You are on page 1of 60

Biochimica et Biophysica Aeta, 472 (1977) 285-344

Elsevier/North-Holland Biomedical Press


BBA 85178

LIPID

PHASE

II. M I X T U R E S

TRANSITIONS

AND

INVOLVING

LIPIDS

PHASE

DIAGRAMS

A. G. LEE

Department of Physiology and Biochemistry, University of Southampton, Southampton S09 3TU ( U.K.)
(Received January l lth, 1977)

CONTENTS
I.

Introduction

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

II.

The theory of lipid mixtures . . . . . . . . . . . . . . .


A. Simple eutectic system . . . . . . . . . . . . . . . .
B. Systems displaying complete miscibility in the solid state
C. Regular solutions . . . . . . . . . . . . . . . . . .
D. Limiting cases of the eutectic and peritectic . . . . . .
E. Sub-regular solutions . . . . . . . . . . . . . . . . .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

. . . . . . .
. . . . . . .
. . . . . . . . .
. . . . . . .
. . . . . . . .
. . . . . . .

286
286
291
296
304
305

III.

Results obtained with phospholipid mixtures

IV.

Mixtures with steroids . . . . . . . . . . . . . . . . . . . . . . . . . . . .


A. Lipid/cholesterol mixtures . . . . . . . . . . . . . . . . . . . . . . . . .
B. Mixtures with other steroids . . . . . . . . . . . . . . . . . . . . . . . .

314
314
317

V.

Mixtures with other hydrophobic molecules . . . . . . . . . . . . . . . . . . .

318

VI.

Mixtures with polypeptides and proteins . . . . . . . . . . . . . . . . . . . .


A. Polypeptides
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
B. Lipid-protein interactions . . . . . . . . . . . . . . . . . . . . . . . . .

320
320
322

VII.

Interaction with small water-soluble molecules . . . . . . . . . . . . . . . . . .

330

VIII.

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

References

. . . . . . . . . . . . . . . . . .

285

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

306

340
341

I. I N T R O D U C T I O N
Those interested in molecular interactions in biological membranes
very fortunate

are in a

p o s i t i o n , s i n c e a c o n v e n i e n t a n d r e l e v a n t m o d e l s y s t e m is r e a d i l y

available -- the lipid bilayer.

Since X-ray diffraction and other studies have shown

t h a t m u c h o f t h e l i p i d i n b i o l o g i c a l m e m b r a n e s is o r g a n i z e d as a b i l a y e r , t h e s t u d y o f
s i n g l e l i p i d b i l a y e r s is a n o b v i o u s first s t e p t o w a r d s a n u n d e r s t a n d i n g o f t h e f u n c t i o n Abbreviation: TEMPO, 2,2,6,6-tetramethylpiperidine-N-oxyl.

286
ing of the biological membrane. In the first part of this review [62], phase transitions
in bilayers of single species of phospholipids were discussed. Here, the discussion is
extended to include lipid mixtures and mixtures of lipids with other molecules.

II. THE THEORY OF LIPID MIXTURES


In recent years the properties of a number of binary and ternary mixtures of
lipids have been reported, and, although such mixtures are very simple by comparison
with those present in biological membranes, they do at least serve to map out some
of the effects that can be expected. All of the studies discussed here will be of lipid
mixtures in the presence of excess water, since this seems to be the only biologically
relevant condition. It is of course possible to study lipid mixtures as a function of
water content, and a variety of structurally complex phases can be observed at low
water content [1].
The properties of mixtures of compounds are usually presented in the form of
phase diagrams. As discussed elsewhere [2], the properties of such phase diagrams
are clearly defined only for compounds which, when pure, undergo first-order phase
transitions. Firstly, therefore, the properties expected for various 'normal' phase
diagrams will be discussed, phases of the diagrams being referred to as liquid and
solid. Discussions of such phase diagrams appear in many textbooks of metal science
(see, for example, refs. 3,4). Secondly the applications of such diagrams to lipid
mixtures will be considered.
IIA. Simple eutectic system
The simplest type of binary system is one exhibiting no solution in the solid
state, so that the only phases are a liquid, which approximates to an ideal solution,
and the solid crystalline components. A temperature versus composition diagram of
this type is illustrated in Fig. 1 : the pressure is maintained constant, and excess water
is assumed present, so that the water phase can be neglected. Since pressure is
assumed constant and the water phase is in excess, the phase rule reduces to
P + F = C + 1. Three phases are involved: liquid, pure solid A and pure solid B.
All three phases will be in equilibrium at a point, the eutectic, of fixed temperature
TE and fixed composition mE. In the phase diagram, there is a region where the liquid
phase alone is stable (marked L). In the region below TE, a mechanical mixture of
the two pure solid phases A and B is stable, the relative proportions of the two
phases being determined simply by the bulk composition of the system. In the twophase areas labelled L A and L + B, pure solid A and pure solid B, respectively,
are in equilibrium with a liquid of variable composition, determined by the temperature of the system. In the two-phase regions, the relative proportions of solid and
liquid phases in equilibrium at any given temperature are given by the lever rule.
Suppose the composition of the sample is given by Xx, and the temperature is Tx.
The compositions of the solid and liquid phases in equilibrium at this temperature

287

rA

.~l.y t. ~

"ra

T
L*B

rE

E
I

A*B

I
I
I
!

lx

xI

xe

B
xB

Fig. 1. The phase diagram for a simple eutectic system of two components.

are given by drawing a horizontal line from the curve TA-E, through Tx until it
intersects another line on the phase diagram, in this case, the temperature axis. Such
a line is known as a tie-line, and its points of intersection with the lines of the phase
diagram give the compositions of the co-existing phases. Thus at the temperature
Tx, the co-existing phases are pure A and liquid of composition Xe. The proportion
of sample in the two phases can then be calculated as follows. If n~ is the sum of the
number of molecules of components A and B in the liquid phase and n~ is the number
of molecules of component A in the solid phase, then for a mixture of mole fraction
Xx at temperature Tx,
Xx(nt +

Ils) = x j nl +

1 " n~

(1)

Thus
Il~

Xl --

XX

Ill

XX --

xy
-

xz

(2)

The proportions of the two phases present are thus inversely proportional to the
distances between their respective compositions and the composition of the total
system along the appropriate isothermal line drawn on the diagram.
The shapes of the curves giving the composition of the liquid in equilibrium
with pure solid A or pure solid B can be calculated if the liquid is assumed to be a
perfect solution. This is important since it allows the study of any experimentally
determined departures from ideality.
Consider a binary mixture containing a mole fraction xA of component A, in
equilibrium with pure solid A at a temperature T K. The chemical potential of
component A in the liquid phase is given by
/~A (L) =

[A (L) - ~ R T l n x A

(3)

288
where #A tE) is a constant, the Gibbs free energy of one mole of pure component A
in the liquid phase at the temperature T. At equilibrium, the chemical potentials of
component A are equal in the liquid and solid phases. Thus
(4)

~A (S) ~ j/~A{L)
= /AA(c) -}- RTInXA

Rearranging,
h'lXA

~A(S) - - ~A(I~)
RT

(5)

It follows that
'61nxA] = --HA(S)'+ HA(L>
,ZST }

(6)

RT 2
( A HA)T
RT 2

where
(A HA)~ : HA(L)- - HA(s)

(7)

and is the heat of melting of pure component A at the temperature T, and Ha (I-)
and HA (s) are respectively the molar enthalpies of pure A in the liquid and solid
phases. Integration between the melting point of pure A (Ta) and the temperature T
gives
InxA = [T (AHA)TdT
TA

(8)

RT 2

The latent heat of melting of pure A at the temperature T is related to that at TA by


T

(AHa)T : (,1H~,)TA I- ~TA z]Cp d T

(9)

where d C, is the difference in specific heat between liquid and solid A. Thus
(AHA)I = (AHA)TA+ A C p ( T - - TA)

(10)

If it is assumed that A C o = 0, which is often reasonable over a small temperature


range, then
lnxA = j,T (AHA)TA
TA

(11)

RT 2

and so
|nxA

- - (AHA)TA ( 1

] )

, T - - T'A

(12)

By analogy, the equation for the curve of the liquid in equilibrium with pure solid B is

289
a
I

T2 . . . .
. . . .

I--

'1"

'E

I \

T3

z~'-

, . . . . . .

xa

I
I

Xy

xE

XB

Fig. 2. Melting and crystallization in a simple eutectic system.

ln(I__XA) --(AI'Ia)T.
(1 - - ~ 1)
R

(13)

where (AHB)v. is the heat of melting and TB is the melting point of pure B. These
Eqns. 12 and 13 are the ones we require for plotting phase diagrams. To reiterate,
Eqns. 12 and 13 define the solubility curves of the components A and B, and are often
referred to as freezing point curves.
A typical pair of curves obtained in this way is shown in Fig. 2. The curve
TA-E represents the solubility curve of A and the curve TB-E represents the solubility
curve of B. This diagram then represents the course of crystallization and fusion for
the mixture. Consider a liquid whose composition and temperature is given by the
point a on Fig. 2. On cooling the liquid, no change of phase occurs until the point x
is reached at the temperature T1. At this temperature, pure solid A will just begin
to crystallize. Obviously, when this occurs, the liquid phase will become slightly
poorer in A. If the temperature is further lowered to a temperature T2 then more
solid A will separate, and so the proportion of A in the liquid phase will decrease,
that is, the composition of the liquid phase moves along the line xy. On further
cooling down to TE, pure solid A continues to separate out and the composition of
the liquid varies along the line x-E. Similarly, if the experiment is started with a
liquid of composition b, cooling has no effect until the temperature T3 is reached,
when pure solid B will start to separate. As cooling is continued, more pure solid B
separates out, and the liquid phase becomes poorer in B moving along the line z-E.
At the point E, a new phenomenon occurs. At the temperature TE, whether we start
with a liquid of composition corresponding to point a or point b, the conditions are
correct for both the components A and B to coexist with the liquid phase. Thus
when we start at the point b, only solid B separates out until the temperature TE is
reached. When this temperature is reached, pure solid A also starts to separate. At

290
this temperature, pure solids A and B co-exist with a liquid of composition x~. Any
attempt at further cooling will lead to complete crystallization: the temperature will
remain constant at TE while the liquid freezes with simultaneous formation of solid A
and solid B. The temperature at which this simultaneous freezing occurs is called the
eutectic temperature, and the composition of the liquid at this temperature is called
the eutectic composition. Further cooling produces no further change in the system.
The solid product is considered thermodynamically to be a mechanical mixture of
pure solid A and B, although it might microscopically consist of an intergrowth of
crystalline A and crystalline B at the eutectic composition.
On heating a mixture of solid A and solid B in the proportions corresponding
to the point c in Fig. 2 there will be no change until the eutectic temperature TE is
reached, at which point the first trace of liquid will appear. The temperature of the
system will remain constant at TE until all of the grains of solid B and sufficient of the
grains of solid A have melted to give a liquid of eutectic composition XE. The temperature of the system is then free to rise as more of the solid phase A melts to give a
liquid whose composition moves along the line E-w. Finally at the temperature T4,
the last trace of A melts to give a liquid of the same composition as the total system.
The two curves from the melting points of the pure components A and B to the
eutectic temperature TE are known jointly as the liquidus curve. It is below the liquidus curve that the solid phase begins to appear. The horizontal line through the
eutectic point is known as the solidus curve and correspondingly represents the
temperature limit above which the liquid phase starts to appear.
Consideration of Eqn. 12 together with some simple physical considerations
considerably limits the shape of liquidus curves that can be expected. For this
purpose it should be noted that Eqn. 12 has a general form of the type
x ~

(14)

A e -ely

For such an equation, plots of x versus 1/y will always have a simple exponential
shape. However, in plots of x versus y, there may or may not be points of inflection,
depending on the signs of c and y. If both c and y are positive, then a point c f
inflection occurs at y = c/2, and if both are negative, then a point of inflection occurs
at y ~ --c/2. If however, y and c have opposite signs, then no points of inflection
occur. For the case of the solubility curve given by Eqn. 12, clearly both c and y are
positive, so that a point of inflection will be expected at least in principle. Whether
or not the inflection point is physically observable depends on whether the corresponding temperature appears above or below the melting point of the pure
component TA. The value of the inflection point temperature TI is given by
(15)

T, = (,,~HA)Ta/2R

and the mole fraction of A in the liquid phase at this temperature is, from Eqn. 12,
InXA - -

("~HA)TA

RTA

(16)

291

t1
t
t2

S
I
I

Xp

XR
XB

X1

Fig. 3. The phase diagram for a binary system with complete miscibility in the liquid and solid
phases.
Clearly the value of XA is limited to the region 1 ~ XA /> O, SO that

(AHO.rA/TA ~ 2R

(17)

If Eqn. 17 is fulfilled, then a point of inflection will be observed experimentally. If,


however,

(Z~/'/'A)TA/TA~ 2R

(18)

then the inflection point will not be observed experimentally. For lipid systems the
inequality 18 will always hold, so that only normal non-inflected liquidus curves
would be expected, that is, curves that are concave with respect to the composition
axis.

liB. Systems displaying complete miscibility in the solid state


The next simplest binary system is that where the two components of the
mixture exhibit complete miscibility in both liquid and solid phases. For this to be
likely, the two pure components would be expected to be chemically similar and to
adopt identical structures in the solid state. This requirement follows from the fact
that as the composition of the mixture is varied, molecules of component B have to
substitute directly for the molecules of component A in the solid, as the composition
varies from a mole fraction of 0 of B to 1. This is only likely if the structures adopted
by A and B in the solid are identical.
As with the previous case, the simplest model for this system is to assume that
both liquid and solid phases are perfect solutions. An analytical expression that can
be used to calculate the phase diagram is easily derived. The expected phase diagram
will generally be of the shape shown in Fig. 3.
At some intermediate temperature T, let XA~s~ and XA~L) be the mole fractions

292
of A in the solid and liquid phases, respectively. The condition for equilibrium at
temperature T is
/[~A($) =

(19)

/~A (L)

and, simultaneously,
#B ~s> = #B <L)

(20)

The chemical potential of A in the liquid phase is given by Eqn. 3, but now a similar
equation has to be written for the solid phase:
RT|nXA tL)

(21)

#A ~s~ = #A Cs) + RTInxA ~s~

(22)

/~A tL) =

]AA(L) +

Therefore, at equilibrium,
In xA~L) -- - - /~A~L) /~A~S)

xA (s~

(23)

RT

And following an argument analogous to that leading to Eqn. 12 we obtain


lnXA<L' _

--(AHA)TA

XA ts)

(1

--

'

(24)

TA'

It is convenient to write this as


XAL)
- - ( A H A ) T A ( 1 1 ')1 = e-A
XA ~s~ -- exp
R
T - - TA,

(25)

Similarly for component B,


xa tL,

(AHB)TB(1

exp ~

1 )]

: e-B

(26)

These equations contain four unknowns, XA~L), XA~s), Xa<L) and xB<s) and to solve them
we need to include the following two relationships
XA <s) :

1 - - XB~s)

(27)

XA~L~ = 1 - - xa (L~

(28)

Substituting for XA<L) from Eqn. 27 into Eqn. 25,


XA (L) :

(1 - - XB (S)) e - A

(29)

and from Eqn. 26


x B ~s) =

X B ~L)/e - B

Therefore,
'XB( L )

(30)

293
and using Eqn. 28
x~(L' = i 1 -- (1_--e_BXA(L')}Ie-A

(32)

Solving for xA~L) gives


xA~L, _ e-A(e-~ -- 1)
e--a __ e--A

(33)

and so using Eqn. 28


x(L )

e-a (e--A__I)
e--A __ e-S

(34)

The mole fractions in the solid phase follow from Eqns. 25 and 26
XA(s)

e -B

--

e - a _ _ e -A

e -A -- 1

xa ts) = e -A __ e_l~

(35)
(36)

These equations can then be solved to give theoretical phase diagrams. The same
equations have been derived by Seltz [5] using a graphical procedure. From Eqns. 25
and 26 it is clear that in the temperature range TB ~ T ~< Ta (and A H > 0), the
ratio XA(L)/XA (S) is less than one and the ratio XatL)/XB is) is greater than one. Thus the
liquid phase is always richer in the lower melting component than the solid phase,
and, conversely, the solid phase is richer in the higher melting component. Within
these limits a number of possible shapes of phase diagram may be obtained, and
these will be described below. First, however, it will be convenient to describe the
properties of the generalized phase diagram shown in Fig. 3.
On cooling a mixture of composition x initially in the liquid phase, there is no
change until a temperature tl is reached, when an infinitesimal amount of solid
appears, with a composition xl. The solid formed therefore contains a larger proportion of the higher melting component B than the original liquid. On further
cooling, there is an interaction between the already crystallized solid and the liquid,
so that more solid is produced, the solid containing a gradually increasing proportion
of the lower melting component A. This can only happen by a preferential absorption of A from the liquid and the diffusion of A into the solid formed previously
at higher temperature. Correspondingly, the amount of liquid is reduced, whilst it
also becomes enriched in the component A. The relative proportion of solid and
liquid present at a temperature T is given by the lever rule
n,~q.fd

nnquld + nsolid

g -- Q
R-- P

(37)

where nna,,ld and nsotid are respectively the total numbers of molecules in the liquid
and solid phases. Finally, the temperature tz is reached at which the amount of liquid

294

70~ A
60
504030
20

0"5

70701 B

60

60-

50

50-

40-

40-

30-

30

20-

o'-5

20x

j-

0"5

Fig. 4. Calculated phase diagrams for a binary system with complete miscibility in the liquid and
solid phases, showing the effects of variation in enthalpies of melting. The melting point of component A is 23.7C and of B, 63C. Enthalpies of melting (kcal tool -1) of components A and B,
respectively: A 3, 3; B 3, 11 ; C I1, 3. Temperatures in C.

remaining becomes infinitesimally small, and at temperatures of t 2 and below the


solid contains crystalline grains of composition x.
Even under these restricted conditions a variety of shapes of phase diagram
can be generated, depending on the transition temperatures and enthalpies of the end
members. As well as the traditional 'lens'-shaped phase diagram (Fig. 3), diagrams
are possible with both the solidus and liquidus curves being either convex or concave
with respect to the composition axis, and inflected solidus curves can be generated.
Fig. 4 shows a set of diagrams generated from Eqns. 33 and 35, assuming melting
points of 23.7 and 63C (the transition temperatures of dimyristoyl phosphatidylcholine and dipalmitoyl phosphatidylethanolamine, respectively), and varying the
transition enthalpies. For diagrams such as these to be established experimentally

295

LI
T2

........

x~3

T3'

..............

