Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

MODEL OF AN EXCITABLE NEURON

1. The excitable cell and the action potential

An important function of the cell is to regulate the potential difference across the membrane by
controlling the ion channels. Several cells, such as neurons and muscle cells, use the membrane
potential as a signal.

In order to better understand cellular electrophysiology, it is appropriate to distinguish cells into


two broad categories. Some cells keep their potential stable: if a current stimulus is applied for a
short time interval, the potential progressively changes and then, as soon as the current is removed,
quickly returns to equilibrium. Such cells are non-excitable, and behave roughly like an integrator
with some losses, that is, as a capacitance that charges and discharges through a resistance. Other
cells, on the other hand, are excitable: following the application of a current of sufficient amplitude,
the membrane potential shows a large excursion, called action potential, before returning to an
equilibrium condition.

The action potential, an example of which is shown in the following figure for a neuron, has
some typical characteristics:

i) presence of a threshold. Only if the membrane potential is perturbed beyond a certain value
(for a typical neuron this is about - 55mv) the action potential is produced. As long as the
membrane potential remains below the threshold value, the cell behaves as if it were non-excitable;

ii) the action potential has a stereotyped form, that is, it is of the "all or nothing" type. This
means that, once the phenomenon is engaged (and therefore once the threshold value is exceeded),
the shape of the action potential becomes independent of the stimulus. As can be seen from the
figure, starting from a rest situation of about - 70mV (typical for a neuron), once the threshold value
is exceeded (around - 55 mV) the action potential rapidly rises to positive potential values,
(showing an overshoot on + 30-40 mV) and then rapidly returns to the rest condition through a
modest undershoot (down to -80-85 mV). The overall duration of the phenomenon is approximately
1-2 ms.

iii) Once the action potential has ended, there is a time interval, called absolute refractory
period, during which the cell behaves as if it were non-excitable; it is no longer possible to generate
the action potential, despite the application of a strong current.

iv) At the end of the absolute refractory period the cell regains the ability to become excited, but
there is a further period of time, called relative refractory time, during which it is necessary to
stimulate the cell with stronger current pulses to generate the action potential. In other words,
during this time interval it is as if the threshold had risen, making it more difficult to trigger the
phenomenon.

1
2. The formulation of the Hodgkin Huxley model and the voltage-clamp experiment

The mechanisms capable of generating the action potential were described by Alan Hodgkin and
Andrew Huxley (with the further collaboration of Bernard Katz) in a series of works that appeared
in 1952. To understand the genesis of the model, we start from the electrical analog of the
membrane in which, for simplicity, all the contributions other than the sodium and potassium ion
channels (including the contributions of chloride and active ion pumps) are incorporated in an
equivalent conductance geq and in an equivalent potential Eeq. Assuming to stimulate the neuron
with a current i (directed from the outside to the inside) we can write the following equation:

 g Na V  ENa   g k V  Ek   g eq V  Eeq   i
dV
C (1)
dt

where V is the membrane potential, and ENa and Ek are the Nernst potentials of sodium and
potassium. Note that, since the sodium and potassium currents (even during the development of an
action potential) are extremely small (order of the influence of Na+ 3.7 pmoles/cm2; order of the
efflux of K+: 4.3 pmoles/cm2) ionic concentrations undergo minimal variations; it is therefore
legitimate to consider the relative Nernst potentials to be constant.

If the neuron behaved according to equation (1), assuming constant conductances and potentials,
the operation would be comparable to that of a capacitance C which is charged and discharged
through conductances; the cell would be non-excitable. Hodkin and Huxley first showed that,
during the appearence of an action potential, variations in the conductivity of some ion channels
occur. In particular, a work by Hodgkin and Katz from 1949 highlighted the role of sodium and
potassium. These authors understood that a temporal change in the conductance of sodium and
potassium could explain, at least qualitatively, the shape of the action potential. In fact, an increase
in gNa is able to shift the membrane potential towards the (positive) value of the sodium Nernst
potential, as it occurs during the first part of the action potential. Conversely, an increase in the
conductance of potassium, gK, is able to shift the membrane potential towards more negative values
(as occurs during the short undershoot with which the action potential ends).

We begin, then, by asking ourselves which law the sodium and potassium channels might follow
during an action potential. The starting point is to imagine that the channels have gates, which
selectively allow the passage of the ionic species, and that these gates can open or close, depending
on the potential of the membrane. Since the opening and closing of these gates is regulated by the
membrane potential, the related ion channels are called voltage-dependent (these are precisely the
sodium and potassium channels object of the model of H. H.).

