Kornberg and
McConnell’s experiment, it has been amply
shown that the 1971 experiment that involved a
lipid probe was quite significant. With a few
exceptions that can be explained on the basis of
the absence of a charged head-group, phospho-
lipid flip-flop in protein-free membranes is a slow
event with a half time of the order of several
hours.
2.2. Protein-induced translocation of phospholipids
in erythrocytes
Using spin-labeled analogs of aminophospho-
lipids(phosphatidylethanolamine
phatidylserine), rapid
translocation of these lipids was found in the
erythrocyte membrane, which could not be ex-
plained on the basis of the physical properties of
the phospholipid analogs (Seigneuret and De-
vaux, 1984; Morrot et al., 1989). Half times of
approximately 60 and 5 min were determined at
37 °C for phosphatidylethanolamine and phos-
phatidylserine (molecules III and IV respectively
in Fig. 2). By contrast in the same cells, phos-
phatidylcholineand
(molecules II and V) diffused as slowly as in
liposomes. With probes II to V, although the
ascorbate technique can be used, the transmem-
brane distribution is easier to determine by back
exchange, a technique introduced by Pagano and
collaborators with fluorescent phospholipids bear-
ing a short sn-2 chain (Sleight and Pagano, 1984).
It takes advantage of a particular feature of the
lipids with a long sn-1 chain and a short sn-2
chain, which is that the lipid analogs can be
washed away from the external leaflet of a cell
membrane with liposomes or with fatty acid free
bovine serum albumin (BSA). Incubating a few
minutes with a buffer containing 1–5% BSA and
spinning down the membranes allows one to de-
termine the fraction of probes in the outer mono-
layer. In the case of spin-labels, it is often
necessary to reoxidize the nitroxide due to chemi-
cal reduction by endogenous reducing agents
present in the cytosol (Fellmann et al., 1994). In
the right hand part of Fig. 3 a back exchange
experiments is schematized. The left hand part of
the same figure shows the ESR spectra of spin-la-
bel II respectively in water, in a lipid membrane,
linked to BSA and after SDS addition. The first
spectrum (spectrum a, Fig. 3) contains a broad
peak due to the strong spin–spin interactions
between molecules in a micelle. The three narrow
lines superimposed are due to a small fraction of
monomers in water experiencing a high frequency
of rotation. The cmc is of the order of a few ?M
and can be determined by ESR spectroscopy
andphos-
lipidoutside–inside
sphingomyelinanalogs

Page 4

P.F. De?aux et al. / Chemistry and Physics of Lipids 116 (2002) 115–134118
Fig. 2. Spin-labeled lipids with a short acyl chain in sn-2
position. Molecules II, III, IV and V are analogs of the four
main phospholipids present in the plasma membrane of eu-
karyotic cells, and correspond to phosphatidylcholine (PC),
phosphatidylethanolamine (PE), phosphatidylserine(PS), sph-
ingomyelin (SM) respectively. Molecules VI and VII are plas-
malogens. Molecule VIII is an analog of lyso-PC and IX is a
PC molecule with a very short sn-2 chain.
ity of the lipid bilayer which is higher than that of
an aqueous medium. To determine the fraction of
probes remaining in the outer monolayer, BSA is
added and the membranes are spun down to
physically separate spin-labels bound to BSA
from spin-labels in the inner monolayer and not
accessible to BSA. This operation is necessary
because spectra b and c (Fig. 3), which corre-
spond to the spin-label in the membrane and
bound to BSA respectively are very similar. To
obtain a more intense signal in the supernatant, it
is convenient to add a detergent such as SDS (1%)
that releases eventually the probe from BSA and
generates more intense peaks (spectrum d, Fig. 3).
The endogenous phospholipase A2that exist in
many cells can be only partially inhibited by
diisopropylfluorophosphate (DFP). Any free fatty
acid that arises from hydrolyzed spin-labeled
phospholipids remains in the supernatant but
does not bind to BSA because short chain fatty
acids do not bind to BSA or to liposomes. Thus,
sometimes a narrow triplet due to spin-labeled
fatty acid tumbling rapidly in water is visible in
the BSA supernatant. This spectral feature is use-
ful because it reports on the probe degradation
and allows a correction to be made (Fellmann et
al., 1994).
Fig. 3. Schematic representation of a back exchange experi-
ment with spin-label lipids. The left hand side of the figure
shows ESR spectra of a spin-labeled analog of phosphatidyl-
choline (molecule II), (a) in water; (b) in a lipid bilayer; (c)
bound to BSA; (d) in the presence of SDS.
(King and Marsh, 1987). As soon as a membrane
suspension is added, the spectrum changes its line
shape (spectrum b, Fig. 3) due to the incorpora-
tion of the probes in a fluid lipid environment.
Spin–spin interactions disappear because the un-
labeled lipids dilute the spin-labeled lipids out,
however the lines are broad because of the viscos-