L4

Tn

.\

x
Fig. 5. The effect of fast cooling on the phase diagram.

it is necessary that the changes in temperature take place very slowly, because a
relatively long time is required at each decrement of temperature to readjust the
compositions of the two phases. This readjustment requires the diffusion of A and B
in both phases, together with the exchange of molecules across the liquid-solid interface. Since diffusion proceeds slowly, especially in the solid state, true equilibrium
is difficult to maintain during cooling.
Reference to Fig. 5 will show the expected result for too rapid cooling. For a
mixture of composition x, freezing will start at temperature 7"1 with the formation of
crystals of composition xl. By 7'2, the liquid composition will have shifted to L2
(because the solid that has crystallised is relatively enriched in the higher melting
component). The solid formed at temperature 7"2 will have the composition x2.
However, if the rate of cooling is too fast, then the solid will not have time to change
its composition from xl to x2, and thus will have some intermediate, average composition represented by x~'. As the temperature drops further to Ts, the average
solid composition x3' will depart more and more from the equilibrium composition
x3. Under ideal conditions, freezing would be completed at temperature Ts. However, under the non-equilibrium conditions, the average solid composition has not
reached the bulk composition x at this temperature, and some liquid must remain.
Freezing therefore continues until the lower temperature Tn is reached.
This process also produces a so-called cored structure. The first particles of
solid produced will have a composition x~. Upon these initial nuclear crystallites
successive layers of the solid phase are deposited, each layer being a little richer than
its predecessor in the low-melting component B. The actual importance of these
effects in lipid bilayers will, of course, depend on the rate of diffusion of lipids in the
solid state. The diffusion coefficient for dipalmitoyl phosphatidylcholine in the gel

296
state has been estimated to be approx. 10 -9 cm 2. s -1 [6]. This value was obtained from
measurements of 31p N M R line widths in sonicated lipid vesicles, and must be taken
very much as an upper limit to the real rate. If the real diffusion coefficient were
l 0 - 9 c m 2 s - 1 , then the root-mean-square displacement of a lipid molecule in the
gel phase would be 2 # m in 10 s. Since the width of a lipid domain appears to be of
the order of a few #m (see Part I, ref. 62), at this rate of diffusion, equilibrium would
be easily maintained. If, however, the rate of diffusion were an order of magnitude
slower, then effects of coring could become important.

IIC. Regular solutions


So far we have only been concerned with ideal solutions, and we have seen that
only relatively simple binary phase diagrams can be generated under these conditions.
However, in general we expect solutions to be non-ideal, particularly so in the solid
state. To be completely miscible in all proportions in the solid state, the structures of
the pure components must be virtually identical. On the other hand, complete immiscibility in the solid state is also unlikely over the complete composition range,
since miscibility would be expected for at least very dilute solid solutions. To allow
for these expectations it is necessary to accept that most solutions are not completely
ideal. For an ideal solution as we have seen (Eqn. 3), the chemical potential of one
component in the solution is related to its mole fraction in the solution
(38)

,UA = # A + R T l n x .

Similarly for the second component


(39)

#a = #B + RTInxR

=/~

+ RTIn(1 - - x . )

Thus the molar free energy of the mixture is


G = XA#. + (l - - XA)I~B + R T ( x A I n x n

+ (l - - XA)ln(l - - XA))

(40)

Since the free energy of a simple mechanical mixture of XA toO1 of A and 1 - - xa tool
of B is x,/~a + (1 - - XA)/~a, the free energy of mixing is
/JGm :

R T ( x A I r l X A ~- (l - - XA)In(I - - Xa))

(41)

Differentiating with respect to T gives the entropy change of mixing,


dSm

= R(xAInxA

-}- (1 --XA)In(1 - - X A ) )

(42)

and the enthalpy of mixing is zero,


dH,~ = 0

(43)

A graph of AGm/RT as a function of x is shown in Fig. 6. For a non-ideal solution,


the chemical potential of A in the mixture can be written in a formal way as
I~A = I~A -~" R T I r l X A j A

(44)

297
0"

~m

RT
-0"5

0 ....
F i g . 6.

0!5 . . . .

Variation

of

AGm/RT w i t h

mole fraction x for an ideal mixture.

wherejA is the activity coefficient of A, and takes account of deviations from ideality.
We can then define the excess chemical potential for A as
/*Ae =

RTJnjA =

/tA - - ~ A - -

RTlnxA

(45)

Similarly, we have an excess free energy for the mixture of


G e = XA /ZA" + X . /ZBe

(46)

The trick is to find manageable expressions for these excess quantities. The simplest
model is to assume that the entropy of the solution approaches the ideal behaviour
of Eqn. 42, but that the enthalpy of mixing is no longer zero, but is given by
AHm :

~o XAXB

(47)

where the constant ~o characterizes the solution behaviour, and can be interpreted
as relating to the difference in energy of A-A, A-B and B-B pair interactions:
po = Z ( 2 U A a -

UAA-

URa)

(48)

where Z is the first coordination number and UAB is the molar energy of an A-B pair
interaction, and so on. It then follows that
/z~,~ :

po xB 2

(49)

~Be = Po XA2

(50)

and

These equations can be applied both to the case where the components are
completely immiscible in the solid phase, and to the case where both the liquid and
solid phases are regular solutions. The first of these is the simplest. In a binary
mixture when the solids are immiscible but mixing in the liquid state is that for a
regular solution, we can write for the liquid state
~/~A = EgA(L)O -~

and

RTInxA + po(l --xA) 2

(51)

298

.-j

330

a,'"b / /
,"//

310-

,:"//

0"5
XDPPE

Fig. 7. Phase diagram for a mixture of dimyristoyl phosphatidylcholine and dipalmitoyl phosphatidylethanolamine, assuming complete immiscibility in the solid state, and curve c, ideal mixing
in the liquid state. Curves a and b show the effects of non-ideal mixing in the liquid state. For a,
po = 1.3 kcal tool -1 and for b po = 1.0 kcal. mol -~ for the solution of phospbahdylcholine in
phosphatidylethanolamine, Temperature m K. The phosphatidylcholine-rich branches of the liquidus
curves a, b and c overlap.

/~B = /z~(L> + RTIn(1 - - x A ) + po(XA)2

(52)

When the solids are immiscible the condition for equilibrium between liquid and
solid becomes, for component A (see derivation of Eqn. 4),
/.~A (S) :

/~A (L)

~-

RTInxA po(1 - - x A ) 2

(53)

Following the derivation of Eqn. 12, we obtain

lnxA+%(I--xA)2
Rr

--(AHA)TA(1
R

l )

7r-- ~

(54)

If mixing had been ideal in the liquid phase, then this same mole fraction of A would
have occurred at some other temperature T la~, given from Eqn. 12 as
lnxA

-- --(AHA)rA ( 1
R
\,7qd~7~t

1 '
TA/.

(55)

Subtracting Eqn. 55 from Eqn. 54 gives


0o(I--XA) z
..... -RT

--(LIHA).rA (1
R
~,

1 )
T-7~t

(56)

This can be rearranged to give


po(1 --

XA) 2

(AHA)ra

--

Tiaeai

(57)

This very simple equation therefore allows phase diagrams to be calculated,

299
340T

330
T

310

300

'

'

\\

/
/

\
|

290

iI

/
/

0"5

2800

0:5

XDpPE

XDPPE

Fig. 8. Phase d i a g r a m s for m i x t u r e s o f dimyristoyl p h o s p h a t i d y l c h o l i n e a n d dipalmitoyl p h o s p h a t i d y l e t h a n o l a m i n e , calculated a s s u m i n g non-ideal m i x i n g in b o t h liquid a n d solid phases.
9o values in kcal m o l - L A. solid line: p(L) = 0.5, 9o(s~ = 0; b r o k e n line: 9otL) = 1.0, 9ots) = 0;
d o t t e d line: 9o(L) ~= - - 1 . 0 , po~S~ = 0. B. solid line: 9o~L) = - - 0 . 8 , 9o ~s~ = + 0 . 8 ; b r o k e n line: 9o(L) =
- - 2 . 0 , 9o (s) = + 1 . 0 ; dotted line: po(L) = + 0 . 8 , 9o(s) = - - 0 . 8 . T e m p e r a t u r e s in K.

taking into account deviations from ideality in the liquid phase. Fig. 7 shows some
diagrams calculated in this way, for various values of the parameter po.
The corresponding calculations for binary mixtures where both solid and liquid
phases are regular solutions is considerably more complicated. In this case, the
chemical potentials in the solid phase are given by
/~A~s~ = FA<s~ + R T I n x A ~s) + po~s) (1 - - XA(S)) 2

(58)

and
/t~ (s) = /~a ~s~ + R T I n ( I - - x A (s~) + po~S)(XA(S))2

(59)

Similarly for the liquid phase


/~A( L ) :

/~A 0(L>

+ R T I n x A (L) + po(L) (1 - - xA(L)) 2

(60)

and
,tta (L) ~ ~n (L) + RTIn(1 --.X'A (L)) + DotE) (XA(L)) 2

(61)

The condition for the liquid and solid solutions to be in equilibrium is


/ ~ A L) ~

~/~AI'S)

a n d ~a (L) = # a fS)

(62)

Therefore,
XA (L)

In ~

Do ( L ) ( 1 - -

XA(L)) 2 --

RT

[~o (S) ( l

--

XA(S)) 2

--

t/~A ( L ) - -

RT

~A ($)

(63)

300

aG m
0

Fig. 9. Variation of free energy of mixing AG,, with mole fraction x for a non-ideal mixture, when
the enthalpy of mixing is negative.
As before (Eqn. 12), this leads to
In ~s3XA~L~+ ~~L~(I -- xA(LO2-- ~~S~( 1 - - XA~S~)Z-- --(AHA)TA
R
( __ ~1)

(64)

and

In 1--___
X-a~S~I--XA(L)+

po~L, (xAL))2RTpo~S)(xAtS))Z _

--(AHB)rBR ('1,7.__Tt~,l)

(65)

Because of their transcendental nature, Eqns. 64 and 65 require numerical methods


for their solution. This is simply done by computer. For a chosen value of T, we
start by assuming a value of XA(L) and then we solve Eqn. 64 by a simple iteration
technique to obtain a value of xn (s). This value is then substituted into Eqn. 65,
which is then solved by an iteration technique to obtain a new value of XA(e) which is
then substituted into Eqn. 64 and so on until the values converge. Solution by the
Newton-Raphson technique is also possible. Results that can be obtained in this
way are illustrated in Fig. 8 for mixtures of dimyristoyl phosphatidylcholine and dipalmitoyl phosphatidylethanolamine. Clearly variation of po~C~and po(S~can produce
radical changes in the shape of the phase diagram. A more extended set of results
is shown in ref. 7. But before proceeding further, it will probably be helpful to look
in a bit more detail at the thermodynamics of such mixtures.
All mixtures can be characterized by their free energy composition diagrams.
That for an ideal mixture is shown in Fig. 6. For non-ideal mixtures, more complex
relationships can be obtained. If the enthalpy of mixing AHm for the non-ideal
solution is negative, then AG,, will be negative at all compositions, since the quantity
--TAS,~ is usually negative. These terms are plotted in Fig. 9. If, however, AHm is
positive, then the shape of the AGm versus xa plot will tend to be more complex,
especially at low temperatures. At high temperatures, the quantity --TASm in the
expression

AGm

AHm -- TAS~

(66)

will be more important than at low temperatures, so that a curve of AGm versus Xn
will be concave downwards over the entire composition range (as in Fig. 9 above).

301

01

k
A

z
XB

Fig. 10. Free energy of mixing vs. composition, when the enthalpy of mixing is positive, allowing
for immiscibility.
However, at lower temperatures, the AHm term becomes more important relative to
the --TASm term, and it becomes possible for the curve of AGm versus xA to be
concave upwards over a range of compositions, giving a curve of the shape shown in
Fig. 10. The central portion of this curve has interesting properties. Thus a mixture
of composition x would not exist as a single homogeneous solution represented by
the point A. The free energy of the mixture would, for example, be lower if the
mixture of this composition were instead to consist of a mixture of two separate
solutions, represented by the points B and C: the total free energy of the mixture
would then be given by the point D. In turn, it is clear that this does not represent
the lowest free energy for the system. The lowest free energy configuration for the
system is that represented by the point G, lying on the common tangent EF. This
most stable configuration consists of two solutions, one of composition y and the other
of composition z. The relative proportions of these two solutions can be obtained
from the lever rule. The fraction of solution y is given by
(z - - x)
(z - - y)
and that of solution z is given by
(x --y)
(z - - y )
Thus we see from Fig. 10 that for all compositions from xB --~ 0 to xB = y, the stable
state of the system is a single solution, as it also is from xB ----z to xB ---- 1. In the range
of compositions from x8 =- y to x~ = z on the other hand, the system consists of two
solutions of composition y and z. This behaviour is referred to as a miscibility gap.
The limiting curve to the two phase region of the miscibility gap is called the solvus.
An illustration of a fairly simple phase diagram of this type is shown in Fig. 11.
This diagram can be understood in terms of the corresponding free energy vs. composition diagrams. At a temperature T1, the free energy vs. composition for liquid

302
r~

r3

XB

Fig. II. A phase diagram showing a miscibility gap.

x
Fig. 12. Free energy of solid and liquid phases corresponding to temperature T1 of Fig. 11.
and solid solutions will be as shown in Fig. 12. Clearly, the stable state is that of a
single liquid solution over the complete range of compositions: this will be true for
any temperature above the melting point of the highest melting component (A in
Fig. 11). At temperature 7"2, the free energy versus composition curves will be as
shown in Fig. 13. In the composition range xB ~ y, the solid solution has the lowest
free energy, and will be the only phase present. Similarly, in the region xa ~ z, only
the liquid solution will exist. In the intermediate region y ~ xa ~ z two phases will
co-exist, the liquid and solid solutions.
At temperature/"3, the free energy versus composition diagram will be similar
to that in Fig. 10 with the co-existence of two solid solutions. With increasing temperature, the compositions of these two solid solutions move closer together and
eventually become equal at the critical temperature T~. The necessary condition for
the formation of two solid solutions and an expression for Tc can be easily derived if

303

i
i

XB

Fig. 13. Free energies of solid and liquid phases corresponding to temperature 7"2 of Fig. 11.