Let us consider a gate relative to the potassium channel. At a certain moment the gate can be
opened (thus letting the potassium ion pass) or closed. We denote by n the fraction of the gates open
at a certain instant t. It must be 0 < n <1. Note that, by the law of large numbers, since there is a
very high number of gates in the membrane of a neuron, this fraction coincides with the probability
that, at a given instant, a single gate is open or closed. The starting point is to imagine that this gate
can open or close with a first order kinetics, regulated by the potential V. We can write the
following differential equation:

2
dn
  n (V )(1  n)   n (V ) n (2)
dt

The previous equation can be interpreted as follows.  n (V ) indicates the opening rate, i.e. the
probability that a closed gate will reopen, while  n (V ) indicates the closing rate, equal to the
probability that an open channel will close again (n being the fraction of open channels and 1 ‒ n
the fraction of closed gates). Note that Eq. (2) derives from similar equations of the kinetics of
chemical substances.

To understand what happens during the development of an action potential, it is necessary to


express the dependence of the opening and closing rates on V, i.e. to express the functions  n (V )
and  n (V ) . It is extremely difficult, however, to find the solution of Eq. (2) in conditions in which
the potential varies freely, being a non-linear equation (in fact, if V varies, the coefficients  and 
also vary over time, making the entire equation non-linear). Conversely, Eq. (2) has a very simple
solution in the linear case, that is, if we consider the coefficients  and  constant. However, this is
only possible under conditions where the membrane potential is constant.

Suppose, then, that at the instant t = 0 the membrane potential has increased from its resting
value V0 (equal to the value of the neuron's equilibrium potential) to a new value, which we will
denote with V and therefore kept constant. We also assume that, at the instant t = 0, the variable n is
in equilibrium at the value n(0) = n0. The solution of the linear equation, assuming V constant, takes
on the known expression

n(t )  n (V )  n0  n (V ) exp t  n (V )  (3’)

where n0 represents the value of the variable n at time 0 (hence the initial state), while n and n
represent, respectively, the asymptotic (equilibrium) value and the time constant. Both depend on
the particular value of V (assumed constant). Moreover, it turns out

 n (V )
n (V )  (3’’)
 n (V )   n (V )

1
 n (V )  (3’’’)
 n (V )   n (V )

Note that, from the knowledge of n (V ) and  n (V ) it is then possible to calculate  n (V ) and  n (V )
solving the (3'') and (3 '''). It turns out in fact:

n (V ) 1 1  n (V )
 n (V )   n (V )    n (V )  (4)
 n (V )  n (V )  n (V )

Hodgkin and Huxley experimentally determined the functions (3'') and (3''') point by point, for
different values of V, through experiments carried out on the squid giant axon, under voltage clamp
conditions. In particular, during these experiments, H. H. have measured the sodium and potassium

3
currents, from which it is possible to trace the conductances of sodium, gNa, and of potassium, gK.
The apparatus used is described in the following figure

Device used in the voltage clamp experiment.

A first electrode, connected to a voltmeter, allows the measurement of the potential difference
between the inside and outside of the axon (with the use of an external reference electrode). This
voltage (denoted Vm in the figure to indicate the membrane potential) is then compared with a
reference value ("command signal"): every time there is a difference between Vm and the desired
value, a current is injected which allows restore Vm to the desired value. Through this feedback
system, with very high gain, the membrane potential is kept practically constant at the desired
value, even when the sodium and potassium conductances vary. Finally, an amperometer measures
the injected current, which coincides with the current that crosses the sodium and potassium
conductances (in fact, being Vm constant, the injected current must necessarily equal the current of
the ionic species). Also note the use of axial electrodes (i.e. electrodes placed along the axis of the
axon) in order to reduce the effects of potential propagation as much as possible. In this way we
tried to limit the dependence of the quantities on space, focusing only on the time variable: the
potential is no longer a function of the distance along the axon but only of the time elapsed since the
application of the stimulus.

4
The first important observation of H.H. was that, following the application of a voltage step, the
current exhibits a trend similar to that shown in the following figure, i.e. the current first shows a
decrease to negative values, followed by an increase (note that a negative current must be
interpreted as a current entering the cell, and vice versa). With various considerations (which we do
not report here) H. H. hypothesized that the initial emtering current was almost entirely attributable
to sodium, while the outgoing current in the subsequent period was attributable to potassium.

