Vous êtes sur la page 1sur 53

CHAPTER 6

Physical principles and formalisms


of electrical excitability
A

*I

A
A

Departments of Physiology and Neurology, Albert Einstein


College of Medicine, New York, New YoFk
Department of Biophysics, The Rockefeller University,
New York, New York

CHAPTER CONTENTS
Membrane Potentials and Ionic Fluxes
Fundamental concepts
Equilibrium state
Nonequilibrium state
Summary
Two important, examples of equilibrium situations
The Donnan equilibrium
Phase-boundary potentials
Electrodes - the measurement of potential differences
Quasi-equilibrium systems
A membrane with a very large fixed-charge density
An oil membrane
Ion transport (the Nernst-Planck flux equations)
Homogeneous uncharged membrane
Homogeneous membranes with special properties
Mosaic membranes
Formal Consequences of Voltage-dependent Conductances
The nature of electrical excitability
Reasons for believing that electrical excitability does not
result from the shifting of ionic profiles
Hodgkin-Huxley equivalent circuit
Current-voltage (I-V) characteristics
Negative-slope conductance
Changing the I-V characteristic without change of the g-V
characteristic
Voltage-dependent Conductance in Thin Lipid Membranes
The unmodified thin lipid membrane
Formation
Permeability and electrical properties
Nonvoltage-dependent modifiers
Carriers
Channel formers
A mosaic membrane formed with two modifiers
Summary
Voltage-dependent modifiers
Monazomycin
Alamethicin
Excitability-inducing material
Single channels
Summary and conclusion
_ _ _ _ ~
IN THIS CHAPTER we intend to elucidate the presently
understood physicochemical principles and formalisms underlying the electrical excitability of biological membranes. The primary analysis is of a model

system, the modified thin lipid (or bilayer) membrane, which illustrates most of the relevant phenomena associated with nerve excitation. Before discussing this model, however, we develop more or less
from first principles the concepts of membrane potentials and ionic fluxes. This forms a rather large part
of the article and may be superfluous for the more
sophisticated reader. Nevertheless we have included
this material because, despite its importance for understanding nerve excitation, it is rather inaccessible
to students and investigators attempting to make
initial contact with the neurophysiological literature.
In the second section we discuss some of the formal
aspects of the behavior of systems containing elements whose conductances are profoundly affected by
the voltage across them. With this as background, we
then discuss the fascinating voltage-dependent phenomena that can arise in suitably doped thin lipid
membranes.
MEMBRANE POTENTIALS AND IONIC FLUXES

Fundamental Concepts

Here we consider two general situations: equilibrium and nonequilibrium states. We analyze the
equilibrium state from both the thermodynamic and
the statistical mechanical viewpoint; the nonequilibrium state is handled by the Nernst-Planck flux
equations.
STATE.
Thermodynamic approach. The
fundamental relation that we need from thermodynamics is that at thermal equilibrium the electrochemical potential, pi,of any species i is the same in
all phases to which the species has access. Thus, for
phases 1 and 2 we can write

EQUILIBRIUM

pi (1) = pi (2)

(1)

provided that species i can move between the two


phases. Equation 1 is the starting point of all our
161

162

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

thermodynamic treatments. For an ideal solution, p,


is given by
p, = p,

+ RT

In X , + P V , + z, F $
relevant work terms

For example, consider particles in a gravitational


field. Then Equation 3 becomes

+ any other

c,,e-lllvs:k?

(4)

(2)

where m is the mass of the particle and g is the


acceleration due to gravity. If there were no thermal
energy, all the particles would sit a t x = 0,the point
of minimum potential energy. On the other hand if
there were no gravitational field and only thermal
energy, the particles would be homogeneously distributed throughout space. Equation 4 is the compromise when both terms operate.
It is important to realize that Equation 3a is equivalent to Equation 1. (Take the logarithm of both sides
of Equation 3a and identify W, as all the terms on the
right in Equation 2a except RT In ci.) Thus the thermodynamic statement that the electrochemical potential of a species is the same in all phases is equivap, = p,+ RT I n c, + P V , + z, F 4 + . . . (2a)
lent to the statistical mechanical statement that the
where c, is the concentration of i. ( p l ( 0in
) Eq. 2a is molecules of the species satisfy the Boltzmann distridifferent from p,in Eq. 2 because the standard state bution.
must be newly defined in going from mole fraction
units to concentration units.)
As indicated in Equation 2, other work terms must NONEQUILIBRIUM STATE. The bases of our discussion of
be included if they contribute to the potential energy ion transport are the Nernst-Planck flux equations
of the species i. For example, if the system is in a
gravitational field (or ultracentrifuge), there will be a
term for the gravitational potential energy. For the
systems we consider, these terms do not arise. In fact,
even the PV term will generally be trivial and therefore not enter into our treatments. Thus, for our
where $,+and $k- represent the flux (in mol/s) of the
purposes, p, can be written as
jth and kth ion, respectively, across unit area a t any
p, = p,(+ RT In c, + z,F $
(2b) point x in the system, u , is the molar mobility of the
Statistical mechanical approach. The basic rela- jth cation, uk is the molar mobility of the kth
tion that we need from statistical mechanics is that at anion, c,+ and ck- represent the concentration of the
thermal equilibrium the particles satisfy the Boltz- jth and kth ions, respectively, a t any point I , and $
mann distribution a t all points in the system to which is the electrical potential at any point x . (Again we
they have access. That is, a t any point x (considering are considering a one-dimensional situation with gradients of concentration and electrical potential occuronly a one-dimensional situation)
ring only in the x direction.)
c , ( x ) = C,,e-%(x)/h7
(3)
Whereas our basic equilibrium equations (Eqs. 1
and
3) have a solid foundation in thermodynamics
where, w ,(XI is the potential energy per particle of
and
statistical
mechanics, the flux equations are less
the ith species a t x , c,, is its concentration a t the point
firmly
grounded
in theory, and in some instances
defined as zero potential energy, and k is the Boltzthey
are
even
grossly
incorrect. We therefore wish to
mann constant. Multiplying the numerator and demake
a
few
points
concerning
them that will give the
nominator of the exponent by N A ,Avogadros numreader
some
feel
for
their
meaning.
ber, we can write Equation 3 in the form
Equations 5 can be written in the form
c,(x) = c,,e-,(s)lRT
(3a)

where p, is the standard chemical potential (in the


particular phase being considered) of the ith species,
X , is mole fraction of ith species, V , is partial molar
volume of ith species, z, is valence of ith species (0,
k l , +2, . . .), P is hydrostatic pressure of the phase, 4
is electrostatic potential of the phase, R is the gas
constant, T is temperature in degrees Kelvin, and F
is the Faraday. (If the solution is nonideal, an activity coefficient, y,, is included with the mole fraction
term. In this article we deal only with ideal solutions.) If species i is dilute, then Equation 2 can be
rewritten as

where W, is the potential energy per mole of the ith


species. If W, is purely electrostatic energy, then
where
Equation 3a becomes
c,(x) =

~ ~ e F-W 2
X ) l~R T

D = uRT

(3a)

The Boltzmann distribution iS a quantitative state-

(7)

some
theoretical problems associated with justifying the use

merit of the balanced but incessant competition be- of the Boltzmann distribution in electrolyte theory do not concern

tween potential energy ( w )and thermal energy ( k T ) . us (67).

CHAPTER

For a nonelectrolyte (z
comes

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

O), Equation 6 simply be-

4 = -D-dc

dx

which is Ficks first law of diffusion. [Equation 7 is


the famous Nernst-Einstein relation between the diffusion constant and the frictional coefficient (13, 14);
the mobility (u)is simply the reciprocal of the frictional coefficient.] Thus the flux of a species is proportional to the negative gradient of its concentration.
On the other hand for an ion at uniform concentration throughout the system (dc/dx = 01, Equation 6
becomes

which is simply the equation for electrophoresis.


That is, the ionic flux is proportional to the electric
field (-d+/dx). Thus Equation 6 says that if there is
both a concentration gradient and an electric field,
the ionic flux is a linear sum of the fluxes that would
arise from each effect alone.
Another way of looking at Equations 5 is to say
that
flux

concentration x velocity

and
velocity

mobility x driving force,2

so that combining we have

flux

mobility x concentration x driving forc:e


4 = u x c x

R T dc
ZF

9)
dx

(5)

The flux equation states that the force on an ion is the


sum of two terms: -RTlc dcidx and -zF d$ldx. The
second term is the electrical force familiar from elementary electrostatic theory, but the first term is
more subtle. It is a phenomenological force resulting
from the random (Brownian) motion of each individual ion. Though the random movements of each ion
are equally likely to be in the positive or negative
direction, statistically it appears as if there is a force
operating in the direction of the concentration gradient (4 1).
Equation 5 is a special case of the more general
expression

which states that the driving force on the ith species


is the negative gradient of its chemical potential.

Substituting Equation 2b, the thermodynamic


expression for the electrochemical potential, into
Equation 8 we obtain Equation 5, the Nernst-Planck
flux equations (if p,( is constant). Note that Equation 8 reduces to the thermodynamic equilibrium
condition, Equation 1, when 4i = 0 at all points; or
Equation 5 reduces to the BoItzmann distribution,
Equation 3ar, when Cbi = 0. [The relationship of the
Boltzmann distribution (equilibrium state) to the
flux equations (nonequilibrium state) is clearly seen
upon differentiating Equation 3a
R T dci
CI

dx

d%
ziF - = 0

du

(which is the same relation as obtained by setting di


= 0 in Equation 5). Thus a t equilibrium the sum of
the electrical force ( - z,F d$ldx) and the diffusional
force (- RT/ci dc,/dx)acting on an ion is zero a t
every point in the system, and hence there is no flux
of matter. When these two forces do not balance,
there is a net driving force on the ion and hence a
flux. I
For our purposes the Nernst-Planck flux equations
(Eqs. 5) are perfectly adequate. Nevertheless the
reader should be aware that there exist situations for
which these equations are insufficient. In particular,
Equations 5 (or more explicitly Eq. 8) say that the
only driving force on a species i is the negative gradient of chemical potential for that species alone; it
neglects the coupling of the gradients of chemical
potentials for other species j to the flux of species i. (A
familiar example of such coupling is solvent drag,
where the gradient of the chemical potential of water
gives rise not only to a flow of water, but also to a flow
of solute dissolved in the water.) Though these couplings, expressed as cross coefficients to forces acting
on other species, are extensively used by those who
deal with the formalism of irreversible thermodynamics (36), they do not concern us here. For the
systems we consider (which are of neurophysiological
interest), they introduce corrections that are a t best
second order.
We consider dilute, ideal solutions of ions
in which gradients of concentration, potential, or
solvent composition occur in only the x d i r e ~ t i o n . ~
The starting point for the treatment of these systems
is one of the following equations
SUMMARY.

4.= I

In a condensed phase such as water, an ion subjected to a


force very quickly accelerates to its terminal velocity. In practice,
therefore (unless we are dealing with very high frequencies),
force gives rise to velocity rather than to acceleration.

163

dc,
d%
uiRT - - ziuiFci- (5)

dx

dx

Transport

For the general case, d/dz is replaced by C in all equations.

164

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I


membrane

CI-

FIG.

This is known as the Gibbs-Donnan condition, and r


is called the Donnan ratio.
From Equations 10a or lob we can directly obtain
the potential difference across the membrane
v = ($,

1. The Donnan system.

Two Important Examples of Equilibrium Situations


Aside from its intrinsic
physiological interest, the Donnan equilibrium illustrates almost all the important concepts and difficulties associated with membrane potentials; we therefore analyze it in some detail. The Donnan equilibrium arises when a membrane separates two solutions containing both permeant and impermeant
ions. For simplicity we restrict our treatment to the
case of an infinitesimally thin membrane separating
two infinite solutions; both solutions contain a single
permeant univalent positive (Na+)and negative (Cl-)
ion species, and, in addition, one solution contains a
univalent impermeant ion, zN, where z = + 1 (Fig.
1). (The mechanism of membrane semipermeability
is irrelevant to our treatment. In practice, the Donnan situation commonly arises when a porous membrane, such as dialysis tubing, holds back a macroion (e.g., protein) which cannot fit through the pores.
Usually the solvent, water, is also permeant.)
Thermodynamic analysis. The conditions that
must be satisfied are, from Equation 1

- J J ~ )=

RT
a+],
In
F
a+],
~

- -

THE DONNAN EQUILIBRIUM.

( l )=

/-%.I

(2)

~ N . I

(9a)

(9b)
which become, upon substituting from Equation 2b
k

&,+ + RT In a+],
=

( 1 ) = P(I (2)

-t F$,

&;;+ + RT In [Na+In+ F I / J ~

(10a)

Note that there is no such equation for N, since we


have specified that it cannot move between the two
aqueous phases. (We have excluded a third equation
equating the chemical potential of water on the two
sides of the membrane. Although the osmotic effect
accompanying the Dorinan equilibrium is of considerable general physiological interest, it is not significant for electrophysiology.) Adding Equation 10a and
10b we obtain

[Cl-ll
RT
In =
[Cl-]?
F
-~

RT
In r
F
-~

(13)

To calculate r, and hence the membrane potential,


we invoke the familiar electroneutrality condition
on both sides of the membrane
[Na+I1= [Cl-1, = c I

(14a)

(electroneutrality condition)
a+],

+ z[N] = LCl-1,

(14b)

(It is to be understood that the electroneutrality condition holds only for remote regions on either side of
the membrane, that is, for macroscopic regions. As
will be seen presently, the alternative treatment by
means of the Poisson-Boltzmann equation establishes
the concentrations and potential as continuous functions ofx, and indeed regions very close to the membrane are not electrically neutral .) Combining Equations 14 and 12 we obtain

Figure 2 is a sketch of r as a function of the impermeant ion (N) concentration for z = +1. As [N+]
increases, r decreases from 1; that is, permeant positive ion concentration decreases on side 2 and permeant anion concentration increases. In the limit of
large +I, there is virtually no Na+ on side 2 and
[Cl-1, = +I. The converse occurs for large -1.
Note also from Equation 13 that the membrane potential, \Ir, will be positive or negative depending on
whether z is + 1 or - 1 and that the absolute value of
\I increases as [N] increases. Figure 3 is a diagram of

or
[Na+;l, - [Cl-1, = r
a+:[,
[Cl-1,
~

FIG. 2. Qualitative plots of the Donnan ratio, r, as a function


of the impermeant ion concentration, IN], for an impermeant ion
of valence +1 or - 1 . (See Eq. 1 5 . )

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

165

and Equation 17 then becomes


dZ$ - 8rrFaCll
dxl
E

(sinh-

RT

2aClI

which, for obvious reasons, is called the PoissonBoltzmann equation. It is convenient to express in
units of RT/F (at room temperature, RT/F = 25 mV)
and write Equation 20 in the form

FIG. 3. Concentrations and potentials (both plotted on ordinate) in compartments 1 and 2 for a Donnan equilibrium.

d2y
~

?.d
the concentrations and potentials in compartments 1
and 2 for a particular concentration of N+.
The thermodynamic analysis, while perfectly correct and providing all the right answers, leaves a
very important question unanswered; namely, what
produces the membrane potential? It is intuitively
clear from Figure 3 that such a potential (of the
polarity shown) must exist to explain why Na+ and
C1- do not diffuse down their respective concentration
gradients across the membrane, but where are the
space-charge regions that generate this potential? To
answer this question, we must turn to the statistical
mechanical treatment of the Donnan equilibrium.
Statistical mechanical analysis (Poisson-Boltzmann equation). Consider again the situation depicted in Figure 1. To simplify the treatment we
assume that N is uniformly distributed on side 2 and
is immobile. (This does not significantly change the
physics; the real system that corresponds to this
model is an extended ion-exchange resin, where a
homogeneous distribution of immobile ions is
achieved with ionized groups covalently bound to a
polymer matrix.) Invoking the Boltzmann distribution for the permeant ions and assuming that the
potential energy term, W,, is purely electrostatic, we
have from Equation 3a (arbitrarily choosing remote
regions to the left in solution 1 as the zero of potential)
[Cl-1 = lCl-]-,ep*
a+]

a+] ,e-FJ

1
~

L,)Z

(sinhy -

z3
2[NaClI

(21)

where
(22)

(23)

L,, is the Debye length of the solution.


Equation 21, the Poisson-Boltzmann equation, can
now be solved, with the appropriate boundary conditions, for y (i.e, $) as a function of x , and then
Equations 16 give [CI-] and a+] as functions of x.
The details of the solution (43) are not important for
our considerations, but the nature of the solution is
crucial to our understanding of membrane potentials.
The concentration and potential profiles are sketched
in Figure 4 for the case of z = + 1; also shown is the
space-charge distribution, p . Let us consider the implications of this figure.
First, it is interesting to see how the Donnan relations for concentration and potential -already obtained above by the thermodynamic approach -come
out of the Poisson-Boltzmann treatment. For remote
regions we have from Equations 16a and 16b

(16a)
(16b)

where [Cl-l-, and a+]-% refer to concentrations (in


mol/cc) a t remote regions to the left where the potential is zero. (Note that [Cl-l-, = a+]. ,= [NaCll.)
From Coulombs law, the electrostatic potential, $,
must also satisfy Poissons equation

a5/J - -4rrp
dx
E

x=o
P

(17)

where E is the dielectric constant of the medium, and


p is the charge density given by

p = F(lNa+l - [Cl-I

+ zlN1)

(18)

Substituting Equation 16 into Equation 18 we have


p =

FLzNl - 2FLNaCll sinh -F+


RT

(19)

FIG. 4. Sketch of concentration and potential profiles (both


plotted on ordinate) for a Donnan equilibrium as determined from
the Poisson-Boltzmann analysis (cf. Fig. 3, which is the result of
the thermodynamic analysis). Note that at any point x the electrical force ( + F d+/& for cations and -F d$/& for anions) and the
phenomenological diffusion force (-RT/INa+l dlNa+]/& and
-RT/[C1 1 dlCl I/&) balance. The space-charge density is shown
in the lower part of the figure.

166

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

and
[Cl-]+,

[Cl~]~,e+~YH

FIG.

which is the Donnan potential as given in Equation


13. Furthermore by equating the two expressions for
q we see that
lCI-I+,lNal,

[Cl-lL,[Na+l-,

which is the Gibbs-Donnan condition as in Equation


2. (Note that bY
Equations 16a and 16b
we have the more general statement
lC1-l[Natl

[C-J-,[Na+]_,

constant

at every point x . )
Second, we now finally see the specific charge separation that gives rise to the electrostatic potential.
There exist narrow space-charge regions on the two
sides of the membrane; in solution 1 the space charge
is negative and in solution 2 it is positive, with of
course

.r:

& =

- J1:pbr

(24)

The potential, $, and the ion concentrations, a+]


and [Cl-I, vary continuously in these regions to their
final values in the remote regions. A comparison of
Figure 3 with Figure 4 clearly illustrates the differences between the thermodynamic and the statistical
mechanical point of view. In the former case, the $
and concentration functions are constant a t their remote values on the two sides of the membrane, with
a discontinuity in these values occurring at the plane
of the membrane; in the latter case these functions
are seen to vary continuously from one remote region
to the other.
The extent of the space-charge regions is determined by the Debye length, L,)(Eq. 23). Roughly
speaking, p falls about e-fold for every Debye length.
For 0.1 M salt solution in water ( E = 801, L,, = 10 A;
thus the region where electroneutrality is significantly violated extends about 40-50 A, too small for
direct sampling with a pipette. Note from Equation
23 that L,, is directly proportional to 4 2 and inversely
Thus in low dielectric conproportional to f l a a
stant media, L,, will tend to be smaller, whereas in
media of low ionic strength, L,, will be larger. We
shall refer back to this point later when discussing
ion distributions in a lipid, or hydrocarbon, phase.
Finally it should be realized that the system is a
typically nonlinear one, which makes it difficult to
describe how the J, and concentration functions are

5. The two-phase system.

generated from a given initial condition. Thus, although the impermeability of N is the ultimate cause
for the membrane potential and ion asymmetries,4
the $ and concentration functions are inexorably
coupled. That is, the $ profile affects the concentration profiles which in turn affect the J, profile, and
so on.
We have dealt in some detail with the Donnan
equilibrium because the general nature of the results
applies to many other equilibrium systems, one of
which we shall discuss shortly. Basically the thermodynamic approach gives the-concentrations and potentials far from the membrane, or interface,
whereas the Poisson-Boltzmann treatment explicitly
describes how the membrane potential arises from
space-charge regions near the membrane; electroneutrality holds only a t remote (many Debye lengths)
regions.
In the Donnan equilibrium, the asymmetry of permeable-ion distributions
(the Donnan condition) and the membrane potential arise because a macro-ion in one of the aqueous
solutions is impermeant. Here we consider the case in
which no impermeant ion is present, but the two
phases are different (for simplicity we consider them
immiscible); for example, one phase is water and the
other is oil (Fig. 5).
Thermodynamic analysis. From Equations 1and 2b
we have
pi;;,+,,+ RT In a+], + F$,
= pl&t)2
+ RT In [Nail2 + F$2 (25a)
PHASE-BOUNDARY POTENTIALS.

+ RT

In [Cl-1, - F$,
= p;:!)l-)s
+ RT In [ClV],

F$2

(25b)

These are the same equations employed for the Donnan equilibrium (Eqs. 10a and lob); this time, however, the p(s are not the same on sides l and 2,
because the solvents are different. The conditions of
electroneutrality for remote regions also give
[Cl-I, = [NaCIl,

(264

[Na+12= [Cl-I, = [NaCll,

(26b)

a+],

Substituting Equation 26 into Equation 25 and adding we obtain


For the true Donnan case, where N is not fixed at a constant
value to the right of the membrane, we would have an Equation
16c, stating t h at N satisfies the Boltzmann distribution for x > 0.
The final result is not much different from t h at shown in Fig. 4,
except t h at N is perturbed upward near the membrane.

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

167

tively. That is, they are the partition coefficients that


would be observed in remote regions i f somehow electrostatic interaction among ions did not occur. Substituting Equation 30 into Equation 27, we have
po/w =

d(PNa+)(PCI-)

(31)

and substituting Equation 30 into Equation 28 gives


Upon substituting Equation 26 into Equation 25 and
subtracting we also obtain
*O/\V

=_ ( $ 2

-,

(32)

$1)

Equation 31 states that the macroscopically observed


partition coefficient for NaCl is the geometric
2F
mean of the individual partition coefficients @) for
Equation 27 is an expression for the oil-water (o/w) Na+ and C1-. Note that if one of these approaches
partition coefficient of NaCl in terms of the standard zero, the partition coefficient for the salt approaches
chemical potentials of Na+ and C1- in the two phases. zero. Equation 32 states that the phase-boundary
Equation 28 gives the phase-boundary potential in potential is determined by the ratio of the individual
terms of these same quantities.
partition coefficients. Note that if these are equal,
It is instructive to transform these expressions into the phase-boundary potential is zero. Also note that
ones containing the intrinsic partition coefficient of in contrast to the Donnan potential, the phase-boundeach ion. At the boundary between the two phases ary potential is not a function of the NaCl concentrathere is a discontinuity in standard chemical poten- tion (provided, of course, the partition coefficients are
tial and hence a discontinuity in ion concentrations. not concentration dependent).
Of course, in (classical) reality there are never actual
Statistical mechanical analysis. Had we stopped
discontinuities; all functions are continuous. Never- our thermodynamic analysis with Equations 27 and
theless, because the boundary between two phases is 28, we would have been in much the same position we
established through short-range van der Waals inter- were in with our thermodynamic treatment of the
actions, the phase transition occurs over molecular Donnan equilibrium; the concentrations of NaCl are
distances (-2 A) and hence can be practically treated different in the two phases, and there is a potential
as a discontinuity, compared to the space-charge re- difference between the two phases (Fig. 6). Again the
gions that can extend over tens or even hundreds of formal expressions are perfectly correct, and again
and hence the ion concen- the origin of the boundary potential is obscure. By
angstroms. Thus the p*.))s,
trations, can change precipitously over a distance extending our thermodynamic analysis to the partiwhere the electrostatic potential, $, remains un- tioning occurring a t the boundary, we have given a
changed.s With this in mind, and assuming that fairly strong hint as to the source of the boundary
there is not a layer of dipoles a t the interface, we can potential. Clearly, applying the Poisson-Boltzmann
write for the conditions at the boundary
analysis with these boundary conditions will show
that space-charge regions exist in both the aqueous
and oil phases. Figure 6 will then be transformed into
the more complete Figure 7. We shall not go through
the formal treatment since, aside from the mathematical details, there are no new physical principles
that have not already been considered for the Donnan
case. The extent of the space-charge regions in the
two phases will depend on their respective Debye
-

(dk+,]
- P$L)J -

(PL:%+)y- PII.l-),)

(28)

Equations 30a and 30b are expressions for the intrinsic partition coefficients G.1 of Na+ and C1-, respec-

ql=O

a Image forces, which make a major contribution to the intrinsic partition coefficient, extend over distances longer than the
phase-transition region, but these are still in general much
shorter than the space-charge regions.

FIG. 6. Concentrations and potentials in the water and oil


phases. The potential, $2, in the oil phase is positive, because we
have assumed that the partition coefficient for sodium, pNa+,
between oil and water is greater than that for chloride, p c , - .