O~G

100

200
0.5

Fig. 14. The development of a miscibility gap in a binary regular solution.


the solutions are assumed to be strictly regular. The molar free energy of mixing of
such a mixture is the sum of that of an ideal solution (Eqn. 41) plus the excess free
energy due to the non-ideality (given by Eqn. 47)
AGm : RT{XAInXA + (1 - - XA)ln(1 - - XA)} -[- poXA(1 - - XA)

(67)

This expression is plotted in Fig. 14 for two different values of 9o. F r o m the symmetry
of this equation, it is clear that the critical point will occur when XA : XB : . The
condition for a point of inflection is given by solving the equation
62AG,.

RT

"~xA2

XA(I --- XA

2po = 0

(68)

Thus the critical temperature is given by


Tc = po/ZR

(69)

It follows that formation of a miscibility gap requires positive values of Po. There
will be no points of inflection if 2RT/po > 1 and two points of inflection symmetrically positioned about XA = when 2RT/po < 1.
Phase diagrams can also be obtained in which the solvus intersects the solidus.
A particularly important case is that where the system exhibits a eutectic, as in Fig. 15.
Previously (p. 286), we have considered the case for a eutectic, where the two

304

XB

Fig. 15. The phase diagram for a binary eutectic mixture with partial miscibility in the solid state.
The metastable extension of the solvus curve above the solidus is represented by the broken line.
The letters a, E and b denote compositions.

compounds are completely immiscible in the solid phase. In the limit, of course, this
is impossible, because complete immiscibility will never occur: all known compounds
are impure to a greater or lesser extent, and one member of the pair will always
dissolve some of the other. This will result in a phase diagram of the type shown in
Fig. 15, where the end members exhibit considerable solid solubility in each other.
In the same way that the curve TA-E represents the variation in liquid phase composition with temperature, the curve TA-a represents the variation of the composition of the solid phase of structure A with temperature. In the spaces bounded
by the lines TA-a-A and TB-b-B, single-phase solid solutions exist, the former having
the structure of solid A and the latter the structure of solid B. In the two-phase
region TA-a-E, a solid solution with the structure of A is in equilibrium with liquid,
and similarly in the region TB-b-E, a solid solution with the structure of B is in
equilibrium with liquid. The two branches of the liquidus curve fall smoothly from
the melting points of the pure components to the eutectic point E, where liquid of
composition xE is in equilibrium with solid solutions of compositions x, and xb.
Below the eutectic temperature TE, the two branches of the solvus fall off from points
a and b towards the pure components. Above TE, the solvus curve is metastable, and
is shown as a broken line.
The last type of diagram we need to consider is that in which the liquid solution
exhibits a miscibility gap. An example of a binary eutectic system showing liquid
immiscibility is shown in Fig. 16.

liD. Limiting cases of the eutectic and peritectic


A fairly common phenomenon in binary mixtures is that in which the eutectic
composition occurs very close to one of the components (Fig. 17). The composition

305

ra

TE~A-~L

~L
A*B

XB

Fig. 16. The phase diagram for a binary eutectic mixture with partial miscibility in the liquid state.

TTA
f

t..s~

.L.S2
Ta

S1.S2

S-~
xB

Fig. 17. The limiting case of a eutectic phase diagram.

range of the terminal solid solution may be so narrow that it cannot be represented
on the phase diagram because its width is less than that of the boundary line used to
enclose it. It should not be assumed, however, that this means that there is no range
of solid solubility, for it is inconceivable that any two compounds could be totally
insoluble in each other at temperatures above absolute zero.
l i E . Sub-regular solutions

The strictly regular solution model is adequate to account for the existence of
both solid and liquid immiscibility, but must always produce a symmetrical solvus.
To explain the appearance of an asymmetry of the solvus it is necessary to consider
more complex deviations from ideality. Rapidly, however, this can degenerate into
mere curve fitting, with relatively little physical reality attributable to the newly
introduced terms. The simplest expedient is to use a two-term expression for the
enthalpy of mixing, so that instead of Eqn. 47
AH~ = XA Xatpo + polXn)

(70)

If this represents a satisfactory approximation to the solution behaviour then the


solution is termed 'sub-regular'.

306
451

,ol

35i

+
o

30

"+

t
+

25

+
/

0"5

XDPPC

Fig. 18. Phase diagrams for mixtures of dimyristoyl phosphatidylcholine and dipalmitoyl phosphatidylcholine, obtained by: 0 , electron spin resonance [8]; , differential scanning calorimetry
[9]; O, fluorescence [10]. Temperatures in C.

III. RESULTS OBTAINED WITH PHOSPHOLIPID MIXTURES

A number of lipid mixtures have now been studied, using differential scanning
calorimetry, electron spin resonance and fluorescence techniques. Although there are
small differences in the shapes of the phase diagrams determined by these various
techniques, the overall agreement is generally good. Fig. 18 compares the results
obtained for mixtures of dimyristoyl phosphatidylcholine and dipalmitoyl phosphatidylcholine. The transition temperatures determined by probe techniques are
consistently lower than those measured calorimetrically, but the shapes of the phase
diagrams are closely similar. Fig. 19 compares the results obtained for mixtures of
dipalmitoyl phosphatidylcholine and dipalmitoyl phosphatidylethanolamine. It is
clear that in this case the agreement between the probe and calorimetric techniques is
relatively poor. One major problem is associated with the width of the phase
transition. As discussed elsewhere [2], the calculation of phase diagrams assumes
that the phase transitions for the individual components of the mixtures are sharp,
occurring over a very narrow temperature range. For lipids, this is not so. The
trouble then is that there is no rigorous way to correct for the finite width of the
phase transition. One can try to correct for the finite width of the phase transition
by assuming that some particular fraction of the separation in temperature for the
'fluidus' and 'solidus' lines in the experimental diagrams is due to the finite width of
the transition for the component lipids. As a very rough working approximation
this contribution can be put equal to the weighted average of the contributions from
the component lipids, so that subtraction of this contribution will give temperatures
which might be more truly plotted as a phase diagram. There is no real theoretical

307
65
+
+
+

oe

60
o

55

+oe

50

+
+
o

45

+
+

350

0"5

X DPPE

Fig. 19. Phase diagrams for mixtures of dipalmitoyl phosphatidylcholine and dipalmitoyl phosphatidylethanolamine, obtained by." O, electron spin resonance [8]; , differential scanning calorimetry [11]; 0 , fluorescence [10]. Temperatures in C.

justification for this approach. Some rough correction of this type is unfortunately
necessary before the experimental phase diagrams can be fitted to any theoretical
model.
It is now clear that for two lipids with the same head group and differing only
slightly in the length of their fatty acyl chains, mixing is close to ideal. This has been
demonstrated for mixtures of dimyristoyl phosphatidylcholine and dipalmitoyl
phosphatidylcholine [8-10]. Fig. 20 shows the phase diagram obtained calorimetrically and that calculated from Eqns. 33 and 35 assuming ideal mixing in both
the solid and liquid phases. Although agreement is quite good, it can be improved
by assuming non-ideal mixing in both phases, as shown in Fig. 20 for the curves
calculated from Eqns. 64 and 65 with po (L) ~ 500 cal" mo1-1 and po(s) ---- 900 cai"
tool -1. Similar mixing has been observed for dielaidoyl phosphatidylcholine/dipalmitoyl phosphatidylcholine [12] and dipalmitoyl phosphatidylcholine/distearoyl
phosphatidylcholine mixtures [8]. These conclusions have been confirmed by freezefracture electron microscopy, since it is possible to observe distinct regions of solidand liquid-phase lipid, in the proportions expected from the phase diagram [13,14].
When the lipids differ in head group, mixing becomes less ideal. Fig. 21
contrasts the experimental phase diagram for mixtures of dimyristoyl phosphatidylcholine and dimyristoyl phosphatidylethanolamine with that calculated for ideal
mixing. Clearly, mixing is now far from ideal. A good fit between theory and experiment can be obtained assuming that both liquid and gel states form regular
solutions. Fig. 22 and 23 show corresponding calculations for mixtures ofdipalmitoyl
phosphatidylethanolamine and dipalmitoyl phosphatidylcholine and dimyristoyl

308
45
T

40

///

//J

30-

25

20

o .5

x DPPC

Fig. 20. Phase diagram for mixtures of dimyristoyl phosphatidylcholine and dipalmitoyl phosphatidylcholine. Q, corrected and uncorrected points determined calorimetrically (from ref. 9);
, transition curves calculated for non-ideal mixing, (parameters in Table I); . . . . , transition
curves calculated for ideal mixing. Temperatures in C.

~0-

40

30

F
20

o.~s

. . . .

X DMPE

Fig. 21. Phase diagram for mixtures of dimyristoyl phosphatidylcholine and dimyristoyl phosphatidylethanolamine. O, corrected and uncorrected points determined calorimetrically (from
ref. 9); - - - , transition curves calculated for non-ideal mixing (parameters in Table I); . . . . , transition curves calculated for ideal mixing. Temperatures in C.

p h o s p h a t i d y l e t h a n o l a m i n e and distearoyl phosphatidylcholine, respectively. Agreement for the first of these mixtures is n o t very good : theoretical curves which give a
reasonable fit at low mole fractions of e t h a n o l a m i n e give a very p o o r fit at high mole
fractions and, conversely, theoretical curves which give a good fit at high mole
fractions of e t h a n o l a m i n e give a poor fit at low mole fractions. However, it is possible that the a p p a r e n t m i n i m u m in the experimental solidus curve at a mole fraction
of approx. 0.2 of e t h a n o l a m i n e is artifactual: as shown in Fig. 19 this m i n i m u m is
only found calorimetrically and does n o t appear in the phase diagrams obtained using

309
TABLE I
PARAMETERS OBTAINED FROM AN ANALYSIS OF EXPERIMENTALLY OBTAINED
PHASE DIAGRAMS
Lower melting component
Dimyristoyl
phosphatidylcholine
Dimyristoyl
phosphatidylcholine
Dimyristoyl
phosphatidylethanolamine
Dlpalmitoyl
phosphatidylcholine
Dimyristoyl
phosphatidylcholine
Dioleoyl
phosphatidylcholine

Higher melting component


Dipalmitoyl
phosphatidylcholine
Dimyristoyl
phosphatidylethanolamine
Distearoyl
phosphatidylcholine
Dipalmitoyl
phosphatidylethanolamine*
Distearoyl
phosphatidylcholine
Dipalmitoyl
phosphatidylethanolamine

(kcal mol-l)

po($)
(kcal mol-~)

0.5

0.9

1.0

1.5

1.15

1.6

0.9

1.3

0.7

1.25

[~o(L)

* See text.
probe techniques. Similar phase diagrams showing partial miscibility in the liquid
and solid phases have also been demonstrated for the following lipid mixtures:
dimyristoyl phosphatidylcholine/dielaidoyl phosphatidylcholine [12], dielaidoyl
phosphatidylcholine/distearoyl phosphatidylcholine [12], dimyristoyl phosphatidylethanolamine/dipalmitoyl phosphatidylcholine [11], and, probably, dipalmitoyl
phosphatidylcholine/dioleoyl phosphatidylcholine and distearoyl phosphatidylcholine/dioleoyl phosphatidylcholine [8,15,16].
For some of these mixtures, almost complete immiscibility in the gel phase
seems likely. Fig. 24 compares the experimental phase diagram for mixtures of
dimyristoyl phosphatidylcholine and distearoyl phosphatidylcholine with that
calculated for complete immiscibility in the gel phase. If ideal mixing is assumed in
the liquid-crystalline phase, then the fit to the observed liquidus curve is close, but
this can be improved by assuming slight non-ideality. The mixtures with dioleoyl
phosphatidylcholine also probably show complete immiscibility in the gel phase.
Fig. 25 compares the calculated phase diagram for mixtures of dioleoyl phosphatidylcholine and dipalmitoyl phosphatidylethanolamine with that observed using ESR
[12].
A more complex phase diagram has been deduced for mixtures of dielaidoyl
phosphatidylcholine and dipalmitoyl phosphatidylethanolamine (Fig. 26). The
horizontal solidus line from approx. 10 to 8 0 ~ ethanolamine indicates that the
ethanolamine is soluble in the phosphatidylcholine up to 10~, and that the phosphatidylcholine is soluble in the ethanolamine up to 20 ~. Beyond those ranges there
is solid-phase immiscibility. Thus far, the phase diagram is conventional (see discussion on p. 303). However, it appears that the lipids are not completely miscible in
the liquid state and there is an area of limited liquid phase miscibility in the diagram,

310
65-

60
T

55-

50-

45-

40- \

0'5

X DPPE

Fig. 22. Phase diagram for mixtures of dipalmitoyl phosphatidylethanolamine and dipalmitoyl
phosphatidylcholine. O, corrected and uncorrected points determined calorimetrically (from
ref. 11); - - ,
transition curves calculated for non-ideal mixing, with ~o(L) = 900 c a l ' mol - t and
po(S) _ 1300 cal" mol-~; . . . . , transition curves calculated for non-ideal mixing, with potL) -- - 5 0 0 cal mol -~ and po(S) = 700 cal tool -1 ; - - -, transition curves calculated for ideal mixing.
Temperatures in C.

:1

3 5 1

O-5

. . . .
X DSPC

Fig. 23. Phase diagrams for mixtures of dimyristoyl phosphatidylethanolamine and distearoyl
phosphatidylcholine. O, uncorrected and corrected points determined calorimetrically (from
ref. 11); -----, transition curves calculated for non-ideal mixing (parameters in Table I); . . . . ,
transition curves calculated for ideal mixing. Temperatures in C.

311

:3201
1

~-o---~

3~~

_~__

2 ~

'~

I-0"5

XDSi~

Fig. 24. The phase diagram for mixtures of dimyristoyl phosphatidylcholine and distearoyl
phosphatidylcholine, assuming complete immiscibility in the solid state.
, assuming ideal
behaviour in the liquid state; . . . . , assuming non-ideal behaviour in the liquid state, with po = 0.7
kcal - tool -1 ; , experimental points determined by ESR (ref. 8). The transition temperatures (K)
are those measures by ESR.