The next step, then, was to try to discriminate between the sodium and potassium currents. For
this purpose, H. H. repeated the experiment in a situation in which the extracellular fluid was
deprived of sodium (instead of sodium they used a substance" choline" that varied the equilibrium
on
potential of the membrane very little). Suppose we indicate with iNa the sodium current measured
off
under normal conditions, following the application of a certain voltage, and with iNa the current
measured in conditions with almost total absence of sodium (even with sodium replaced, a
minimum current was still present). A fundamental hypothesis of H. H.'s model was that the
conductances of sodium and potassium (and, therefore, their respective currents) had a temporal
trend that depended only on the applied voltage. This is reflected in two sub-hypotheses: i) the
on
iNa
temporal shape of the sodium currents in the two cases is identical, and therefore off
 c (where c
iNa
on
indicates a constant); ii) the potassium current iK is the same in the two cases. Indicating with iTot
off
and iTot the total ionic current in the two cases (measured with the voltage clamp) we obtain then:
on
iTot  iNa
on
 ik  c iNa
off
 ik off
and moreover iTot  iNa
off
 ik from which iTot
on
 iTot
off
 c iNa
off
 iNa
off
and

5
on
iTot  iTot
off
i 
off
. We then arrive at the expressions of the sodium and potassium currents, in the
c 1
Na

conditions of normal behavior, starting from the total currents measured:

c iTot
on
 iTot
off
 c iTot
on
 iTot
off
  ciTotoff  iToton
on
iNa  i k  iTot
on
 iTot
off
 iTot
on
 (5)
c 1 c 1 c 1

Once the sodium and potassium currents have been obtained for a certain voltage value, it is
possible to derive the conductances, since the voltage difference across the membrane is known.
Let's start by considering the conductance of potassium. The trends shown in the figures below
resulted from the voltage clamp experiments. The first figure refers to a single step increase, with
subsequent return to the base value. The second figure shows the trend over time of the conductance
gK following different voltage steps,

Trend of the conductance of potassium over time, in response to a single step increase in voltage
and a subsequent drop in voltage to the quiescent value.

Time course of the conductance of potassium, in response to voltage steps of different amplitude.
6
The next step, which allows the estimation of the functions  n (V ) and  n (V ) is to express the
conductance of potassium as a function of the variable n. Indicating with gkmax the maximum
conductance of potassium, obtainable when all the channels are open, and remembering that, in the
cell membrane, the channels are in parallel with each other, we can write

g k (t )  g k max pk (t ) (6)

where pk denotes the probability that, at the instant t, a potassium channel is open. In fact, for pk = 0
all the channels are closed and the conductance is zero; for pk = 1 all the channels are open and the
conductance is at the maximum; for intermediate values of pk the conductance linearly depends on
the fraction of open channels (the probability coincides with the fraction for the law of large
numbers).

We assume now that there is only one gate for each channel, and that it can be open or closed. In
this case we would have pk(t) = n(t). Then, replacing the expression (3 ') in (6) results:

gk (t )  gk max n(t )  gk max n (V )  n0  n (V ) exp t  n (V )

If we fix n0 and gkmax with a certain “cleverness” (these parameters must be the same for all
experiments), the goal is to derive n (V) e  n (V) for the particular value of V used during the
experiment and then derive n and n for the same value of V.

However, H.H. observed that, with the previous expression, the fitting was inadequate. To
resolve this discrepancy, we can then assume that there are multiple gates in a single potassium
channel. If there are υ gates in a channel, and each of them has the same probability n of being open
or closed at a certain instant, and the gates are comparable to independent random variables of the
binary type (1 = open, 0 = closed) then the probability of opening of a channel will be:

pk (t )  n(t) (7)

Hodgkin and Huxley tried to fit the eq. (7) to the experimental data, using the lowest value of υ
that provided a good estimate. That value turned out to be four. Therefore, the potassium
conductance equation takes the following form (in which υ = 4 was used and Eq. (3 ') was replaced
ny Eq. (7))

g k (t )  g k max n(t )4  g k max n (V )  n0  n (V )  exp t  n (V ) 


4
(8)

The fitting of equation (8) allows the derivation of the values n (V ) and  n (V ) for different
values of V, from which it is then possible to derive the functions  n (V ) and  n (V ) points by
points, using equations (4).