I
I

168

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

Jm,

lengths. Since LD =
there tends to be a
compensating effect between dielectric constant and
salt concentration. Thus a low dielectric constant by
itself would lead to a smaller Debye length, but in
general a low dielectric constant is accompanied by a
small partition coefficient of the salt between water
and the low dielectric constant phase, which leads to
a larger Debye length. In the cases we consider later
of membranes with an essentially hydrocarbon interior, the concentration term strongly predominates,
so that the space-charge region in the membrane
interior is much more extended than the one in the
aqueous phase. Of course, regardless of the extent of
the space-charge regions, overall charge conservation
must always obtain

Solution 1

Solution 2

FIG. 8. Method for measuring the potential difference across a


membrane.

cerned not with the technical questions of which amplifiers to use or what brand of oscilloscope is best,
but rather with an important theoretical question
(which also happens to have important practical implications). The basic problem is the following: in
order to measure the potential difference across a
membrane, we must insert a pair of electrodes into
(24) the system - one electrode on each side of the mempdx = pdx
brane (Fig. 8). By necessity there will exist at each
We might also note that a truly continuous treat- solution-electrode (soln/elec) interface a potential,
ment would not show the discontinuities in ion con- generally called an electrode potential. Thus the pocentrations at x = 0, as is depicted in Figure 7, but tential that we measure, Y,,,,,~i5u,.,,c,,
is in principle the
rather would show a continuous transition over a algebraic sum of three potentials: the membrane podistance of a few angstroms. Such a treatment would tential (the quantity of interest) plus two electrode
require a "van der Waals, image force, . . . -Boltz- potentials
mann" analysis of this region, in analogy to the Pois(33)
q'measured - Vrrnernbrane 4- qeleclsoln I + q s o i n 2leier
son-Boltzmann treatment. Needless to say, the much
more complex nature of these forces makes such an How then do we make contact with the solutions so
analysis extraordinarily difficult (if not impossible). that the sum of the last two terms in Equation 33 is
negligible? This is crucial, for when we go to measure
a
membrane potential, we want to measure a quanElectrodes - the Measurement
tity that is a unique property of the ionic system and
of Potential Difference
not a quantity that is dependent on the particular
Up to this point in our discussion of potentials pair of electrodes we happen to choose.
associated with ionic systems, we have not described
To illustrate more concretely the problem of meashow one goes about measuring these potentials. Be- urement, consider again the Donnan system of Figfore considering other examples of membrane poten- ure 1. With two theoretical formulations we have
tials, we must discuss this problem. We are con- shown that

1:

il',

*",,.ml,r;,nC.

, q
Notl
I

*-----

$/'

*O/

RT
[Cl-1,
-- In ~F
[Cl-1,

(13)

and the question now is, can we measure it? Suppose


we use reversible Ag-AgC1 electrodes. [The electrode
reaction is AgCl + e + Ag")' + C1- (soln).] What then
is ~ , , , ( . ; l S U l . ( . t , ? The answer turns out to be zero. Let us
see why. The "electrode potential" of a Ag-AgC1 electrode in contact with a solution containing C1- is

where qois the so-called standard potential. (In this


case it is the potential of the half cell when the
solution is one molar in chloride.9 If we keep our sign
FIG. 7. Sketch of concentration and potential profiles for a
phase-boundary equilibrium a s determined from the PoissonBoltzmann analysis (cf. Fig. 6, which is the result of the thermodynamic analysis; a s in Fig. 6, P , ~ . > pci-). The space-charge
density is shown in the lower part of the figure.

It is interesting to note that Eq. 34 can be derived by exactly


the Same methods we have employed in treating our pure ionic
systems. Thus, since the system is in equilibrium and C 1 ~can
move between the two phases (solid and solution), we have from
our thermodynamic relation

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

convention straight we then have

(*(,

RT

- pIn [Cl-I,)

r
*i*Ivr/w1n

The electrode potentials exactly cancel the membrane potential, and operationally we measure no
potential difference at all. We certainly made a poor
choice of electrodes!
We could have predicted this result purely from the
second law of thermodynamics without going
through the above algebra. Since each electrode is in
equilibrium with its solution and the solutions are in
equilibrium (Donnan) with each other, the entire
system must be in equilibrium. If there indeed were a
potential difference between the electrodes, we could
construct a perpetual motion machine of the second
kind. That is, we could connect a load between the
two electrodes and do work. At one electrode, C1would go into the solution, and at the other electrode,
C1- would come out; the overall chemical composition
of the solutions would not change. If the Donnan
condition became perturbed by the transport of NaCl
from one solution to the other, we could merely pause
for a while and let the Donnan condition reestablish
itself (utilizing, of course, only thermal energy).
When one electrode becomes almost depleted of AgC1,
we merely switch the electrodes from one solution to
the other, a process that, in principle, requires negligible work. Thus we could indefinitely convert thermal energy into work without any other change in
the universe -a clear violation of the second law of
thermodynamics.
Since a reversible pair of electrodes will always
give
= 0, we must try something different.
Why not use a pair of stainless-steel wires? We might
indeed measure the correct Donnan potential, but
then again we might not. The problem is that there is
not a well-defined process dominating the potential of
or

kI

(solid) = k 1(soh)

169

the steel wires. What is the dependence of this potential on Na+ and C1- (or in the more general case of the
Donnan equilibrium, on any other permeant ions
present in the system) concentration? What effect
does the macro-ion N have on this potential? It is
possible that none of these have significant effect on
the electrode potential, and therefore the two electrode potentials will be equal and cancel each other
out, leaving the Donnan potential as the only quantity measured. But we cannot be sure, since we have
no theory to work from.
It turns out that we can measure the membrane
potential by introducing appropriate salt bridges. Instead of putting the Ag-AgC1 electrodes directly into
the solutions, we place them into 3 M (or saturated)
KC1 and make contact to the solutions through the 3
M KCl (Fig. 9). Now a t first glance it might appear
that things are even worse, because the measured
potential is now the sum of five potentials instead of
three:

In addition to the two electrode potentials, there are


now two liquid junction potentials -one between 3 M
KCl and solution 1 and one between 3 M KCl and
solution 2. On reflection, however, it is clear that
things are not worse than before, for since the AgAgCl electrodes are in identical solutions (i.e., 3 M
KCl), the electrode potentials must be equal and
hence cancel. This then leaves the two liquid junction
potentials with 3 M KC1, and it turns out that these
are small, both because K+ and C1- have the same
mobility and because their concentration is large
(40). Given a salt bridge with a large concentration of
a salt whose ions have equal mobilities, it can be
shown that the liquid junction potential between this
bridge and any "reasonable" solution is small (40).
(KCl happens to be the most convenient salt to use.)
Thus the only term left in Equation 35 that is either
not negligible or does not cancel out is 9m,,ml,ri,n,,,
the
quantity we wish to measure.
Note that, although solutions 1 and 2 in our Donnan example contain C1-, the presence of this particular ion is not relevant to our reason for using 3 M
KCl; these bridges would be equally effective with
chloride-free solutions. Also note that since the metal
Ag/AgCI,

,Ag/AgCI

&? (solid) - FJlplrc


= &',' (sold + RT In ICI-I - F&,"

and rearranging, we obtain Eq. 34 where


and

~ ' , 4 , . , h < , l "= ($elPC

'Po = j j

- JlWd

(d'i)(solid) - &i' (soh))

Similarly a Poisson-Boltzmann analysis would show an extended


space-charge region in the solution near the electrode and a very
narrow one in the electrode itself.

rne m i ro n e
Method of measuring membrane potential by making
contact with the solutions through 3 M KCI junctions.
FIG. 9.

170

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

electrodes are in identical solutions, any identical


pair of electrodes in theory could be used (even stainless-steel wires). For practical reasons of stability and
convenience, the most commonly used electrodes are
either Ag-AgC1 or calomel (Hg-Hg,Cl,).
It should also be clear that the measurement, by
use of the arrangement shown in Figure 9, of a finite
potential difference for a Donnan equilibrium is not a
violation of the second law, since with the salt
bridges present, the complete system of membrane
and electrodes is not in equilibrium.
Finally, we must warn the reader who wishes to
pursue this subject of the measurement of membrane
potentials further that he will come across the view of
certain purists who claim that, since one can only
measure ~ l , l , , l , , , ~ , . , l l , , . through some such artifice as the
introduction of liquid junctions, it is not sensible to
even talk about membrane potentials. That is, it is
impossible to assign values to the individual terms in
Equation 35, and hence one can only talk about
*,)
,),,,,,
In fact, there are some ingenious arguments that prove that most of the measured Donnan potential occurs not a t the membrane but at the
contact between salt bridges and solution. It is not
our purpose to contribute to these polemics. We hope
that the Poisson-Boltzmann analysis of the Donnan
equilibrium will convince anyone that indeed there is
a membrane potential intrinsic to that system and
that our treatment of the phase-boundary potential
(and other systems we shall discuss shortly) will also
convince the skeptic of the reality of membrane potentials. It might also be germane to point out that if
electrophysiologists had taken the purists critique of
membrane potentials seriously, the subjects of electrical excitability, receptor potentials, and postsynaptic potentials would never have gotten off the
ground, and most neurophysiologists today would be
out of business.

the concentration of fixed charge is large compared to


the concentration of NaCl in the two compartments;
that is, [Nl + [NaCI],, [NaCll,. If the membrane
thickness is large compared to the Debye length in m,
then with solutions of equal concentration on the two
sides, there exist two Donnan potentials of the same
magnitude, as shown in Figure 1lA (where for the
sake of concreteness we take z = +l).Figure 1lA is
simply a symmetrical duplication of Figure 3. At each
interface, there is a large Donnan potential jump
between solution and membrane, but there is no
potential difference across the membrane, because
these two jumps are equal.
Now consider the case in which the concentrations
of NaCl in solutions 1 and 2 are not equal (e.g.,
[NaCll, < [NaCl],). This system is no longer in equilibrium, and NaCl will diffuse slowly from solution 1
to 2; within the membrane there is a concentration

(1)
FIG.

(2)

10. An ion-exchange membrane separating two solu-

tions.

Concentration

/i

Potential
profile

Quasi-equilibrium S,ystems
Before we take up the problem of diffusion potentials involving the flux equations, we consider two
nonequilibrium systems that are sufficiently close to
equilibrium that they can be treated with good accuracy by the methods already employed. The considerations developed here are particularly relevant to our
future discussions of bilayer membranes.

63

I + \I

profiles

A MEMBRANE WITH A VERY LARGE FIXED-CHARGE DEN-

SITY.Suppose that we modify the Donnan system in


Figure 1 so that the solution with the macro-ion
separates two solutions of NaCl (Fig. 10). Furthermore, let the macro-ion, N, be uniformly distributed
and immobile (as in the ion-exchange resin we discussed previously (see subsection Statistical mechanical analysis (Poisson-Boltzmann equation). Then
the middle compartment (m) of Figure 10 is a membrane (in point of fact, an ion-exchange membrane)
separating solutions 1 and 2. Let us also assume that

Concentration

(1)
O
i

(m)

Pot en t i a I
orofile

&

+2

FIG. 11. A: concentration and potential profiles for a n ionexchange membrane of large positive fixed-charge density separating two solutions of equal concentration of NaCI. B: same as
A , except t h a t NaCl concentrations in compartments 1 and 2 a r e
unequal. (The Na and C1- profiles within the membrane have a
small negative slope t h a t is not clearly seen in t h e figure.)

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

gradient of Na+ and C1- -the concentration profiles


within the membrane have a finite slope. However,
the high concentration of fixed charge, N+, permits
only a very small concentration of Na+ in the membrane; thus the salt concentration gradient will be
quite shallow, and diffusion of NaCl across the membrane will be very slow indeed. To a first approximation we can therefore neglect the finite slopes within
the membrane and the slow flux of NaCl across the
membrane, and treat the system by our equilibrium
methods. Thus, as in the case when the concentrations of NaCl in solutions 1 and 2 were equal, we
again have two Donnan equilibria. This time, however, they are not identical, and there exists across
the membrane a potential difference (Fig. 11B).
For highly charged membranes, the counter ion is
virtually the only mobile ion present within the
membrane. Thus the membrane is "permselective"
for C1- (if z = -1, it is permselective for Na+). Note
that this permselectivity arises purely from the Donnan effect and is not dependent on any steric factors.
On the basis of the permselectivity for C1-, we can
immediately calculate the membrane potential from
our thermodynamic equilibrium criterion
kT(1) = k T ( 2 )
pi!;'

+ RT In [Cl-I,

FICII

= &!;)-

+ RT In [Cl-I,

FI,!I~

If instead the membrane contained a negative fixed


charge, it would be permselective for Na+, and by the
same argument as above we would obtain

RT
Y=+-lnF

"af],
"a+],

In general, for a membrane that is permselective for


an ion, i, of valence z (regardless of the mechanism
for the permselectivity), we obtain, by equating the
electrochemical potential of the ion on the two sides
of the membrane
(37)

Equation 37 is often called the Nernst equation, and


v' is called the Nernst potential.
It is instructive to see how the membrane potential
can also be derived from the algebraic sum of the two
Donnan potentials. Since N+ is very large, within the
membrane [Cl-1, -- "+I. The Donnan potential between solution 1 and the membrane (m) is

Combining we have

which is Equation 36. Thus the transmembrane potential is made up of the difference between two large
Donnan potentials.
We have been assuming that the membrane thickness is large compared to the Debye length within it,
and we have therefore been able to draw the concentration and potential profiles as in Figure 11, without
worrying about the very thin space-charge regions. If
the membrane thickness and the Debye length were
comparable, the continuous profiles in the spacecharge regions would have to be explicitly calculated.
Also it would now be meaningless to speak of the
transmembrane potential as the sum of two Donnan
potentials at each interface, since the distinction between interface and electroneutral membrane interior no longer exists; the space-charge region extends
throughout the entire membrane. To explicitly calculate the membrane potential would require the solution of the Poisson-Boltzmann equation. It turns out,
however, that even in such a thin membrane, virtually the only mobile ion present is C1-, for which
the thin membrane is still permselective. Thus our
equilibrium results are still applicable, and the membrane potential will still be the Nernst potential for
c1-.
Let us now, in the same way that
we extended the Donnan system of Figure 1 to make
an ion-exchange membrane in Figure 10, extend the
water-oil system of Figure 5 to make an oil membrane bounded by water phases as in Figure 12.
Suppose, to start with, that the membrane is thick
compared to the Debye length within it and that the
NaCl concentrations on the two sides differ. If we
assume that the partition coefficients of one or both of
the ions is very small, then again we can neglect the
small gradients of NaCl within the membrane and
the slow flux of NaCl across the membrane and treat
the system as being at equilibrium. The concentration and potential profiles are shown in Figure 13,
where for the sake of concreteness we have made the
Na+ partition coefficient considerably larger than
that of C1-. We see that there are two large positive
AN OIL MEMBRANE.

(H,O)

;(oil); (H,O)

(1) ( m )
FIG.

and between solution 2 and the membrane

17 1

tions,

(2)

12. An oil membrane separating two NaCl aqueous solu-

172

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

Thus for the thick membrane (membrane thickness


LJ the membrane potential is zero, whereas for
Concentration
(oil) I
the
thin membrane (membrane thickness < L,J it is
I Na+.CIprof I les
the Nernst potential for Na+. Clearly, for intermediate thicknesses, the membrane potential will lie
somewhere between these values. For the oil membrane, the membrane potential is a function of mem+m
brane thickness.
Potential
I
That phase-boundary potentials manifest themI
1
I
I
profile
selves across a thin oil phase is of great importance
I
I
I
for the thin lipid membranes we consider later. The
antibiotic valinomycin, for example, is a n uncharged
molecule that is very lipid (hydrocarbon) soluble and
FIG. 13. Concentration and potential profiles for a thick oil
membrane separating two NaCl aqueous solutions. The potential has a high affinity for K. Thus, in the presence of
within the membrane ($,) is positive, because we have assumed valinomycin, the partition coefficient of K+ into a
that Na+ partitions better into the membrane than C1-.
lipid phase is much larger than that of any other ion
normally present. In the presence of a KCI concentraphase-boundary potentials, but they are equal and tion difference across a thin lipid membrane, theregiven by Equation 32
fore, one expects to record, and indeed does, a membrane potential given by the Nernst potential for K+.
For a thick oil membrane, however, this potential is
not manifested for precisely the reasons we have
given above.
As we pointed out earlier, the phase-boundary potential is not a function of the NaCl concentration in the Ion Transport (the Nernst-Planck
aqueous phase. Even though Na+ is much more fa- Flux Equations)
vored in the oil phase than C1- and consequently
We now shall discuss nonequilibrium situations,
there are large phase-boundary potentials, the total
membrane potential is zero. This result contrasts that is, those where we can no longer approximate
sharply with that for the high-density fixed-charge the situation with our equilibrium criterion, but
membrane we discussed earlier, across which the must deal directly with the flux of ions across the
membrane. This necessitates the use of Equations 5a
Nernst potential appears.
The above analysis was predicated on the mem- and 5b- the Nernst-Planck flux equations. We begin
brane being thick compared to the Debye length
I
within it. Consider now a membrane of thickness
comparable with the Debye length. Instead of Figure
I
13, we must now draw the complete ionic profiles
including the space-charge regions, since there is no
electroneutral region within the membrane. The profiles in Figure 14 are an extension of those in Figure
7. For comparison we have redrawn Figure 13 in
Figure 14A, exaggerating the space-charge regions
for a thick membrane; in Figure 14B we have expanded the scale, since the space-charge regions extend through the entire membrane.
To calculate the membrane potential we cannot use
Equation 32, but must solve the Poisson-Boltzmann
equation. By inspection of the profiles, however, we
see that Na+ is virtually the only ion in the membrane; that is, the membrane is permselective for
Na+. Consequently the membrane potential must be
given by the Nernst potential for Na+
Not 1-1-

2
1

This is true provided the mobilities of Na+ and C1- in the


membrane are equal. If they are not, then, as we shall see in a
later section, there will be a diffusion potential due to the asymmetry in mobility. Nevertheless our main conclusion continues to
hold; namely, the phase-boundary potentials do not contribute to
the membrane potential.

FIG. 14. A: a redrawing of Fig. 13 for the concentration profiles of a thick oil membrane, with the space-charge regions
shown (exaggerated). Br concentration profiles for a thin oil
membrane. Note that there is no region within the membrane
where electroneutrality (a+] = [CI-I) holds. The entire membrane thickness corresponds to the regions near the boundaries in
Fig. 14A.

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

with a homogeneous uncharged membrane, which


elucidates most of the major concepts involved in
transport. We then briefly discuss some slightly more
complicated membranes and conclude with a treatment of nonhomogeneous (mosaic) membranes, with
particular emphasis on how they differ from homogeneous membranes.

173

where

u c ujcj+
j

ukckk

The definition of electric current, Z, given above is the


traditional one, namely, the algebraic sum of the
positive and negative ion fluxes. (Positive ions flowing from left to right make a positive contribution to
the current, whereas negative ions flowing in the
same direction make a negative contribution.) Multiplication of the fluxes by the Faraday (F) converts the
current into the conventional units of C/s (amp) from
those of mol/s. In free diffusion, C.j4j+= Z,&
and
therefore Z = 0. Dividing through by F Z (U + V) and
integrating across the total membrane thickness
(from 0 to 6) yields

HOMOGENEOUS UNCHARGED MEMBRANE. General case.


a) The integrated form of the flux equations. Consider
a convection-free, homogeneous, uncharged membrane of unit area separating two infinitely large
aqueous solutions of univalent ions, both solutions
being continuously stirred to maintain uniform conditions (Fig. 15). We further assume that the membrane is freely permeable to all ions and water, that
water flow can be neglected, and that the mobilities
of the ions though different for different ions are
constant throughout the membrane; that is, they are
not a function of concentration. By a homogeneous where
membrane, we mean one whose properties do not
vary within any plane parallel to the membrane sur= ($1 - $2)
faces. Thus there are no regions showing special
Equation 38 is a fundamental relationship that
permeabilities to certain ions; this is in contrast to follows directly from summing and then integrating
the mosaic membrane which we discuss later. A the flux equations. It has an interesting interpretamembrane that is convection free and freely permea- tion. The term F2(U + V)is simply the conductivity of
ble to ions and water is one with properties no differ- the solution a t any point x in the membrane, and 1/
ent from those of a n equal thickness of free solution, F2(U + V) is thus the specific resistivity a t this point.
except that it is possible to stir on both sides of it Therefore dx/Fz(U + V) is the resistance (R) of a slab
without perturbing the situation within the mem- of solution of dx thickness within the membrane, and
brane; also there are no special partitioning effects of consequently
the ions at the interfaces. A real membrane that
approximates the above model is very coarse, uncharged filter paper.
PY?+
V)
Starting with the flux equations
is the total integral resistance (R,,,)of the membrane.
The first term on the right in Equation 38 therefore
represents the ZR drop across the membrane. Equation 38 states that the membrane potential is made
dx
up of a n ZR drop plus another term called the diffusion emf (*J. (WhenZ = 0,
becomes the only term
and is often called the diffusion potential across the
summing over all species, and subtracting we obtain membrane. On the other hand, in the absence of the
second term on the right, Equation 38 is simply
Ohms law.) That is

il:

=]&I

*/I

(38)

where
Recording
electrodes
Stirnuloti
elec tro

FIG. 15. A membrane separating two infinite, well-stirred


aqueous solutions. Stimulating and recording electrodes are
shown.

(39)

(40)
(U + V)
From Equation 38 we can represent the electrical
characteristics of the membrane by an equivalent
circuit consisting of an emf and a resistance in series
(Fig. 16). Both the resistance element and the emf
may be voltage dependent (as we shall shortly see),
and thev are so indicated in the eauivalent circuit.

174

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

In his original paper (63), Planck showed that for


L,J, the combination of the
thick membranes (6
flux equations with the electroneutrality condition
(Eq. 41) yields concentration and potential profiles
experimentally indistinguishable from the exact result [see also (311. We shall not attempt to reproduce
the arguments for this conclusion; we merely appeal
t
I
to the readers intuition that macroscopic samples
(i.e., thickness s L,)) of solution will never, upon
FIG. 16. Equivalent circuit for a homogeneous, uncharged
membrane separating two solutions of arbitrary ionic composi- chemical analysis, reveal violations of electroneutraltion.
ity. Note well, however, that the point in question is
essentially a mathematical one; is Equation 41 mathEquation 38, although extremely important, has ematically an appropriate replacement for Equation
only formal significance until the concentration pro- 17? From the physical standpoint, Equation 41 is
files within the membrane are known; that is, to use ridiculous, for if electroneutrality were satisfied a t all
the equation (to evaluate the two integrals on the points in solution, there would be no mechanism for
right) we must know how the concentrations of all an electrostatic potential to develop in or across the
ions vary with x. The flux equations (Eqs. 5a and 5b) membrane. In using Equation 41, the reader should
by themselves are not sufficient to specify these (just also understand that it, and not Poissons equation,
as the Boltzmann distribution is not sufficient to must be used with the flux equations; the electroneuspecify the concentration profiles for the equilibrium trality condition cannot be used directly in conjuncsituation). We need another condition. This clearly tion with Poissons equation. To substitute Equation
must be Poissons equation (Eq. 17). Therefore the 41 into Equation 17 would lead to the conclusion that
general mathematical problem for nonequilibrium d+/dx is a constant, whereas, as we shall see in our
systems is to solve the combined flux equations plus examples, this is generally incorrect. The proper
Poissons equation with appropriate boundary condi- mathematical point of view is that the concentration
tions. In fact, if we combine Equations 5 and 17 we and potential profiles obtained by solving the flux
have a Poisson-Flux equation for nonequilibrium sys- equations in combination with the condition of electems in complete analogy with the Poisson-Boltz- troneutrality are (for thick membranes) an excellent
first approximation. [To find p , one can now substimann equation for equilibrium systems.
tute the potential profile into Poissons equation with
b) T h e condition of electroneutrality and the con- appropriate boundary conditions (311.
stant-field condition. We now come to an issue that
For a thin membrane the electroneutrality condican be a source of some confusion. In general the tion is not a good approximation and, in general,
Poisson-Flux equation is analytically intractable and Poissons equation must be used directly. In many
requires numerical methods for its solution. In these cases of interest with thin lipid membranes, however,
days of computers there can be worse happenings. It it turns out that the concentration of ions within the
is nonetheless very helpful to have analytical forms membrane is so low that 0 is a good approximation for
that are good approximations to the exact solutions. p in Equation 17 (59, 77). In other words, the conThere are two useful approaches to this end. One is to stant-field assumption is a good approximation to the
replace Poissons equation with the electroneutrality field within the membrane.
condition
c) Time-dependent solutions and steady-state solu(41) tions. The flux eauations actuallv should be written
and then use this in conjunction with the flux equa- with partial derivatives, since in general the concentions to obtain the concentration and potential pro- trations and the potential are functions not only of x
but also of time ( t ) . That is, when the voltage or
files. The other is to assume that p is so small that it
current across the membrane is changed by a n extercan be approximated by 0 in Poissons equation. Upon
nal (voltage or current) source, the rearrangement of
one integration of Equation 17 we then get
ionic Drofiles is not instantaneous. From conservation of mass, we write the continuity equation for
constant-field condition (42) each ion
&
6
dc,
-- -continuity equation (43)
that is, the electric field within the membrane is
dt
ax
constant. Equation 42 is then substituted into the
flux equations, and the concentration profiles are which when combined with Equations 5 yields
then determined. Under what circumstances are
dC,+
d%
Equations 41 and 42 appropriate replacements of
= u J R T 4 + u,F
(44a)
Poissons equation?
dt
dXL

*---

CHAPTER

6:

175

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

(A
Equations 44 are to Equations 5 what Fick's second
law is to Fick's first law; in fact, forgetting the second
term on the right, Equation 44 is Fick's second law

+ B) = 2RT
(c, - c,)
S
~

and substituting this back into Equation 46 we get


(47)

Note that the concentration profile of the ions is


linear and that it is not a function of voltage (Fig. 17).
This simple result holds only for the single-salt case
where
(for a thick membrane); as we see in our second
D = uRT
(7) example, the concentration profiles are generally
Although we are interested in the time transients, we nonlinear and voltage dependent. Note also that,
shall not deal explicitly with the solution of the time- upon substituting Equation 47 into either Equation
dependent equations (in conjunction with Poisson's 45a or 45b and integrating, we find that i,b is a logaequation or one of its substitutes), since the algebra is rithmic function of x ; that is, the field is not constant
analytically intractable. Rather, we shall concen- within the membrane.
We can now calculate our terms in the equivalent
trate exclusively on steady-state solutions to the flux
equations and then argue qualitatively how the sys- circuit. In fact, we can obtain q,)even without the
tem passes from one steady-state situation to another use of Equation 47. We have immediately from Equation 40
when the voltage (or current) is suddenly changed.
In the steady state, the flux (4) of each ion is constant at every point in the membrane. From the
continuity equation it immediately follows from this
that iklat vanishes, and hence, the system is not time
In "; (48)
dependent.
With this background we now consider two specific
examples illustrating the important properties of a To calculate R,,,, we need the concentration profiles
homogeneous membrane. In particular we focus on (Eq. 47); combining this with Equation 39 we have
the implications of the equivalent circuit of Figure
16. Both our examples are confined to thick membranes, where we can apply the condition of electroneutrality. We choose the thick membrane, because
we think it is instructive for the reader to see the
complexities that can arise with "simple" filter paper.
The qualitative features of the results obtained are
equally valid for thin membranes.