340320"
T 300280

260

240

O.F;

XDpPE

Fig. 25. The phase diagram for mixtures of dtoleoyl phosphatidylcholine and dipalmitoyl phosphatidylethanolamine. - - - , assuming ideal behaviour in the liquid state; . . . . , assuming non-ideal
behaviour in the liquid state, with 9o = 1.25 kcal' tool-L. , experimental points determined by
ESR (ref. 12). The transition temperatures (K) are those measured by ESR.

where two liquids o f different composition are in equilibrium. In principle, diagrams


o f this type might be calculable assuming non-ideal mixing in both the gel and liquid
crystalline phases (see ref. 7).
A n interesting possibility discussed by W u and McConnell [12] is that the
liquid-liquid immiscibility f o u n d for mixtures o f dielaidoyl phosphatidylcholine and
dipalmitoyl phosphatidylethanolamine will lead to an asymmetrical bilayer membrane.
Unfortunately little is yet k n o w n about the trans-bilayer symmetry of these lipid
mixtures, although it is generally assumed and seems likely that if lipids in one
particular region o f one side o f the bilayer are in say the gel phase, then the iipids in
the corresponding region o f the other half o f the bilayer will be as well. However,

312

T~
70

60

i
I

DEPC-DPPE

L1 , L 2 i

50

*C

40

30

LI*-~

2O

1c~---i

%A

SA'B
50
MOLE

100"/.
(DPPE)

Fig. 26. The phase diagram for mixtures of dielaidoyl phosphatidylcholine and dipalmitoyl phosphatidylethanolamine (from ref. 12).
when two immiscible liquid phases are present, then it is possible that one will be
predominantly on one face of the bilayer, and the other on the other face. It is then
likely that such a chemically asymmetrical membrane will be non-planar. Putting it
the other way round, if the membrane is curved (as it will be in a biological cell), and
if the lipids show liquid phase immiscibility, then it is likely that the lipid composition
of the two halves of the bilayer will differ [12]. This concept is rather different from
the asymmetric distribution of lipids found in small sonicated lipid vesicles, which can
be attributed to differences in packing density in the two highly curved surfaces
[17,18].
It is likely that immiscibility for lipids with different head groups is associated
with the organisation of these different head groups. Unfortunately, there is at
present considerable confusion in this area. Originally it was suggested that the head
group of phosphatidylcholines was oriented parallel to the surface of the bilayer in the
gel state but perpendicular to the surface in the liquid-crystalline state. On the other
hand, the phosphatidylethanolamine head group seemed to be oriented parallel to
the surface in both states [2]. In the X-ray crystal structure of a phosphatidylethanolamine, the head group has in fact been found to be perpendicular to the fatty
acyl chains with considerable intermolecular interaction as previously suggested [19].
However, neutron diffraction studies of bilayers of egg lecithin in the liquid-crystalline
state also suggest that the head groups lie parallel to the bilayer surface, in contradiction to earlier models [20]. Some N M R experiments (on the nuclear Overhauser
effect) have also been interpreted in terms of a head group conformation parallel to

313
the surface [21,22], although others (on paramagnetic shifts) suggest a conformation
perpendicular to the surface [23]. Obviously, these disagreements need to be sorted
out before it is possible to usefully discuss changes in conformation caused by mixing
with other lipids. Nevertheless, Yeagle et al. [22] have presented evidence that
suggests that in mixed regions of phosphatidylethanolamine and phosphatidylcholine,
intermolecular hydrogen bonding occurs between the different head groups, which
would certainly have important consequences for the mixing of lipids.
The presence of Ca 2 causes a considerable enhancement of the phase separation of negatively charged lipids, and, since this occurs at molar ratios of Ca 2 to
negatively charged lipid of up to 0.5: 1, it is likely to be due to a 'bridging' action of
Ca 2, as discussed in Part I [62]. Such effects have been observed in mixtures of
phosphatidylcholines with phosphatidic acids [24-26], phosphatidylserine [10,26-29]
and phosphatidylglycerol [30,31]. The Ca2-bound negatively-charged lipid forms
rigid domains embedded in the otherwise liquid-crystalline lipid. Further evidence
for a direct chemical binding of Ca 2 comes from the commonly observed differences
between the effects of Ca 2 and Mg2+: if only shielding effects were important, all
divalent metal ions would, of course, have the same effect at any given concentration.
However, Ohnishi and Ito [27] found that 2 mM CaC12 in the presence of 100 mM
KCI caused a phase separation in mixtures of phosphatidylserine and phosphatidylcholine, that Ba 2+ and Sr 2+ had some effect, although much tess than Ca 2, and that
Mg z seemed to cause no phase separation at all. Papahadjopoulos et al. [29] also
report that Mg 2 has no segregating effect in mixtures of these two lipids. In mixtures
of phosphatidylcholine and phosphatidic acid, phase separation was caused by all the
divalent metal ions, although the effects of Mg 2 were smaller than for the other
ions [25].
Mixtures of phosphatidylcholines and lysophosphatidylcholines have been
studied using differential scanning calorimetry, but no clear picture has yet emerged
[32,33]. With some lipids there appears to be little effect, whereas with others there
is evidence for two different lipid phases below the transition temperature, although
it is not known what these two phases might be.
Studies with more complex lipids will obviously be of interest, but only one has
so far been reported and that is a study of mixtures of phosphatidylcholine and the
glycolipid called globoside (2-acetamide-2-deoxygalactopyranosyl-fl-(l ~ 3)-galactopyranosyl-a(1 -+ 4)-galactopyranosyl-fl-(1 ~ 4)-glucopyranosyl-(l -+ 1)ceramide). It
was shown that whereas the globoside formed a 'tubular' structure in water, when
mixed with phosphatidylcholine up to 50 mole percent, a lammellar phase formed:
with higher proportion of globoside, a tubular structure again formed [34]. Possible
phase separations in this mixture have not yet been studied.
In summary, it is clear that the mixing properties of lipids depend both on the
nature of the fatty acyl chains and on the nature of the head groups. In particular,
it is interesting that unsaturated and saturated lipids appear to mix particularly
badly in the gel state, presumably as a result of poor packing characteristics. McCammon and Deutch [35] have shown in a highly simplified model based on the

314
Bragg-Williams approximation that the phase behaviour of mixtures of dipalmitoyl
phosphatidylcholine and distearoyl phosphatidylcholine can be interpreted in terms
of the interactions between fatty acyl chains. Although for simplicity they had to
assume that the disorder in fatty acyl chains in the liquid-crystalline phase was
largely due to 'kink formation' (which is probably not valid, see ref. 36), the model
does emphasize the importance of steric interactions between fatty acyl chains in
determining the mixing properties of lipids.

IV. MIXTURES WITH STEROIDS


IVA. Lipid~cholesterol mixtures
Despite much work, there is still considerable uncertainty about the nature of
lipid/cholesterol mixtures [37]. It is known that hydrated phosphatidylcholines can
take up cholesterol to a molar ratio of 1 : 1, beyond which pure cholesterol separates
out [38]. Raman spectroscopy of a 1:1 mixture of cholesterol and dipalmitoyl
phosphatidylcholine shows that at temperatures below approx. 20C the lipid fatty
acyl chains are in a largely trans configuration (atlhough less so than in the absence of
cholesterol) and that a transition to a principally random configuration of the palmitic
acid chains occurs over a broad temperature range up to approx. 80C [39]. Importantly, this gradual transition appears to have no associated latent heat of transition [40]. Change in this system therefore seems to be definitely not of the firstorder type. This is hardly surprising, since the presence of 50 mole percent of
cholesterol will prevent a highly co-operative transition for the fatty acyl chains, as
long, that is, as the two remain mixed in the gel phase - - if on forming the gel phase,
separate phospholipid and cholesterol phases were to form, then a normal phase
transition would be observed, associated with the phospholipid phase.
These results are consistent with observations that in the presence of 50 mole
percent of cholesterol, the lipid fatty acyl chains are more disordered below the lipid
phase transition temperature than they are in the absence of cholesterol, but are more
ordered above the temperature of the phase transition [37].
Controversy however, relates to the organisation in mixtures containing less
than an equimolar proportion of cholesterol. Three possibilities have been mooted:
firstly that a 1:1 lipid" cholesterol complex is formed, secondly that a 2:1 lipid"
cholesterol complex is formed, and thirdly that no complex is formed and that the
cholesterol is freely dispersed. On the basis of studies using X-ray diffraction [41]
and differential scanning calorimetry [40] it has been proposed that in mixtures of
dipalmitoyl phosphatidylcholine and cholesterol containing less than 33 mole
percent cholesterol, two phases are present: a pure lecithin phase and a mixed phase
of lecithin and cholesterol at a molar ratio of 2:1. These conclusions have been
criticized by Phillips and Finer [42], who, on the basis of P M R data, suggest that in
mixtures containing less than equimolar amounts of cholesterol, discrete regions of
1 : 1 lipid " cholesterol complex separate out. Similarly, fluorescence experiments are

315

20

0-I

0.2 o 3

0.4 0.5

Xchol
Fig. 27. Phase diagram for the free phospholipid component in mixtures of cholesterol and
dipalmitoyl phosphatidylcholine. Up to a cholesteroh phospholipid molar ratio of 0.2, temperatures
corresponding to both the end of the main transition and the lower end of the pre-transition are
visible (from ref. 43). Temperatures in C.
consistent with a I : 1 complex, and phase diagrams have been presented for the pure
lipid phases presumed to be present in such mixtures [43] (Fig. 27). These diagrams
suggest that there is little miscibility between pure lipid and the lipid" cholesterol
complex in the gel phase, but that in the liquid-crystalline phase, there is miscibility
up to approx. 35 mole percent cholesterol. Interestingly, other studies show discontinuities at approx. 35 mole percent cholesterol, including changes in waterbinding [44] and in the condensing effect of cholesterol on lipid molecular area [37].
The lifetime of the lipid cholesterol complex has not been definitely established, but
Phillips and Finer [42] suggest that the exchange rate between complexed and uncomplexed lecithin is slower than 0.03 s. These results are also broadly in agreement
with the studies of Hu; and Parsons using electron microscopy [45]. In mixtures of
dipalmitoyl phosphatidylcholine and cholesterol at 11 C, an electron diffraction ring
characteristic of crystalline fatty acyl chains was seen until the molar cholesterol ratio
reached 1 : 1. In mixtures containing less than 50 mole percent cholesterol, a ribbonlike structure was seen (Fig. 28), each ribbon being less than 100 nm wide. These
separate domains could correspond to areas of pure lipid and lipid" cholesterol
complex. Hui and Parsons [45] have pointed out the similarity between this ribbon
structure and that of the lowest energy configuration of systems of two immiscible
components, such as eutectic solids [46] or magnetic domains in super conductors [47].
On increasing the temperature beyond the transition temperature, both the diffraction
ring due to the fatty acyl chains and the ribbon pattern disappear.
It is possible that to speak of a 1 : 1 lipid cholesterol complex in these systems
is misleading, since formation of a complex is generally considered to involve some
specific bonding interaction between the two components of the complex. In this
case it might simply reflect some packing constraint of the general type suggested by
Engelman and Rothman [41]. Thus it might be necessary to prevent the formation
of certain cholesterol-cholesterol contacts, which could be done if a 'line' of cholesterol molecules were separated by a line of phospholipid in the gel phase. This
would then give rise to an apparent 1:1 lipid" cholesterol 'complex'. A line of

316

Fig. 28. Dark-field electron micrograph of a hydrated single bilayer of a cholesterol:phospholipid


mixture at a molar ratio of 0.65 at 11 C (from ref. 45).

phospholipid molecules surrounded by lines o f cholesterol molecules on either side


would be unable to take part in a co-operative phase transition as observed. In the
liquid-crystalline state, a phospholipid adjacent to a cholesterol molecule would be
relatively immobilised, and so exclude T E M P O or chlorophyll a, again as observed.
In contrast to the above reports, Shimshick and McConnell [48] have concluded from a study of the partitioning o f T E M P O into lipid cholesterol mixtures,
that no complexes are formed and that a phase diagram can therefore be drawn for
the total lipid in the sample. However, as discussed elsewhere [43], their published
data is, in fact, not inconsistent with the formation of a 1 : 1 complex.
Interactions between cholesterol and phosphatidylcholines show a certain
selectivity. Thus in mixtures o f phosphatidylcholines showing phase separation,
cholesterol interacts preferentially with the lipid with the lowest transition temperature [43,49]. This can be explained simply as a preferential exclusion from lipid in
the gel phase. Mixtures of cholesterol and lysophosphatidylcholine have also been
studied, and a stable lamellar phase o f 1 : 1 molar ratio is formed [50].

317
IVB. Mixtures with other steroids
Although much less is known regarding the interaction between lipids and other
steroids, there appear to be three requirements for a 'strong' lipid-steroid interaction :
a planar nuclear ring system, a cholestane-type sidechain at C-17, and a 3/3-hydroxyl
group [37]. This has been demonstrated, for example, in monolayer studies, where
it has been shown that the molecular areas of many steroid/lecithin mixtures show
deviations from simple additivity (the condensation effect). This effect is particularly
marked for cholesterol, but is much less marked for a non-planar steroid such as
coprostanol, in which steroid rings A and B are cis oriented. A hydrocarbon chain
at C-17 is important, since androstan-3/3-ol shows no condensation with phosphatidylcholines. Again, epicholesterol with a 3a-OH group shows little interaction [51].
Similar effects have been observed in differential scanning calorimetric studies.
Addition of 20 mole percent epicholesterol to phosphatidylcholine has virtually no
effect on either the temperature or the heat of the transition. Both androstan-3fl-ol
and the keto-steroid cholest-4,6-dien-3-one, however, appear to broaden the transition
and shift it to lower temperatures [52]. It has also been established that whereas the
3fl-hydroxysterols such as cholesterol, cholestanol and coprostanol mix with phosphatidylcholines up to a molar ratio of 1:1, other steroids do not mix as well, the
upper limits to the steroid : lecithin molar ratio being put at 0.7: 1 for androstan-3/3-ol,
0.3:1 for epicholesterol and 0.55:1 for cholest-4,6-dien-3-one [53]. Cholesteryl
linolenate has also been reported to have a limited solubility in hydrated lecithin
bilayers [54]. Similar effects were found in a study of the ordering of a spin-labelled
cholestane derivative in bilayers of beef brain lipid: strong ordering effects were
observed when the added steroid had a hydroxyl group in the 3/3 position and a
hydrocarbon chain at C-17 [55].
The requirements for a planar steroid ring and a hydrocarbon chain at C-17
is explicable if the steroid is to be oriented with its long axis perpendicular to the
bilayer surface, since interaction with the neighbouring fatty acyl chains will then be
maximised. The requirement for an equatorially oriented 3-hydroxyl group suggests
the formation of stable hydrogen-bonded or dipole-dipole interactions with an
acceptor group such as the negatively charged phosphate oxygen atom of the
phosphatidylcholine head group [51,56]. However, de Kruyff et al. [57] have now
examined the interaction between cholesterol and a variety of lipids and shown that
the simple explanation for the effect of the 3fl-OH group cannot be true. Thus the
interaction between cholesterol and a lipid lacking the two oxygen atoms of the ester
bond at the 2-position of phosphatidylcholine is the same as that between cholesterol
and a phosphatidylcholine, so that they cannot be important in complex formation.
Similarly, the oxygen atoms connecting the phosphorus atom with the glycerol and
the choline part of the molecule are not important, and nor are the choline or
phosphate groups [57]. It has been suggested that the requirement for a 3/3-OH
group is connected in some way with the water of hydration associated with the
bilayer [44,57].
Oldfield and Chapman [58] have reported that phase transitions in sphingo-

318
CH 3

ON

C~O

H
(I)

(~)

myelin and cerebroside are removed by addition of an equimolar proportion of


cholesterol, as with phosphatidylcholines. A rather detailed study of the interaction
between lipids and a spin-label derivative of 5/3-androstan-17/3-ol-3-one (I) has been
reported by Trauble and Sackmann [59]. They suggest that above the temperature
of the phase transition the steroid and phospholipid molecules form an ideal mixture.
Below the transition, however, a mosaic structure was formed, with small domains
of steroid embedded within the lipid matrix. It remains to be seen how well these
observations apply to non-spin-labelled androstane derivatives.
Interaction between lipids and physiologically important steroids such as
oestrogens and progesterones does not appear to have been studied in detail, although
Willmer [60] has discussed some of the possibilities. Some work has, however, been
reported on the effects of anaesthetic steroids. In particular, Connor et al. [61] have
found using differential scanning calorimetry that incorporation of alphaxalone (lI)
into liposomes of dipalmitoyl phosphatidylcholine causes a marked decrease in
transition temperature. Surprisingly, however, in studies of the same system using
chlorophyll a as fluorescence probe no effect of alphaxalone could be detected (Lee,
A. G. (1976) unpublished observations).