Let us now consider the case of sodium conductance. The conductance values shown in the
following figure resulted from the voltage clamp experiment:

7
Time course of sodium conductance, in response to voltage steps of different amplitude.

As can be seen, the trend of sodium is more complicated than that of potassium: a very rapid
initial increase in conductance, which occurs within 1 ms, is followed by a progressive reduction in
conductance. This pattern has been explained with the presence of two different types of gate: a
gate responsible for the activation of the sodium channel, whose probability of opening will be
indicated below with m, and a gate responsible for the inactivation of the sodium channel, whose
probability of opening is indicated in followed with h. Equations similar to (2) hold for both
variables. Then

dm
  m (V )(1  m)   m (V ) m (9)
dt

dh
  h (V )(1  h)   h (V ) h (10)
dt

An excellent fit was obtained assuming the presence of three activation gates, and only one
inactivation gate, for each sodium channel. The sodium conductance equation therefore takes the
form:

g Na (t )  g Na max m(t ) 3 h(t ) 


(11)
 g Na max m (V )  m0  m (V )  exp t  m (V )  h (V )  h0  h (V )  exp t  h (V ) 
3

from which the functions  m (V ) and  m (V ) ,  h (V ) and  h (V ) could be obtained point by point.

Finally, the functions  s (V ) and  s (V ) obtained point by point (with s = n, m, h) have been
interpolated with algebraic functions. This leads to the complete model of Hodgkin Huxley,
8
consisting of four differential equations (in the state variables V, n, m, h) and six algebraic
equations. The complete set of equations is shown below

3. The complete model and the genesis of the action potential

State equations:

 g Na max m 3hV  E Na   g k max n 4 V  Ek   g eq V  Eeq   i


dV
C
dt

dm
 m (V )(1  m)   m (V ) m
dt

dh
  h (V )(1  h)   h (V ) h
dt

dn
  n (V )(1  n)   n (V ) n
dt

Algebraic equations:

 40  V
 m (V )  0.1
  40  V 
exp  1
 10 

  65  V 
 m (V )  4 exp 
 18 

  65  V 
 h (V )  0.07 exp 
 20 

1
 h (V ) 
  35  V 
exp  1
 10 

 55  V
 n (V )  0.01
  55  V 
exp  1
 10 

  65  V 
 n (V )  0.125 exp 
 80 

with the following values for the parameters

9
gNamax = 120; gkmax = 36; geq = 0.3; ENa = 55; Ek = -77; Eeq = -54.4. (voltages are in mV,
conductances are in mS/cm2, times are in ms; all other dimensions are congruent).

To understand in a qualitative way the behavior of the model when the voltage V is free (therefore
no longer in the voltage clamp conditions) it is interesting to observe the following figure, which
shows the values of the variables n, m, h in stationary conditions, and the respective time constants.
We can note that in conditions of strong membrane repolarization (V < ‒70 mV) the sodium
channels are almost completely closed (in fact the variable m is just above zero) the potassium
channels are also largely closed (the variable n is equal to about 0.1-0.2) while variable h is at a
high value (therefore, if there were no m, the channel would be open). We also note that the variable
m has a time constant much lower than the time constant of the other two variables (less than 1 ms,
compared to 6-8 ms of the other ones).

Patterns of thr time constants and of the variables m, n and h at equilibrium, obtained from the
voltage clamp experiments

The trend of the variables m, n and h is able to explain the mechanisms that determine the action
potential. Imagine perturbing the neuronal cell with a current impulse such as to open a sufficient
number of sodium channels (in fact, the variable m has a much faster dynamics than the other two
variables, and therefore the sodium channels open, following the depolarization of the membrane,
with times less than 1 ms). The opening of the sodium channels causes an entry of Na+ ions from
the outside to the inside of the membrane, which is followed by a further increase in the membrane
potential (i.e. a further depolarization). A positive feedback is then engaged, which induces a strong
and rapid increase in the membrane potential. If the conductance of sodium became much greater
than the conductances of all the other ion channels, the membrane potential would be brought to the
value of the Nernst potential of sodium, therefore about 30-40 mV.

However, as time progresses (i.e. after 1-2 ms) two other phenomena occur: potassium channels
begin to open (variable n) and sodium channels begin to close (variable h). Note, in fact, from the
previous figure, that these variables have a time constant of a few ms. Consequently, the membrane

10
potential starts to decline (greater output of K+ ions from the inside to the outside of the membrane,
less input of Na+ ions). The membrane potential transiently moves towards the equilibrium potential
of potassium, causing an undershoot. At this point, due to the decrease in the potential V, the
phenomena described above trigger in the opposite direction (m rapidly decreases, n decreases and
h increases, both with a slower dynamics) until the potential returns to the equilibrium.