The single-salt case. In this case, the solutions on


the two sides of the membrane consist of a single salt
(e.g., NaC1) a t concentrations c , and c2. We now wish
to solve for the concentration profiles so that we can
in Equation 38'. Writing the flux
evaluate R,,, and q,)
equations and remembering that the condition of
electroneutrality holds at any point in the membrane
(namely c+ = c - = c ) , we have
-A

-B

4+
dc - F
_ _ = -RT d+
U
dx
4-- = -RT &
dc+ F c dlCl
U
dx

cz

Thus, for the single-salt case, Equation 38 becomes

6
_-____-__
In
F'(u + u)(c~- c,)

Cn
-

C,

+ (U - U ) RT
-~
1n c2

(u + u ) F

~~

c,

(50)

The equivalent circuit is shown in Figure 18. Neither


q,,
nor R,,, are voltage dependent, which we knew a
(45a) priori from the fact that the concentration profiles do
not change with voltage. In Figure 19 we plot the
(45b)

and adding
(A

dc
+ B) = 2RT dx

Integrating Equation 46 across the membrane thickness (6)gives

FIG. 17. Concentration profile within a thick, homogeneous,


uncharged membrane separating two NaCl solutions at different
concentrations.

176

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

the slower diffusing species would be speeded up.


We can see this coupling more precisely. Adding
the flux equations (Eqs. 45a and 45b) for the case of
I = 0 (++ =
= 4) we have

--==

+-

(u-V)

RT

(u+v)

..--

c2

+=--

In cI

2uv
dc
RT u+v
&

and we see that the salt diffuses as a single species


according to Ficks law, but the diffusion constant

D=- 2uv R T

18. Equivalent circuit for the single-salt case for a thick,


homogeneous, uncharged membrane.
FIG.

u+v

(52)

is a mean of the individual diffusion constants

D+ = uRT

D-

vRT

The electric field coupling the ions is hidden in this


form of writing the flux.
Note the limiting cases for q,,.
When u = u, TI,
= 0,
as we would intuitively predict from the above argument concerning the origin of the diffusion potential.
u or u + u, q,,reduces to the
For the extremes u
Nernst potential for the more mobile ion. This is
again intuitively obvious, since if the mobility of one
ion is much less than the other, the membrane is, to a
first approximation, permselective to the more mobile ion.

19. Current-voltage characteristics for the single-salt


case for a thick, homogeneous, uncharged membrane. These
characteristics follow from the equivalent circuit in Fig. 18. (Note
that these are drawn for c1 > c,; if c , > c p ,V,,(the potential a t I =
0 ) would change sign. That is, the plots for u > u and u < u would
be interchanged.
FIG.

current-voltage characteristic of this circuit for u = u,


u > u, and u < u. The current-voltage characteristic is
linear.
It is instructive to consider the membrane emf (or
diffusion potential), V/,, which, for a given concentration ratio, is determined by the relative magnitude of
the cation mobility ( u ) and anion mobility (u). The
diffusion potential physically arises from the separation of charge that is set up as one ion tends to diffuse
faster than the other. In Figure 20 we consider what
happens if a t t = 0 there is a sharp boundary in salt
concentration, going from c , to c2, and imagine that
the ions are uncharged and have different diffusion
coefficients (i.e., mobilities). Each species will diffuse
independently to form the typical Gaussian concentration profile. The faster diffusing species would
have the concentration profile marked fast, and the
slower diffusing species would have the concentration
profile marked slow. If the particles were now
charged, as indicated, it is clear that a space-charge
region would result and give rise to an electric field
coupling the motion of the species together. The
faster diffusing species would be slowed down, and

The bi-ionic case. In this case there are three ion


species present - for concreteness we choose two cations (Na+and K+)and one anion (W)- and the total
concentrations of ions on the two sides of the membrane are equal. Then the flux equations become

(53a)

Cl

slow ionFIG. 20. How a n electric field would develop, if a t t = 0 there is


a sharp boundary in salt concentration from c , to c 2 and the
mobilities (i.e., diffusion coefficients) of the ions are different.
The profiles shown are those that would exist a t some short time
later if the ions were uncharged. We see, however, that since the
ions are charged, this creates a space-charge region t h a t will give
rise to a n electric field that tends to couple the motion of the ions
together.

6:

CHAPTER

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

with the electroneutrality condition being


Letting A

A,

+ AN;, we obtain
dc+
+ Fc+ d4J
A = RT
dx
dx

dl-

RT-

dx

d$
Fc--

dx

Substituting Equation 54 into Equation 55b and adding this to Equation 55a we obtain

(A

+ B) = 2RT dc+
dx
~

(56)

Thus the total cation (and anion) concentration varies


linearly with z. (This incidentally is true for all cases
of the homogeneous, thick, uncharged membrane. 1
Since, however, c I += c2+, this means that c+ (andc-)
are constant throughout the membrane. Using this
fact we can immediately calculate qIl

is not voltage dependAs in the single-salt case,


ent.
If we subtract Equation 55b from Equation 55a we
obtain
d$
2F-

CLC

(A - B)

177

centration profiles for the open-circuited situation


(i.e., no current being passed across the membrane).
Now when side 1 is made positive with respect to side
2, Na+ is moved into the membrane and K+ moved
out (Fig. 21B). Since Na+ is a less mobile ion than K+,
clearly the membrane resistance will increase. In the
limit of large positive currents, Na+ has completely
replaced K+, and the membrane resistance is that of
a membrane bathed on both sides with 0.1 M NaC1.
Conversely, for negative currents, K+ replaces Na+
(Fig. 2 1 0 and the resistance falls; in the limit of
large negative currents, the membrane is flushed
with 0.1 M KC1. Figure 22 is a qualitative plot of the
steady-state current-voltage characteristic for this
system. This nonlinear q - Z characteristic for the biionic case should be contrasted to the linear characteristics for the single-salt case (Fig. 19).
From this example we see that the basis for the
voltage dependence of resistance is the shifting of the
ionic concentration profiles within the membrane, so
that ions of different mobilities replace each other. It
should also be clear from the expression for PI, (Bq.
40) that in general this shifting of ionic profiles
should also lead to changes in qI,.It is fortuitous that
in the bi-ionic case the steady-state value of q,,is not
does change during the tranvoltage sensitive. [q,>

FIG. 21. Concentration profiles within the membrane for the


bi-ionic case of a membrane separating 0.1 M NaCl from 0.1 M
KCI. A, I = 0; B , I is large and positive; C, I is large and negative.

10.1 M NaCl

C+

and since C + is a constant, this means that d+l& is a


constant. Thus, for the bi-ionic case the electric field
within the membrane happens to be a constant, in
contrast to the result obtained for the single-salt
case. We now use this fact to solve for the concentration profiles and hence evaluate Ft,,,,. The important
result of this calculation, which is rather messy (20),
is that
is a function of voltage. Thus the equivalent circuit for the bi-ionic case consists of a voltagedependent resistance and a voltage-independent emf.
For more general cases, the emf is also voltage dependent, and the equivalent circuit is that shown in
Figure 16.
We now want to see intuitively why the membrane
resistance is voltage dependent. Consider the special
case of the membrane separating 0.1 M NaCl (side 1)
from 0.1 M KC1 (side 2). Figure 21A shows the con-

en,

FIG. 22. Sketch of the steady-state current-voltage characteristic for the bionic case of Fig. 21. Note the characteristics (dashed
lines) for a membrane separating either symmetrical 0.1 M NaCl
solutions or symmetrical 0.1 M KCl solutions, which the bi-ionic
characteristic approaches asymptotically for large positive and
negative potentials, respectively. Also shown are the chord and
slope resistance lines at point P .

178

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

FIG. 23. Nonlinear steady-state current-voltage characteristic


with chords drawn at ( I , , 9,)
and ( I x , qJ.If current is stepped
f r o m l , toZ,, t h e voltage originally attains t h e value a t P and then
increases with time along t h e vertical line from P until i t reaches
the value q 2 .Similarly, if current is stepped from I , to I,, t h e
voltage originally attains the value a t Q and then moves along
the vertical line from Q until it reaches t h e value 9,.

sient from one steady state to another (81.1


The current-voltage characteristic in Figure 22 also
illustrates several concepts that will be important for
our later discussion of excitable systems: the system
shows rectification; that is, for voltages (i.e., values of
q of the same magnitude but opposite sign, the
(absolute values of the) currents are not equal. Also,
as for any circuit element with a nonlinear currentvoltage characteristic, the term resistance is ambiguous and has a t least two reasonable meanings. One is
the chord resistance given by

resistance at I , , since the ionic profiles have not had


time to change. As they rearrange themselves, the
resistance rises (Na+ is replacing K + in the membrane) and the voltage increases until it attains the
steady-state value for I,. Similarly, if the current is
suddenly stepped from I , to 14, the instantaneous
voltage will be given by I , times the chord resistance
at I,, and then the (absolute value of the) voltage will
decrease as the ionic profiles rearrange. The responses are schematized in Figure 24.
We see that the membrane resistance is not only
nonlinear, it is also time uariant. The time dependence enters because of the time required for the ionic
profiles to shift from one steady-state configuration to
another. As the profiles are shifting, the integral
resistance is continuously changing. (Y,, is also continuously changing, but we are neglecting that in
this discussion.) Note also that the response in Figure
24A is phenomenologically similar to the response of
the RC network in Figure 25A, whereas the response in Figure 24B is phenomenologically similar
to the response of the RL network in Figure 2%. It
can be shown formally that the AC-impedance characteristics of time-variant resistances can display
phenomenological capacitances and inductances (42).
To summarize the time-variant aspect of the system:
the instantaneous response to a change in current (or
voltage) is along the chord resistance; the voltage (or

(58)

In this case, since \If,,is not voltage dependent, the


chord resistance is identical to the integral resistance. Another equally valid meaning for resistance is
the slope resistance given by
(59)

Both of these terms are illustrated in Figure 22 at a


given value of q .In discussing membrane resistance
for a nonlinear system, one must always specify the
resistance referred to. Of course, for linear currentvoltage characteristics, such as in Figure 19, the two
resistances are identical, and no distinction need be
made.
It is very instructive to consider how the system
passes in time from one steady-state point to another
(Fig. 23, which is a redrawing of the characteristic in
Fig. 22). Suppose we are sitting at the point (II,'PI)
and suddenly step the current up to I,. The instantaneous voltage will be given by I , times the chord

FIG. 24. The change of voltage with time in response to steps


of current. These responses follow from t h e analysis of Fig. 23 as
given in t h e legend and text.

FIG. 25. RC (A) and RL ( B ) networks t h a t would give responses similar to those shown in Fig. 24.

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

179

current) then relaxes to the point on the steady-state


current-voltage characteristic appropriate for the
new current (or voltage).
HOMOGENEOUS MEMBRANES WITH SPECIAL PROPERTIES.

We have been discussing the behavior of a membrane


whose only property is that it allows stirring on both
sides for maintenance of boundary conditions; otherwise the membrane imposes no new restrictions on
ion movement. Our real example has been coarse
filter paper. We now wish to consider two examples
in which the membrane phase is not merely a confined salt solution of composition grading between
that of the two aqueous phases bathing the membrane. After discussing the two examples, we shall
comment on the general situation of membranes with
special properties.
Fixed-charge membrane. We have already considered the quasi-equilibrium situation that arises when
the fixed-charge density of a membrane is very large
and the membrane separates a single salt a t two
different concentrations. We now extend the treatment of fixed-charge membranes to the case of lower
charge densities. Since the co-ion is no longer excluded, transport (when Z = 0) of ions and salt cannot
be neglected. (Even in membranes of large fixedcharge density, significant fluxes of counter ions occur; e.g., for the bi-ionic case of NaCl vs. KCl separated by a high-density negatively charged membrane, there is, a t Z = 0, a flux of Na+ from compartment 1 to compartment 2, and an equal and opposite
flux of K+ from 2 to 1. Our treatment will of course
also apply to these cases.) We do not go through the
details of the analysis, but merely point out the highlights of the principles involved.
The general analysis of fixed-charge membranes is
a straightforward combination of the Donnan equilibrium and the flux equations (71). It is assumed that
the interfacial processes are rapid compared to transport through the membrane interior; hence one assumes that equilibrium is maintained a t each interface a t all times (even in the face of current flow). For
the fixed-charge membrane, this means that the
Donnan equilibrium is satisfied a t each interface.
Thus the concentrations of ions just within the membrane satisfy the Donnan condition with respect to
the ions just outside the membrane. Subject to these
new boundary concentrations, the ions then move
according to the flux equations. Essentially then the
fixed charge has established new boundary conditions for the Nernst-Planck regime of ions. Current
flow will shift the concentration profiles within the
interior of the membrane, but it is assumed that the
concentrations Ijust inside the membrane are not
perturbed by the current. Figure 26A illustrates the
concentration profiles for the single-salt case in a
positive fixed-charge membrane. If +I = 0, the
system reduces to the single-salt case for an uncharged membrane, and if +I % [NaCll,,,, the sys-

(1)

(m)

(2)

FIG. 26. Concentration profiles (A) and potential profile ( B )for


a positive fixed-charge membrane separating NaCl solutions at
different concentrations. Note the Donnan jumps in concentrations and potential at the interfaces.

tem approaches the quasi-equilibrium case of a membrane with a very large fixed-charge density, described earlier. Note that in this latter case the membrane is almost exclusively permeable to C1- not
because the mobility of chloride in the membrane is
so much larger than that of sodium, but because the
concentration of chloride is so much greater than that
of sodium. This illustrates an important general principle applicable to both electrolytes and nonelectrolytes: the permeability of a membrane for a species is
generally dependent on two factors. One is the mobility of the species in the membrane phase, and the
other is the ability of the species to enter the membrane in the first place. Unless one has other information, it is impossible to tell which is responsible for
a given molecule being poorly permeable.
For a fixed-charge membrane, the total membrane
potential is the algebraic sum of three terms: two
Donnan potentials a t the interfaces, generally called
rr, and rr2, and a Nernst-Planck diffusion potential
within the membrane, called (GI, - $2m). Thus
= TI

+ ~2 +

($1,

( 60)

- $2m)

(The potential profile is illustrated for the single-salt


case in Fig. 26B.) To find
- $2m) one must solve
the flux equations subject to the new boundary conditions established by the Donnan equilibria at the
interfaces (71). The electroneutrality condition
within the membrane now takes the form

C cj+ + ZN - X C ~ -= 0

(61)

Note that for the bi-ionic case, the Donnan potentials


(mland rr2) are equal and hence cancel, so that the
membrane potential in Equation 60 is given just by
the internal diffusion potential
Oil membrane. The analysis of this system is completely analogous to that of the fixed-charge membrane, but the equilibria a t the boundaries are now
partition equilibria, not Donnan equilibria. Subject
to the new concentrations established just within the
membrane, the ions once again diffuse according to
the Nernst-Planck flux equations. (The concentration
profile for the single-salt case is illustrated in Fig.
27.) Once again the membrane potential is the algebraic sum of three potentials: two phase-boundary
potentials and an internal diffusion potential. Thus

T(o/u)a

T(u/o)z

($Im

$2m)

( 62)

180

THE NERVOUS SYSTEM I

HANDBOOK OF PHYSIOLOGY

As we pointed out earlier (see subsection Quasi-eguiZibrium Systems. AN OIL MEMBRANE), the phaseboundary potentials are equal for the singe-salt case
and hence cancel. Thus, despite the relative solubilities of the anion and cation in the oil phase, the
membrane potential will only be a function of their
relative mobilities in this phase. Therefore, for the
single-salt case, Equation 62 reduces to

qJ

u R T cZrn
In u + u F
c,,,
u

=--

On the other hand, the phase-boundary potentials


will in general be different for the bi-ionic case, and
all three terms in Equation 62 will contribute to II/.

General considerations. The two examples we have


just given illustrate the general approach taken with
membrane transport. One assumes that the membrane can conceptually be divided into three regions:
two interfacial regions and an interior region. It is
assumed that interfacial processes are so fast that
equilibrium conditions prevail there. These processes
establish new boundary conditions Ijust within the
membrane, and transport then takes place through
the membrane interior according to the flux equations. (This analysis incidentally is equally applicable to nonelectrolyte transport.) The membrane potential is the algebraic sum of two equilibrium potentials and one diffusion potential. Since the membrane
substance is discontinuous at the interface, there
exist discontinuities in concentrations and electrical
potential a t each interface, but the assumption of
equilibrium is equivalent to assuming that the electrochemical potential of each species is nevertheless
continuous across the interface. [For example, in a
fixed-charge membrane there are jumps in concentration and electrical potential at the interface (see
Fig. 261, but the electrochemical potential of each ion
on the two sides of the interface is the same. In fact,
these characteristic discontinuities of the Donnan
equilibrium were derived on the basis of equality of
electrochemical potential.] Of course, one could be
more sophisticated and expand these discontinuities
into the actual space-charge regions present a t each
interface.
There are two points worth noting about this analysis of membrane transport and potentials. First, it is
possible that situations might be found where the
assumption of very fast interfacial processes relative

I C 1 rn

(1)

(m)

(2)

FIG. 27. Concentration profile for an oil membrane separating a single salt at two different concentrations.

to diffusion within the membrane begins to break


down. It would then no longer be possible to assume
equilibrium a t the boundaries, and an explicit analysis of the boundary kinetics would have to be included
in the overall transport. The previous type of analysis, however, still gives a good qualitative picture
even in this instance. For example, suppose that
interfacial events were not fast enough to maintain
the Donnan conditions. Nevertheless the sign of the
change of ion concentrations across the interface can
be obtained from the Donnan analysis. That is, for a
positively fixed-charge membrane, the anion concentration will be elevated across the interface and the
cation concentration depressed. The interfacial potentials will then be somewhat reduced from their
equilibrium values.
The second point is that the analysis presupposes
the possibility of dividing the membrane up into interfacial regions and interior. Clearly this becomes a
very tenuous assumption when the membrane is of
the order of 50-p\ thickness, as with the lipid bilayers.
In fact, the analysis breaks down. It is no longer
possible to break up the membrane potential as in
Equation 62, since the space-charge regions extend
throughout the membrane. Here, an explicit analysis
combining the flux equations and Poissons equation
with the appropriate boundary conditions must be
used. It is interesting that for nonelectrolyte transport across the bilayers (e.g., isotopic water flux), the
procedure of dividing the membrane into three regions leads to predictions in good agreement with
experimental results (18). It appears that partition
equilibrium is attained rapidly with respect to diffusion even through a very thin (-50 A) region.
MOSAIC MEMBRANES.
We have been considering membranes whose properties do not vary in a plane parallel to the membrane surfaces; that is, the membranes
are homogeneous. We now wish to comment on mosaic membranes consisting of regions with different
permeability characteristics. For example, the membrane may contain some regions permselective for
Na+, other regions permselective for K + , and still
others relatively nonselective among univalent ions.
As we shall see later, such membranes are of direct
relevance to biological membranes, particularly excitable membranes, and can be experimentally realized with modified thin lipid membranes. At this
point we are not concerned with the mechanisms for
the selectivities, but accept them as given. (If the
membrane consisted of cation and anion selective
regions, we could imagine that each region was a
patch of high fixed-charge density ion-exchange
resin; such membranes can be fabricated with ionexchange beads imbedded in an inert matrix.)
For a membrane with regions in parallel, the
equivalent circuit has the form shown in Figure 28.
(For completeness we have included the membrane
capacitance, which we discuss in a later section.)

6:

CHAPTER

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

181

It should be understood that the dissipative processes for the homogeneous and mosaic membranes
are quite different. In the former the ions flow
through a common region, and hence there are no
local current flows. In the mosaic membrane ion
movement is through local currents. Despite these
physical differences, it happens that the steady-state
properties of a homogeneous membrane can also be
formally represented by the circuit in Figure 28 (20).

'.iI
"7
I
t

1
1

FIG. 28. Equivalent circuit for a mosaic membrane. The individual elements are shown to be ideally selective for a given ion,
but this need not necessarily be the case.

FORMAL CONSEQUENCES OF VOLTAGEDEPENDENT CONDUCTANCES

The Nature of Electrical Excitability

In the subsection Ion Transport (the NernstPlanck Flux Equations) we developed the properties
Here we have indicated that in principle the conductof a homogeneous membrane and showed that even
ances (g,)can be voltage dependent. (In treating parwith
such a simple membrane as coarse filter paper,
allel circuits, it is more convenient to use conductis
possible
to observe rectification, nonlinearities,
it
ances than resistances.) The emfs, however, are inand
time
transients
in the membrane potential. Since
variant, being determined by the concentrations of
these properties are also characteristic of electrically
ions on the two sides of the membrane and the permeexcitable biological membranes, the question arises
ability characteristics of the regions. If the regions
are permselective, as in Figure 28, each emf is the whether the mechanisms operating in the simple
Nernst potential for the permeant ion of that region. systems we discussed are also responsible for generThis circuit should be contrasted to the one for the ating action potentials and bioelectric phenomena.
homogeneous membrane given in Figure 16. In that We remind the reader that the nonohmic behavior we
case there is a single conductance and one emf, both have described is due to the shifting of the ionic
of which can be voltage dependent. For the mosaic profiles of mobile ions within the membrane in the
membrane, we have from elementary circuit theory face of an applied voltage or current; this change in
the ionic profiles leads to changes in the membrane
z, = gj (9- Ej)
(63) conductance and the membrane emf. The question
then is whether such a mechanism could account for
for each element of the circuit. Summing over all action potentials of nerve and muscle.
elements, we have from Kirchhoff s law
It is very difficult to give a definitive answer to this
question. It can probably be proved that it is impossible to reproduce electrical excitability with a filter
paper membrane, even with the most bizarre mixture
where
of ions on the two sides of the membrane. This does
not preclude the possibility that a membrane with a
z = TZI
(65)
fixed-charge density varying as arc coth x3", or one
Equation 64 bears the same relation to the mosaic with some continuously varying function of the diemembrane as Equation 38 bears to the homogeneous lectric constant of its ''oil" interior, or a combination
membrane. Comparing the two, we see that each of these structures could not generate action potenequation has an Ohm's law term and also another tials. Although no one has demonstrated, either theoterm that may be called the diffusion potential. In a retically or experimentally, that one can develop a n
formal sense the two expressions are quite similar.
excitable membrane simply from electrophoretically
Although the flux equations describe the move- moving ionic profiles within the membrane, neither
ment of ions across the membrane, the equivalent has anyone presented proof that it is impossible to do
circuit of Figure 16 does not indicate that there is ion so. [In fact, if one allows the ionic profiles to be
flux when Z = 0. In contrast, the equivalent circuit of shifted not only by the electric field (i.e., electrophoFigure 28 depicts the local currents that flow through resis) but also by solvent drag produced by electrothe elements of the membrane even when no current, osmosis and hydrostatic pressure gradients, then one
I , is being passed across the membrane. Under these can indeed develop, with a membrane made merely of
circumstances, the membrane potential is given ex- sintered glass, oscillatory behavior and complex allclusively by the second term in Equation 64, whereas or-none responses that are in some ways phenomenoEquation 65 becomes
logically similar to excitable tissue (72, 73).] There
are nevertheless strong indications for believing that
zzj
= 0
such a mechanism cannot be the basis for biological
j

182

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

excitability, and indeed there is compelling evidence


that excitable membranes are not homogeneous, but
rather mosaic structures. We shall discuss the bases
for the above assertions and a t the same time present
the current view of the nature of biological excitable
systems.

ability changes significantly. In fact, Hodgkin and


Huxley's analysis of excitability in the giant axon
was possible only because they could clearly resolve
and separate the time course of the sodium and potassium conductance changes (29). Since then, the belief
that these components are separate has been reinforced with the discovery of pharmacological agents
REASONS FOR BELIEVING THAT ELECTRICAL EXCITABIG
that inhibit one, but not both, conductances. For
ITY DOES NOT RESULT FROM THE SHIFTING OF IONIC
example, tetrodotoxin (TTX), the puffer fish poison,
PROFILES. T h e magnitude of rectification. In excitable
can completely and reversibly block the sodium concells such as the squid giant axon, the ratio of the ductance changes without affecting the magnitude
conductance a t large depolarizations to the conduct- and the time course of the potassium conductance
ance a t large hyperpolarizations can be several transients (49). Conversely, tetraethylammonium
hundred to one. Since the major conducting ions are (TEA) reduces the potassium conductance without
Na+ (on the outside) and K+ (in the axoplasm), with significantly affecting the sodium transients (2).
approximately equal total salt concentration on each
It is difficult t o imagine a homogeneous regime
side of the membrane, the situation is approximately
of ions where the permeability changes for one ion are
bi-ionic. In such a case, rectification is achieved by not intimately coupled with those of the other ions. In
exchanging in the membrane a more mobile ion with the bi-ionic case that we considered, for example, the
a less mobile ion, depending on the sign of the poten- movement of Na+ into or out of the membrane was
tial. Assuming that the contribution of all other ions accompanied by the obligatory and opposite moveto the membrane conductance can be neglected (an ment of K+. It is even more difficult to conceive of an
extreme case), the limiting rectification ratio is given agent, acting in a homogeneous membrane, that
by the ratio of the sodium and potassium mobilities in could inhibit transients associated with one ion and
the membrane (i.e., u K / u N a which
),
is approximately yet not affect those associated with the other ions in
1.5 in free aqueous solution. If this phenomenon ac- the membrane. For these reasons, virtually all eleccounted for the observed rectification ratio of several trophysiologists believe that excitable membranes
hundred, either the potassium mobility within the are mosaic structures of the form described in the
membrane must be abnormally large, or the sodium subsection MOSAIC MEMBRANES, and although Hodgmobility must be abnormally small, or both.
kin and Huxley did not explicitly state this in their
The time scale of the action potential. In our discus- basic papers (29-331, they clearly had this picture in
sion of transient behavior during ionic profile shifts, mind. Certainly the simplest way in which the sowe omitted the time scale. Since the relaxation of the dium and potassium conductances could be functionprofiles is basically a diffusion process, the times ally independent is for the regions of sodium and
involved will be roughly those required for the root potassium permeability to be physically separated.
mean square displacement (?)I/* of an ion to be equal
to the membrane thickness. This is given by the well- HODCKIN-HUXLEY EQUIVALENT CIRCUIT.
Analysis of
known result from Brownian motion theory (13, 14) electrical excitability of the squid giant axon mem-

brane, and any other excitable membrane, starts


with the equivalent circuit for a mosaic membrane of
In aqueous solution, the diffusion coefficient, D, for the type shown in Figure 28. The ions involved in the
small ions such as Na+ and K+ is of the order
circuit can vary depending on the particular biologicm2/s.Taking the membrane thickness as 100 A (lo-'; cal membrane or the ions in the solution that bathes
cm), we find from Equation 66 that the time involved the membrane. Because much is known about the
for the displacement of concentration profiles is of the squid axon, we use this system as representative of
s, that is, 0.1 ps. The events associated the others, with the understanding that the princiorder of
with a n action potential, occurring as they do on the ples operative there are generally believed to apply to
millisecond time scale, are four orders of magnitude all excitable membranes. For the squid giant axon in
slower. We would therefore have to postulate that the normal seawater, the equivalent circuit of Figure 28
diffusion constants (i.e., mobilities) of ions within the takes the particular form shown in Figure 29 (32).
membrane are a t least a factor of lo4 times smaller The sodium and potassium conductances are voltage
than the mobilities of ions in free solution.
dependent, whereas the leakage conductance (g,) Dissociation of sodium and potassium conduct- probably a lumping together of several ion permeaances. One of the distinguishing features of the bilities - is constant. The sodium and potassium
permeability changes occurring during an action po- emf's are, of course, given by the Nernst potentials
tential (or under voltrage-clamp conditions) is the dif- for these ions, and are assumed to be constant, beferent time course of these changes for different ions. cause the concentrations of these ions in seawater
In the squid giant axon, for example, sodium permea- and axoplasm are constant (under the experimental
bility rises and begins to fall before potassium perme- conditions). Since [Na+loutside> [Na+linsideand
x2 = 2Dt

(66)

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

Outside ( s e a water)

183

ion gradients. (It is probably also responsible for the


long-term maintenance of the membrane structure.)
The reader should clearly understand that at present
the dependence OfgK and g N , on membrane potential
must be taken as empirically established relations;
what he must then appreciate is how, given these
relations, the action potential follows from purely
analytical considerations. This is a major concern of
the remainder of this chapter starting with the next
section Current-Voltage Characteristics. Before beginning this treatment, however, we wish to say just
a little more about what possibly might be behind the
voltage dependence of the conductances.
We have already indicated that for various reasons
it is rather unlikely that the conductance changes
can be due to shifting of ionic profiles. The most
popular idea is that the total sodium conductance (for
example) is made up of many separate sites (possibly
channels) which can be in a conducting or nonconducting state (open or closed). The membrane potential then controls the fraction of sites which are open
or closed. (Or, for an individual site, the membrane
potential determines the probability of its being
opened or closed.) The dependence of g,, on voltage
thus results not from changing the conductance of an
individual sodium-permeable site, but rather from
changing the number of sites that are conducting. Of
course, this still leaves unanswered the mechanism
by which voltage controls the opening and closing of
sites. The above picture is a n attractive one and is,
again, the simplest imaginable given the data. We
shall see later that this model can be realized in thin
lipid membranes that have been appropriately doped.