V. MIXTURES WITH OTHER HYDROPHOBIC MOLECULES


The interaction between phospholipids and a variety of small hydrophobic
molecules has been studied.
Mixtures of phosphatidylcholines and phosphatidylserines with chlorophyll a
have been studied as models for the photosynthetic membrane. At molar ratios
greater than approximately 1 chlorophyll a per 70 moles of dipalmitoyl phosphatidylcholine, the lipid pre-transition is abolished and the phase transition broadened so
as to occur over the range approx. 42-31C [16]. Chlorophyll b also broadens the
phase transition of dipalmitoyl phosphatidylcholine [63].
In phosphatidylcholine mixtures with fatty acids, X-ray diffraction evidence
suggests that the carboxyl groups of the free fatty acid chains tend to be located
mainly in the region between the carbonyl and the glycerol group of the phosphatidylcholine [64]. Fluorescence experiments suggest a high degree of miscibility for
phosphatidytcholines and fatty acids, both in the gel state and in the liquid-crystalline

319

:8:-::::~::~

60-

40

.4,' ' ~ / ~
f/~] r"

//

//

20

o.~

0.4

o.~

o-8

1.o

Xdodecanol
Fig. 29. Phase diagrams for mixtures of dodecanol with: - - - - , dipalmitoyl phosphatidylcholine;
. . . . , dimyristoyl phosphatidylcholine; . . . . . . . .
dipalmitoyl phosphatidylethanolamine. Temperatures in C.

state. Addition of divalent metal ions, however, causes a separation of the fatty
acids [65].
Addition of myristic acid to dipalmitoyl phosphatidylcholine has been shown
to cause an increase in transition temperature [65], and studies using differential
scanning calorimetry have shown that this is true for all n-carboxylic acids containing
more than 12 carbons, whereas decanoic and octanoic acids cause a decrease [66].
Similarly, alcohols up to n-decanol cause a decrease in phase transition temperature
for dipalmitoyl phosphatidylcholine whereas alcohols from C-12 to C-18 cause an
increase [66,67]. For dimyristoyl phosphatidylcholine the break point comes at
n-octanol, with longer chain alcohols causing an increase in transition temperature,
and n-octanol and shorter chain alcohols causing a decrease [67,68]. As will be
explained in Section VII, a decrease in transition temperature would be expected for
the shorter chain alcohols and acids. The reason for the increase in transition
temperature for the longer chain alcohols and acids is not yet clear. Partial phase
diagrams for mixtures with dodecanol are shown in Fig. 29. A lamellar structure is
maintained up to a molar ratio of approx. 0.8 of dodecanol [66,67]. The results could
be explained if dodecanol were acting effectively as another lipid, with a phase
transition at approx. 55C. Although the bulk melting point of dodecanol is 24C,
there is no reason to expect that the 'melting point' in a lamellar phase would be
equal to that in the bulk phase. Further, if n-dodecanol were simply acting as another
lipid, then addition of n-octanol to mixtures of lipid and n-dodecanol should cause a
decrease in transition temperature, as is observed for other lipid mixtures (see p. 336)
this has in fact been observed to occur [67]. However, if this were true, then it would
be expected that the 'transition temperatures' of tetradecanol and octadecanol should
be higher than that of dodecanol in the bilayer, and hence that these longer chain
alcohols should cause a greater increase in the lipid transition temperature. Experimentally, the longer chain alcohols are observed to cause approximately the same
increase in transition temperature as does n-dodecanol [66]. An alternative ex-

320
planation is that mixing in the liquid-crystalline and gel states is far from ideal, and
approaches that in Fig. 8B.
Lastly, dipalmitoyl phosphatidylcholine bilayers have been reported to take
up tocopheryl acetate to 40 mole percent, beyond which tocopheryl acetate separates
out [69]. From the phase diagram derived from this mixture, the components were
suggested to be completely miscible in both the gel and liquid-crystalline phases, but
surprisingly whereas tocopheryl acetate caused a marked decrease in the temperature
of the completion of gel phase formation, it had very little effect on the temperature
of the start of gel phase formation. Such behaviour is not easily explained.

VI. MIXTURES WITH POLYPEPTIDES AND PROTEINS

VIA. Polypeptides
Although there are a number of reported studies of interactions between
phospholipids and polypeptides, no very clear picture has yet emerged. The interaction between phosphatidylcholines and the cyclic polypeptide alamethicin has been
studied by Hauser et al. [70] in 1970, and more recently Lau and Chan [71] have
shown that interaction occurs primarily in the region of the lipid head group. Of
more interest is the antibiotic chlorothricin which has been postulated to interact
primarily with the lipid fatty acyl chains. The antibiotic causes a reduction in the
heat of the lipid phase transition up to a molar ratio of 1:1, but with no apparent
change in the mid-point temperature of the transition [72]. The most likely explanation for the reduction in the heat change associated with the transition is that
there is an interaction between the chlorothricin and the lipid fatty acyl chains, so
that lipid in direct contact with chlorothricin can no longer take part in a co-operative
phase transition with the rest of the lipid. X-Ray data suggests that at high molar
ratios of chlorothricin, the structure is no longer lamellar [72]. The apparently odd
observation is that there is no effect on the transition temperature. Adding a foreign
compound to the bilayer would normally be expected to change the transition
temperature. If the antibiotic is immiscible with lipid in the gel state (as seems likely),
but is completely miscible with it in the liquid-crystalline state, then the antibiotic
will cause a decrease in transition temperature, given by Eqn. 81. This would predict
a decrease in transition temperature of at least 10C for a 2:1 molar ratio of dipalmitoyl phosphatidylcholine to chlorothricin: the decrease will be even larger if chlorothricin is considered to be present as a lipid - antibiotic 'complex'. The fact that no
such decrease is observed experimentally might suggest that the chlorothricin or
chlorothricin lipid complex is immiscible with free lipid in both the gel and liquidcrystalline phases. This conclusion has to remain tentative at the present time.
The actual state of the lipid in contact with the antibiotic is not clear: it could
logically either be in a more disordered or in a more ordered state than the bulk lipid.
The only requirement is that interaction with the antibiotic prevents co-crystallization
with the bulk lipid. Presumably with decreasing temperature, the fluidity of the contact

321
lipid will decrease, but in a gradual fashion, such that there is no discontinuous change,
and certainly involve no latent heat of transition. In other words, the transition will
no longer be first order.
The interaction of gramicidin S has been suggested to be with the lipid head
groups only, although addition of the antibiotic causes a gradual decrease in the
temperature of the lipid phase transition [73]. However, interaction with gramicidin
A shows very similar effects in differential scanning calorimetry, and in a more
recent publication [9] has been suggested to interact with the lipid fatty acyl chains.
In this case it appears that phospholipid in contact with antibiotic is again not
allowed to take part in a phase transition, but now the presence of the antibiotic'lipid
'complex' lowers the transition temperature of adjacent lipid. The published results
are rather surprising in that the transition temperature for dipalmitoyl phosphatidylcholine is found to be 46C, compared to the accepted value of 41.7C [9]. However,
at a molar ratio of dipalmitoyl phosphatidylcholine to gramicidin A of 91:9, the
transition temperature of the lipid is lowered by 3.5C. The expected decrease from
Eqn. 81 is approx. 3C, so that in this case it seems that there is miscibility in the
liquid-crystalline state. Polymyxin B has been suggested to interact with both the
polar groups of the lipid and the fatty acyl chains [73].
Rather more is known about the interactions between phosphatidylcholines and
melittin, the polypeptide which is the main cytolytic component of bee venom. The
first twenty amino acids of this peptide are mostly apolar whereas the C-terminal
hexapeptide is polar and highly basic [74]. At molar ratios ofpeptide to phosphatidylcholine greater than approx. 1 : 250, the peptide removes the lipid pre-transition, and
this has been shown to be due to the hydrophobic fragment 1-19 rather than to the
basic fragment 20-26 [75]. The peptide also reduces the heat content of the main gel to
liquid-crystalline phase transition, and broadens it slightly; but without any shift in
temperature. It has been estimated that approx. 10 phosphatidylcholine molecules
are prevented from participating in the phase transition per melittin molecule.
Interestingly however, it was found that the hydrophobic fragment 1-19 had no
effect on the heat of the transition, whereas the basic fragment 20-26 reduced it,
although not as effectively as melittin itself. This indicates therefore that it is the
polar rather than the hydrophobic segment of melittin which is responsible for the
interaction with phosphatidylcholine molecules, preventing them from cooperating
in the melting of the bulk lipid. On the basis of ESR data with spin-labelled fatty
acids, it seems that the fatty acyl chains in contact with melittin are more disordered
and less fluid than the bulk lipids [76,77]. Here, disorder means that the fatty acyl
chains adopt a wider distribution of angles with respect to each other than they would
in a normal bilayer structure. This presumably is caused by the steric packing
requirements of the melittin. The localized disruption of the membrane is presumably
sufficient to explain the increased ion permeability observed when melittin is added
to liposomes at low concentration [78]. Raman studies show that with increasing
temperature the fluidity of the fatty acyl chains in contact with melittin increases,
but in a somewhat complex manner. Thus there is a sharp increase in fluidity for

322
some of the methylene groups near the carboxyl end of dimyristoyl phosphatidylcholine bound to melittin at approx. 34C, whereas for the methyl terminal end, an
increase in fluidity occurs at approx. 27C. In the absence of melittin, the transition
for both ends of the chain occurs, of course, at the same temperature, as expected for
a 'melting' of the chains [79]. It is the lack of co-operativity within the chain when
bound to the polypeptide which must account for the fact that no transition is detected
in differential scanning calorimetry for the bound lipid.
The results with melittin are a very important aid in understanding the studies
performed on membrane protein systems. At low temperatures, dimyristoyl phosphatidylcholine in contact with melittin is immobilised, in the sense that the fatty
acyl chains are in an extended, all-trans configuration. However, although this will
probably maximise the enthalpy of interaction between melittin and the fatty acyl
chains, it will lead to a loss of conflgurational entropy for the chains. As the temperature is increased, the configurational entropy term becomes increasingly important,
so that rotation about C-C bonds appears, despite the decreased interaction with
melittin which must inevitably follow. This occurs first at the methyl-terminal end
of the chain, which can be considered to be the first part of the chain to 'come loose'
from the polypeptide. With further increases in temperature, the chain will increasingly come away from the polypeptide, until it is free along most of its length.
Even at high temperatures, however, steric restrictions would be expected to impose a
restriction on rotational freedom about C-C bonds in the chains, so that fluidity will
be less than for bulk lipid. The extent to which the lipids adjacent to melittin can be
considered to mix with bulk lipids will depend on the steric arrangement imposed on
the annular lipids by the melittin. If the annular lipids are still basically in a lamellar
kind of arrangement, then they would be expected to mix well with bulk lipid: the
presence of melittin and its annular lipids will then cause a decrease in the transition
temperature for the bulk lipid (Eqn. 81). If, on the other hand, the annular lipid has
been organized to be far from lamellar, then mixing with bulk lipid will be poor, and
so the polypeptide and its associated lipid will be immiscible with bulk lipid, in both
the gel and the liquid-crystalline states.

V1B. Lipid-protein interactions


Lipid-protein interactions would be expected to show the same kind of effect
outlined above for lipid-polypeptide interactions. Few systems have as yet been
studied in any depth, so that no clear picture can be painted. However, the studies
that have been reported do at least map out some of the effects that can be expected.
Novosad et al. [80] have reported a detailed study of the interaction between the
apolipoprotein A p o C - l l l isolated from human very low density lipoproteins and
dimyristoyl phosphatidylcholine. The interaction leads to a loss of bilayer structure
and a lipid protein complex could be isolated containing 34 mol of lipid per mol of
apoprotein. The lipid-protein interaction leads to a reduction in fatty acyl chain
motion as detected by ESR. Increasing temperature leads to increasing fluidity for
the chains, but the change was very gradual as shown by changes in the solubility of

323
0.6

0.5

~x"~
~.s'~. .,''Ar~

DMPC
0.4

_-

oC 0.3

DMPC:ApoC-III
0.2

50:1

/
O.1 j

10

20

30

4O

50

~0

T
Fig, 30. The effect of Apo C-Ill on dimyristoyl phosphatidylcholine as determined by di-tert-

butyl nitroxide. The parameter a is approximately equal to the ratio of label in the fluid lipid phase
to that in the bulk aqueous phase. Closed symbols, increasing temperature; open symbols, decreasing temperature (from ref. 80). Temperatures in C.

T E M P O (Fig. 30). This very gradual increase in fluidity with temperature probably
means that the co-operativity of the transition is low, both with regard to intermolecular and intra-molecular effects: it seems very unlikely that the transition can
be regarded as a first-order transition.
To confuse this picture somewhat, the interaction between ApoC-III and egg
phosphatidylcholine has been shown to be very different to that with dimyristoyl
phosphatidylcholine: in particular, with egg phosphatidylcholine, a bilayer structure
appears to be maintained [80]. Unfortunately, effects of this sort are very difficult
to interpret, since it is unclear to what extent they merely represent experimental
problems. For meaningful comparison, it is essential that the systems should have
reached their thermodynamically most stable configurations, otherwise rather than
reflecting differences in lipid-protein interactions, the results might only be showing
differences in the rates of change under some particular experimental condition.
A second system that has been studied in some detail is the (Ca2+,Mg2+) dependent ATPase from sarcoplasmic reticulum that has been isolated and reconstituted with a variety of phospholipids [81,82]. Enzyme activity is maintained as
long as there are at least approximately 35 lipids per ATPase, so that it has been
suggested that this is the number of lipids which surround the penetrant part of the
protein in the membrane. Interestingly, it appears that the activity of the ATPase
is controlled by these nearest neighbour, or 'annular' lipids [83] and that lipid
further from the protein has little effect [81,84]. Thus in dipalmitoyl phosphatidyl-

324
choline, the ATPase has a very low activity at temperatures below approx. 32C,
but at this temperature activity appears and increases with increasing temperature:
no breaks could be seen in Arrhenius plots at 42C, the transition temperature for
bulk dipalmitoyl phosphatidylcholine. The increase in activity above 32C correlates
with changes in ESR spectra indicating an increased motion for the fatty acyl chains
of the annular lipids. The picture that emerges is as follows. Below approx. 32C,
the fatty acyl chains of dipalmitoyl phosphatidylcholine in the annulus around the
ATPase are highly immobilized, and the rigidity of the lipid microenvironment
prevents the normal operation of the ATPase. Above 32C, the fluidity of the
annular lipids begins to increase in what is probably a very uncooperative fashion,
and the increasing fluidity allows increasing ATPase activity. Interestingly, the
perturbation of the dipalmitoyl phosphatidylcholine phase transition does not extend
significantly beyond the annulus, because a phase transition at approx. 41 C could be
detected as soon as extra-annular lipid was present in the system [84]. This is rather
surprising: the presence of an 'impurity' in the shape of the lipid protein complex
would be expected to cause a decrease in the phase transition temperature for the
bulk lipid. Thus at a lipid:protein molar ratio of 67: l, the phase transition for the
non-annular lipid would be expected to be between approx. 0.5 and 1 C lower than
that in the absence of protein. The failure to observe such an effect might be attributable to experimental difficulties, or it might be due to the failure of the assumption of
complete miscibility between the lipid protein complex and free lipid in the liquidcrystalline state. If the annular lipid is bound to the ATPase to give some relatively
long-lived complex, with a decrease in fluidity for the annular lipid, then it is quite
possible that the bilayer structure would be more-or-less perturbed in the annulus.
If this were so, then mixing with lipid in a bilayer structure would be either poor or
impossible: in other words, lipid and lipid " protein complex will be immiscible. This
immiscibility would lead to an increased probability of defects in the region between
the annular lipids and the next lipid shell, as described by Hesketh et al. [84] and
which could explain the increased permeability of membranes to ions caused by the
presence of the ATPase [85].
It is important to realise that immiscibility of the type suggested here is not
universal. Thus Grant and McConnell [86] have reconstituted glycophorin with
dimyristoyl phosphatidylcholine and have observed a small decrease in lipid phase
transition temperature at a lipid:protein molar ratio of 120: 1. The calculated decrease for this system is approx. 0.3C. Indeed, it is possible that immiscibility in the
(Ca2+,Mg2+)-ATPase system might be a function of the lipids used in the reconstitution. Thus, although no net accumulation of Ca 2+ occurred when the ATPase
was reconstituted with dioleoyl phosphatidylcholine, net uptake was found when the
system was reconstituted with lipids extracted from the sarcoplasmic reticulum
membrane itself [87].
For the (Ca2+,Mg2+)-ATPase complex with dimyristoyl phosphatidylcholine,
ATPase activity appears at temperatures above approx. 24C, which is very close to
the temperature of the gel to liquid-crystalline phase transition for this lipid [81].

325
[

(b)

20

4o

Fig. 31. Calculated order parameters for annular lipid around a protein as a function of the strength
of the lipid-protein interaction (described by the parameter V~p). The dotted line shows the change
calculated for lipid alone (from ref. 89).

This would seem to be rather in the nature of a coincidence. Presumably, a lipidprotein complex is formed as with dipalmitoyl phosphatidylcholine, and, because of
its shorter length, interaction of the C-14 fatty acyl chain would be expected to be
weaker than for the C-16 chain, so that an increased fluidity can occur at a somewhat
lower temperature. Kleeman and McConnell [88] have shown that the T E M P O
solubility of the (CaE+,Mg2+)-ATPase complex with dimyristoyl phosphatidylcholine
starts to increase at approx. 22C.
Some of these features have been nicely reproduced in theoretical calculations
of the interaction between proteins and lipid fatty acyl chains performed by Marcelja
[89]. Lipids in the annulus around a protein are calculated to be more ordered than
bulk lipids, because of steric restrictions to their rotational freedom. Second and
third lipid neighbours will also be affected, but to a much smaller extent. The
phase transition for the annular lipid is predicted to become broad, and depending on
the strength of the Van der Waals interactions between lipid and protein, the transition
temperature may be increased or decreased - the weaker the lipid-protein interaction,
the lower the transition temperature (Fig. 31). Kimelberg and Papahadjopoulos [90]
have reconstituted a delipidated ( N a + + K+)-ATPase from rabbit kidney outer
medulla with a variety of phosphatidylglycerols. They found breaks in Arrhenius
plots of ATPase activity at 20, 31 and 44C for enzymes reconstituted with
dimyristoyl-, dipalmitoyl- and distearoylphosphatidylglycerol, respectively. These
temperatures can be compared with the phase transition temperatures of the pure
lipids of 21-23, 38-40 and 52-54C, respectively. Again there was no exact equality
between the temperatures for changes in enzyme activity and the temperatures of the
lipid phase transitions. Again the ATPase activity appears to be controlled by the
fluidity of the lipid annulus, this in turn being controlled by the nature of the inter-

326

v\

-,.