The phenomena described above are summarized in the following figure.

Conductance trend of sodium and potassium, able to explain the genesis of the action potential.
Note the presence of a positive feedback in the dynamics of sodium, and the stabilizing effect of the
dynamics of potassium.

Finally, note that the presence of a period of time in which m has returned to the equilibrium value
(close to zero) while the variables h and n are still at an altered value (h close to zero, n very high,
due to the their slow dynamics) is able to explain the presence of the refractory times. In fact,
during this time interval, the dynamics of sodium is hindered by the variable h, while the potassium
channels, still open, keep the action potential at a more negative value. It is therefore very difficult
to alter the membrane potential, and induce an action potential, through an external stimulus.

11
The propagation of the action potential along the axon

In the previous discussion we imagined that the membrane potential was the same throughout the
cell, thus having a lumped constant model. In reality, the membrane potential can vary in a very
significant way along the cell membrane, especially for those neurons characterized by very long
axons, and considering the rapid variations of the potential. The membrane potential along an axon
must therefore be described more accurately through a distributed constant model, which considers
not only the dependence of the potential on time, but also on the spatial coordinates.

In the following we will assume a neuron with a long axon of longitudinal shape, and we will
concentrate on the phenomenon of potential propagation. In particular, we will assume that the axon
is comparable to a cylinder of radius a. Furthermore, we consider that the potential variations along
the other two directions are negligible compared to the variations along the longitudinal coordinate,
so we will consider the membrane potential as a function only of the variables time, t, and of the
longitudinal coordinate, x.

The construction of the model follows a treatment similar to that used in the theory of cables. Let
us consider an infinitesimal portion of the axon with length dx. The electrical analog of the
membrane is shown in the following figure, where we use lowercase letters to indicate quantities
expressed per unit of length (if longitudinal) or per unit of surface (if transversal), while capital
letters are used for quantities expressed in whole units.

The menaing of the symbols is the following one:

ri: longitudinal resistance per unit length, due to the intracellular fluid (in KΩ/mm);

re: longitudinal resistance per unit length, due to the extracellular fluid (in KΩ/mm);

c: membrane capacity per unit surface;

V(x, t): membrane potential at the coordinate x and at time t;

12
Ii (x, t), Ie(x, t): longitudinal, intracellular and extracellular currents at time t and at the coordinate
x.

im (x, t): membrane current per unit surface at the coordinate x and at time t.

ie (x, t): current injected (from teh external to the internal) per unit surface at the coordinate x and
at time t.

The first equation we can write concerns the balance of flows at the x coordinate. However, it
should be remembered that the capacitance c, and the two currents im and ie, are expressed per unit
of surface area. The set of these three terms represents the transverse admittance of the membrane.
This transverse admittance represents the membrane circuit that we used in the previous paragraphs.
To obtain the values relating to an infinitesimal portion with length dx, these terms must be
multiplied by the surface of the membrane. A section with length dx will have an area equal 2  a
dx. We can then write the following balance equation at the node:

 V ( x, t ) 
2 a dx c  im ( x, t )  ie ( x, t )  I i ( x, t )  I i ( x  dx, t )
 t 

which, dividing by dx, gives:

 dV ( x, t )  I ( x, t )
2 a c  im ( x, t )  ie ( x, t )   i (12)
 dt  x

The current Ii,, in turn, is connected to the intracellular potential (which we will call Vi(x,t)) by
the following equation, which is obtained by calculating the infinitesimal voltage drop along the
resistance ri, that is

ri dx I i ( x, t )  Vi ( x, t )  Vi ( x  dx, t )

in which the resistance has been multiplied by dx, since the parameter ri is a resistance per unit
length. The previous equation can thus be written as follows

1 Vi ( x, t )
I i ( x, t )   (13)
ri x

A similar equation also holds as to the current Ie,, which flows externally to the membrane, using
however the extracellular potential, and the relative extracellular resistance per unit of length:

1 Ve ( x, t )
I e ( x, t )   (14)
re x

In the eqs. (13) and (14), however, the new variables, Vi and Ve. appear. We want to express
these variables using only the membrane potential V. To achieve this goal, we can note that the
following balance relationship holds between the two currents

13
I i ( x, t )  I i ( x  dx, t )  I e ( x  dx, t )  I e ( x, t ) (15)

This equation was obtained, simply, by expressing the current that passes through the transverse
admittance (ie the right-hand member of Eq. (12)): the current entering from above must obviously
be equal to the current going out from below.