7 c

Inside ( o x o p l a s m )
FIG. 29. The Hodgkin-Huxley equivalent circuit for the squid
giant axon membrane in normal seawater.

[K+loutside < [K+linside> ENa and EK are of opposite


polarity. Ultimately this system would run down
(Na+coming in and K + going out), unless some counterbalancing process exists. This energy-requiring
process is the role assigned to metabolism. The leakage emf (E,)
is, like the leakage conductance, probably a lumped parameter. Although the leakage element modifies the behavior of the system somewhat,
the dominating factors are the voltage-dependent sodium and potassium conductance elements; for the
most part, therefore, we confine our comments to
these and ignore the leakage element.
The Hodgkin-Huxley description of the action potential is briefly summarized as follows: in the restso that the resting potential sits
ing state, g, >> gNa,
near EK. At threshold depolarization, g N a has increased enough that the membrane further depolarizes. This leads to a further increase in g N a , which
leads to further depolarization, and so on. The net
result of this process is the rising phase of the action
potential, which tends toward EN, a t its peak. The Current-Voltage (I-V) Characteristics
falling phase results from a turning off of g,, (called
In our treatment of homogeneous membranes we
sodium inactivation) and a turning on of g K ; both
were able to deal explicitly with the physical basis for
processes act to move the membrane potential back to nonlinearities and time transients in the voltage or
EK. Regenerative behavior develops because the
current response. The present state of knowledge
membrane potential is determined by the relative
does
not permit us to do so for the voltage-dependent
conductances (permeabilities) of the membrane to
conductances
of excitable tissue. Our aim must be the
potassium and sodium [if we neglect the capacitance
more
modest
one of deriving the analytical consecurrent, the membrane potential is given at all times
quences
of
having
elements with conductances that
by the last term in Equation 64, which in this case is
are steep functions of the membrane potential. Nev(gNaEXa + g,EK)/(gNa + gK)],while a t the same time ertheless, a n understanding of these consequences is
these conductances are functions of the membrane essential,
both because of the physiological implicapotential. The functional dependence ofg,, and g, on
tions (e.g., will a fiber produce an action potential or
membrane potential is determined in voltage-clamp
not?), and because it allows us to distinguish the
experiments. That is, the membrane potential is held
physical problems from the mathematical sequelae.
fixed (via external electrodes) a t various levels and
the transient changes in the sodium and potassium NEGATIVE-SLOPE CONDUCTANCE. The existence of a
conductances recorded (29-31, 33).
region of negative-slope conductance9 in the I-V
The physics underlying the voltage-dependent con- characteristic plays a prominent feature in the elecductances is still not understood; it is a major un- trophysiological literature. [Often one sees in the
solved problem in electrophysiology. It is fairly well
" Throughout the remainder of this chapter we use the more
established that metabolism is not directly involved
common electrical notation and call the membrane potential V
in excitability [at least in the giant axon (3411. The rather than T.
role assigned to metabolism in the excitable mem!' Because we are dealing with parallel circuits (e.g., Fig. 29),
brane is the establishment and maintenance of the we talk mostly about conductance rather than resistance.

184

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I


9

literature the term negative conductance or negative


resistance, when what is really meant, of course, is
negative-slope conductance (or resistance). The omission of the word slope is a careless usage, but since
the concept of a truly negative conductance is so
appalling, there is little chance of misunderstanding.] We want to use an elementary example to indicate why this is such a crucial feature of excitability.
Consider the simple circuit of Figure 30. The
steady-state current-voltage relation is given by
Ohms law

I =gv
(67)
where g is the steady-state conductance. If g is constant, that is, not voltage dependent (Fig. 3U),
then
the I-V characteristic is linear (Fig. 3lA).Suppose
instead, that g is an increasing function of V such as
in Figure 31B. The 1-V characteristic shows simple
rectification (Fig. 31R). Now imagine that the g-V
characteristic of Figure 31B is shifted to the left (Fig.
31C), so that the increase of conductance with voltage
occurs in the second quadrant instead of the first. If
the dependence of g on V is not too steep, the I-V
characteristic still shows simple rectification (Fig.
31C),but if g increases steeply enough with V, then
there is a region in the I-V characteristic where the
slope is negative (Fig. 31C). This is because the
product g x V in Equation 67 actually can decrease
(become more negative) in a region where g increases
markedly with V. By the same reasoning it is apparent that if g is a decreasing function of V, the I-V
characteristic can show only rectification if that decrease occurs in the second quadrant, but it may
show a negative-slope region if the decrease occurs in
the first quadrant.
In analytical terms, we have upon differentiating
Equation 67

.
FIG. 31. Possible steady-state g - V characteristics and their
corresponding I-V characteristics for the circuit of Fig. 30.

dg
V-c-g
dV

(69)

(Note that if g is voltage independent, Equation 68


reduces to dlldV = g, which is just a simple, linear
ohmic device with the characteristic shown in Fig.
3U.)
We have just seen that whether the I-V characteristic is a rectifying one or one with a negative-slope
region depends on the quantitative question of how
(68) steeply g changes with V and in which quadrant this
occurs. What then is so special about having a negative-slope region? Simply this: in all I-V curves of
and the existence of a negative-slope region is as- Figure 31,if V is the independent variable, then only
sured if there exists ia voltage range where
one value of I exists for each value of V, and Z
changes smoothly with increasing V. But suppose I is
the independent variable. For the case of a negativeslope region, the voltage response will be given by the
dashed curve in Figure 31C. Starting a t I = - m , V
increases continuously until the point A where dlldV
= 0. An infinitesimal increase in I forces V to move to
point B on the other limb of the curve. The system
has a threshold and an all-or-none response! Similarly, as current decreases from values greater than
a t B, voltage continuously decreases until point C
where dlldV = 0, a t which point a further infinitesimal decrease in current forces the voltage to move to
point D on the first limb.
If we make the reasonable assumption that the
time dependence of g on V is monotonic (i.e., in
FIG. 30. A circuit consisting of only a voltage-dependent conmoving from one point on the steady-state g-V charductance element.

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

acteristic to another the conductance either increases


or decreases but does not oscillate), then the negative-slope region between A and C in Figure 31C is
unstable under current-clamp conditions. Suppose
that we have gotten to point P in Figure 32 by voltage
clamping (i.e., having voltage as the independent
variable) and then suddenly switch to current clamp
so that we are still a t point P, but now current is the
independent variable. Now imagine that a small,
sudden increase in current occurs (from I , to I z ) ,
either spontaneously or by design. Initially the voltage goes to that of point P, since instantaneously one
moves along the chord conductance. (The instantaneous Z-V characteristic is linear, because the membrane events underlying the conductance changes
take time to occur.) But at this increased voltage, g is
no longer in a steady state, and, according to the
steady-state characteristic in Figure 31C, it will begin to increase. This increase in g will lead to a
further increase in V, and so on, so that V will
continuously increase (move to the right) until it
reaches point Q . Similarly a small decrease in Z from
I , to I:, will initially move the voltage to P and
ultimately over to point S . Thus points in the negative-slope region are unstable under current control,
and hence with current the independent variable,
this region is inaccessible. This, by the way, is one of
the main reasons for using the voltage-clamp technique to study excitable tissue. It enables one to gain
access to regions of the I-V characteristic that are not
accessible under current control.
The situation in nerve is much more complicated,
because one does not deal with steady-state I-V
curves, but rather with a series of I-V curves at
different times. (Also there is more than one voltagedependent element, and there are emf s in series with
these elements.) Nevertheless, the elementary example we have presented contains the essence of the
importance of negative-slope conductance. The existence of such a region allows the possibility of instabilities and discontinuities in the voltage response.
Such characteristics are required for an excitable
system with a threshold.
CHANGING THE

1-vCHARACTERISTIC WITHOUT CHANGE


I

185

FIG. 33. Two circuits (A and C) with the same steady-stateg-V


characteristic ( B ) .

FIG. 34. A: the steady-statel-V characteristic for the circuit of


Fig. 33A, given the steady-stateg-V characteristic of Fig. 33B. B:
the steady-state I-V characteristic for the circuit of Fig. 33C,
given the same steady-stateg-V characteristic (i.e., Fig. 338).

OF THE g-V CHARACTERISTIC. We noted above that if


g increases with V in the first quadrant (Fig. 33B),
the I-V characteristic for the circuit in Figure 33A is
monotonic (Fig. 34A). Now suppose we place a positive emf of magnitude E in series with this conductance element, so that the equivalent circuit is now
that of Figure 33C. (In terms of a membrane system,
Fig. 33A corresponds to the situation in which the
concentration of the penneant ions is the same on
both sides of the membrane, whereas Fig. 33C corresponds to the case of a concentration difference of the
permeant ions across the membrane.) Also assume
that the existence of the emf, E , does not alter the
dependence ofg on V; that is, Figure 33B continues to
hold independent of the value of the emf. We see from
Figure 34B that the I-V characteristic now has a
region of negative slope. This arises because atZ = 0,
V = E . But, when V = E , the system is already a t a
high conductance. If we go to values of V less than E ,
then conductance decreases, and it decreases steeply
enough to produce a region of negative slope.
Analytically, instead of Equation 67

z = gv

(67)
which is Ohms law for the circuit of Figure 33A,
Ohms law for the circuit of Figure 33C takes the form

Z = g(V - E )
From Equation 67 we had
32. A steady-state I-V characteristic with a region of
negative slope. This is an enlargement of Fig. 31C, illustrating
why any point P in the region of negative slope is unstable under
current-clamp conditions (see text).
FIG.

-dl= g + v dV

dg
dV

(70)

(68)

and since the g of Figure 33B is increasing with


voltage (dg/dV > 01, dZldV in Equation 68 will never

186

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

be negative for positive values of V."' From Equation 70, however, we have

dl

dg
E )(71)
dV
The second term is now negative for 0 < V < E , so
that if dgtdV is large enough, the slope dlldV in this
region can be negative. Thus simply with the insertion of a positive emf in the circuit, the rectifying
characteristic of Figure 34A is converted to a characteristic with a region of negative-slope conductance
(Fig. 3 B ) , without an,y change whatsoever i n the g-V
characteristic. Similarly the Z-V characteristic with a
negative-slope region shown in Figure 31C" (based
on the g-V characteristic of Fig. 31C) can be converted to a rectifying characteristic by the insertion
of a negative emf into the circuit of Figure 30.
The example just given is illustrative of a very
important fact of electrophysiology, well known to
investigators in the field, which the reader must
understand if he is to appreciate the level of certain
problems: the physics underlying excitable phenomena are contained i n the g-V characteristic. That is,
the physical question is: what makes g such a steep
function of V over certain voltage ranges? As we have
stated earlier, this is a t present an unsolved problem.
Given this g-V characteristic, whether or not the Z-V
characteristic has a region of negative slope can depend simply on the presence and magnitude of an emf
(concentration gradient of permeant ions) in the system. The Z-V characteristic can be switched from a
rectifying one to an excitable one merely by changing
the permeant ion concentration on one side of the
membrane without affecting the g-V characteristic. ' I
Which Z-V characteristic obtains is a trivial mathematical consequence of Ohm's law. If the investigator
is unaware of this, he may believe, in viewing such
different I-V characteristics as in Figure 3 4 4 and B ,
that he is dealing with two completely different physical phenomena, whereas in point of fact the underlying membrane events are identical. Whether or not a
particular cell can be excited may be determined by
ion concentrations in the media that have no direct
effect on underlying mechanisms. (We shall see with
excitable thin lipid membranes that divalent cations
can have profound effects merely because of changes
they produce in the surface potential.)
Although the presence or absence of a region of
negative-slope conductance in the Z-V characteristic
is a trivial one, if one is concerned about underlying
physical mechanisms, it is, of course, of paramount
-

dV

+ (V

"' We have drawn the g-V characteristic so that for negative


values of V, g is a constant. Therefore the second term in Eq. 68
vanishes, and dlldV cannot be negative in the third quadrant
either.
' I We are assuming that the introduction or removal of the
concentration gradient (i.e.,the E ) does not affect theg-V characteristic. In our examples with excitable lipid membranes, we give
examples where this is true and also an example where this does
not hold.

importance to the cell itself and to the neurophysiologist who studies it, for it determines whether the cell
can generate and conduct action potentials under
normal circumstances, with all the implications that
follow from that. We have tried to stress the conceptual level at which the problem exists. If one is studying the development of embryonic excitable tissue,
and the onset of excitability occurs at some critical
point in time, this may mean that macromolecules
crucial for voltage dependence have been inserted
into the membrane. It may also mean, however, that
the internal sodium or potassium levels have
changed somewhat. To the biochemist or the biophysicist concerned with the mechanism of excitation,
this is a crucial issue. To the neurophysiologist interested in the integrative behavior of the tissue, the
mechanism for the onset of excitability is of peripheral interest. Our remarks about the presence or
absence of excitability also extend to the shape and
duration of action potentials, when these occur. Profound changes in kinetics can be realized by seemingly minor changes in ionic environment that do not
affect the time dependence of g on V. Again the
importance attached to the mechanism of these
changes depends on the interests of the investigator.
VOLTAGE-DEPENDENT CONDUCTANCE
I N THIN LIPID MEMBRANES

From the time of the classic observation of Gorter


& Grendel(25) that the extracted lipids from erythro-

cyte membranes cover twice the surface area of the


cells, a major tenet of membrane physiologists has
been that the plasma membrane contains (among
other components) a bilayer of lipid. The DavsonDanielli paucimolecular model of the plasma membrane (91, which is the model most in harmony with
optical, biochemical, and physiological data, consists
of a bilayer of phospholipid (and cholesterol) with
protein both adsorbed a t the surfaces and interpenetrating the lipid to provide polar pathways for hydrophilic molecules. The membrane, as visualized by
physiologists, is a mosaic structure consisting of a sea
of lipid bilayer, itself an impermeant barrier to ions
and polar nonelectrolytes, in which are immersed
specialized pathways. (Biochemists, with a more recent interest in cell membranes, seem to have rediscovered and reapproved this model for themselves. 1
Since the advent of the Davson-Danielli model,
attempts have been made to reconstruct its major
structural characteristics in vitro, in order to study
its permeability properties explicitly. In 1963 Mueller
et al. (56) described for the first time a convenient
method for forming a phospholipid bilayer, separating two aqueous phases, whose permeability and
electrical properties could then be studied. Subsequently, Mueller and Rudin discovered several molecules which, when introduced into the bilayer, pro-

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

duced ion-conducting regions exquisitely sensitive to


membrane potential; indeed, given the appropriate
ion concentrations on the two sides of the membrane,
such systems can develop full-fledged action potentials and rhythmic oscillations (51, 53, 54). It is the
description and analysis of these doped membranes
that we wish to present in the context of electrical
excitability in cell membranes. Before doing this,
however, we first describe the formation and properties of unmodified phospholipid bilayers and briefly
mention the characteristics of some nonvoltage-dependent modifiers that can be introduced into these
membranes as background information with which to
assess and appreciate the excitability phenomena
presented in the subsection Voltage-Dependent Modifiers.
The Unmodified Thin Lipid Membrane
FORMATION. The membrane is formed across a hole
(approximately 1 mm2) in a plastic partition (e.g.,
polyethylene, Teflon) separating two aqueous salt
solutions. The hole provides the only connection between the two aqueous phases, and one interposes
the membrane by painting under water an appropriate lipid solution across the hole with a small
sable-hair brush (Fig. 35). Initially the brush deposits
a thick film which, when viewed in reflected light,
shows beautiful interference colors. This film spontaneously thins to an optically black structure that is
the thin lipid membrane of interest. [It is optically
black because its thickness (-75 A) is much less than
the wavelength of visibile light (-3,000 A); hence the
reflected light from the front face, which undergoes a
180 phase change on reflection, destructively interferes, almost completely, with reflected light from
the rear face, which undergoes no phase change (74).1
Most of the deposited lipid solution ends up in a thick
torus around the edge of the hole. The final thin
structure covering most of the area is a phospholipid
bilayer, the backbone of the Davson-Danielli model
(Fig. 36).
The electrical properties of the film are convenELECTRODE

PARTITION
ELECTR0 DE

CLEAR
PLASTICWINDOW

FIG. 35. The arrangement for forming and studying bilayers


on a planar plastic partition separating two aqueous phases.

I
I

I
Aqueous
phase

I Hydrocarbon

187

phase

I
I

Aqueous
phase

FIG. 36. Schematic drawing of a phospholipid bilayer. Filled


circles, polar ends of the phospholipids; wavy lines, their fatty acid
chains. The latter make up the hydrocarbon interior of the membrane, whereas the former anchor this region to the two aqueous
phases.

iently measured by either passing steps of current


across the membrane and recording the voltage responses (current clamping), or by applying steps of
voltage and recording the current responses (voltage
clamping). Usually a single pair of electrodes is used
for both stimulating and recording, but, if necessary,
separate electrode pairs can be used. Permeability
measurements are performed by introducing a radioactive isotope of the species of interest into one
aqueous phase and recording its rate of appearance in
the other. Despite the thinness of these membranes,
they can be stable for many hours and withstand
many experimental manipulations without breaking.
Membranes can be formed from a variety of amphipathic molecules dissolved in appropriate organic solvents. A very simple membrane-forming solution is a
1-3% solution of synthetic dioleoylphosphatidylcholine in n-decane. A much more complex solution consists of a mixture of the total Folch lipid extract of
kosher ox brain white matter plus dl-a-tocopherol
dissolved in ch1oroform:methanol. Cholesterol is often included in the membrane-forming solution in
various proportions. Whereas the chloroform:
methanol solvent diffuses away into the aqueous
phases, some hydrocarbon (e.g., decane) does remain
in the film. Since only a minute fraction of the total
solution applied across the hole ends up in the thin
membrane (the rest going into the torus), one cannot
assert a priori that the proportions of the various
components in the film are the same as those in the
membrane-forming solution.
PERMEABILITY AND ELECTRICAL PROPERTIES. In response to a step of current, the voltage across the
membrane rises exponentially to a steady-state value
(Fig. 37A ). The behavior is that of the simple parallel
RC circuit shown in Figure 37B. The capacitance is
about 0.5 pF/cm2, which is consistent with an approximately 50-A-thick region of low dielectric constant
hydrocarbon ( E = 2) separating two conducting

188

HANDBOOK OF PHYSIOLOGY

-c

THE NERVOUS SYSTEM I

t ,

FIG. 37. Diagrams of the response of a parallel RC circuit ( B )


to a step of current ( A ) o r a step of voltage (0.
(The voltageclamp response, C , is actually that which occurs if there is a small
resistance in series with the circuit inB; without that resistance,
the current response is a 8 function.) These are the responses of
an unmodified lipid bilayer.

(aqueous salt solution) phases; that is, the two solutions form a parallel plate capacitor. The resistance,
obtained by dividing the steady-state V by the applied I, is remarkably high, being about lon a*cm2.
The resistance is purely ohmic up to at least *75 mV.
Above this value it may fall, and at voltages larger
than 150 mV, the membranes often break. (The voltage at which membranes break varies considerably
with the membrane composition.) Breakage probably
results from dielectric breakdown, caused by the
intense fields (several hundred thousand volts per
centimeter) that exist across the membrane when
moderate potential differences are applied across
such a thin structure.
In response to a step of voltage across the membrane, there is a surge of capacitance current (CdVI
dt) which rapidly decays exponentially (Fig. 37C).
[Ideally the response of the circuit of Fig. 37B to a
step of voltage is an infinite capacitance current lasting for zero time, since dVldt is infinite for zero time
and is subsequently zero; this current surge then
charges the capacitor to the potential V. In practice,
however, there is always a small resistance in series
with the membrane (produced by the resistance of the
aqueous solutions in contact with the membrane),
and this causes the charging time of the membrane
capacitance to be finite, since dVldt across the capacitor is now finite.] The small steady-state current I
that remains, after the capacitance surge, flows
through the membrane resistance, which is calculated by dividing the applied voltage V by I.
The large electrical resistance reflects the virtual
impermeability of the membrane to small ions, which
is confirmed by the absence of significant tracer ion
flux across the membrane. Tracer measurements also
reveal a very low permeability to polar nonelectrolytes (e.g., urea, sugars) but a very high permeability

to lipophilic solutes (e.g., indole, butanol). All these


results are consistent with the view that the membrane behaves simply like a thin layer of hydrocarbon
(18). Molecules traverse the membrane by a solubility
diffusion mechanism; that is, they partition into the
hydrocarbon region of the membrane and then diffuse through it, The impermeability to ions and the
poor permeability to hydrophilic nonelectrolytes reflect their low partition coefficient between hydrocarbon and water. Conversely, nonpolar molecules are
highly permeant because of their solubility in hydrocarbon. Thus the permeability and electrical properties of unmodified thin lipid membranes are determined almost exclusively by the hydrocarbon tails of
the lipids (the region included between the two
dashed lines in Fig. 36);the polar head groups serve
merely to anchor this region between two aqueous
phases.
This description of the mode of transport through
lipid bilayers is best illustrated by the water permeability data. The permeability coefficient for water is
semiquantitatively predicted from the solubility and
diffusion coefficient of water in bulk hydrocarbon
such as hexadecane (18).This also illustrates the very
important concept that the hydrocarbon interior of
the film is fluid and is much more like liquid hydrocarbon than like solid wax. Macroscopically this is
obvious from the ability of the films to withstand
insertion and withdrawal of a micropipette without
breaking or developing a leak; the membranes have a
self-sealing property, which, not so incidentally, is
also characteristic of plasma membranes.
In accordance with this model, membrane permeability to water (and other nonelectrolytes) can be
increased or decreased by varying the lipid composition. In particular, cholesterol, which has a wellknown condensing effect on lipid monolayers, decreases the permeability to water and other nonelectrolytes (18). In terms of the solubility-diffusion
mechanism of transport, this condensing action leads
to an increased viscosity of the hydrocarbon region
with a consequent decrease in diffusion coefficients
within this phase. It probably also leads to decreased
partition coefficients into the hydrocarbon region.
The very high resistance and simple ohmic behavior of lipid bilayers make them relatively uninteresting to physiologists, particularly electrophysiologists, as models for plasma membrane function. If
one believes, however, that these membranes are the
backbone of the Davson-Danielli model and that this
model is a good one for plasma membranes, then local
modifications of the bilayer structure must be introduced to create permeability shunts for ions and
small polar nonelectrolytes. In plasma membranes
these shunts are presumably created by proteins (although there may be exceptions to this). As of now,
attempts to introduce into the films proteins that
produce interesting permeability characteristics
have not been very successful. However, several

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

189

small (mol wt ca. 1,000) molecules, some of which are


peptides, have been discovered that dramatically increase the membranes permeability to ions by mechanisms that may be directly related to those operating biologically. In a few cases, the ion permeability
induced by such agents is strongly voltage sensitive;
these examples are considered in the subsection Voltage-dependent Modifiers. In the other instances the
ion conductances induced are essentially ohmic. They
are nevertheless of sufficient interest, even to electrophysiologists, that we shall now describe and discuss
some examples of these modifiers.

Nonuoltage-dependent Modifiers
The reason for the virtual impermeability of lipid
bilayers to ions is the immeasurably small partition
coefficient of small ions (e.g., Na+, K+, and C1-)
between hydrocarbon and water. The dominating factor for this small partition coefficient is the enormous
electrostatic energy required to transfer a charge of
radius 2 A (the approximate size of the small ions)
from a medium of high dielectric constant (i.e., water) into one of low dielectric constant (i.e., hydrocarbon). From electrostatic theory this energy, U ( r ) ,is
related to the radius, r , of the ion and the dielectric
constants E , and E, of the two media by the formula

FIG 38. Schematic diagram of a cation carrier, such as valinomycin. The carbonyl oxygens provide a polar environment for the
cation, and in fact substitute for the first hydration shell normally surrounding the cation in water. The exterior aspect of the
molecule is nonpolar, thus making it compatible with the hydrocarbon interior of the membrane.

phobic exterior of the structure shown in Figure 38.