3b

,2-o

7'

,'o

Terr~erot uee [ ~ ]

Fig. 32. The effect of temperature on the fluorescence intensity of cytochrome bs. O, pure protein;
(3, dipalmitoyl phosphatidylcholine' cytochrome b5 complex; ~ , ovalbumin in the presence of
dipalmitoyl phosphatidylcholine (from ref. 91).
actions between the lipid and the protein. Increasing the length of the fatty acyl chain
increases the strength of the interaction, but effects of the lipid head group must also
be considered. Thus when reconstituted with beef brain phosphatidylserine, the
discontinuity in Arrhenius plots of the (Na q- K)-ATPase activity occurred at
15C, although the temperature for the phase transition in the pure lipid occurred
approximately between 5-13C [90].
This type of model for lipid-protein interactions has been confirmed in a very
direct fashion by a study of the complex of cytochrome b5 and dipalmitoyl phosphatidylcholine [91]. This protein shows a strong fluorescence from a tryptophan
residue in a hydrophobic environment, which can be used as a probe for changes in
protein conformation. As shown in Fig. 32, there is a change in protein conformation
starting at approx. 32C and ending at approx. 40C. Thus changing fluidity of the
annular lipid can bring about changes in protein conformation and so changes in
protein activity.
The mechanism of the (Ca2+,Mg2+)-ATPase has been shown to involve initial
binding of Ca 2 and ATP, followed by the formation ofa phosphorylated intermediate
and the release of ADP. Calcium is then translocated through the membrane,
presumably by some conformational change, and then finally the calcium is released
and the protein dephosphorylated. The principal effect of annular lipid has been
shown to be on the processes leading to dephosphorylation, and probably on the step
involving the translocation of calcium [92]. Certainly, if Ca 2 transport occurs by a
gated-pore mechanism of the type suggested by Martonosi [93], then it would be
expected to rely on a fairly fluid lipid micro-environment. The importance of fluidity

327
11
6.0

logtlU/mg)
0

-2

3:2

'

3.~,

3-6

I/T' 103
Fig. 33. Arrhenius plots of ATPase activities of the dipalmitoyl phosphatidylcholine (Ca2+,Mg2+) ATPase complex in the absence ( O ) and in the presence ( 0 ) of 50 mM benzoyl alcohol (,from ref. 84).
Temperatures in K.

is emphasized by the effect of benzyl alcohol on the activity of the ATPase reconstituted with dipalmitoyl phosphatidylcholine [84]. Addition of 50 mM benzyl
alcohol causes an approximately 2C drop in the temperature of the break in
Arrhenius plots of activity, and maintains a higher activity for the enzyme down to at
least 20C (Fig. 33): the effect of adding 50 mM benzyl alcohol at a given temperature
is equivalent to increasing the temperature by approx. 2-3C.
The fact that all three enzymes discussed above when reconstituted with
dipalmitoyl phosphatidylcholine show breaks in activity at approx. 32C is of interest.
If the fatty acyl chain-protein interaction was relatively non-specific, simply involving
a generally hydrophobic protein surface, then interactions would be expected to be
similar for all proteins. Some specificity might, however, be expected with regard to
lipid head groups or as regards phospholipids versus cholesterol. Indeed Warren et al.
[83] have presented evidence that suggests that cholesterol might be excluded from
the lipid annulus around the (Ca2+,Mg2+)-ATPase, when both cholesterol and
phospholipid are present. There is also evidence for some specificity in phospholipidprotein interactions. Such effects were found in a recent study of the activity of
fl-hydroxybutyrate dehydrogenase in mixtures of phosphatidylcholines and phosphatidylethanolamines [94]. Activity was higher in the presence of the ethanolamine.
Interestingly, it was found that on cooling the system reconstituted with a 1 : 1 molar
ratio of phosphatidylcholine and phosphatidylethanolamine, gel-phase lipid started
to separate at 25C, whereas a break in the Arrhenius plot of enzyme activity did not
occur until 20C. This was attributed to lipid phase separation, with the enzyme
preferentially partitioning into lipid in the liquid-crystalline phase [94]. An alternative
explanation along the lines proposed above would be that the phase transition for the
annular lipid started at a lower temperature than for the bulk lipid. These possibilities
could be distinguished by studying fl-hydroxybutyrate dehydrogenase reconstituted
with only sufficient lipid to form the annulus.

328
The preferential exclusion of protein from lipid in the gel phase has been
demonstrated for glycophorin, the major glycoprotein of human erythrocytes. In
freeze-fracture electron microscopy of lipid mixtures containing glycophorin, the
glycophorin is almost exclusively present in regions of liquid-crystalline-phase lipid,
under conditions where both gel and liquid-crystalline-phase lipid are present [86].
Similar effects have been observed for the (Ca2+,Mg2+)-ATPase [88].
A number of other studies support the general picture of lipid-protein interaction presented above. Using differential scanning calorimetry it has been shown
that the hydrophobic protein lipophilin extracted from brain myelin interacts with
dipalmitoyl phosphatidylcholine, causing no effect on the temperature of the midpoint of the lipid phase transition. There is, however, an increase in the temperature range for the transition and a decrease in enthalpy. It was concluded that an
annulus of approximately 15 lipid molecules was formed around the protein, and that
these could take no part in the phase transition. The non-annular lipid was unaffected [95,96]. Similar conclusions were derived from an ESR study of the same
system [97]. The presence of immobilized lipid has also been suggested around
cytochrome oxidase [98] and cytochrome-b5 [99].
In all of these studies it is important to be clear what is meant by the word
immobilized. The term is generally used to mean that annular lipid has relatively
little rotational freedom, and does not necessarily mean that the lipid-protein interaction is long lived. From ESR experiments it is often difficult to distinguish between
exchanging and non-exchanging annular lipid. To be immobilized in the sense of
lacking rotational freedom it is only necessary that the rotational correlation time be
in the slow tumbling region, fro, ~ 3 10-8 s, which is significantly less than the jump
lifetime of approx. 10 7 s for lipids in a bilayer [100]. In fact, most experiments seem
to be consistent with exchange between annular and non-annular lipid at a rate equal
or greater than approx. 106 S - 1 [97-102]. In N M R studies of the retinal rod membrane, Brown et al. [101,102] have taken the analysis of 'immobilization' one step
further. They conclude that immobilization of lipid in the annulus around rhodopsin
corresponds to a decrease in the high-frequency rotations about C-C bonds, with
little effect on slower large amplitude segmental excursions. The restriction on highfrequency rotation would account for the observed increase in order parameter for
spin-labelled phospholipids caused by rhodopsin [102]. The slower, large amplitude
angular motions could still occur if during their execution, the annular phospholipid
could diffuse away from the rhodopsin. The N M R results for this system show that
this is reasonable. Rotations about C-C bonds occur at relatively high frequency with
the fatty acyl chain in a particular lattice site, so that a large effect of protein-lipid
interaction would be expected. Large amplitude segmental motions however occur
with a much lower frequency and if coupled to lateral jumps to adjacent lattice sites,
then these motions would be relatively uninhibited by interaction with protein [100].
All the proteins discussed so far have been hydrophobic proteins. A rather
different pattern of effects can be expected for hydrophilic proteins. Pancreatic
ribonuclease has been shown to interact with negatively charged bilayers such as

329
those of phosphatidylglycerol, but with no change in transition temperature, although
there was a slight increase in the enthalpy of the transition. There was no interaction
with bilayers of phosphatidylcholine [96]. Polylysine also absorbs onto bilayers of
phosphatidylglycerol, and although the effects are complex, it can at least be said
that the interaction leads to an increase in transition temperature. In this it is like
the interaction of C a 2+ o r M g 2 with negatively charged lipid.
Effects of other positively charged proteins are more complex. The A1 basic
protein from myelin has been reported to increase the phase transition temperature
of dilauroyl phosphatidylglycerol, as does Ca 2 [103]. However, the same protein
causes a decrease in phase transition temperature for dipalmitoyl phosphatidylglycerol at high protein concentrations [96]. There is no obvious simple explanation
for these results. The amino acid sequence of the protein has been reported [104],
and contains a rather random distribution of positively charged residues, which could
well interact electrostatically with negatively charged lipids. At the same time however, there are regions of nonpolar residues which could penetrate into the hydrophobic portions of the membrane. This type of interaction between lipid and Am
basic protein has been suggested on the basis of monolayer studies at the air/water
interface. Interestingly, in these studies, the strongest interaction with lipids typical
of the myelin membrane occurred with cerebroside sulphate [105].
There have been a large number of studies of the interactions between cytochrome c and negatively charged lipid, all of which make the interaction electrostatic.
The topic has been reviewed by Nicholls [106] who points out two major experimental
problems. Firstly, at high molar ratios of cytochrome c, is a lamellar structure
maintained? Secondly, in the very acid environment of the phospholipid head group
region of a bilayer of negatively charged lipid, there is the possibility that the cytochrome c conformation will change. Chapman et al. [9] have observed that
cytochrome c reduces the phase transition temperature of phosphatidylserine and
Papahadjopoulos et al. [96] have also observed that it reduces that of phosphatidylglycerol. Presumably this results from a disordering of the lipid fatty acyl chain
packing, necessary to allow optimal interaction between the lipid head groups and
suitable positively charged groups on the protein. Such an effect would only really
become apparent at the temperature where gel-phase lipid was being formed, and
Van and Griffith [107] have shown by ESR studies that cytochrome c has no significant
e~ect on the fluidity of lipids in the liquid-crystalline phase. It has also been shown
that cytochrome c can induce a phase separation in mixtures of steroid and
phosphatidylglycerol [108].
Binding of negatively charged groups in proteins to negatively charged
phospholipids is also possible through Ca 2* o r M g 2 bridging. Although this has
not yet been much studied, it is known that binding of prothrombin to phosphatidylserine involves Ca 2 [109].

330
vii. INTERACTION WITH SMALL WATER-SOLUBLE MOLECULES
The interaction of most water-soluble organic compounds with lipid bilayers is
controlled by relatively non-specific charge-charge and hydrophobic interactions.
Thus Cerbon [110] studied the interaction between both tetracaine and butacaine and
bilayers of egg phosphatidylcholine by proton NMR, and observed a significant
immobilization for the hydrophobic chains of the anaesthetics, although it should be
noted that Hauser et al. [111] found no evidence for the immobilization of tetracaine
in this system, also using NMR. Incorporation of a charged molecule such as
tetracaine into a neutral phosphatidylcholine bilayer will produce a positive charge at
the bilayer surface. This has been demonstrated by studying the displacement of
trivalent metal ions bound to the surface [112]. Interaction of a number of other
drugs with phosphatidylcholine bilayers have also been studied using proton N M R
and immobilization observed [113].
If the lipid is negatively charged, then stronger interaction with positively
charged drug molecules will be possible. Hauser et al. [111] have reported on interactions between anaesthetics and phosphatidylserine, and Crompton et al. [114] have
studied interactions between butacaine and mitochondria.
Abood and Hoss [115] have shown that morphine interacts with phosphatidylserine at an air/water interface. Interestingly, the binding shows stereospecificity in
that levorphanol decreases the binding of morphine at lower concentrations than
does its enantiomer dextrorphan [115]. Similarly, it has been shown that the binding
of etorphine and naloxone to cerebrosides and cerebroside sulphates are prevented
by levorphanol but not by dextrorphan [116,117]. Cater et al. [118] have studied the
effects of some of these compounds on the phase transition in dipalmitoyl phosphatidylcholine. Morphine itself had no effect, suggesting that its interaction with charged
lipids must have been largely because of the presence of the charged lipid head groups.
Various morphine derivatives containing hydrophobic groups did decrease the
transition temperatures, but there was no obvious correlation between the structure
of the derivative and the extent of the decrease in transition temperature [118].
Binding of serotonin to gangliosides has been measured, and shown to be
mainly electrostatic [119]. The interaction has been characterised by PMR [120].
Lastly, an interesting interaction between negatively charged phospholipids and
dimethyl sulphoxide has been observed [121 ]. This compound increases the transition
temperature of dimyristoyl phosphatidylglycerol and also leads to the appearance of
a discrete new transition at high temperature. The new upper peak was observed to
disappear after melting and so possibly results from a metastable state and super
cooling. The nature of the interaction in this system has not yet been established.
The effects of neutral molecules such as the alcohols can be analysed according
to a theory presented by Hill [122]. At equilibrium a solute will partition between the
aqueous and lipid phases. The solute within the lipid phase will be partly in regions
of lipid in the liquid-crystalline phase and partly in regions in the gel phase. Let the
g
At
mol fraction in the liquid-crystalline phase be Xsolute
l
and in the gel phase Xsolute.

331
equilibrium, the chemical potential of the solute in the gel phase, /[Zsolute
g
must equal
that in the liquid-crystalline phase, ~solute,
1
g
!
~solute ~ ~solute

(74)

Similarly, the chemical potentials of the lipid in the two phases must be equal,
g

(75)

/Zlipid ~ /~lipid

The simplest case is that considered by Hill [122], where the solute is practically
insoluble in gel-phase lipid so that Xsolut~
g
= 0. As long as the concentration of solute
is low, it is reasonable to write
H 1
= H lO)
lipid
lipid

(76)

and
1
S lipid ~

l(O)
t
S lipid - - RlnXlipid

(77)

where ~l(O)
~" lipid and .~(o)
Olipid are respectively the molar enthalpy and entropy of lipid
in the liquid-crystalline state and Xlipi
d l is the mole fraction of lipid in the liquidcrystalline phase.
Equilibrium between the gel and liquid-crystalline phases is achieved at a
temperature T such that
1
~ H l
1
= H g
__
ktlipid
lipid - - TSlipid
lipid

TSgpid

(78)

Thus we obtain,
Tro
~

1(0)
lipid-

n,

1
mnxlipid-

S g
--- H g
l i p i d ) - H l(0p)id
lipid

(79)

Since for (pure) gel-state lipid


H lipid
g
= H g()
lipid anft"

S glipid =

S g()
lipid

(80)

we obtain the familiar relationship for the depression of the freezing point (Eqn. 12)
1

1 = R

Tc

A Hc

lnx 1

(81)

lipid

Here T is the transition temperature in the absence of solute and AH is the enthalpy
of transition,
A/ate ~

Hll:Id--

Hg(0)lipid

(82)

For dilute solutions, where the depression of the transition temperature is small,
Eqn. 81 can be approximated as
A T

T - - T

RT~2

A He

solute

(83)

332
This equation then shows that for a one-degree depression in the transition temperature of dipalmitoyl phosphatidylcholine, the concentration of the solute in the
membrane has to be 0.044 tool per tool of lipid. It is now necessary to consider the
equilibrium distribution of solute between water and the (liquid-crystalline) lipid
phase. The chemical potential of the solute in the aqueous phase, ~ q can be expressed in terms of the chemical potential of the pure solute (if a liquid)
(84)

,uaq ~ ,u + R T l n c

It can also be expressed in terms of the chemical potential of a water-saturated


solution,
(85)

/~.q = ,u '~'t + R T l n r

where r is the fraction of saturation, and is given by


r =

(86)

C./S

Here Cw is the aqueous concentration of the solute and S is the concentration at


aqueous saturation, both expressed in terms of moles of solute per mole of water
[122]. Differentiating Eqn. 83 with respect to r gives the dependence of the depression
of the transition temperature on the aqueous concentration expressed as the fraction
of saturation,
6,1T

61"

RT2

6 I

Xslu~e
6r

,~1H~

(87)

However, since
6 I

Xsolute

Xsolute

6r

(88)

~Cw

it follows that
6 -Xsolule
.!
. . . .
e~r

P" S

(89)

where P is the partition coefficient of the solute between the two phases. Thus,
~ j T = RT_~t'. s
6r

(90)

.IHc

Hill [122] has shown that the analysis can be taken one step further. At equilibrium.
the chemicial potential of the solute in the two phases will be equal, so that
,u ~"t + R T l n r

Ilrl(O)

r'solute

~(- R T I n x I .