Page 5

P.F. De?aux et al. / Chemistry and Physics of Lipids 116 (2002) 115–134119
Fig. 4. Kinetics of the outside-inside redistribution of spin-la-
bel analogs of PS, PE, PC and SM (molecules II, III, IV and
V) introduced in the outer leaflet of human erythrocytes at
37 °C. On the right hand side, the percentage of endogenous
phospholipids in the inner leaflet are indicated (reproduced
from Devaux, 1991)
modifications of the head-group on PS or PE
spin-labels, such as the addition of an extra CH2
or esterification of the carboxyl of PS or methyla-
tion of PE, it was shown that the outside–inside
transport was very sensitive to subtle modifica-
tions of the head-group. Spin-labeled alkenyl-acyl
phospholipids (or plasmalogens) with a short (C5)
sn-2 chain bearing a five member oxazolidine
nitroxide ring were also synthesized. It was found
that spin-label plasmalogen PE (molecule VI) was
transported in human erythrocytes membrane al-
most at the same rate as diacyl PE (Fellmann et
al., 1993). On the other hand, plasmalogen PC
(molecule VII) was not transported. Spin-labeled
analogs of lyso-PC and of lyso-PS (molecule VIII)
were practically not transported. However, sur-
prisingly a mere acetylation (molecule IX) allowed
the PS analog to be recognized by the aminophos-
pholipid translocase and to be transported rapidly
from the outer to the inner monolayer of the
plasma membrane (Morrot et al., 1989).
Thus, these spin-labeled lipids with a short sn-2
chain contain the crucial part of a lipid that is
recognized by the aminophospholipid translocase.
Data with fluorescent probes will be discussed
below but first, we shall recall other advantages of
spin-labeling experiments.
To overcome completely the problem of spin-
labelhydrolysis,ether
(molecules X, Fig. 5) were synthesized and their
translocation rates in erythrocyte tested (Fell-
mann et al., 2000). The presence of an ether bond
in sn-2 position does not modify the rate of PS
translocation, as indicated by the translocation of
a 1-ether-2-ester PS analog (Fig. 5). On the other
hand, the di-ether PS is translocated in erythro-
cytes at a slower rate (half time of 40 min at
37 °C instead of 5–10 min for the diester deriva-
tive) but the plateau is the same as that of the
diester spin-label PS. Thus, the ether bond is
indeed a slight perturbation. This confirms once
more that phospholipid translocation is sensitive
to details of the chemical structure of the phos-
pholipid but that a nitroxide on the sn-2 chain
does not impede the rapid translocation of
aminophospholipids.
Finally, another interesting possibility with the
spin-labeled probes described here is that two
spin-labeled lipids
Fig. 4 shows the passage from the outer to the
inner monolayer of four spin-labeled analogs of
the main phospholipids present in the human
erythrocyte membrane (molecules II to V). At
equilibrium, the transmembrane distribution of
the spin-labeled probes is remarkably similar to
that of endogenous phospholipids, as determined
with phospholipase A2and sphingomyelinase in
the same cell (see right hand part of Fig. 4). Thus,
it can be inferred that these probes, in spite of the
nitroxide moiety and in spite of the short sn-2
chain are good reporters of the behavior of en-
dogenous long chain phospholipids. The inhibi-
tion of the active transport by cytosolic calcium,
by N-ethyl-maleimide and by vanadate and the
strict lipid selectivity led to the proposal of a
putative selective transporter named ‘aminophos-
pholipid translocase’.
It can be inferred from Fig. 2 that the head-
group of the phospholipid analogs (formula II to
V) is not perturbed by a probe on the acyl chain.
Similarly, the glycerol moiety or ceramide back-
bone, in the case of the sphingomyelin, is also
unaffected. Any small perturbation, that could
arise from the short sn-2 chain, would be present
in the four probes. Thus, it would be hard to
explain how the nitroxide ring on the sn-2 chain
could determine by itself the selectivity of the lipid
transport. In fact, by doing various chemical