In this way partition coefficients of
or larger can
be obtained for small ions. Such partition coefficients
result in large enough concentrations of the complex
in the membrane to yield significant conductances,
assuming a reasonable diffusion coefficient for the
complex in the membrane.
Another way to reduce the energy barrier for ions
(72) is to introduce a polar channel into the membrane, so
that an ion entering the membrane now moves from
For the case of an ion such as K+ (r = 2 A) being one high dielectric constant region (bulk aqueous
transferred between water ( E ~= 80) and hydrocarbon solution) into another (the polar environment of the
(E, = 2), U ( r )= 1.5 eV. From the Boltzmann distribu- channel). Since this channel is embedded in the low
tion, the partition coefficient, P h r l H z ( ) , between hydro- dielectric constant hydrocarbon of the membrane, it
carbon (hc) and water is then given by (remembering still requires energy to effect the transfer from solution into the channel (601, but not nearly so much as
that at room temperature, kT/q = 25 mV)
that given in Equation 72.
Both ion transport mechanisms just discussed
- e--liUirllk? = e - l . 5 0 0 i 2 S
10-26 (73)
phclHd)c h c (transport
via a carrier or through a polar channel)
CHzO
have been realized experimentally in lipid bilayers.
which is a very small number indeed. The above Since such mechanisms have traditionally been procalculation, which neglects possible short-range posed by physiologists to account for ion conductances
forces and structural considerations of the solvents, is in plasma membranes, it is particularly pertinent to
clearly a n oversimplification quantitatively. How- see how such systems can act in a n artificial memever, even if these other considerations modify the brane of similar dimensions, chemical composition,
basic electrostatic term by a factor of say lo6, the and structure.
partition coefficient into the hydrocarbon phase is
CARRIERS.
Of the molecules now known to function as
still miniscule.
There are at least two ways to reduce drastically ion carriers, the most interesting is the class of cyclic
this electrostatic energy barrier. One is effectively to antibiotics that transport alkali cations. Examples of
increase the size of the ion by enclosing it in a spheri- these are the depsipeptides [e.g., valinomycin and
cal dielectric cavity, with the inside of the cavity the enniatins (52)] and the macrocyclic tetralides (70)
lined by polar groups to provide a favorable environ- (e.g., nonactin, monactin). They are all cyclic molement for the ion (Fig. 38). An increase of r to 6 A, cules that have many oxygen atoms, either carbonyl
although only a factor of 3, introduces this factor into or ether (Fig. 39). These oxygens form the lining of
the exponent of Equation 73, so that the partition the cavity for the prototypic structure shown in Figcoefficient increases by a factor of
This number ure 38, and in essence they substitute for the H,O
can further be increased by many orders of magni- oxygens in the first hydration shell of the ions in free
tude, because of the intrinsic solubility of the hydro- solution. That is, the naked alkali cation exchanges
~

190

HANDBOOK OF PHYSIOLOGY

fl

THE NERVOUS SYSTEM I

NONACTIN
ENNlATlN B
FIG. 39. Structural forrnulas for the cation carriers nonactin
(11) and enniatin B (64).

its hydration shell for these carbonyl and ether oxygens. Although these molecules are cyclic, the ion
does not sit in the hole of the doughnut. Rather, the
molecule folds around the ion to create the spherical
cavity (10, 37, 62). In so doing, the polar oxygens
surround the cation and are in turn "buried" within
the complex; the external aspect of the complex is
completely nonpolar, making it very lipid soluble.
One of the most interesting aspects of these molecules is that they can markedly discriminate among
the alkali cations whose membrane permeability
they increase. Valinomycin, for example, has the
selectivity order: Rb+ > K + > Cs+ > Na' = Li+, with
K+ permeability a t least 300 times greater than that
of Na+ (52). The order of selectivity is established
either by comparing conductances of valinomycintreated membranes in Na+ and K+ media (all concentrations being identical), or by measuring bi-ionic
potentials between KC1 and NaCl across valinomycin-treated membranes. The order and degree of selectivity reflect the magnitude of the difference in
free energy between the ion in its hydrated form in
water and the ion "hydrated by the oxygens of the
antibiotic.
The conductance increases produced by these antibiotics are obtained by adding the antibiotics in micromolar or smaller quantities to one or both aqueous
phases. Membrane conductance increases linearly
with both the antibiotic concentration and the cation
concentration (461." Both of these results are consistent with the view that the current-carrying species is
of the form: (antibiotic-cation)+.In symmetrical salt
solutions, the Z-V characteristic is linear up to -+50
mV; in salt gradients, rectification is obtained consistent with the shifting of ionic profiles in the membrane (52). (Since the membrane is thin, a nonlinear
I-V characteristic obtains even in the single-salt
case - the electroneutrality condition not holding. 1
Clearly one is dealing with a carrier system. The
antibiotic forms the complex with the cation at one
interface, diffuses across the membrane, and then
releases the cation a t the other interface.
I 2 The linear dependence of conductance on cation concentration is obtained on membranes made with lipids having no net
charge. We shall see later that the surface potential, associated
with charged lipid membranes, can complicate these results.

The ability of these antibiotics to pull alkali cations


into an organic phase is not confined to lipid bilayers.
The same phenomenon is observed in bulk partitioning experiments between water and organic solvents,
provided that a soluble anion is present to follow the
complex and thus preserve electroneutrality (16, 66).
As we saw in our earlier discussion of partitioning of
ions between oil and water, the macroscopic partition
coefficient is proportional to the square root of the
product of the individual partition coefficients (see
Eq. 31). If the anion's partition coefficient is virtually
zero, there will be no movement of salt into the
organic phase regardless of how favored the antibiotic-cation complex is in this phase.
In this context the reader should also note that,
although a lipophilic anion is needed to demonstrate
bulk partitioning, such an anion is not needed for the
effects produced by the antibiotics in the bilayers.
Because of the thinness of the hydrocarbon region,
relatively low concentrations of antibiotic-cation
complex in this phase (of the order lopnM) will produce significant membrane conductances. The Debye
length is considerably larger than the membrane
thickness, so that the space-charge region extends
throughout the membrane phase. In fact, until the
ion concentration reaches approximately
M, one
can neglect any space-charge limitation in conductance. That is, one can assume that the space-charge
density, p , in Poisson's equation is zero and that
consequently the constant-field approximation is
valid (59, 77). The validity of this statement is supported by the fact that conductance varies linearly
with antibiotic and cation concentration over a considerable range. In any case, the point that we wish
to stress is that the complexing and partitioning
properties of these antibiotics is an inherent property
of the molecules themselves, and they do not require
a bilayer structure for their action to be manifest. For
these molecules to increase significantly the conductance of a membrane, however, a thin structure is
necessary. As we shall see in the next section, the
action, and indeed the existence, of channels requires
a bilayer.
CHANNEL FORMERS. T h e polyene antibiotics nystatin
and amphotericin B. The polyenes are also cyclic
antibiotics, but they differ drastically in chemical
structure from the molecules just considered [(38);
Fig. 401. They have a lactone ring, many hydroxyl
groups, and a polyenic chromophore of from 4 to 7
double bonds. The two we are considering have in
addition one carboxyl group and one amino sugar;
between pH 5 and 8 they are zwitterions with no net
charge. Nystatin and amphotericin B are quite similar in both their structure and in their action on
bilayers; for the purposes of this discussion we do not
distinguish between them. Chemically the major difference is that amphotericin B is a heptaene, whereas
in the nystatin molecule a tetraene and diene are
present separated by a saturated carbon-carbon link-

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

191

FIG. 40. Structural formula of amphotericin B (47). Nystatin


differs from amphotericin B in that it contains a tetraene and a
diene in place of the heptaene chromophore; in other respects the
molecules are very similar.

age. Aside from the details of the chemistry, the most


striking difference between these molecules and the
carriers is that the latter are very lipophilic and thus
quite soluble in hydrocarbon. On the other hand,
nystatin and amphotericin B are not very soluble in
either water or hydrocarbon, as one might guess from
their formulas. They are amphipathic molecules,
with a hydrophilic portion consisting of the hydroxyl
groups, the amino sugar, and the carboxyl group, and
a hydrophobic portion made up of the polyene chain.
The segregation of these two aspects of the molecule
is most strikingly seen in the space-filling molecular
model of amphotericin B (Fig. 41).
In amounts micromolar or less, nystatin (or amphotericin B) greatly increases the conductance of
sterol-containing (usually cholesterol) bilayers (6).
[This is consistent with their action on biological
membranes. Polyenes will cause leakage of small
molecules through plasma membranes containing
sterol. They are ineffective on bacterial and mitochondrial membranes - membranes that are sterol
free (381.1 The conductance increase is primarily due
to increased univalent anion permeability. Concomitant with the conductance increase is a n increase in
membrane permeability to water and small nonelectrolytes (35). The permeability to nonelectrolytes decreases with increasing Stokes-Einstein radius of the
molecule; molecules larger than glucose (radius = 4
A) are virtually impermeant. To obtain these effects
on conductance and nonelectrolyte permeability, it is
usually necessary to add the antibiotic to the aqueous
compartments on both sides of the membrane (6). In
contrast to the linear dependence of conductance on
antibiotic concentration observed with the carriers,
nystatin-induced conductance is proportional to a
large power of the antibiotic concentration [6th-l2th
power, depending on the lipid (6)l.
All these results are consistent with the idea that
many nystatin (or amphotericin B) molecules interact with membrane sterol to form aqueous channels
in the membrane. In fact, given that the polyene can
span half of the bilayer and that maximal effectiveness is achieved by adding antibiotic to both sides of
the membrane, the clear implication is that the com-

FIG. 41. Molecular model (CPK) of amphotericin B. In A the


completely hydrophobic face of amphotericin B is seen; in B the
molecule has been rotated 180 about its long axis to reveal the
opposite face with its many hydroxyl groups. Note that the molecule consists of two chains: a polyene chain (seen on the right in A
and on the left in B ) and an amphipathic chain. The hydrophobic
and hydrophilic faces of the amphipathic chain are seen in A and
B , respectively. At the bottom of the figures are the polar amino
sugar and carboxyl groups; at the top is a single hydroxyl group
(seen most clearly in B ) . In C, a CPK model of lecithin is shown
for comparison with amphotericin B. [From Finkelstein & Holz
(19).reprinted by courtesy of Marcel Dekker, 1nc.l

plete channel is formed from two half channels


interacting from opposite sides. Figure 42 is a suggested model of the channel. Note that the exterior
aspect of the channel is completely hydrophobic,
making it compatible with the bilayer interior,
whereas the interior of the channel is copiously lined
by hydroxyl groups. The ion selectivity presumably
results either from the partial substitution of these
hydroxyls for the anion hydration shells or from a n
orientation of the OH dipoles that makes the interior of the channel positive.
As with the carriers, the polyene-induced conductance is ohmic with identical solutions on the two
sides of the membrane. In the presence of a singlesalt gradient or bi-ionic situation, nonlinearities in
the Z-Vcharacteristic occur, consistent with a priori
notions of shifting ionic profiles within the channel
(6).
It is interesting that, despite the complex multimolecular structure of the channels, they are not permanently fixed regions in the membrane. Removal of
the antibiotic from aqueous solution causes the conductance to fall with time, until it returns to the low
value characteristic of unmodified membrane. This
clearly implies that in the presence of antibiotic there
is a dynamic equilibrium between molecules in
aqueous solution and those in the membrane and
that therefore channels must be continuously forming and breaking up. This further illustrates the fluid
nature of the bilayer system. Significantly, the
permeability increases produced by nystatin and am-

192

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

photericin B in plasma membranes (e.g., red cells)


are also readily reversible (5).

HCO-L-Val-Cly-L-Ala-D-Leu-L-Ala-D-Val-L-Val-D-Val-L-Trp
D-Leu-L-TrpD-Leu-L-Trpo-Leu-L-TrpNH.
CH,CH,OH

Gramicidin A . This antibiotic is a linear pentadecapeptide (Fig. 43) which a t concentrations of


M
or less significantly increases the conductance of bilayers to univalent cations (22, 23). The antibiotic is
effective from one side of the membrane, but acts
much more rapidly when added to both sides (personal observation). It is insoluble in water and hydrocarbon, but extremely surface active; its effect on
membrane conductance is not readily reversed by
removing the molecule from aqueous solution. For
practical purposes the association of gramicidin A
with the membrane is irreversible (26). The selectivity between univalent cations and anions is complete,
but, unlike the carriers, there is very little discrimination among the alkali cations. The order and degree of selectivity is approximately the same as that
of the sequence obtained from the mobilities of the
ions in free solution (28). As with the other molecules
we have discussed, gramicidin A-treated membranes
are ohmic in symmetrical solutions.
One of the fascinating observations made with this
system is that, if sufficiently small amounts of gramicidin A are added to solution, it is possible to see
fluctuations in membrane conductance (28). There
appears to be a unit fluctuation (whose size is about 5
x lo- 0- in 0.1 M NaCl), and larger fluctuations

FIG. 43. Structural formula of valine-gramicidin A (68).(Isoleucine gramicidin A has t-Ile in place of the first L-Val.)

are multiples of this size (Fig. 44). The frequency and


duration of these fluctuations are not voltage dependent.13The clear implication is that the unit conductance step is the conductance of a single site. From its
magnitude, it is almost certain that the ions are
moving through a channel, rather than being transported by a carrier (28). The suggested structure of
the channel is shown in Figure 45. Two monomers of
gramicidin-A hydrogen bond to form a helical structure that spans the membrane (76). The outer aspect
of the helix is hydrophobic, allowing it to sit in the
bilayer, whereas the interior (r = 2 A) is lined by
carbonyl oxygens to form a polar pathway for cations.
These oxygens are assumed to completely or partially
substitute for the inner hydration shell of the cations.
This is the channel analogue for the cation carriers
discussed earlier.
The general feature of the model in Figure 45 is
supported by a number of observations (although the
details of the model might ultimately have to be
modified). Kinetically the rate of conductance increase is proportional to the square of the antibiotic
concentration, suggesting that two gramicidin molecules are necessary to form a conduction path (22).
The increased efficacy of antibiotic added to both
sides of the membrane is also consistent with this
picture. Also the length of the channel model in
Figure 45 is about 30 A. Since the thickness of the
bilayer can be varied by changes in the membraneforming solution, it is gratifying that gramicidin A is
more active on thinner membranes (thicknesses near
30 A) and that the lifetime of a single channel, as
measured by the fluctuations, increases with decreasing membrane thickness (28). Finally, malonylgramicidin A, the dimer formed by chemically linking two
monomers head to head, is active, and in this case the
rate of conductance increase is linear with concentration (76).
The conductance fluctuations presumably represent formation and dissociation of single dimeric
channels. It appears that the total number of gramicidin A molecules in the membrane is much greater
than the number of channels and that a channel
FIG. 42. Half of a n amphotericin B-created pore. Each amphotericin B molecule is schematized a s a plane with a protuberance forms when two monomers associate to form a dimer
and a solid dot. The shadtd portion of each plane represents the (28); the dissociation of the dimer might then be the
hydroxyl face of the amphipathic chain, the protuberance represents the amino sugar, and 0 represents the single hydroxyl event that closes the channel. In this case, a memgroup a t the nonpolar end of the molecule. The aqueous phase is brane with either one or zero conducting units at any
at the bottom of the figure and the middle of the membrane is a t one time consists not of one channel opening and
the top. We see that the interior of the pore is polar, whereas the closing repeatedly, but rather of a different channel
exterior is completely nonpolar; there is also a wedge in the
(i.e., a different pair of monomers) which forms and
exterior of the pore, between each pair of amphotericin B mole-

U Y P

cules, that can accommodate a sterol molecule. Note the ring of


hydroxyl groups in the middle of the membrane that can hydrogen bond with a n identical structure from the other side to form a
complete pore. [From Finkelstein & Holz (19).1

In some membranes the membrane thickness decreases with


increasing voltage. In these cases the frequency and duration of
the fluctuations increase with membrane potential.

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

the present instance the conductances are not voltage


dependent; they are only functions of the concentrations of the antibiotics. A membrane treated with
both nystatin and valinomycin is also interesting in
that it has ion transport occurring both through
channels (the chloride conductance) and via carriers
(the potassium conductance).
For the mosaic membrane of Figure 46A, there is
clearly only one possible stable value for the membrane potential (in the absence of an applied current)

V=
FIG. 44. Conductance fluctuations for a membrane treated
with a small amount of gramicidin A. (Actually one is seeing the
current fluctuations in the face of a constant voltage across the
membrane.) The membrane was formed from glyceryl monooleate dissolved in hexadecane and separates 0.5 M NaCl solutions
at 23C. The events marked a and p on the left-hand side of the
record occurred very infrequently. [From Hladky & Haydon (28).1

193

+ gCIECl
g K + gC1

gKEK

where g K and g,, are functions of valinomycin and


nystatin concentrations, respectively. Although it
adds little to the description of this system, it is
useful for our future discussions of voltage-dependent
systems to demonstrate this graphically. By definition, in the absence of a n applied current

breaks up anew each time. The observation that malonylgramicidin A channels have much longer lifetimes (personal observations) supports this picture of
channel formation and disassembly.
The proposed size of the gramicidin A channel (2-A
radius) is smaller than that postulated for the amphotericin B channel (-4-A radius). The observation
that gramicidin A channels are permeable to water
but impermeable to urea, in contrast to the significant urea permeability associated with the polyene
channels, is consistent with this estimate of size (17).
A MOSAIC MEMBRANE FORMED WITH TWO MODIFIERS.

Implicit in our discussion of carriers and channels is


that, even if membrane conductance has been increased by a factor of lo6 from that of the unmodified
membrane (i.e., from
to
flR-*.cm-2
), most of
the bilayer remains unmodified. For gramicidin A,
where the conductance per channel has been measured and channel size is fairly well known, one can
explicitly calculate the fraction of the membrane area
occupied by (open) channels a t a given conductance
and show that this is indeed a small number for any
practical conductance that can be achieved. The most
clear-cut and significant illustration of this concept,
however, is the demonstration that one can introduce
two modifiers into the membrane and obtain results
predictable from the hypothesis of parallel, independent action of the two systems.
This is seen in a membrane treated with both
valinomycin and nystatin (18). The former makes the
membrane cation selective, whereas the latter makes
it anion selective. In the presence of a KC1 gradient,
the membrane potential can assume all values between EK and (approximately) Ecl,depending on the
relative amounts of the two antibiotics added. The
equivalent circuit for this system is shown in Figure
46A, which is a particular example of the general
mosaic membrane circuit of Figure 28. Note that in

FIG. 45. Proposed structure of the gramicidin A channel. A:


side view of a CPK molecular model of two molecules of gramicihelix conformation, linked by three hydrogen
din A in the
bonds through their formyl ends. B: end view of the same model
showing the central channel 2 A in radius. [From Urry (75).]

194

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

FIG. 46. A: equivalent circuit for a membrane treated with


both nystatin and valinornycin and separating KCl solutions at
different concentrations. (The membrane capacitance is shown
for completeness.) B: plot of I, and I , . , as a function of V for the
elements of the circuit in A . The slopes of these lines are g, and
g,.,,respectively. At point.A, IK= -I,.,; the value of V at this point
is the membrane potential (in the absence of an external current
source). Note that A must lie between E , and El,,;its exact
position is determined by the slopes ofIKandI,.,ii.e.,gKand g , . , ) .

fiers is reasonably advanced. There still remain important questions of detail (e.g., how do the ions lose
their hydration shell; what is the rate-limiting step
in transport?) that will continue to intrigue physiologists and physical chemists, but it is probably correct
to say that the general pictures of the mode of transport induced by the various modifiers we have discussed are essentially correct. This happy state of
affairs does not hold at present for the voltage-sensitive modifiers presented in the next section. Consequently, although we shall include some modest speculations about the mechanisms of the voltage dependencies, our major emphasis is on the phenomenology of the conductance changes and on the experimental realization of the formal consequences (which
we discussed somewhat in the subsection CurrentVoltage Characteristics)of having voltage-dependent
conductance elements. Specifically, we shall see how
electrical excitability can be produced in thin lipid
membranes with these molecules.
Voltage-dependent Modifiers

Although it has been possible to generate currentvoltage characteristics that bear certain resemblances to those of excitable membranes by means of
polarization effects in the unstirred aqueous layers
z =- z, + = 0
on the two sides of the membrane, the resemblance is
superficial and clearly not relevant to biological phewhere
nomena (55).On the other hand, three modifiers have
been discovered which confer on bilayer membranes
voltage-dependent conductances that bear striking
similarities to those seen biologically: 1) monazomycin, 2) alamethicin, and 3) excitability-inducing maIn Figure 46B we plot ZK and Z(.,,for arbitrary values terial (EIM).
Monazomycin (1, 48) is an antibiotic of mol wt
ofg, andg,.,, as a function of V. Note that both curves
1,200
produced by a Streptomyces. Its empirical forare straight lines (the gs are constant), and hence
and its functional groups inthere is only one point, A, where the algebraic sum of mula is CwtH,,902,,N,
clude
a
sugar,
numerous
hydroxyl groups, and a sinthe currents is zero. The value of V a t point A is the
gle
amino
group
which,
below
pH 9, confers on the
membrane potential that one records across the openmolecule a charge of + 1. Probably it is structurally
circuited membrane.
similar to the polyene antibiotics, but it does not
SUMMARY. The modifiers discussed in this section are
contain a polyenic chromophore. Alamethicin is a
interesting models for ion transport in plasma mem- large cyclic polypeptide antibiotic produced by Trichbranes. Both carriers and polar channels have been oderma uiride; its structural formula is given in Figinvoked to account for biological ion permeability, ure 47. Excitability-inducing material is a peptide of
and both of these are realized in the bilayer mem- unknown chemical composition, produced by Aerobrane. It is of course intriguing to speculate about the bacter cloacae (American type culture #961) (541,
relation of these modifiers to the real transporters whose molecular weight is uncertain because of its
in cell membranes. Whether molecules quite similar tendency to aggregate in solution.
to these, or regions of larger proteins bearing a reWe shall present in some detail the phenomenologsemblance to these structures, function in plasma ical properties of monazomycin-treated membranes,
membranes will probably not be known until chan- particularly the steady-state properties, as they illusnels and carriers are extracted from cell membranes trate rather clearly the similarities of these artificial
and reconstituted in t.he bilayers. There is some rea- systems to biological excitable membranes. Also we
son to believe that a t least the principles involved in have personally had more experience with this molethe bilayers are applicable to plasma membranes.
cule than with the other two. We shall then discuss
In this regard it is worth noting that our under- the actions of alamethicin and EIM. A t the outset, we
standing of the mechanism of action of these modi- wish to stress that we do not claim that these mole-

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

cules, or molecules closely related to them, are present in excitable membranes; that question is still
moot. We do feel that a n understanding of the mechanisms operating in these systems (mechanisms
that have not yet been worked out) is not only relevant to our understanding of biological mechanisms,
but also perhaps directly applicable. The excitable
bilayer membranes are also very useful pedagogically in illustrating the contributions of strongly
voltage-dependent conductance elements to membrane behavior, and this usefulness is one of the
major reasons for discussing them in this article.
Before beginning our description of the voltagedependent systems, it is important to establish a
convention about the direction of membrane current
in terms of positive and negative current. If we
call one aqueous compartment the inside (in) and
the other the outside (out), and if furthermore we set
V,,,, = 0, then the value and sign of Vi, gives the
potential difference across the membrane. For example, V = -60 mV means that the solution in the
inside compartment is -60 mV with respect to the
solution in the outside compartment. With this convention, the direction of current flow across the membrane for an applied positive V is from the inside
compartment to the outside, that is, in the outward
direction. Thus positive values of Z are referred to as
outward current and are shown above the abscissa in
the figures; negative values of I are referred to as
inward current and are shown below the abscissa in
the figures. These conventions are those used by most
electrophysiologists, where inside and outside refer
to the cytoplasmic and extracellular space, respectively.

195

0.2 sec

FIG. 48. Voltage-clamp responses of monazomycin-treated


phosphatidylethanolamine membranes. Membranes (area = 1
mm) were formed at room temperature in 0.1 M KC1; monazomycin was then added to t h e inside chamber (33 j ~ g / m in
l A and
17 pglml in B , C, and D ) . The arrows designate the I = 0 level,
and the uertical blips in A . B, and D m a r k t h e onset of stimulation. A: successive current responses to positive rectangular voltage pulses of 25, 29, 33, 37, and 41 mV. B: record illustrating t h e
difference in kinetics for turning on and turning off conductance.
The initial stimulus is +60 mV; after t h e current reaches a steady
state, the stimulus is switched to -60 mV. Cr higher time resolution of t h e conductance turnoff. The current has already reached
a steady-state value for a voltage of +65 mV, and the stimulus is
then switched to -65 mV. D: further illustration of difference in
kinetics for conductance turnon and turnoff. The initial stimulus
is +55 mV; after the current reaches a steady state, the voltage is
reduced to +50 mV. Note t h a t the initial drop in current when the
voltage is reduced from +55 to +50 mV is much greater than the
initial rise i n current (hardly visible) when the voltage is increased from 0 to +55 mV. This reflects t h e much larger steadystate conductance a t +55 mV than at 0 mV. IFrom Muller &
Finkelstein (57).J

MONAZOMYCIN.
Response to steps of voltage i n symmetrical salt solutions. When monazomycin is added
in small amounts (micromolar or less) to one side,
which we take as the inside, of a membrane separating identical salt solutions (e.g., 0.1 M KCl), the
small signal characteristics are those of unmodified
membrane; that is, the membrane can be represented
by the simple parallel RC circuit of Figure 37B, with
an ohmic resistance of approximately 10iL.cm2.This
resistance remains ohmic and large for negative potentials of all magnitudes. For larger positive potentials (whose magnitude depends on the amount of
monazomycin added, see Relation 74 below), the responses are those shown in Figure 48A .I4 The current
rise is S-shaped to a final steady-state value. Upon

Aib- 12

Gly-I0
Leu-1 1

FIG.