SOlUte

(91)

I f the concentration of water in pure water-saturated solute is small, then the chemical
potential of the pure solute and of the water-saturated solute will be very close,
#~.t ~ #o

(92)

333
Thus from Eqn. 91,
RTln(xlsolute ) : #o _ _ t~solute"l(0)

(93)

The right-hand side of Eqn. 93 is the difference in chemical potential of one mol of
pure solute and of one mol of the solute dissolved in one mol of the membrane. This
is given by the entropy of mixing which is Rln2. Thus,
RTIn

x1
solute _
t"

Rln2

(94)

or using Eqn. 86
RTIn

xI
s?lute. S =

Rln2

(95)

cw

Therefore,
x1

solute . S : P" S = 2
Cw

(96)

Substituting into Eqn. 90 we obtain Hill's expression,


6AT
i)r

--

RTc 2
'
A Hc

(97)

Agreement between theory and experiment is good for the short-chain alcohols
[67,122]. Fig. 34 shows the effect of n-pentanol on the temperatures of the upper and
lower ends of the phase transition in dipalmitoyl phosphatidylcholine. Effects are
obviously linear, and the value of P" S obtained using Eqn. 89 is 2, as predicted
theoretically. For longer-chain alcohols, however, agreement with theory is not so
close (Table II). Thus the partitioning of the alcohols appears to become less ideal
as the chain length increases, at least in these experiments where the transition
temperatures were detected from variations in the fluorescence of chlorophyll a
incorporated into the lipid [67]. This contrasts with the results obtained by Hill [122]
using light scattering, who found that, although partitioning of the odd-numbered
chain length alcohols was non-ideal, that of the even-numbered chain length alcohols
up to 1-octanol was close to that expected for ideal behaviour. The reason for this
difference has not been established. A more important observation, however, is that
the long-chain alcohols cause an increase in width for the lipid phase transition (see
Fig. 35). This has been seen both in studies using chlorophyll a as a fluorescence
probe [67] and in studies using differential scanning calorimetry [123].
For the short chain alcohols, packing with the lipids in the gel state is likely to
be poor, so that solid solution is unlikely as assumed in the above calculations. For
the longer-chain alcohols, however, packing is likely to be considerably better in a
mixed bilayer in the gel state, and part of the deviation from theoretical expectation
can be attributed to this cause. In this case there is no reason to modify Eqns. 76

334

pentanoS?l I u
mM

~
40
3(1

t
1

't0"'lll
30

40

50

Fig. 34. The effect of n-pentanol on the temperatures of the upper (u) and lower (1) ends of the
phase transition in dipalmitoyl phosphatidylcholine. Temperatures in C.

TABLE II
EFFECT OF n-ALCOHOLS ON LIPID PHASE TRANSITIONS
Data from refs. 67 and 124.
Lipid

Alcohol

Mid-point
transition

Upper end of
transition

Lower end of
transition

~1T/r

~1T/r

A T/r

P" S

xg
solute
X1
solute

Apparent
X~olute
solute

Dipalmitoyl
phosphatidylcholine

Butanol
43
Pentanol 40
Hexanol 31
Heptanol 29
Octanol
19

2.1
2.0
1.5
1.4
1.0

43
40.5
29
20
12

0
0
0.3
0.5
0.7

43
40.5
32
33
24

0
0
0.2
0.2
0.4

Dimyristoyl
phosphatidylcholine

Octanol

11

0.4

10

0.8

25

0.6

Dipalmitoyl
phosphatidylethanolamine

Octanol

10

0.5

0.8

0.8

and 77 for the liquid-crystalline phase, but Eqn. 80 will no longer be true. The m o s t
obv i o u s a s s u m p t i o n is to write for the gel phase, by an al o g y with the liquid-crystalline
phase,

Hlipld

= H lipid
g()

(98)

335

octanol
raM2.0

!,

1'5

, ,

~o

~io

4'o T--~O

Fig. 35. The effect of n-octanol on the temperatures of the upper (u) and lower (1) ends of the phase
transition for dipalmitoyl phosphatidylcholine (solid lines). The broken lines show the lines calculated from Raoult's Law. Temperatures in C.
and
g
=
Slipid

(99)

s glipid
(0) __ Rlnx~ipi d

so that Eqn. 81 becomes

R lnfXllipid~

n-~= ~

(lOO)

\ ..ld

If both Xsolute
i
and Xsolut
g
are small then,

R,n (x',,o,;/

\x~ipi al ~-" -e(Xsolute - - X~olute)

(t01)

so that

( xg )

RTc2 x I
l
A T "~ -A-H
solule

solute

--

xV---

"

(lo2)

solute

Following the assumptions that led to Eqn. 97, we obtain [124]


~A T

RTe2 2 { l

x- gSOlute ~

(lo3)

W h e n the ratio of solute partitioned into gel-phase lipid is very m u c h less than that
into liquid-crystalline-phase lipid, then, o f course, Eqn. 103 reduces to Eqn. 97.
Fitting Eqn. 103 to the experimental results gives a ratio o f X~olute/Xlso~ute,and these
are listed in Table IL The decreasing effect o f the alcohols on the upper transition

336
temperature with increasing chain length can be attributed to an increasing partitioning into the gel-phase lipid. It would however be unreasonable to attribute the
decreased effect of the alcohols on the lower end of the transition to a decreased
partitioning into the gel state, since it has been postulated in Part I [62] that the
upper end of the transition corresponds to transition between 'bulk' gel and liquidcrystalline phases, whereas the lower end of the transition corresponds to this
transition in the region of vacancies or dislocations and the alcohol would presumably
be more soluble in the disordered gel-state regions, not less. However, it is the disorder of the gel-state lipid that provides an answer. Experimentally, it seems that the
Rlnxgpid term in Eqn. 99 is having less than its expected effect: that is, the entropy
of mixing in the gel state is considerably less than that expected for a random distribution of solute and lipid. This would, of course, follow if the solute were preferentially partitioning into disordered regions of the gel state - the solute would then be
effectively ordered into relatively small regions of the gel-state lipid, rather than being
randomly distributed throughout the whole of the gel [124]. The depression of the
transition temperature of the lipid in the region of the vacancies would then be
greater than for the bulk lipid, and an increase in width of the transition with increasing solute concentration will result, as observed experimentally. This is perhaps
the strongest piece of evidence for the picture of the lipid phase transition presented
in Part I [62]. If the lipid sample were physically homogeneous, then it is difficult
to see how addition of a foreign compound could lead to an increase in width for the
transition. If, on the other hand, vacancies and dislocations are present, then there
is a natural explanation for an increase in width of the transition.
The effects of alcohols on lipid mixtures have also been studied [67]. The
primary effect of 1-octanol on the phase diagram for mixtures of dimyristoyl phosphatidylcholine and dipalmitoyl phosphatidylcholine is to shift both the upper,
fluidus curve and the lower, solidus curve to lower temperature (Fig. 36). The shape
of the phase diagram also changes because of the greater effect of 1-octanol on the
width of the phase transition for dimyristoyl phosphatidylcholine. The overall effect
is to increase the amount of fluid lipid present.
Addition of l-octanol to a mixture of dimyristoyl phosphatidylcholine and
dipalmitoyl phosphatidylethanolamine also lowers both the fluidus and solidus curves
(Fig. 37). However, addition of alcohol in this case removes the horizontal portion
of the solidus curve, and could be taken to show that the lipids are now completely
miscible in the gel phase. Such a conclusion would probably not be justified, for the
following reasons. As shown in Table II, the effects of 1-octanol on the temperatures
of onset and completion of gel-phase formation are equal in dipalmitoyl phosphatidylethanolamine: in other words, 1-octanol causes no increase in width for the phase
transition. Similarly, it is observed that for mixtures containing a mole fraction of
dipalmitoyl phosphatidylethanolamine of 0.2 or greater, addition of l-octanol has
an almost equal effect on the onset and completion temperatures. This is not so for
the pure phosphatidylcholines, where the effect on the temperature of completion of
gel-phase formation is much more marked, and it is the large effect on this temper-

337
50-

10

o-6

o.8

I-o

XDPPC

Fig. 36. The effect of octanol on the phase diagram for mixtures of dimyristoyl phosphatidylcholine
and dipalmitoyl phosphatidylcholine:
, no octanol; - - -, 1.5 mM octanol. Temperatures in C.

70-

5G

30

10
0

o.6

o.8

XO~E

1.o

Fig. 37. The effect of octanol on the phase diagram for mixtures of dimyristoyl phosphatidylcholine
and dipalmito2yl phosphatidylethanolamine: - - , no octanol; - - - , 2.3 mM octanol. Temperatures in C.

ature for dimyristoyl phosphatidylcholine which results in the disappearance of the


horizontal portion o f the phase diagram. As described elsewhere [2], these diagrams
are experimentally simple diagrams of the temperatures o f onset and completion of
gel phase lipid formation, and it is an extrapolation to treat them like normal phase
diagrams.
Charged molecules have especially interesting effects on mixtures containing
charged lipids. Thus, in a mixture o f phosphatidylserine and phosphatidylcholine in
the presence o f Ca z+, there is a phase separation o f phosphatidylserine. Addition of
the positively charged anaesthetic tetracaine, however, displaces Ca 2+ from the

338
2
mM
Dibucain~

'!

30

35

40

45

Fig. 38. The effect of dibucaine on the (mid-point) phase transition temperature of (broken line),
dipalmitoyl phosphatidylcholine; (solid line), dipalmitoyl phosphatidylcholine containing 11 mol
myristic acid at pH 7.2. Temperatures in C.

bilayer, reversing the phase separation. Presumably charge effects of this kind will
follow from the binding of any positively charged drug to a membrane.
For molecules very unlike the lipids in structure, it is likely that mixing will be
non-ideal in the liquid-crystalline state, and an analysis of depression of freezing
point will be even more complex than that for the alcohols outlined above. Jain et al.
[125] have studied the effects of a series of derivatives of adamantane on the transition
temperature of dipalmitoyl phosphatidylcholine, and failed to find any clear relationship between depression of freezing point and estimated hydrophobicities.
The effect of charged molecules will be more complex because of the build-up
of charge on the liposomes. Thus effects will depend in a non-linear way on concentration, as shown in Fig. 38 for dibucaine. The binding constant in this case can
be written as
K = K o exp(F~Po/kT)

(104)

where Ko is the binding constant at high ionic strength, F is the Faraday and +o is
the surface potential, which can be estimated from the Guoy-Chapman theory.
McLaughlin and Haray [126] have shown that an equation of this kind can be used to
obtain a good fit to the experimentally determined binding of 2,6-toludinyl napthalene
sulphonate (TNS) to phospholipid bilayers. If binding is limited by the build-up of
positive charge, then incorporation of negative charge into the bilayers should increase the effect of dibucaine: as shown in Fig. 38, this is indeed the case. Further,
the effect of the negative charges can be completely masked by addition of Ca 2+ or
Mg 2, which bind to the negative charges [65]. Papahadjopoulos et al. [127] have
also shown that partitioning of dibucaine into phosphatidylserine is considerably
greater than that into phosphatidylcholine. Effects of dibucaine on dipalmitoyl
phosphatidylethanolamine are considerably less than on the phosphatidylcholines,
presumably as a result of tighter packing in the ethanolamine [65].

339
12Drug
\

mM

\
\\

6-

\
\
\

\
o\ \

30

S'5

40

Fig. 39. The effect of pentobarbitone on the phase transition temperature of dipalmitoyl phos-

phatldylcholine: - - ~,at pH 8,5; . . . . , at pH 7.3. Temperatures in C.


Although the Gouy-Chapman theory may provide an adequate explanation
for some aspects of the binding of charged molecules to bilayers with a high charge
density, it is unlikely to do so for bilayers with a large spacing between charges. In
the latter case, discrete charge effects will become important. In particular, in the
nerve membrane it has been postulated that negatively charged molecules may be
concentrated in proximity to the sodium channel (see ref. 128). If this is so, then
positively charged drug molecules will partition preferentially into regions close to
the negative charges.
Negatively charged pentobarbitone also reduces the phase transition temperature of dipalmitoyl phosphatidylcholine in a non-linear fashion, in this case because
of the build-up of negative charge (Fig. 39). The relative proportion of ionised
([A-]) and unionised ([A]) pentobarbitone present at any p H value can be calculated
from the Henderson-Hasselbalch equation,
log [A-] = pKa - - pH
[A]

(105)

Since the pKa value of pentobarbitone is 8.0 [129], 76 ~ of the pentobarbitone will
be ionised at p H 8.5, and only 17 ~ at pH 7.3. Pentobarbitone therefore has a greater
effect at p H 7.3, because of the slower build-up of negative charge (Fig. 39). The
effects of barbiturates can be related to their hydrophobicities [129]. Similar relationships have been shown for chlorpromazine and the fl-blockers propranolol and
practolol [130].
These results have been incorporated into a model for the action of local
anaesthetics, since it has been observed that there is a correlation between the concentration of drug required to produce a significant decrease in transition temperature and that required to block the sodium current in nerve [128]. The lipid surrounding the sodium channel in nerve has been postulated to be in a rigid or gel-like
state, and it is the rigidity of the lipid micro-environment that keeps open the channel
through the nerve. Addition of anaesthetic fluidises this annular lipid, allowing the
sodium channel to relax into an inactive configuration in which the channel is closed.

340
VIII. CONCLUSION
In the first part of this review it was established that the lipid gel to liquidcrystalline phase transition was in many ways like a normal melting transition, but
with some important differences. Firstly, it is a first-order transition since there is an
observable enthalpy of transition. On the other hand, even for an ideal sample the
transition would not be discontinuous, because gel and liquid-crystalline phases
would necessarily have to co-exist within a single bilayer during the transition. In
practice, of course, no lipid sample will be chemically pure or physically homogeneous,
and this will lead to a further increase in width for the transition. The relatively broad
transitions observed experimentally can probably be best considered as the sum of a
number of sharper transitions, each taking place within discrete regions or domains
of the bilayer. In such a system, hysteresis effects can be expected and these have,
indeed, been observed.
If phase transitions could only be triggered by changes in temperature, then they
could be only of little importance in the functioning of biological membranes, since
most biological systems maintain a constant temperature. In fact, however, phase
transitions for charged lipids can be triggered by changes in the concentrations of
metal ions in the aqueous phase, and phase changes for all lipids can be triggered by
addition of a variety of hydrophobic drugs. Just how important these phenomena
are remains to be seen. Nevertheless, it is clear that they could serve as a convenient
means of amplifying a small biological signal into a large membrane effect. For this
to be so, target proteins in the membrane would probably have to be surrounded
by lipid in a gel or solid-like state, unlike the bulk of the membrane lipid which seems
generally to be in a liquid-crystalline state. Such solid-like lipid has been postulated
to surround the sodium channel in nerve [128], the cytochrome P-450 system in
microsomes [131] and bacteriorhodopsin in the purple membrane of Halobacterium
halobium [132]. For a system operating as a channel through the membrane, the
advantage of a rigid micro-environment would be to keep open the channel. For a
pump protein, on the other hand, a rigid micro-environment would most likely
prevent operation, by preventing the necessary conformational changes. This could
then act as an important control mechanism, with some suitable change in, say,
divalent metal ion concentration releasing the pump from inhibition. It may be
that the microsoma[ cytochrome P-450 system operates in this kind of way, being
released from inhibition after its hydrophobic substrates trigger a phase change in
the surrounding lipids to a fluid, liquid-crystalline-like state.
The rigid micro-environment around such proteins could be established either
by selection of specific lipids or by the strength of the lipid-protein interaction (or by
a combination of both). Certainly, it would be misleading to think of the phase
'transitions' of these annular lipids as being similar to those of bulk lipids. Since the
transition has no associated enthalpy of transition it cannot be a first-order transition,
but must be a non-cooperative-type of change. If this is so, then the temperature of
the mid-point of the transition of the annular lipid cannot be expected to be equal to

341

the transition temperature of the bulk lipid: the transition temperature will depend on
the relative strength of lipid-protein and lipid-lipid interactions. Unfortunately, this
makes the interpretation of thermal effects in biological membranes even more
complex than it seemed in the past.