47. Structural formula of alamethicin (61). A i b , a-amino-

isobutyric acid.

l 4 In response to a step of voltage, the first occurrence is a


surge of capacitance current, a s discussed for the unmodified
membrane. This occurs too rapidly to be seen in the records of
Figure 48. I t should be understood t h a t throughout t h e remainder of this article, when we speak of the instantaneous current
t h a t is observed after a sudden change in voltage, we a r e talking
about the current t h a t exists after t h e capacitance current has
disappeared. Ideally the capacitance current is infinite and of 0
time duration.

196

HANDBOOK O F PHYSIOLOGY

THE NERVOUS SYSTEM I

removal of the voltage step, the current immediately


g x [K+][monazo+]Se~/k~
(74)
falls to zero. If the sign of the voltage is reversed after
the steady state has been reached, the current instan- where s and n are approximately equal to 5 (at least
taneously observed is of the same magnitude (but of in membranes made from bacterial phosphatidylethcourse of opposite sign) as prior to reversal; this cur- anolamine (PE) or from bacterial phosphatidylglycrent then decays rapidly (in comparison to the cur- erol (PG) (57). In Relation 74 we have written [K+]
rent rise) in an exponentiallike manner (Figs. 48, B rather than [KCl], since the monazomycin-induced
and C). Similarly, reduction of the voltage to a conductance element is selective for univalent catsmaller positive value leads to a n instantaneous cur- ions (57). This selectivity is somewhat paradoxical,
rent that is directly proportional to the voltage; this since monazomycin (monazo) is itself positively
current then decays further in an exponentiallike charged. The selectivity is not ideal; a 10-fold salt
fashion (Fig. 480). The behavior is that of a voltage- gradient yields potentials of about 50 mV, instead of
dependent conductance with no emf in the circuit the Nernst potential of 58 mV. Unlike the potassium
conductance of nerve, the monazomycin-induced con(Fig. 301, and the current a t all times is given by
ductance discriminates poorly among the alkali catI = gv
(67) ions. The simple linear dependence of conductance on
cation concentration given in Relation 74 holds only if
where g is a function of both voltage and time. [The
the membrane is made from a lipid having no net
fact that the instantaneous current is linear with charge (e.g., PE); for a negatively charged membrane
voltage means that the instantaneous I is determined
(e.g., PG), conductance decreases markedly with inby the chord conductance existing prior to the change
crease of univalent cation concentration (see the folin V (Fig. 491.1
lowing subsection The effect of surface charge on the
In most of the following we are concerned with the
steady-state g-V characteristic).
steady-state properties of the membrane. The steadyThe most interesting aspect of Relation 74 is the
state conductance is obtained by dividing the final steep exponential dependence of conductance on voltcurrent value by the applied voltage; Equation 67 age. Since n lies between 4 and 6, conductance inremains applicable, but with the understanding that creases e-fold every 6 to 4 mV (kTlq = 25 mV). Such
g is not time dependent. The records in Figure 48 are a n extraordinary dependence ofg on V is characterisstrikingly similar to voltage-clamp records for the tic of the sodium and potassium conductance elepotassium conductance element in the squid axon ments in nerve. It is instructive to consider a model
(29). The major difference is that the bilayer response that can produce this large dependence of g on V,
occurs over seconds, whereas the axonal response although the specifics may not be completely applicaoccurs in milliseconds.
ble to monazomycin.
Suppose that monazomycin is a rod-shaped moleSteady-state conductance-voltage (g- V) characteristic. For a membrane separating identical KCl (or cule (e.g., like amphotericin B) of sufficient length to
NaCl) solutions, the steady-state dependence of con- span the membrane and that its positive charge is
ductance on voltage, monazomycin concentration, located a t one end of the rod. Suppose further that
and salt concentration is summarized by the relation either the rod can lie adsorbed (on the side to which it
was added) parallel to the membrane surface, or
that it can be oriented perpendicular to the surface
I
and span the membrane, with its charged end located
4
a t the opposite side. This orientation can be produced
either by thermal energy (kT)or by electrical energy
(qV). Finally, assume that n such oriented rods can
aggregate to form a conducting channel. If N is the
total number of monazomycin molecules associated
with the membrane (at a given concentration of monazomycin in solution) and p is the number of these
molecules that are perpendicularly oriented (i.e.,
V
span the membrane), then from the Boltzmann disv,L
v.
-1
tribution

49. Schematic drawing of t h e nonlinear steady-state I-V

(75)

characteristic of a monazomycin-treated membrane separating


identical solutions. At V , , the value of t h e current is t h a t corresponding to P . Any change in potential (to V, for example) leads
to a n instantaneous value of the current a t t = 0 corresponding
to a point on the chord drawn between P and the origin. The
current subsequently relaxes (along t h e dashed uertical line) t o
its final value (at t = I) on t h e steady-state characteristic.

where E , represents the chemical energy difference


between the parallel and perpendicular orientation.
For monazomycin, E , is large and positive; that is, a t
V = 0,most of the rods have the parallel orientation.
If the equilibrium for the reaction to form a channel

FIG.

CHAPTER

6:

197

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

brane has a negative surface charge, the potential GI,


a t the surface will be negative with respect to remote
points in solution. This means that [K+],, and
e-l nE,-nqL)IkT
[monazo+],,,
the concentrations of potassium and
[1
+
e-lb,,-ilV)/kT
n
(76)
g = giipn x gap" = go
monazomycin at the interface, are greater than their
1
For values of V such that qV e E , % 0, Relation 76 concentrations, [K+], and [monazo+],, in bulk solureduces to Relation 74 insofar as the functional de- tion. In fact, from the Boltzmann distribution we can
write
pendence of conductance on voltage is concerned.
If N is directly proportional to [monazo+], the con[i+],, = [i+],e-c1%/k7
(78)
centration of monazomycin in solution, then we also
have from Relation 76 that g
[monazo+I, which where i+ is any univalent cation, including monazoagrees with Relation 74 if s = n (as indeed is approxi- mycin itself. (Similarly, [it],, is depressed from its
mately the case). In the context of this model, the concentration in bulk solution. The situation is com[K+]term in Relation 74 means that the conductance pletely analogous to the Donnan equilibrium (Fig. 41,
per channel, go, is a linear function of the concentra- except that the fixed charge is concentrated a t a
tion of univalent cation, the current-carrying species surface rather than being distributed throughout a
volume.)
in the channel.
For the particular case of one univalent cation
The effect of surface charge on the steady-state g-V (excluding monazomycin, whose concentration is
characteristic. If the terms in Relation 74 have any negligible), one univalent anion, and one divalent
physical meaning at all, they must refer to quantities cation, the solution of the Poisson-Boltzmann equaa t the membrane surface or across the membrane tion leads to the following relationship between u and
proper. That is, [K+] and [monazo+] must be the $0 (58)
concentrations of potassium and monazomycin at the
membrane surface; it is only these concentrations
(79)
that the membrane sees. Similarly V must be the
+ [i2+]r(2e~lq,/kT
+ e-211q,/kT - 3)
potential difference directly across the membrane
from interface to interface; it is only this potential For membranes made from phosphatidylglycerol,
difference that can drive monazomycin molecules u == 1 charge per 60 k2(58). From Equation 79, this
through the membrane. For a membrane made from means that in 0.01 M KC1, $I, = -180 mV; in 0.1 M
a lipid having no net surface charge (e.g., PE), the KC1, &, = -120 mV; and in 1 M KCl, +,, = -60 mV.
above statement is somewhat gratuitous. The con- Note that the potential changes 60 mV for a 10-fold
centrations of potassium and monazomycin a t the change in KCl concentration. This follows intuitively
interface are the same as in bulk solution, and all the from the fact that a t large negative surface potenpotential, V, appears directly across the membrane. tials, with only KC1 present, the double layer near
When, however, the lipid is negatively charged (e.g., the interface is virtually exclusively made up of K+.
PG), the distinction between bulk and surface con- Thus this region is formally like a potassium-seleccentrations takes on real significance.
tive membrane, and hence the potential a t the surIf the membrane has a negative surface-charge face follows the Nernst relation.
density, u, and the concentration of mobile ions
As a consequence of this change in GI, with KC1
within the membrane is essentially zero, then in concentration, interfacial monazomycin concentrasolution there will be a space-charge region with a tion decreases with increasing KC1 concentration.
positive charge density, p , with
Thus, from Equation 78, a 60-mV decrease in the
(absolute) magnitude of +,, (produced by increasing
pdx=-u
(77) [KCl] from 0.01 to 0.1 M) results in a 10-fold decrease
of [monazo+],,. Since g a [monazo+],,J, where s = 5 ,
that is, the total charge in solution balances the conductance (at a given voltage) should decrease by a
surface charge. (There is of course such a space- factor of lo5 when the KC1 concentration is raised 10charge region a t each interface.) The ions in solution fold. This is indeed seen experimentally (58).
satisfy the Poisson-Boltzmann equation, and the
Of even greater interest is the effect of divalent
mathematical problem is then, as we indicated for cations when +,, is large. Equation 79 predicts then
the Donnan equilibrium, one of solving for the con- that small concentrations of divalent cation ( 2 1/100
centration and potential profiles in solution. For our [K+]) control +(,. This is a direct consequence of the
present purposes, however, it is only necessary to Poisson-Boltzmann analysis, without any assumpconsider the surface potential, +[,, that is, the poten- tion about binding of divalent cations to the negatial a t the membrane interfacek). Without for the tive surface. Thus, in 100 mM KC1, a symmetrical
moment worrying about the quantitative aspects of increase of [Ca2+]or [MgZ+]from 0 to 1mM decreases
the problem, it is intuitively clear that if the memby -15 mV (for IT = 1 charge per 60 A2). This in

(np % pn) lies far to the left, and we call g,, the
conductance per channel then

il,

198

HANDBOOK OF PHYSIOLOGY

0.001

THE NERVOUS SYSTEM I

10

20

30

40

50

60

70

80

90

100

MEMBRANE

outside (Fig. 5lB). In the latter case, the difference in


surface potentials results in a potential difference
between the two interfaces not measurable by the
recording electrodes, which are a t infinity. This
potential difference, however, establishes an electric
field within the membrane that is seen by monazomycin. As far as monazomycin is concerned, the situation is the same as if a negative potential difference
were applied across the membrane through a n external pair of electrodes. This then is the basis for the
shift in the g - V characteristic. In summary, the g-V
characteristic has shifted because operationally one
always measures the potential difference across the
membrane a t points distant from the membrane,
whereas the potential difference that is relevant to
the monazomycin system is that existing between the
two interfaces. Similar shifts in the characteristics of
the sodium and potassium channels of nerve are observed upon changing Ca2+or Mg2+concentrations on
FRONT

POTENTIAL (mv)

REAR

._

FIG. 50. The effect of divalent cation added to the outside


solution on the steady-state g-V characteristic of a membrane
with monazomycin in the inside solution. The phosphatidylglycerol membrane was formed at room temperature in 0.01 M KCl.
Monazomycin was then added to t h e inside chamber, and after
approximately 20 min, theg-V characteristic labeled IMP+] = 0.0
rnM was taken. MgSO, was then added to t h e outside chamber to
the concentrations indicated on the curves. I t required only 1 min
(the time to stir in MgSO,) for the new g-V characteristic to be
established. Note the large shift of t h e g - V characteristic to the
right along t h e voltage axis. Similar results a r e obtained with
Ca+. Slope of lines = e-fold conductance change per 4.0 mV;
monazornycin concentration = 0.5 pg/ml; membrane area = 1
rnm. [From Muller & Finkelstein (581.1

FRONT

REAR

60

turn reduces monazomycin concentration at the interface and hence causes a large reduction of conductance at a given voltage.
So far, the consequences of negative surface charge
we have described have been confined to cases of
symmetrical changes in ion concentration. Such symmetrical changes alter interfacial monazomycin concentration, with a consequent change in conductance. l 3 Interesting effects also occur, however, if divalent cation concentration is changed only on the
outside (i.e., the side without monazomycin).
Addition of divalent cation such as Mg2+or Ca2+to
the outside of a monazomycin-treated membrane
shifts the g-V characteristic to the right along the
voltage axis (Fig. 50). What is the explanation for
this? Figure 51 depicts the potential profiles for the
case of symmetrical salt solutions (Fig. 5lA) and for
the case of finite Ca2+ concentration only on the
I The same phenomenon operates with the cation carriers
discussed earlier. A change in [Ca+l or [Mg+l alters interfacial
Inonactin-K+l,for example, with a consequent change in conductance (45). The conductance changes in the monazomycin system
are much larger, because conductance varies with the 5th power
of the antibiotic concentration, rather t h a n linearly a s with the
simple carrier systems.

0
-60
-120
-180

/O01M KCI

CoCl2

FIG. 51. Potential profiles for a phosphatidylglycerol membrane separating t h e salt solutions indicated. At large (greater
than 100 A) distances from the membrane intmfaces, the potential on both sides is zero (in the open-circuited condition). The
quantity of interest is &,, the value of the potential a t t h e two
interfaces. A: symmetrical salt solutions. $,, is the same at each
interface, and there is no electric field within the membrane. B:
divalent cation is present only in the REAR (outside) solution.
Because of the asymmetry in surface potentials, there exists a
potential difference across the membrane proper (though none
measurable between the two solutions) and hence a field within
the membrane. The sign of the potential difference is such a s to
turn off t h e monazomycin-induced conductance. As far as the
field t h a t monazomycin sees, the situation in B is identical to
having no Ca+ present but instead passing sufficient current
across the membranes to make the inside compartment -60 mV
with respect to t h e outside. This is the basis for t h e shifts of t h e g V characteristics to the right in Fig. 50. [From Muller & Finkelstein (58).1

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

the outside of the fiber (21), and the physical basis for
these shifts is precisely the same as that described for
the model membrane system.
Development of negative-slope conductance. At a
given monazomycin and KC1 concentration, the relationship between membrane conductance and voltage
can be written as
g

gunino,lifit*,l

c enuV/A"T

(80)

where the conductance of the unmodified membrane


is included as a parallel element to the monazomycininduced conductance and C is a constant dependent
on monazomycin and cation concentration. In symmetrical KC1 solutions we have

gv

(67)

or

Substituting Equation 80 into Equation 68 we obtain

For there to be a region with negative-slope conductance, dlldV must somewhere equal zero. From Equation 81 this occurs when

For typical concentrations of monazomycin, C <


that is, for V 5 0,g = guiiin,,,lin,,,l.Therefore,
Equation 82 will never be satisfied, even for negative
voltages. The I-V characteristic will be monotonic,
showing, however, strong rectification properties.
If we now increase the KCl concentration on the
outside, thus introducing a positive emf, E, into the
system, we then have

gunin<l<lific.il;

g(V - E)

(70)

which becomes, upon substituting in Equation 80


V - E ) + C ens"i'c'''(V - E ) (83)
I =
For large V, the monazomycin-induced conductance
is large compared to the conductance of the unmodified membrane, and Equation 83 becomes

L'//<7'(

E)

Therefore

We see, then, that dZIdV

0, when

(834

199

and there now exists a region of negative-slope conductance. (Of course, E must be sufficiently large so

kT

, the monazomycin-induced
nq
conductance is greater than that of unmodified membrane. Otherwise, our approximation (Eq. 83a)
was not justified.) We have thus realized, with an
explicit g-V relation (i.e., Eq. 801, the concept developed earlier in the subsection Current-Voltage Characteristics. Namely, we have converted a simple rectifying I-V characteristic into one with a negativeslope region merely by introducing a salt gradient
across the membrane, without changing the g-V
characteristic. 'Ii Figure 5% shows the I-V characteristics of a monazomycin-treated membrane with and
without a salt gradient, and Figure 523 shows the
corresponding g-V characteristics. Note that our theoretical prediction is experimentally realized. The IV characteristic drastically changes, while the g-V
characteristic essentially remains the same.
Response to steps of voltage in the presence of a salt
gradient. We have seen how a positive emf affects the
steady-state I- V characteristic of a monazomycintreated membrane (Fig. 52A). It is also interesting to
consider how the voltage-clamp response is modified.
Figure 53 shows the current responses to positive
steps of voltage, all starting from V = 0, for V < E, V
= E , and V > E , as well as the response when the
membrane potential is returned to V = 0 from these
values. These records are best understood with the
aid of the schematic drawings in Figure 54. The two
lower drawings in each column show schematically
the oscillographic recording of the membrane potential and the corresponding current response (the capacitive current transients are not shown). The uppermost drawing in each column complements the
lower two, in that the path of the current and potential is shown in the "I-V plane"; that is, a t any
moment in time a set of I-V values is indicated by a
point in the I-V plane.
Consider first the left-hand column of Figure 54,
where the voltage step, Vb, is less than E . At V = 0,
there is a small inward current given, from Equation
70, by -g$, where g , is the resting conductance a t
zero membrane potential. With the onset of the step
change of the potential to Vb, the current changes
instantaneously to the value g,(V, - E ) . That is, a t
time zero the conductance remains a t the resting
value, g,, since, as noted before, g needs time to
change. In the I-V plane this response appears as a
sudden movement from point a to point b along the ZV line for the resting conductance, g,. The inward
current then proceeds along an S-shaped time course
to the steady-state value c. Note both the similarity
and difference between this response and those
that when V

I" Note that this is possible with an uncharged lipid membrane


but not with a charged lipid membrane. In the latter case, increasing the KCI concentration on one side changes the surface
potential on that side and hence shifts the g-V Characteristic.

200

HANDBOOK O F PHYSIOLOGY

THE NERVOUS SYSTEM I

If V < E , Equation 70 demands that Z must be negative (inward). Thus identical changes of g with time
a t a given positive voltage will always lead to outward currents in the absence of a salt gradient, but to
inward currents in the presence of a salt gradient, if
V <E.
Upon return of the membrane potential to zero, the
(67) inward current increases instantaneously from the
value c to the higher value, d. (At c, the current is
given by g J V h - E ) , where g,. is the steady-state
(70) conductance for V = Vh. When the voltage is returned to zero, the instantaneous current is given by
-g,E, since the conductance has not yet changed
from its steady-state value.) The current then relaxes
along a n exponentiallike time course back to the
resting current, a. This tail of the current - the term
used in describing the voltage-clamp data obtained
on the squid giant axon (30) -results from the relaxation of the conductance back to the resting value a t
zero potential. (By comparing the value of the current
at a with that a t d, we have a direct measure of the
increase in the conductance, since Z, = g& and I , =
g&.) Note that in Figure 4&4 there is no tail of
current. This is because when V = 0, Z must be zero
(from Eq. 67), regardless of what changes occur in g.
In Figure 48B-D,however, where the voltage is
switched from a positive value to a negative, or less
positive, value, a tail of current is seen, as the conductance relaxes to the value appropriate t o the new
voltage.
Now consider the middle column of Figure 54,
[KCd IN
where the voltage step, V,, is equal to E . Of necessity
COMPARTMENTS
the current goes to zero, since ( V - E ) = 0. Moreover,
FRONT
REAR
0.020 M
0.020 M
as long as the membrane potential is maintained a t
E , it is obvious from Equation 70 that current must
remain zero, despite the rise in conductance resulting
@ 0.153 M
0.153 M
from the increase in membrane potential from zero to
V,. The dramatic conductance increase that occurred
while the membrane potential was clamped a t V = E
is revealed by the large inward current that flows
when the potential is returned t o zero. This current is

shown in Figure 4&4 for a membrane with E = 0. In


both cases the current changes sigmoidally, because
g increases sigmoidally; but whereas when E = 0, Z is
always outward, when E is positive, Z is inward for
voltages less than E . This is simply a consequence of
the fact that in one case

-=gv

whereas in the other case

== g(V

BO-

7060-

50

2010-

-10-

MEMBRANE

-20-

\
-3040-

-50-60-

~-

10

///
// 8 i:;;
[KCg

1.0

IN COMPARTMENTS :
REAR

FRONT

0.1

-1

I0

I5

20

25

30

35

MEMBRANE POTENTIAL

0.020 M
0.153 M
0153 M

I
40

(mv)

45

I
50

I
55

...