REFERENCES
1 Luzzati, V. (1968) in Biological Membranes (Chapman, D., ed.), pp. 71-123, Academic Press,
London
2 Lee, A. G. (1975) Progr. Biophys. Molec. Biol. 29, 3-56
3 Gaskell, D. R. (1973) Introduction to Metallurgical Thermodynamics, McGraw-Hill, New York
4 Reisman, A. (1970) Phase Equilibria, Academic Press, New York
5 Seltz, H. (1934) J. Am. Chem. Soc. 56, 307-311
6 Cullis, P. R. (1976) FEBS Lett. 70, 223-228
7 Pelton, A. D. and Thompson, W. T. (1076) Progress in Solid State Chemistry, 10, 119-155
8 Shimshick, E. J. and McConnell, H. M. (1973) Biochemistry 12, 2351-2360
9 Chapman, D., Urbina, J. and Keough, K. M. (1974) J. Biol. Chem. 249, 2512-2521
l0 Lee, A. G. (1975) Biochim. Biophys. Acta 413, 11-23
l l Blume, A. and Ackerman, T. (1974) FEBS Lett. 43, 71-74
12 Wu, S. H. and McConnell, H. M. (1975) Biochemistry 14, 847-854
13 Ververgaert, P. H. J. Th., Verkleij, A. J., Elbers, P. F. and van Deenen, L. L. M. (1973) Biochim.
Biophys. Acta 311, 320-329
14 Grant, C. W. M., Wu, S. H. and McConnell, H. M. (1974) Biochim. Biophys. Acta 363, 151-158
15 Phillips, M. C., Hauser, H. and Paltauf (1972) Chem. Phys. Lipids 8, 127-133
16 Lee, A. G. (1975) Biochemistry 14, 4397-4402
17 Michaelson, D. M., Horwitz, A. F. and Klein, M. P. (1973) Biochemistry 12, 2637-2645
18 Litman, B. J. (1974) Biochemistry 13, 2844-2848
19 Hitchcock, P. B., Mason, R., Thomas, K. M. and Shipley, G. (1974) Proc. Natl. Acad. Sci. U.S.
71, 3036-3040
20 Worcester, D. L. and Franks, N. P. (1976) J. Mol. Biol. 100, 359-378
21 Yeagle, P. L., Hutton, W. C., Huang, C. and Martin, R. B. (1975) Proc. Natl. Acad. Sci. U.S.
72, 3477-3481
22 Yeagle, P. L., Hutton, W. C., Huang, C. and Martin, R. B. (1976) Biochemistry, 15, 2121-2124
23 Hauser, H., Phillips, M. C., Levine, B. A. and Williams, R. J. P. (1976) Nature, 261, 390-394
24 Galla, H.-J. and Sackmann, E. (1975) Biochim. Biophys. Acta 401,509-529
25 Ito, T. and Ohnishi, S.-I. (1974) Biochim. Biophys. Acta 352, 29-37
26 Ohnishi, S. I. and Ito, T. (1973) Biochem. Biophys. Res. Commun. 51, 132-138
27 Ohnishi, S. I. and Ito, T. (1974) Biochemistry, 13, 881-887
28 Ito, T., Ohnishi, S. I., Ishinaga, M. and Kito, M. (1975) Biochemistry 14, 3064-3069
29 Papahadjopoulos, D., Poste, G., Schaeffer, B. E. and Vail, W. J. (1974) Biochim. Biophys. Acta
352, 10-28
30 Jacobson, K. and Papahadjopoulos, D. (1975) Biochemistry 14, 152-161
31 Van Dijck, P. W. M., Ververgaert, P. H. J. Th., Verkleij, A. J., van Deenen, L. L. M. and de
Gier, J. (1975) Biochim. Biophys. Acta 406, 465-478
32 Klopfenstein, W. E., de Kruyff, B., Verkleij, A. J., Demel, R. A. and van Deenen, L. L. M.
(1974) Chem. Phys. Lipids. 13, 215-222
33 Blume, A., Arnold, B. and Weltzien, H. U. (1976) FEBS Lett. 61, 199-202
34 Tinker, D. O., Pinteric, C., Hsia, J. C. and Rand, R. P. (1976) Can. J. Biochem. 54, 209-218
35 McCammon, J. A. and Deutch, J. M. (1975) J. Am. Chem. Soc. 97, 6675-6681
36 Lee, A. G. (1977) in Receptors and Recognition (Cuatrecasas, P. and Greaves, M. F., eds.),
Vol. 5A, Chapman and Hall, London, irt the press
37 Demel, R. A. and de Kruyff, B. (1976) Biochim. Biophys. Acta 457, 109-132
38 Lecuyer, H. and Dervichian, D. G. (1969) J. Mol. Biol. 45, 39-57
39 Lippert, J. L. and Peticolas, W. L. (1971) Proc. Natl. Acad. Sci. U.S. 68, 1572-1576

342
40
41
42
43
44
45
46
47
48
49

Hinz, H.-J. and Sturtevant, J. M. (1972) J. Biol. Chem. 247, 3697-3700


Engelman, D. M. and Rothman, J. E. (1972) J. Biol. Chem. 247, 3694-3697
Phillips, M. C. and Finer, E. G. (1974) Biochim. Biophys. Acta 356, 199-206
Lee, A. G. (1976) FEBS Lett. 62, 359-363
Newman, G. C. and Huang, C. H. (1975) Biochemistry 14, 3363-3370
Hui, S. W. and Parsons, D. F. (1975) Science 190, 383-384
Kraft, R. W. (1966) J. Metals 18, 192-202
Farber, T. E. (1958) Proc. Roy. Soc. A. 248, 460-481
Shimshick, E. J. and McConnell, H. M. (1973) Biochem. Biophys. Res. Commun. 53, 446-451
De Kruyff, B., van Dijck, P. W. M., Demel, R. A., Schuijff, A., Brants, F. and van Deenen,
L. L. M. (1974) Biochim. Biophys. Acta 356, 1-7
50 Rand, R. P., Pangborn, W. A., Purdon, A. D. and Tinker, D. O. (1975) Can. J. Biochem. 53,
189-195
51 Demel, R. A., Bruckdorfer, K. R. and van Deenen, L. L. M. (1972) Biochim. Biophys. Acta
255, 311-320
52 De Kruyff, B., Demel, R. A. and van Deenen, L. L. M. (1972) Biochim. Biophys. Acta 255,
331-347
53 Demel, R. A., Bruckdorfer, K. R. and van Deenen, L. L. M. (1972) Biochim. Biophys. Acta
255, 321-330
54 Janiak, M. J., Loomis, C. R., Shipley, G. G. and Small, D. M. (1974) J. Mol. Biol. 86, 325-339
55 Butler, K. W., Smith, I. C. P. and Schneider, H. (1970) Biochim. Biophys. Acta 219, 514-517
56 Green, J. R., Edwards, P. A. and Green, C. (1973) Biochem. J. 135, 63-71
57 De Kruyff, B., Demel, R. A., Slotboom, A. J., van Deenen, L. L. M. and Rosenthal, A. F.
(1973) Biochim. Biophys. Acta 307, 1-19
58 Oldfield, E. and Chapman, D. (1972) FEBS Lett. 21,303-306
59 Trauble, H. and Sackmann, E. (1972) J. Amer. Chem. Soc. 94, 4499-4510
60 Willmer, E. N. (1961) Biol. Rev. 36, 368-398
61 Connor, P., Mangat, B. S. and Rao, L. S. (1974) J. Pharm. Pharmacol. 26, 120P.
62 Lee, A. G. (1977) Biochim. Biophys. Acta 472, 237-281
63 Nicholls, P., West, J. and Bangham, A. D. (1974) Biochim. Biophys. Acta 363, 190-201
64 Podo, F. and Blasie, J. K. (1976) Biochim. Biophys. Acta 419, 1-18
65 Lee, A. G. (1976) Biochim. Biophys. Acta 448, 34-44
66 Eliasz, A. W., Chapman, D. and Ewing, D. F. (1976) Biochim. Biophys. Acta 448, 220-230
67 Lee, A. G. (1976) Biochemistry 15, 2448-2454
68 Hui, F. K. and Barton, P. G. (1973) Biochim. Biophys. Acta 296, 510-517
69 Schmidt, D., Steffen, H. and yon Planta, C. (1976) Biochim. Biophys. Acta 443, 1-9
70 Hauser, H., Finer, E. G. and Chapman, D. (1970) J. Mol. Biol. 53, 419-433
71 Lau, A. L. Y. and Chan, S. I. (1974) Biochemistry 13, 4942-4948
72 Packe, W. and Chapman, D. (1972) Biochim. Biophys. Acta 255, 348-357
73 Packe, W., Chapman, D. and Hillaby, R. (1972) Biochim. Biophys. Acta 255, 358-364
74 Haverman, E. (1972) Science, 177, 341-322
75 Mollay, C. (1976) FEBS Lett. 74, 65q58
76 Williams, J. C. and Bell, R. M. (1972) Biochim. Biophys. Acta 288, 255-262
77 Verma, S. P., Wallach, D. F. H. and Smith, I. C. P. (1974) Biochim. Biophys. Acta 345, 129-140
78 Sessa, G., Freer, J. H., Colaccio, G. and Weissman, G. (1969) J. Biol. Chem. 244, 3575-3582
79 Verma, S. P. and Wallach, D. F. H. (1976) Biochim. Biophys. Acta 426, 616-623
80 Novosad, Z., Knapp, T. D., Gotto, A. M., Pownall, H. J. and Morrisett, J. D. (1976) Biochemistry 15, 3176-3183
81 Warren, G. B., Toon, P. A., Birdsall, N. J. M., Lee, A. G. and Metcalfe, J. C. (1974) Biochemistry 13, 5501-5507
82 Warren, G. B., Toon, P. A., Birdsall, N. J. M., Lee, A. G. and Metcalfe, J. C. (1974) FEBS
Lett., 41, 122-124
83 Warren, G. B., Houslay, M. D., Metcalfe, J. C. and Birdsall, N. J. M. (1975) Nature 255,
684-687
84 Hesketh, T. R., Smith, G. A., Houslay, M. D., McGill, K. A., Birdsall, N. J. M., Metcalfe, J. C.
and Warren, G. B. (1976) Biochemistry 15, 4145-4151
85 Martonosi, A., Jilka, R. and Foster, F. (1974) in Membrane Proteins in Transport and

343
Phosphorylation (Azzone, G. E., Klingenberg, M. E., Quagliariello, E. and Siliprandi, N.,
eds.), pp. 113-123, North-Holland Publ. Co., Amsterdam
86 Grant, C. W. M. and McConnell, H. M. (1974) Proc, Natl. Acad. Sci. U.S. 71, 4653-4657
87 Warren, G. B., Toon, P. A., BirdsaU, N. J. M., Lee, A. G. and Metcalfe, J. C. (1974) Proc. Natl.
Acad. Sci. U.S. 71,622-626
88 Kleeman, W. and McConnell, H. M. (1976) Biochim. Biophys. Acta 419, 206-222
89 Marcelja, S. (1976) Biochim. Biophys. Acta 455, 1-7
90 Kimelberg, H. K. and Papahadjopoulos, D. (1974) J. Biol. Chem. 249, 1071-1080
91 Dufourcq, J., Faucon, J. P'., Lussan, C. and Bernon, R. (1975) FEBS Lett., 57, 112-116
92 Nakamura, H., Jilka, R. L., Boland, R. and Martonosi, A. N. (1976) J. Biol. Chem. 251, 54145423
93 Martonosi, A. N. (1975) in Calcium Transport in Contraction and Secretion (Carafoli, E., ed,),
pp. 313-328, North-Holland Publ. Co., Amsterdam
94 Wilschut, J. C., Regts, J. and Scherphof, G. (1976) FEBS Lett., 63,328-332
95 Papahadjopoulos, D., Vail, W. J. and Moscarello, M. (1975) J. Memb. Biol. 22, 143-164
96 Papahadjopoulos, D., MoscareUo, M., Eylar, E. H. and Isac, T. (1975) Biochim. Biophys.
Acta 401,317-335
97 Boggs, J. M., Vail, W. J. and Moscarello, M. A. (1976) Biochim. Biophys. Acta 448, 517-520
98 Jost, P. C., Griffith, O. H., Capaldi, R. A. and Vanderkooi, G. (1973) Proc. Natl. Acad. Sci.
U.S. 70, 480-484
99 Dehlinger, P. J., Jost, P. C. and Griffith, O. H. (1974) Proc. Natl. Acad. Sci+ U.S. 71, 2280-2284
100 Brown, M. F., Miljanich, G. P. and Dratz, E. A. (1977) Biochemistry 16, 2640-2648
101 Brown, M. F., Miljanich, G. P., Franklin, L. K. and Dratz, E, A. (1976) FEBS Lett. 70, 56-60
102 Hong, K. and Hubbell, W. L. (1972) Proc. Natl. Acad. Sci. U.S. 69, 2617-2621
103 Verkleij, A. J., de Kruyff, B., Ververgaert, P. H, J. Th., Tocanne, J. F. and van Deenen, L. L. M.
(1974) Biochim. Biophys. Acta 339, 432-437
104 Eylar, E. H. (1970) Proc. Natl. Acad. Sci. U.S. 67, 1425-1431
105 London, Y., Demel, R. A., Geurtz van Kessel, W. S. M., Vossenberg, F. G. A. and van Deenen,
L. L. M. (1973) Biochim. Biophys. Acta 311, 520-530
106 Nicholls, P. (1974) Biochim. Biophys. Acta 346, 261-310
107 Van, S. P. and Griffith, O. H. (1975) J. Memb. Biol. 20, 155-170
108 Birrell, G. B. and GriMth, O. H. (1976) Biochemistry 15, 2925-2929
109 Nelsestuen, G. L. (1976) J. Biol. Chem, 251, 5648-5656
110 Cerbon, J. (1972) Biochim. Biophys. Acta 290, 51-57
111 Hauser, H., Penkett, S. A. and Chapman, D. (1969) Biochim. Biophys. Acta 183, 466-475
112 Fernandez, M. S. and Cerbon, S. (1973) Biochim. Biophys. Acta 298, 8-14
113 Seydel, J. K. and Wassermann, O. (1976) Biochem. Pharmacol. 25, 2357-2364
114 Crompton, M., Bamtt, G. J., Bradbury, J. H. and Bygrave, F. L. (1976) Biochem. Pharmacol.
25, 2461-2464
115 Abood, L. G. and Hoss, W. (1975) Eur. J. Pharmacol. 32, 66-75
116 Loh, H. H., Cho, T. M., Wu, Y.-C. and Way, E. L. (1974) Life Sci. 14, 2231-2245
117 Cho, T. M., Cho, J. S. and Loh, H. H. (1976) Life Sci. 18, 231-244
118 Cater, B. R., Chapman, D., Hawes, S. M. and SaviUe, J. (1974) Biochim. Biophys. Acta 363,
54-69
119 Ochoa, E. L. M. and Bangham, A. D. (1976) J. Neurochem. 26, 1193-1198
120 Krishnan, K. S. and Balaram, P. (1976) FEBS Lett. 63, 313-315
121 Lyman, G. H., Papahadjopoulos, D. and Preisler, H. D. (1976) Biochim. Biophys. Acta 448,
460-473
122 Hill, M. W. (1974) Biochim. Biophys. Acta 356, 117-124
123 Jain, M. K., Wu, N. Y.-M. and Wray, L. V. (1975) Nature 255, 494-496
124 Lee, A. G. (1977) Biochemistry, 16, 835-841
125 Jain, M. K., Wu, N. Y.-M., Mbrgan, T. K., Briggs, M. S. and Murray, R. K. (1976) Chem.
Phys. Lipids 17, 71-78
126 McLaughlin, S. M. and Haray, H. (1976) Biochemistry 15, 1941-1948
127 Papahadjopoulos, D., Jacobson, K., Poste, G. and Shepherd, G. (1975) Biochim. Biophys.
Acta 394, 504-519
128 Lee, A. G. (1976~ Nature, 262, 545-548

344
129
130
131
132

Lee, A. G. (1976) Biochim. Biophys. Acta, 455, 102-108


Lee, A. G. (1977) Molec. Pharm. 13, 474-487
Stier, A. and Sackmann, E. (1973) Biochim. Biophys. Acta 311,400-408
Blaurock, A. E. and Stoeckenius, W. (1971) Nat. New Biol. 233, 152-156

You might also like