~~

.~

~~~~~

F I G . 52. A ; steady-state I-V characteristics of a monazomycintreated phosphatidylethanolamine membrane: development of a


negative-slope conductance region by t h e addition of a salt gradient. The membrane was formed a t room temperature in 0.02 M
KCl; monazomycin was then added to t h e FRONT (inside) chamber and approximately 15 min later the I-V characteristic labeled
0was obtained. A KCI gradient was established by raising IKCI I
to 0.153 M in t h e REAR (outside) chamber. This gave rise to a
diffusion emf of 42 mV. The I-V characterist.ic labeled @ was
taken approximately 1 min after the gradient was established.
The gradient (along with the emf) was then abolished by raising
IKCI] in the front (inside) chamber to 0.153 M. T h e I - V characteristic labeled @ was taken approximately 1 min after the gradient
was abolished. Rc the steady-state g - V characteristics of a monaand @) and preszomycin-treated membrane in the absence (0
ence @ of a diffusion emf created by a salt gradient. The g-v
characteristics have been calculated from the corresponding I-V
and by
characteristics in A by Equation 67 (for curves @ and 0)
Equation 70 (for curve @) with emf = 42 mV. IFrom Muller &
Finkelstein (57).I

CHAPTER

6:

201

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

is voltage dependent. Suppose that a t V = Ecl,gcl %Then a t Z = 0, the system is stable at a potential
near Ecl, since the negative membrane potential
keeps g N a turned off. On the other hand, if sufficient
monazomycin has been added, it is possible that a t V
= EN^, g N a
gel. In that case at = 0 the system is
also stable at a potential near ENa, since the positive
turned on. We thus see intuitively
potential keeps gNa
that it is possible to have two stable states for such a
membrane, and it should be possible to pulse the
system from one state to the other. This is seen
experimentally in Figure 56 (where KC1 happened to
be used instead of NaCl).
Analytically the membrane potential a t Z = 0 is
given by

gNa.

FIG. 53. Voltage-clamp responses of a monazomycin-treated


membrane with a positive emf created by a salt gradient. The
inside compartment, which has monazomycin (5 pglml), contains
0.01 M NaC1, and the outside compartment contains 0.077 M
NaCI. The resulting emf is approximately 40 mV. Membrane
area = 0.01 mmy. The membrane was formed from a mixture of
ox-brain lipid and tocopherol. See text and Fig. 54 for a discussion
of these records.

and the mathematical problem is to find conditions


permitting two values of V that satisfy Equation 86,
given gel, C, Eel, and ENa. (Actually there are either
one or three values, since for the bistable system
there must exist, as we shall see below, one metastable point.) Rather than pursuing this line, however,
we show graphically how bistability arises from the
shown in Figure 54 by the deflection from e to f of the steady-state Z-V characteristics of the two elements.
current tracing (see also the line extending from e to f Figure 57 is a plot of the steady-stateZ-V characterisas shown in the Z-V plane). At f, Z = g,E, whereg, is tics of the nystatin and monazomycin elements. We
the steady-state conductance for the voltage V,. seek those points a t which total current, Z, is zero;
Again, as in the previous experiment, the current
subsides to the resting level, a, as the conductance
wanes to its resting value, g,.
When the membrane potential is stepped up to a
value greater than E (3rd column of Figure 541, the
current changes instantaneously from inward to outward and then rises in an S-shaped curve to a larger
outward value. Again the return of the potential to
zero gives rise to a large inward tail of current. In the
Z-V plane, the beginning and end of the step change,
V,, appear as instantaneous excursions along the line
a to g and along the line h to i, respectively.
All the features described schematically in Figure
54 are seen experimentally in Figure 53. We have
ir
spent some time discussing these records and those in
Figure 48, because they are so similar to those found
for the sodium system (neglecting inactivation) and
the potassium system in the squid giant axon; it was
from the analysis of records such as these that Hodgkin and Huxley experimentally established a family
of conductance vs. time curves for both the rising and
decaying phases of the conductance (30).

IX

Creation of a bistable system. Consider a membrane treated with nystatin on both sides and monazomycin only on the inside. Let it separate NaCl
solutions, with [NaCll,,,, > [NaC1li,. The equivalent
circuit for this system is shown in Figure 55. The
nystatin-induced conductance, g(.l, is constant,
whereas the monazomycin-induced conductance, g N a ,

:-.E
L C
0

F,.
f - 1.

v,

- 0

O C

$2

.g

time

FIG. 54. Schematic drawings (lower two) of the records in Fig.


53 along with the steady-state I-V characteristic of the membrane
(upper drawing). See text for a discussion of these drawings.

202

HANDBOOK O F PHYSIOLOGY

THE NERVOUS SYSTEM I

or

Outside

kg

I,,

z<y

Na (rnonazo-(
mycin)

-c dV
dt

-~

(88a)

For convenience we refer to the sum of the ionic


currents, I,, + Z,.l, by Zi.
By definition, the system is in a steady state when
A
Inside
FIG. 55. Equivalent circuit for a membrane treated with both
nystatin and rnonazomycin and separating NaCl solutions with
I NaCl I,,,,L,,t,c > I NaCIJ,,,,,,,,.(The membrane capacitance is shown
for completeness.)

1;

that is, a t what potential(s) the system can sit in the


open-circuited state. Since

z = INa + ZCI

(87)

the graphic solution is simply those values of V where


I,, and ZcI are of equal magnitude but of opposite
sign." In Figure 57 there are three such points: A, B,
and M; we shall shortly show that only A and B are
stable points and that M is metastable. Notice that if
there were too much nystatin in the system, the slope
of the chloride current would be much steeper, and
there would exist only one stable point near E,.,.
Conversely, if there were too much monazomycin in
the system, the only stable point would be near Exa.
It is also obvious from contemplating the graph that a
region of negative slope in a t least one of the characteristics is a necessary (although not sufficient) condition for the existence of more than one point a t
whichZ = 0.
Let us now discover intuitively from the steadystate characteristics the dynamics of the system. It is
pedagogically useful to imagine that the kinetics of
the monazomycin system are extremely fast; that is,
g N a changes virtually instantaneously with voltage,
rather than along the time course shown in Fig. 48.
Thus no matter how fast V varies, IN, is always
determined from the steady-state characteristic of
Figure 57. (At the end of our discussion, we indicate
how the analysis is modified if this assumption of
rapid kinetics is not satisfied.) It is also useful to
include the membrane capacitance in the description.
It should be clearly understood, however, that the
phenomenon we are discussing does not depend on
the existence of a membrane capacitance, although in
both bilayers and plasma membranes this element is
always present.
At any moment in time the total current Z through
the membrane is the sum of the two ionic currents,
Z x a [ = g,, (V - E d 1 and ZCI[= g d v - E d ] , plus the
capacitance current, I,.(= C dVldt). If there is no
external current source (that is, the membrane is in
the open-circuit condition), then

li

Remember t h a t I,,

F&

and I,,

-F&

,.

FIG. 56. Bistable property of a membrane treated with both


nystatin and rnonazomycin. The membrane was formed at room
temperature in a solution containing 0.01 M KCI and 0.001 M
MgSO,. Nystatin was added to both chambers to a concentration
of 5 pgiml. After approximately 20 min t h e membrane reached a
steady conductance of about 10 (I-'. At this point t h e KCl
concentration was raised in the outside chamber to 0.077 M. This
produced a membrane emf of -41 rnV. Monazomycin was then
added to the front chamber to a concentration of 3 pgiml. This
produced no significant change in the emf. After approximately
15 min, t h e above record was obtained. ( I n the record the horizontal line is V = 0; above the line are negative potentials, and below
the line a r e positive potentials.) At a , a positive step of current =
3.75 nA was applied and a t b i t was removed. The potential
returned toward the original -41 mV. At c. the current was again
applied, and a t d it was removed. This time the potential continued increasing and achieved a new stable state of +41 mV. At r , a
negative step of current = -4.05 nA was applied, and a t f i t was
removed. The potential returned to +41 mV. A t g t h e c u r r e n t was
again applied and a t h i t was removed. This time the potential
continued to decrease and flipped back to the original state of -41
rnV. Note the much smaller ZR drops at e and g compared with
t h a t a t a , thus demonstrating t h a t t h e conductance is much
higher a t +41 rnV (where the monazornycin-induced conductance
is turned on) than a t -41 mV (where only the nystatin-induced
conductance is significant). I From Muller & Finkelstein (57).J

'r
L

(rnonazornycin)

y/

_ _ --

:+
in

,V

FIG. 57. Schematic plots of the steady-stateI-V characteristics


for the nystatin- and monazomycin-induced conductance elements of the circuit in Fig. 55. The slope of the dashed line is t h e
monazomycin-induced conductance, g,,, in the resting state ( a t
point A ) . (The slope of the line drawn betweenE,, and a n y point
on the I,, characteristic gives t h e value of g,, at t h e voltage
corresponding to t h a t point.)
I

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

203

dVldt = 0, that is, when there is no capacitance the dashed chord line shown in Fig. 571, the exact
current. Therefore, from Equation 88, in the steady value being determined by the kinetics of the system.
state Ii = I N a + I,, = 0, as we have indicated previ- These kinetics are determined both by the intrinsic
ously. If dVldt is positive, then (from Eq. 88a) Ii is kinetics of the monazomycin system and by the memnegative (inward); conversely, if dVldt is negative, Ii brane capacitance, through the time constant of the
is positive (outward). Inward (ionic) current causes circuit in Figure 55. Under these circumstances it is
charge to flow onto the capacitance in a direction to rather tricky to define the exact threshold for flipping
make the membrane potential more positive; out- between states, since this can also depend on the
ward (ionic) current causes charge to flow onto the duration of the stimulating current. Intuitively, howcapacitance in a direction to make the membrane ever, we expect the threshold potential to be in the
potential more negative. Thus changes in membrane vicinity of M.
We have seen that with one voltage-dependent syspotential can be followed conveniently by the direction of current flow with respect to the capacitance; it tem (monazomycin) and one ohmic system (nystatin)
is for this reason that we include the capacitance in in parallel, it is possible to have a membrane that can
flip between two states. In other words, the memthe following analysis.
Imagine that the membrane potential is sitting a t brane can generate half an action potential. To generpoint A in Figure 57 (where IN, = -Icl) and that we ate a full-blown action potential, we need either a
apply a pulse of outward current through the mem- voltage-dependent mechanism that turns off the
brane (by means of current-passing electrodes) of monazomycin-induced conductance after it develops
such magnitude and duration that the membrane (i.e., inactivation of the conductance) or two, not just
potential (and hence the membrane capacitance) is one, voltage-dependent conductance elements in parbrought to a point C to the left of M. After this allel. Both situations can be realized in the bilayer
current shock, we see from the figure that Icl is systems. We briefly describe the phenomenon of ingreater than I N a , and hence the resulting net ionic activation in monazomycin-treated membranes and
current ( I N a + Z ( ~ , )is positive (outward). This acts to show how action potentials could, in principle, be
recharge the membrane capacitance t o the original generated with this system. In the subsection EXCITAresting state; that is, after the stimulus, the mem- BILITY-INDUCING MATERIAL. Action potentials, we see
brane potential returns to A, since Ii remains out- how action potentials are in fact generated with two
ward until point A, where it is zero.
voltage-dependent conductance systems in parallel.
On the other hand, if our current shock is of SUEInactivation. In the records of Figure 48A, we see
cient magnitude to bring the membrane potential to
that
in response to a step of positive voltage, the
a point D to the right of M, thenIN,is greater than IcI,
current
rises to a maximum level and remains there.
whereupon the direction of net ionic current is inThis
behavior
can be modified by adding micromolar
ward and acts to discharge the capacitance further.
amounts
of
dodecyltriethylammonium
bromide (CIS),
As the membrane potential increases further (moves
CH:j(CH2),IN(CPHT,):j+,
to
the
monazomycin
side of
to the right), IN,remains larger than I,, and hence
the
membrane.
The
responses
of
Figure
48A
are
then
continuously moves the potential to the right until
58;
pheconverted
into
responses
such
as
in
Figure
point B, where again IN, + I,, = 0. By the same
arguments, it is clear that once the system is a t B, it
can be driven back to A by applying an inward current stimulus of sufficient strength to shift the potential to the left of M.
In principle the system could also remain a t point
M, since there tooI,, + I,., = 0. However, the slightest fluctuation in the potential would inevitably
cause the system to settle down either at A or at B,
because of the mechanisms we have outlined above.
The point M, then, is not stable, but rather defines a 0
threshold potential for a system capable of residing
at two different potential levels.
FIG. 58. Inactivation of monazomycin-induced conductance by
In the above analysis, we assumed that the steadybromide (C,z). A membrane formed
state IN,-V characteristic applies a t all times. Since dodecyltriethylammonium
from phosphatidylglycerol and cholesterol separates symmetrical
g N a does not reach the steady-state value correspond0.1 M KCl solutions and contains in the inside compartment 1 pg/
ing to a given voltage instantaneously, this is some- ml monazomycin and 1.3 x lo-. M dodecyltriethylammonium
what unrealistic, and the situation now becomes bromide. The current record (traced from the oscillograph recordmore complicated. If the system is initially a t A, then ing) is in response to a voltage step (applied a t the upward arrow
removed a t the downward arrow) of 88 mV. Membrane area
at the voltage reached following the outward current and
= 1 mm2. Vertical line, 200 nA; horizontal line, 5 s. Note the
stimulus, IN,will lie somewhere between the steady- contrast of this record with Fig. 48 A , where C,, is absent. [From
state value and the instantaneous value (given from Heyer (27).1

204

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

nomenologically the monazomycin-induced conductance undergoes inactivation. The mechanism for this
effect is that, as the monazomycin-induced conductance turns on, C,, passes through the membrane and
is adsorbed onto the opposite interface (27). This reduces the surface potential on that side and thus
creates a field within the membrane of the sign (negative) that turns off the conductance, just as if the
actual clamping voltage were reduced. Consequently,
instead of increasing monotonically to a final steadystate level, the conductance rises to a peak and then
falls to a low value.
Consider now a membrane treated with valinomycin on both sides and monazomycin plus C,, on one
side, with gradients of both KC1 and NaCl as shown
in Figure 59A. The valinomycin system is K+ selective and voltage independent; the monazomycin system is cation selective with little discrimination
among the cations. The equivalent circuit for the
membrane is shown in Figure 59B.
Suppose that the relative amounts of valinomycin
and monazomycin are so adjusted that gKS g,.,, for V
= E
K and g,.,, gK for v = E,,,. The system is then
stable (at Z = 0) both near V = EK apd near V = E,.,,.
Thus, neglecting the presence of C,, for the moment,
the membrane treated with valinomycin and monazomycin can be bistable for precisely the same reasons that the nystatin-monazomycin-treated membrane is bistable. We could again plot steady-state ZV characteristics for each element as in Figure 57.
With C,, present in sufficient quantity, however, the
state near V = Era,is not stable, because the monazomycin-induced conductance is inactivated after it
turns on, and therefore the steady-state condition of
gCat gK is not realized. Transiently, however, this
state can be achieved prior to inactivation. Therefore,
if the kinetics are proper, it should be possible to
pulse the potential from point A to a value comparable to point M in Figure 57, a t which point the system
would spontaneously (in the absence of applied current) move toward point B (for the reasons we discussed in connection with Fig. 57) but would then
return toward A as the monazomycin-induced conductance inactivated. (In other words, the steadystateZ-V curve for the monazomycin system with C,,
would have only one stable point in combination with

Monazomyci

A
FIG. 59. A: membrane treated with valinomycin on both sides,
monazomycin and C,, on one side, and separating NaCl- and KClcontaining solutions as shown. Bc equivalent circuit for the situation depicted in A .

the valinomycin system, namely A.) In short, a fullfledged action potential occurs.
The analytical basis for the action potential in this
system is similar to that in the frog node of Ranvier
and the squid giant axon. There, excitability results
primarily from activation followed by inactivation of
the sodium conductance (although changes in the
potassium conductance also affect the kinetics).
Clearly, whether or not an action potential is possible
depends crucially on the kinetics of inactivation. If
they are too rapid, the state grat gK is never
reached. (This has been the situation to date with the
monazomycin + C,,-treated membrane, and therefore action potentials have not yet been achieved with
this system.) Graphically (Fig. 60), it is easiest to
imagine that the monazomycin-induced conductance
develops completely before any inactivation occurs
(solidZN,curve).'" If the system is pulsed from point A
to a potential to the right of M, it flips into a second
"stable" state a t B by the same mechanism discussed
in connection with Figure 57. As inactivation develops, ZNa decreases, resulting in outward current (since
now ZK > ZNa) that continually moves the potential t o
the left, until it returns to A. Note that with the
steady-state ZN,-V characteristic (dashed curve),
there is only one point, A, whereZK + I,, = 0. This is
the only stable point for the system in the steady
state.

ALAMETHICIN. Of the three molecules that induce


substantial voltage-dependent conductances in bilayers, alamethicin is the only one whose complete
structural formula is presently known (Fig. 47). Unfortunately, to date there does not exist a realistic
physical model for either the channel or the voltage
dependence. Nevertheless, if this system of known
components cannot be understood eventually, then
the task of understanding the mechanisms of voltagedependent conductances in excitable biological membranes is close to hopeless.
Steady-state properties. In certain respects the
steady-state g-V characteristic of alamethicin-treated
membranes resembles that of monazomycin-treated
membranes. When alamethicin is added in micromolar (or less) amounts to one side of the membrane, the
membrane conductance shows an exponential dependence on voltage (53). As with monazomycin, this
system "turns on" for positive potentials, that is,
when the side containing alamethicin is made positive with respect to the other side. There is, however,
a similar "turn on" of conductance for somewhat
larger voltages of opposite polarity (53). Thus the
Z-V characteristic can be more or less symmetrical,
as shown in Figure 61. The interpretation of this behavior is that alamethicin, unlike monazomycin, is
somewhat lipid soluble (53); though added to only one

InTo stress the analogy to t h e axonal membrane, we have


called the current carried through the monazomycin system I , ,
rather than I, Under t h e conditions depicted in Fig. 59, most of
the current would indeed be carried by Na+.

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

205

This second difference probably arises from the


following considerations: monazomycin channels are
formed from interactions among monazomycin+ ions;
this is the species moved by the applied electrical
field. Alamethicin, on the other hand, has one ionizable carboxyl group, so that it can exist either as
alamethicin- or alamethicin". At low pH most of the
antibiotic is uncharged, yet the system displays the
same voltage dependence as at high pH (7). This
observation demonstrates that the charge on alamethicin is not essential for its activity. From the
sign of the rectification in the I-V characteristic (see
Fig. 611, it appears that, as with monazomycin, a
positive species is being driven into, or through, the
membrane. Thus the voltage-dependent species appears to be the complex between uncharged alamethicin and one cation; that is, (alamethicin-cation)+.
FIG. 60. Schematic plots of the I-V characteristics for the valiIf this is so, and the equilibrium-alamethicin +
nomycin- and monazomycin-induced conductance elements of the cation @ (alamethicin-cation)+-lies far to the left,
circuit in Fig. 59. (To stress the analogy both to the axonal
membrane and to Fig. 57, we have called the monazomycin- then if conductance is proportional to [(alamethicininduced emf and current EN, and I,,, respectively, rather than cation)+Is, it will be proportional to (cationY (alaE,.,, and I?=,.) The solid I,, characteristic is the sodium current methicin)", as shown in Equation 89 with r = s. In
that would develop in the absence of inactivation (i.e., in the other words, whereas monazomycin is already posiabsence of C12).The dashed I,, characteristic is the steady-state
current that develops in the presence of C12.Note that a s drawn tively charged, alamethicin must complex with a
cation to produce the positively charged, voltagethere is only one point (A) i n the steady state where I K= -I,,,
whereas there are three points (A, M, and B ) that can satisfy this dependent species.
condition in the absence of inactivation.
Apparently alamethicin can form a complex not
only with an alkali cation such as K+, but also with
aqueous phase, it can nevertheless diffuse across the divalent ions such as Ca2+or even trivalent A13+(7).
membrane and achieve a n effective concentration at As one might anticipate, n in Equation 89 should
the opposite interface.
In addition to its steep exponential voltage dependI
I
ence, alamethicin-induced conductance is also a
highly nonlinear function of both antibiotic concentration and salt concentration (53). The steady-state
-5
I
I
dependence of conductance on all these variables, for
an uncharged membrane separating identical salt
solutions (e.g., NaCl), is given over a considerable
range by
g = IKC11' [alamethicinl"e"'~~"~'~~'(89)

where r, s, and n are all equal to about 6 (53). Comparing this with the steady-state behavior of a monazomycin-treated membrane

Membrane potential

--I-

xp

-200

[KCl][monaz~+]~e~~'~~''~~'
(74)
where s and n are approximately equal to 5, we see
two major differences. First, alamethicin-induced
conductance increases with both positive and negative potentials, whereas monazomycin-induced conductance increases only for positive potentials. This,
as indicated above, is probably because alamethicin
diffuses across the membrane and achieves a significant concentration on the side opposite from the original addition. Secondly, alamethicin-induced conductance is proportional to the salt concentration raised
to a large power (=6), whereas monazomycin-induced
conductance is linearly dependent on salt concentration.

lmvl

x'

I
100

yx-

200

I
X

-I 0Ji
I

-20

FIG. 61. Steady-state I-V characteristic for an alamethicintreated membrane. The membrane (formed from sphingomyelin
and tocopherol) separates 0.1 M KCI solutions buffered to pH 7
with 5 m M histidine chloride. Alamethicin is present only in the
inside aqueous compartment a t a concentration of
g/ml.
[From Mueller & Rudin (53).]

206

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

increase proportionally with the valence of the ion,


and this is indeed the case. Thus, in CaCl,, n = 12,
and in AlCl:{,n = 18 i(7).
It is interesting to compare the behavior of alamethicin with that of valinomycin or others of the
cation carriers we considered earlier. Both form complexes with alkali cations through their carbonyl
groups, which substit.ute partially or completely for
the hydration shell of the cation (65). The carriers we
discussed earlier are completely cation selective and
discriminate markedly among the alkali cations,
whereas alamethicin is poorly selective for cations
over anions (approximately a 30-mV potential difference is created by a 10-fold concentration difference).
Alamethicin furthermore shows very little discrimination among the alkali cations. The other striking
difference is that cation carriers such as valinomycin
and nonactin are not voltage dependent, whereas
alamethicin shows marked voltage-dependent phenomena. The reason for this is not yet clear, but it
may be because of functional side groups in alamethicin that allow for interaction among molecules; valinomycin and nonactin have no such groups. The
structure of the channel is still not known, and although it is possible that the hole in the cyclic molecule provides the conducting pathway, that is, the
alamethicin molecules form a doughnut stack, probably the channel forms from a space created by aggregated alamethicin molecules. [We are reminded
of nystatin and amphotericin B, also cyclic molecules,
that form a channel not by stacking but by arranging themselves in a c,ylinder (see Fig. 421.1
The unusual effect.s of adding small amounts of
basic proteins such as protamine, histones, and
spermine to the aqueous phases are further interesting features of alamethicin-treated membranes (53).
These molecules appear to interact with the alamethicin channel and change its selectivity from cationic
to anionic. This change in selectivity apparently does
not result from a more positive surface potential,
produced by a general adsorption of positively
charged protein molecules onto the bilayer. Instead,
what appears to happen is that a second population of
anion-selective channels is created (53). The relative
number of cation- and anion-selective channels is
determined by the amount of basic protein added to
the system. Without protamine, all channels are cation selective, whereas in excess protamine, all channels are anion selective. By careful titration, the
relative numbers of these two populations can be
controlled.
By analogy to the membrane treated with both
nystatin and monazomycin, a bistable situation can
arise in the presence of a salt gradient, in which the
membrane can be pulsed between a cationic and an
anionic resting potential. The present system has the
additional feature, however, that both the cationand anion-selective channels are voltage dependent
(whereas in the nystatin-monazomycin system, only

the cation channel was voltage dependent). Thus fullblown action potentials are not only possible, they
have been experimentally realized (53). We do not
discuss this phenomenon here, but defer an analysis
of such action potentials t o the protamine-EIMtreated membrane, where essentially the same situation arises.
Kinetics. The response of an alamethicin-treated
membrane to voltage steps resembles the response of
a monazomycin-treated membrane, but with two major quantitative differences. [We might add that the
kinetics of all three voltage-dependent systems discussed in this paper - monazomycin, alamethicin,
and EIM-are dependent on the lipid composition of
the membrane. In this chapter we can only indicate
general differences in behavior; under given conditions the kinetics of one system can be made to resemble those of another (55).] The current (conductance) rise does not have as marked a n S shape as in
the monazomycin-treated membrane. In fact, the
current rise appears exponentiallike (44) (cf. Fig. 62
with Fig. 48A) unless the system is examined carefully a t short times. The other major difference is the
much faster turn off of the alamethicin-induced conductance when the potential is reduced or reversed
(44).

The kinetics of both monazomycin- and alamethicin-treated membranes are certainly most fascinating and undoubtedly reflect cooperative interaction
among several antibiotic molecules. The quantitative
analysis and interpretation of the kinetics, however,
are very difficult, and little has been published. One
complicating factor is that antibiotic in the membrane is in dynamic equilibrium with antibiotic in

FIG. 62. Voltage-clamp responses


of alamethicin-treated
membrane. A lecithin-cholesterol membrane separates 0.5 M
NaCl solutions with alamethicin in t h e inside aqueous compartment. Note t h e exponentiallike rise in current here, in contrast to
the marked S-shaped rise seen with monazomycin in Fig. 48 A .
[From Mauro et al. (44). I

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

solution. This is fortunate for analyzing steady-state


data, since it means that the system can reach a true
equilibrium after each addition of antibiotic to the
aqueous phase. It is a distinct disadvantage in the
study of the kinetics, however, for it is difficult to
determine whether rate-limiting steps are truly
membrane-associated phenomena, or whether they
are related to partitioning of the antibiotic at the
interfaces or even to diffusion of antibiotic through
the unstirred aqueous layers at each interface.
EXCITABILITY-INDUCING MATERIAL. General phenomenotogy. A membrane with EIM on one side (the
inside) and separating identical salt solutions sits
in a high-conductance state at V = 0. This contrasts
with monazomycin- and alamethicin-treated membranes, which are at low conductance at V = 0. Over
certain ranges of both positive and negative potentials, conductance decreases steeply with voltage and
remains in a low-conductance state for larger voltages (54). A typical steady-state g-V characteristic
and the consequent I-V characteristic are shown in
Figure 63.
Membranes treated with EIM are highly cation
selective (541, developing membrane potentials
greater than 55 mV in 10-fold NaCl or KCl salt
gradients; they do not display significant specificity
among cations. The cation selectivity can be converted to anion selectivity by the addition of protamine to the system (54). If the relative amounts of
EIM and protamine are properly adjusted, the membrane can simultaneously have populations of cationand anion-selective channels. Such a membrane can
display action potentials [(51, 54); see subsection Action potentials, which follows].
The phenomenology of protamine interaction with
EIM-treated membranes is very similar to the effects
of basic proteins on alamethicin-treated films. There
are, however, two differences: a variety of basic proteins is effective with alamethicin, but only protamine (or arginine-rich polypeptides) is effective with
EIM (54). Furthermore, sphingomyelin must be present in the membrane for good coupling of EIM and
protamine, whereas the alamethicin-basic protein
coupling occurs with a variety of lipids (54).
To the uninitiated, the behavior of EIM might
appear quite different from that of alamethicin and
9

FIG. 63. Schematic drawing of the steady-state g-V characteristic and the consequent I-V characteristic for an EIM-treated
membrane separating symmetrical solutions.

207

monazomycin, because in symmetrical solutions conductance decreases with voltage whereas with the
other two antibiotics it increases. This difference,
however, is more one of detail rather than of fundamental significance. As pointed out in discussing the
monazomycin system, the resting conductance can be
shifted simply by applying a salt gradient or by adding divalent cations or positively charged adsorbates
to one side of the membrane. The EIM resting conductance can be similarly shifted. In short, for any of
the three voltage-dependent modifiers we have discussed, the conductance of the membrane is determined by two factors: chemical forces and electrical
potential. The former has to do with the specific
chemical free energy difference between the conducting and nonconducting state of the antibiotic. For
alamethicin and monazomycin (with most lipids), the
nonconducting state has a lower free energy; for EIM
the converse is true. The electrical potential difference across the membrane adds another term to the
free energy and, depending on its sign, shifts the
conductance state. Note also that the electrical potential difference may be created across the membrane
by divalent cations or absorbates and not be measured by the recording electrodes. (We discussed this
phenomenon with respect to the action of divalent
cations on monazomycin-treated membranes. In
summary then, whether conductance increases or decreases with voltage is a detail dependent on such
extrinsic factors as the choice of lipid, ions in solution, and other adsorbates. What is common and
most important to all three of the voltage-dependent
molecules discussed is the extraordinary steep dependence of conductance on membrane potential over
certain voltage ranges.
Action potentials. At proper concentrations of protamine in an EIM-treated membrane, both cationand anion-selective voltage-dependent channels exist. (This is also true, as mentioned earlier, of alamethicin combined with basic proteins, and this discussion is equally applicable to that system.) In the
presence of a KC1 gradient, the equivalent circuit of
the membrane is given in Figure 64. This circuit is
similar to that of a membrane treated with both
monazomycin and nystatin (Fig. 551, except that the
anion conductance produced by nystatin is not voltage dependent. In that case bistability could be
achieved, and we demonstrated this analytically,
graphically, and experimentally. We shall follow the
same approach in this instance and see how action
potentials, rather than just bistability, can occur.
Suppose that the steady-state g- V characteristics of
the cation- and anion-selective channels are as shown
in Figure 65. (We exclude from consideration the
region at large positive potentials where g falls, as
shown in Fig. 63.) We have assumed that they are
identical in shape (i.e., protamine has not changed
the steady-state voltage dependence of the channels),
but that gK (steady state) is greater than g,, (steady

208

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

Outside

Inside
FIG. 64. Equivalent circuit for a membrane treated with both
EIM and protamine, separating KCl solutions a t different concen trations.

gK

EK

Ec I

FIG. 65. Hypothetical steady-state g-V characteristics of g,


and g, I for the circuit o f Fig. 64. We have assumed that the
relative amounts of EIM and protamine are such that there are
many more potential K+-conducting channels than Clk-conducting channels. Note that [KCll,,,,,,~ > [KCI],,,,,,,, so that E K is
negative and E ( , is positive.

state), because there are more cation than anion


channels. Figure 65 also shows EK and Eel, which are
equal in magnitude but opposite in sign. [For simplicity we have assumed that the channels are ideally
selective for either cations (K+) or anions (Clk).] The
resting potential, V, of the system is given by

should be apparent that if they are properly adjusted,


a full action potential will develop.
Let us now analyze the system graphically. Figure
66 diagrams (among other things) the steady state I-V
characteristics of the two channels, based on the g-V
Characteristics of Figure 65. Since the ionic current,
I , through the membrane is given by
the necessary condition for stability with no applied
current is that I, = -Icl.
We see from the figure that
this occurs only a t point A; thus, a t rest, V = EK. Now
if we assume that g,., changes rapidly compared with
g,, then in response to a stimulus, the chloride system moves along the steady-state I-V characteristic,
while the potassium system moves along the instantaneous characteristic (chord conductance, as shown
by the dashed line). By the same arguments used to
discuss the monazomycin-nystatin system (Fig. 571,
M is a metastable point [Ic., (steady state) = -IK
(instantaneous)], and, for stimuli that bring the potential to the right of M, the potential continues t o
move over to point B ( V = E c J , where again Icl
(steady state) = -I, (instantaneous). But, unlike the
monazomycin-nystatin system, this system cannot
remain at B, because I , slowly increases as gK begins
to increase (toward the steady-state value appropriate for any given voltage). The resulting outward
recharges the membrane capacicurrent ( I , >
tance and continues to move the potential to the left
until it returns to point A. Thus a complete action
potential is produced. This is shown experimentally
in Figure 67.
The above analysis describes one way action potentials can arise in the EIM-protamine system. Actually, because the steady-state g-V characteristic of
EIM is more complicated than given in Figure 65 (see
Fig. 63), there are a multitude of ways of generating
I,

and since gK/gCl is a constant much greater than 1,


there is only one stable resting value of V, which is
approximately equal t o EK. (Note that this is in contrast t o the monazomycin-nystatin system, where
there were two stable states, one near ENaand one
near Eel.)
Imagine now that the kinetics of the chloride conductance change are much faster than those of the
potassium conductance change. Then if the system is
driven over to a potential near Eel, it remains there
for a while, since for short times, g,, 9 gK. As g,
begins to increase toward its steady-state value, however, the potential returns back toward EK. This return is caused both by the turn on of& and by the
simultaneous turn ot'f of g,, as V declines. (Of course,
as the potential falls back toward E,, gK will eventually stop increasing and also begin to decrease.)
The details of these processes are clearly dependent
on the kinetics of the conductance changes, but it

(steady state)

t /

FIG. 66. Schematic plots of the I-V characteristics for the elements of the circuit in Fig. 64 based on theg-V characteristics of
Fig. 65. Solid curves, steady-state characteristics; dashed line.
instantaneous K + characteristic in the resting state (i.e., with the
system sitting a t point A ) .

CHAPTER

6:

209

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

EIM, particularly on oxidized cholesterol membranes, are especially clear-cut and instructive.
When very small amounts of EIM are present, the
current fluctuates in discrete steps in response to a
constant applied potential (12, 39); a representative
record is shown in Figure 68 (cf. gramicidin A record
of Fig. 44). The conductance of an EIM channel is
considerably larger than that of a gramicidin A channel, the former being 3 x lo-'" R-l in 0.1 M KC1,
whereas the latter is 1 x lo-" h-I. A much more
significant difference, however, is that the fraction of
time that an EIM channel spends in its conducting
state is voltage dependent; the duration of a gramiciFIG. 67. Action potential produced with EIM and protamine.
din A channel is not voltage dependent. It is fascinatThe membrane (formed from a mixture of sphingomyelin, phosphatidylserine, cholesterol, and tocopherol) was formed in 3 mM ing that if one plots the fraction of time a single
KCl buffered t o pH 7 with 5 mM histidine chloride; EIM was then channel stays open versus voltage, one obtains the
added t o the inside solution to bring the resistance down to about same curve as the conductance versus voltage relalo5 a m 2 . The KCl concentration in the inside compartment was tion for a membrane containing many channels (Fig.
raised to 100 mM, and this produced a resting potential of -50
mV. Protamine was then added to the inside compartment to a 69). Thus the voltage dependence of conductance of an
concentration appropriate to produce the record shown. (Recon- EIM-treated membrane is a reflection of the number
structed from a motion picture by P. Mueller and D. 0. Rudin.) of channels that are open at a given voltage. [The
conductance of a membrane with a large number of
channels is equal to the conductance of an open chanaction potentials with this system (54). We chose this nel times the fraction of time a single channel is open
one particular case, because of its analogy to the multiplied by the number of channels. In other
squid giant axon; in our example the chloride system words, the ensemble average (the fraction of channels
corresponds to the squid's sodium system without
e 0.2
inactivation.
We have discussed two ways to obtain action potentials in modified lipid bilayer membranes: one is with
a membrane containing a voltage-independent conductance plus a voltage-dependent conductance that
inactivates (the valinomycin-monazomycin, C12treated membrane), and the other is with a membrane containing two voltage-dependent conductances, neither of which inactivates (the EIM-protamine-treated membrane). Interestingly, the squid 0
5
10 TIME, sec 15
20
'
25
giant axon contains both mechanisms. The rising ' L o 0
phase of the action potential is due to turning on of
FIG. 68. Current versus time record showing discrete current
gNa;
the falling phase is due both to inactivation ofg,, jumps for a single EIM channel in an oxidized cholesterol memand to turning on ofg,. In principle, either one of brane. The membrane separates 0.1 M KC1 solutions buffered to
these last two operations would produce the falling pH 7 with 5 mM histidine chloride. Upper truce, current record;
phase. In actuality, it is the inactivation ofg,, which lower truce, applied voltage. [From Ehrenstein e t al. (12).1
is of greater significance; the combination accelerates
the process.

l Z L I
Y

The multimolecularity implied


from both the steady-state g-V characteristics and
the kinetics strongly suggests that all three of the
voltage-dependent modifiers discussed in this article
function as channel formers rather than as carriers.
This assumption is confirmed by the occurrence
of single-channel events, such as described earlier
for gramicidin A-treated membranes, in membranes
treated with either alamethicin or EIM.Ig
Excitability-inducing material. The results with
SINGLE CHANNELS.

The failure, so far, to record single-channel events in monazomycin-treated membranes may possibly be due to a relatively
small conductance per channel, a short lifetime per channel, or
both.
Iu

-----------.
'.
\

'I-

G 0.6

'\.

\\

IL
n

20

40
60
VOLTAGE, mV

\\

80

--

FIG. 69. Voltage dependence of the fraction of time a channel


is open. Dashed curue, relative conductance of a many-channel
membrane; 0, determined from one-channel experiment; A, determined from two-channel experiment; m, determined from fourchannel experiment. [From Ehrenstein et al. (12).]

210

HANDBOOK OF PHYSIOLOGY

THE NERVOUS SYSTEM I

open) of a number of channels equals the time average (the fraction of time a channel stays open) of a
single channel (ergodic property). Put another way,
the channels act independently and do not cooperatively affect each others behavior even in a membrane containing many channels.]
In the section FORMAL CONSEQUENCES OF VOLTAGEDEPENDENT CONDUCTANCES, we discussed the possibility that strongly voltage-dependent conductances
might arise from shifting of ionic profiles and gave
several reasons for believing this mechanism unlikely in biological excitable membranes. The EIM
system confirms this expectation in the model system.
Given that a single channel is open, the current
through it is directly proportional to the voltage
across the membrane (12, 39). Thus the conductance
of a single channel is ohmic and not voltage dependent. What is voltage dependent is the state of the
channel; that is, voltage determines if (or rather the
probability that) the channel is open or closed. The
entire macroscopic voltage-dependent conductance
follows from this gating property of the channel.
We wish t o add, by way of completeness, that oxidized cholesterol membranes happen to give especially simple results - namely, the EIM channel exists in only two states. In other lipids (e.g., sphingomyelin) there are several conducting states, the highest one being comparable to that found in oxidized
cholesterol and the lowest one being essentially zero
(4). The conductance ratio between the on and off
state in an oxidized cholesterol membrane is only
about a factor of 5, so that even an off channel has a
significant conductance. It appears that in oxidized
cholesterol the channel can assume only two of several possible conductance configurations and that the
channel can never be fully turned off (nonconducting). These single-channel findings are again in harmony with the macroscopic conductance data. Thus,
in an oxidized cholesterol membrane containing
many channels, the ratio of membrane conductance
at voltages where the system is fully turned on to
membrane conductance a t voltages where it is fully
turned off is only about 5, whereas in sphingomyelin
membranes this ratio is several hundred, and the off
state is indistinguishable from unmodified membrane (4).
Alamethicin . The single-channel data for alamethicin-treated membranes are somewhat different from
those for EIM-treated membranes. When a channel
turns on, its conductance fluctuates rapidly among
five or six distinct levels (15, 24), as shown in Figure
70. Neither the time that the system spends at each
level nor the total duration that the channel remains
on is strongly voltage dependent (151, in contrast to
the results just described for EIM. What is under
voltage control is the frequency with which channels
turn on; that is, the duration of the quiescent period
between bursts of activity is under voltage control; as
the potential is increased, frequency of channel occurrence increases, and the channels eventually be-

gin to overlap in time as more are recruited.


There is an interesting correspondence between
the voltage-independent pair malonylgramicidin Agramicidin A and the voltage-dependent pair EIMalamethicin. The former of each pair creates a (more
or less) permanent channel in the membrane that
opens and closes with a certain mean frequency; in
the case of EIM, this process is under voltage control.
The latter of each pair creates a channel by the aggregation of monomers within the membrane, and
channel closure is actually a disassembly of the
aggregate. With gramicidin A the number of monomers in the membrane is a function of the concentration of the antibiotic in solution, whereas with alamethicin, this number is also a function of voltage.
In both cases there is a large excess, or pool, of
monomers compared to the number of channels.
Hence, one does not see a single channel opening and
closing, but rather a certain number of individual
monomers aggregating and disaggregating, and then
at a later time the same number of, but different
individual, monomers aggregating and disaggregating, and so on.
SUMMARY AND CONCLUSION. Our description of bilayers treated with voltage-dependent modifiers has proceeded on two different levels. On one plane we have
presented the phenomenology of the voltage-dependent conductances and have indicated possible physical mechanisms responsible for them. These mechanisms are poorly understood a t present, and their
investigation constitutes a currently active area of
research. The similarity in the phenomenology of
these systems and of biological systems suggests that
there may also be a similarity in underlying physical
mechanisms. On the other plane, we have used these
model systems as pedagogical tools to illustrate the
theme developed in the section FORMAL CONSEQUENCES OF VOLTAGE-DEPENDENT CONDUCTANCES.
Namely, we have seen experimentally how, given a
system with a strongly voltage-dependent conductance, the shape of the I-V characteristic depends on
such extrinsic variables as the existence of a resting
potential (created by gradients of permeant ions) and
the presence of multivalent ions or charged adsorbates on one side of the membrane. We have also
demonstrated bistability or full-fledged action potentials using two conductance elements in parallel, one
having a negative slope region in its I-V characteristic. We hope that this presentation has illustrated the
point made in the subsection Current-Voltage Characteristics. Namely, the physics of the system is contained in the dependence of conductance on voltage
and time. Given this, the excitability properties
follow as analytical consequences.

** For simplicity, and in order to keep the article to manageable


length, we have focused almost exclusively on the steady-state
dependence o f g on V. This unwarranted slight of the kinetics in
no way implies that the kinetic aspects of the conductance change
are unimportant or trivial. On the contrary, they are so difficult
that at present there is no easy way of handling them.

CHAPTER

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

211

FIG. 70. A: fluctuations in the current through a membrane, formed from glycerol monooleate,
in the presence of a very small amount of alamethicin. The aqueous phase was 2 M KCI, and the
applied potential 210 mV. The baseline a t each end of the group of fluctuations corresponds to the
conductance of the pure lipid membrane. B: current fluctuations for the same system as in A recorded over a considerable period on a storage oscilloscope ( 1 large division = 5 x lo-"' A). The
intensities of the lines represent the probabilities of finding the various conductance levels. This
sort of record makes clear that within a group of fluctuations such as in A, the current tends to
take certain well-defined values. [From Gordon & Haydon (24).]

REFERENCES
1. AKASAKI,
K., K. KARASAWA,
M. WATANABE,
H. YONEHARA,

H. UMEZAWA.
Monazomycin, a new antibiotic produced
by a streptomyces. J . Antibiotics, Tokyo, Ser. A 16: 127-131,
AND

1963.
C . M.,
2. ARMSTRONG,

A N D L. BINSTOCK.
Anomalous rectification in the squid giant axon injected with tetraethylammonium chloride. J. Gen. Physiol. 4 8 859-872, 1965.
3. BASS,L. Potential of liquid junctions. Trans. Faraday SOC.6 0

1914-1919, 1964.
4. BEAN,R. C. Protein-mediated mechanisms of variable ion

conductance in thin lipid membranes. In: Membranes. Lipid

Bilayers and Antibiotics, edited by G. Eisenman. New York:


Dekker, 1973, vol. 2, p. 409-477.
5. CASS,A,, AND M. DALMARK.
Equilibrium dialysis of ions in
nystatin-treated red cells. Nature 244: 47-49, 1973.
A N D V. KRESPI.
The ion permeabil6. CASS,A., A. FINKELSTEIN,
ity induced in thin lipid membranes by the polyene antibiotics nystatin and amphotericin B. J. Gen. Physiol. 56: 100-124,
1970.
7. CHERRY,
R. J., D. CHAPMAN,
AND D. E. GRAHAM.
Studies on

the conductance changes induced in bimolecular lipid membranes by alamethicin. J. Membrane Biol. 7: 325-344, 1972.

212

HANDBOOK OF PHYSIOLOGY

fl

T H E NERVOUS SYSTEM I

8. COHEN,H., AND J. W. COOLEY.


The numerical solution of the
time-dependent Nernst-Planck equations. Biophys. J . 5: 145162, 1965.
9. DAVSON,H. Growth o f the concept o f the paucimolecular
membrane. Circulation 2 6 1022-1037, 1962.
M., J . D. DUNITZ,AND J . KRAJEWSKI.
Structure of
10. DOBLER,
the K complex with enniatin B, a macrocyclic antibiotic
with K- transport properties. J . Mol. B i d . 42: 603-606, 1969.
11. DOMINGUEZ,
J . , J. D. DUNITZ,H. GERLACH,
A N D V .PRELOG.
Stoffwechselprodukte von Actinomyceten 32. Mitteilung.
Uber die Konstitution von Nonactin. Helv. Chim. Acta 4 5
129-138, 1962.
G . , H. LECAR,
AND R. NOSSAL.
The nature of the
12. EHRENSTEIN,
negative resistance in bimolecular lipid membranes containing excitability-inducing material. J . Gen. Physiol. 55: 119133, 1970.
A. On the movement o f small particles suspended
13. EINSTEIN,
in a stationary liquid demanded by the molecular-kinetic
theory o f heat. (Translated from A n n . Phys. 17: 549-560,
1905.) In: Investigations On the Theory of the Brownian Mouement by Albert Einstein, edited with notes by R. Furth.
(Translated by A. D. Cowper.) New York: Dover, 1956, p. 118.
14. EINSTEIN,
A. The elementary theory o f the Brownian motion.
(Translated from Z . Eiektrochem. 14: 235-239, 1908.) In: Znuestigations on the Theory o f t h e Brownian Mouement, by Albert
Einstein, edited with notes by R. Furth. (Translated by A. D.
Cowper.) New York: Dover, 1956, p. 68-85.
M. Voltage gateable ionic pores in15. EISENBERG-GRUNBERG,
duced by alamethicin in black lipid membranes (Ph.D. thesis). Pasadena: California Institute of Technology, 1972.
16. EISENMAN,
G . , S. CIANI,AND G. SZABO.
The effects of the
macrotetralide actin antibiotics on the equilibrium extraction o f alkali metal salts into organic solvents. J . Membrane
B i d . 1: 294-345, 1969.
A. Aqueous pores created in thin lipid mem17. FINKELSTEIN,
branes by the antibiotics nystatin, amphotericin B, and
gramicidin A: implications for pores in plasma membranes.
In: Drugs and Transport Processes, edited by B. A. Callingham. New York: Macmillan, 1974, p. 241-250.
18. FINKELSTEIN,
A., A N D A. CASS.Permeability and electrical
properties of thin lipid membranes. J . Gen. Physiol. 52: 145s172s, 1968.
19. FINKELSTEIN,
A., AND R.HOLZ.Aqueous pores created in thin
lipid membranes by the polyene antibiotics nystatin and
amphotericin B. In: Membranes. Lipid Bilayers and Antibiotics, edited by G . Eisenman. New York: Marcel Dekker, Inc.,
1973, vol. 2, p. 377-408.
A., A N D A. MAURO.Equivalent circuits a s re20. FINKELSTEIN,
lated to ionic systems. Biophys. J . 3: 215-237, 1963.
21. FRANKENHAEUSER,
B.. A N D A. L. HODGKIN.The action o f
calcium on the electrical properties of squid axons. J . Physiol.
London 137: 217-244, 1957.
22. GOODALL,
M. C. Structural effects in the action o f antibiotics
on the ion permeability o f lipid bilayers. 111. Gramicidins A
and S, and lipid specificity. Biochim. Biophys. Acta 219:
471-478, 1970.
M. C. Thickness dependence in the action o f grami23. GOODALL,
cidin A on lipid bilayers. Arch. Biochem. Biophys. 1 4 7 129135, 1971.
L. G. M., A N D D. A. HAYDON.
The unit conductance
24. GORDON,
channel of alamethicin. Biochim. Biophys. Acta 255: 10141018, 1972.
E., AND F. GRENDEL.
On bimolecular layers o f lipoids
25. GORTER,
on the chromocytes o f the blood. J . Exptl. Med. 41: 439-443,
1925.
26. HAYDON,
D. A,, AND S. B. HLADKY.
Ion transport across thin
lipid membranes: a critical discussion of mechanisms in selected systems. Quart. Rev. Biophys. 5: 187-282, 1972.
27. HEYER,E. J. Two physical methods o f producing inactivation in monazomycin.-treated thin lipid membranes (Ph.D.
thesis). New York: Albert Einstein College of Medicine of
Yeshiva University, 1974.

28. HLADKY,
S. B., A N D D. A. HAYDON.
Ion transfer across lipid
membranes in the presence o f gramicidin A. 1. Studies on the
unit conductance channel. Biochim. Biophys. Acta 274: 294312, 1972.
29. HODGKIN,
A. L., AND A. F. HUXLEY.Currents carried by
sodium and potassium ions through the membrane o f the
giant axon ofloligo. J . Physiol. London 116: 449-472, 1952.
30. HODGKIN,A. L., AND A. F. HUXLEY.The components o f
membrane conductance in the giant axon of Loligo. J . Physiol. London 116: 473-496, 1952.
31. HODGKIN,
A. L., AND A. F. HUXLEY.
The dual effect o f membrane potential on sodium conductance in the giant axon o f
Loligo. J . Physiol. London 116: 497-506, 1952.
A. L., AND A. F. HUXLEY.
A quantitative descrip32. HODGKIN,
tion o f membrane current and its application to conduction
and excitation in nerve. J . Physiol. London 117: 500-544,
1952.
A. L., A. F. HUXLEY,
AND B. KATZ.Measurement
33. HODGKIN,
of current-voltage relations in the membrane of the giant
axon ofLoligo. J . Physiol. London 116: 424-448, 1952.
A. L., A N D R. D. KEYNES.Active transport of
34. HODGKIN,
cations in giant axons from Sepia and Loligo. J . Physiol.
London 128: 28-60, 1955.
The water and nonelectrolyte
35. HOLZ,R., AND A. FINKELSTEIN.
permeability induced in thin lipid membranes by the polyene
antibiotics nystatin and amphotericin B. J . Gem Physiol. 56:
125-145, 1970.
36. KATCHALSKY,
A., A N D P. F. CURRAN.
Nonequilibrium Thermodynamics i n Biophysics. Cambridge, Mass.: Harvard
Univ. Press, 1965, p. 1-248.
37. KILBOURN,
B. T.. J . D. DUNITZ,L. A. PIODA,A N D W. SIMON.
Structure of the K - complex with nonactin, a macrotetrolide
antibiotic possessing highly specific K transport properties.
J . Mol. B i d . 30: 559-563, 1967.
38. LAMPEN,
J . 0. Interference o f polyene antifungal antibiotics
(especially nystatin and filipin) with specific membrane functions. In: Biochemical Studies of Antimicrobial Drugs, edited
by B. A. Newton and P. E. Reynolds. Cambridge, Mass.: SOC.
Gen. Microbiol., 1966, p. 111-130.
39. LATORRE,
R., G . EHRENSTEIN,
A N D H. LECAR.Ion transport
through excitability-inducing material (EIM) channels in
lipid bilayer membranes. J . Gen. Physiol. 60: 72-85, 1972.
40. MACINNES,
D. A. The Principles of Electrochemistry. New
York: Dover, 1961, p. 220-245.
41. MAURO,A. Some properties o f ionic and nonionic semipermeable membranes. Circulation 21: 845-854, 1960.
42. MAURO,A. Anomalous impedance: a phenomenological property o f time-variant resistance. An analytic review. Biophys.
J . 1: 353-372, 1961.
43. MAURO,A. Space charge regions in fixed charge membranes
and the associated property o f capacitance. Biophys. J . 2: 179198, 1962.
44. MAURO,A,, R. P. NANAVATI,
A N D E. HEYER.Time-variant
conductance o f bilayer membranes treated with monazomycin and alamethicin. Proc. Natl. Acad. Sci. US 69: 3742-3744,
1972.
S. G . A., G. SZABO,
A N D G . EISENMAN.
Diva45. MACLAUGHLIN,
lent ions and the surface potential of charged phospholipid
membranes. J . Gen. Physiol. 5 8 667-687, 1971.
S. G. A,, G . SZABO,G. EISENMAN,
AND s. M.
46. MCLAUGHLIN,
CIANI. Surface charge and the conductance o f phospholipid
membranes. Proc. Natl. Acad. Sci. US 67: 1268-1275, 1970.
47. MECHLINSKI, c . P. SCHAFFNER, P. GANIS,AND G. \VITABILE.Structure and absolute configuration o f the polyene
macrolide antibiotic amphotericin B. Tetrahedron Letters 44:
3873-3876, 1970.
48. MITSCHER,
L. A,, A. J . SHAY,
AND N. BOHONOS.
LL-A491, a
monazomycin-like antibiotic. Appl. Microbiol. 15 1002-1005,
1967.
N. C. ANDERSEN,
A N D T.
49. MOORE,J . W., M. P. BLAUSTEIN,
NARAHASHI.
Basis of tetrodoxins selectivity in blockage o f
squid axons. J . Gen. Physiol. 5 0 1401-1411, 1967.
P., AND D. 0. RUDIN.Induced excitability in recon50. MUELLER,

w.,

CHAPTER

51.

52.

53.
54.
55.

56.
57.
58.

59.
60.
61.
62.
63.
64.

6:

PHYSICAL PRINCIPLES AND FORMALISMS OF ELECTRICAL EXCITABILITY

stituted cell membrane structure. J . Theoret. Biol. 4: 268-280,


1963.
MUELLER,
P., AND D. 0. RUDIN.Action potential phenomena
i n experimental bimolecular lipid membranes. Nature 213:
603-604, 1967.
MUELLER,P., AND D. 0. RUDIN.Development of K+-Na+
discrimination in experimental bimolecular lipid membranes
by macrocyclic antibiotics. Biochem. Biophys. Res. Commun.
26: 398-404, 1967.
MUELLER,
P., AND D. 0. RUDIN.Action potentials induced in
bimolecular lipid membranes. Nature 217 713-719, 1968.
MUELLER,
P., AND D. 0. RUDIN.Resting and action potentials
in experimental bimolecular lipid membranes. J . Theoret.
BzoZ. 1 8 222-258, 1968.
MUELLER,
P., AND D. 0. RUDIN.Translocators in bimolecular
lipid membranes: their role in dissipative and conservative
bioenergy transductions. Current Topics Bioenerg. 3: 157-249,
1969.
MUELLER,
P., D. 0. RUDIN,H. T. TIEN,A N D W. C . WESCOTT.
Methods for the formation of single bimolecular lipid membranes in aqueous solution. J . Phys. Chem. 67: 534-535, 1963.
MULLER,
R. U . , AND A. FINKELSTEIN.
Voltage-dependent conductance induced in thin lipid membranes by monazomycin.
J. Gen. Physiol. 6 0 263-284, 1972.
MULLER,R. U., AND A. FINKELSTEIN.
The effect of surface
charge on the voltage-dependent conductance induced in thin
lipid membranes by monazomycin. J. Gen. Physiol. 6 0 285306, 1972.
NEUMCKE,
B., A N D P. LAUGER.
Space charge-limited conductance in lipid bilayer membranes. J . Membrane Biol. 3: 54-66,
1970.
PARSEGIAN,
A. Energy of a n ion crossing a low dielectric
membrane: solutions to four relevant electrostatic problems.
Nature 221: 844-846, 1969.
PAYNE,
J. W., R. JAKES,AND B. S. HARTLEY.
The primary
structure of alamethicin. Biochem. J . 117: 757-766, 1970.
PINKERTON,
M., L. K. STEINRAUF,
AND P. DAWKINS.
The
molecular structure and some transport properties of valinomycin. Biochern. Biophys. Res. Commun. 3 5 512-518, 1969.
PLANCK,
M. Ueber die Erregung von Electricitat und Warme
in Electrolyten. A n n . Phys. Chem. Neue Folge 39: 161-186,
1890.
PLATTNER,
P. A., K. VOGLER,
R. 0. STUDER,
P. QUITT,A N D W.
KELLER-SCHIERLEIN.
Synthesen in der Depsipeptid-Reihe. 1.

65.
66.
67.
68.
69.

70.

71.
72.

73.
74.
75.
76.

77.

2 13

Mitteilung. Synthese von Enniatin B. Helv. Chim. Acta 46:


927-935, 1963.
PRESSMAN,
B. C . Ionophorous antibiotics a s models for biological transport. Federation Proc. 27: 1283-1288, 1968.
PRESSMAN,
B. C . , E. J . HARRIS,W. S. JACGER,
A N D J . H.
JOHNSON.
Antibiotic-mediated transport of alkali ions across
lipid barriers. Proc. Natl. Acad. Sci. US 58: 1949-1956, 1967.
RUTGERS,
A. J. Physical Chemistry. New York: Interscience,
1954, p. 392.
SARGES,
R., AND B. WITKOP.Gramicidin A. V. The structure
of valine- and isoleucine-gramicidin A. J. A m . Chern. SOC.87:
2011-2020, 1965.
SHEMYAKIN,
M. M., E. I. VINOGRADOVA,
M. Y. U. FEIGINA,
N. A. ALDANOVA,
N. F. LOGINOVA,
I. D. RYABOVA,
A N D I A.
.
PAVLENKO.
The structure-antimicrobial relation for valinomycin depsipeptides. Experientia 21: 548-552, 1965.
SZABO,G., G. EISENMAN,
AND S. CIANI.The effects of the
macrotetralide actin antibiotics on the electrical properties of
phospholipid bilayer membranes. J. Membrane Biol. 1: 346382, 1969.
TEORELL,
T. Transport processes and electrical phenomena in
ionic membranes. Progr. Biophys. Biophys. Chern. 3: 305-369,
1953.
TEORELL,
T. Electrokinetic membrane processes in relation to
properties of excitable tissues. I. Experiments on oscillatory
transport phenomena in artificial membranes. J. Gen. Physiol. 42: 831-845, 1959.
TEORELL,
T. Electrokinetic membrane processes in relation to
properties of excitable tissues. 11. Some theoretical considerations. J . Gen. Physiol. 42: 847-863, 1959.
TIEN,H. T. Thickness and molecular organization of bimolecular lipid membranes in aqueous media. J . Mol. B i d . 16: 577580, 1966.
URRY,D. W. Protein conformation in biomemhranes: optical
rotation and absorption of membranes suspensions. Biochim.
Biophys. Acta 265: 115-168, 1972.
URRY,D. W., M. C. GOODALL,
J . D. GLICKSON,
A N D D. F.
MAYERS.
The gramicidin A transmembrane channel: characteristics of head-to-head dimerized ?ilL., helices. Proc. Natl.
Acad. Sci. US 68: 1907-1911, 1971.
WALZ,D., E. BAMBERG,
A N D P. LAUGER.
Nonlinear electrical
effects in lipid bilayer membranes. I. Ion injection. Biophys.
J. 9: 1150-1159, 1969.

Vous aimerez peut-être aussi