By inserting (13) and (14) into (15), dividing by dx, and assuming that ri and re are not functions
of x (we consider an axon with constant diameter), we obtain:

I i ( x, t ) 1  2Vi ( x, t ) I e ( x, t ) 1  2Ve ( x, t )
    
x ri x 2 x re x 2

By remembering that V(x, t) = Vi(x, t) - Ve(x, t), the previous equation can be rewritten in the
following form

I i ( x, t ) 1  2Vi ( x, t ) 1  2Vi ( x, t ) 1  2V ( x, t )
    
x ri x 2 re x 2 re x 2

from which we obtain:

 1 1   2Vi ( x, t ) 1  2V ( x, t )
    
 ri re  x re x 2
2

and therefore

I i ( x, t ) 1  2V ( x, t )
  (16)
x ri  re x 2

By now replacing (16) in (12), we obtain the expression of the membrane potential along the axon:

 V ( x, t )  1  2V ( x, t )
2 a c  im ( x, t )  ie ( x, t )  (17)
 t  ri  re x
2

The solution of Eq. (17) depends, of course, on the expression of the current flowing through the
membrane per unit area. We know that this current is mainly determined by ionic currents and that,
as seen in the Hodgkin-Huxley model, it can take on a very complex course over time. In particular,
this current will also be a function of the potential V.

Equation for the propagation along the axon: linear case

Let's first consider the case in which the neuron is close to its resting potential V0, and we consider
small variations in V, such as not to trigger the action potential. A similar treatment would also
apply, more generally, to a non-excitable cell. If the membrane does not deviate from linear
behavior, the membrane current can be written in the form

14
im ( x, t )  gm V ( x, t ) V 0  (18)

where gm, called membrane conductance, represents the conductance per unit of surface in resting
conditions. We substitute (18) in equation (17) and, remembering that V0 is constant (i.e.
independent of t and x), we pass to the variation system, using the new variable v = V – V0. We
obtain

 v ( x, t )  1  2 v ( x, t )
2 a c  g m v ( x, t )  ie ( x, t ) 
 t  ri  re x
2

We call rm = 1/gm the membrane resistance (which will be expressed in MΩ ∙ mm2). Multiplying all
the members of the previous equation by rm, and dividing by 2a, we get:

v( x, t ) rm  2 v ( x, t )
rm c  v( x, t )  rmie ( x, t ) 
t 2 a ri  re  x 2

Let's set now:

 m  rm c : time constant of the membrane

rm
 : space constant
2 a ri  re 

Let's check the units of measurement. Since c is a capacitance per unit area (for example c = 10
nF/mm2), and rm being the reciprocal of a conductance per unit area (for example rm = 1 MΩ ∙ mm2)
we obtain a time constant of the order of 1-10 ms. Finally, being a in m (for example an axon can
have a radius of 2 m), the longitudinal resistances are expressed per unit of length (for example:
ri  re = 40MΩ /mm) it follows that  has the dimension of a length, typically: (1 MΩ ∙ mm2/[(40
MΩ /mm) ∙ (2 m) (2)]]1/2 = [1 mm3/ (40 ∙ 4  10-3 mm)]1/2 ~ 2 mm.

v ( x, t )  2 v ( x, t )
m  v ( x, t )  rmie ( x, t )  2 (19)
t x 2

Remember that, in writing Eq. (19), we considered rm as the membrane resistance in resting
conditions and we assumed that its value remains constant, as in the case of a non-excitable cell.

The solution of equation (19) requires a specification not only of the initial state (at time t = 0), but
also of the spatial boundary conditions (it is in fact a partial differential equation). There are several
ways to specify these conditions. With regard to space, the boundary conditions must be specified at
the terminal points of the axon, or at the branches. For example, at a branch point it will be
necessary to specify the continuity of the potential and the conservation of the total current. In the
following we will consider only the simpler case of an axon of infinite length. Obviously such a
case cannot occur in practice, but it can be considered a good approximation of a very long axon.

15
I example: stationary condition in the presence of a constant current.

Suppose we stimulate the axon with a constant current, and that the electrode is point-like (so the
current is null everywhere except for x = 0). In particular, suppose we start with an electrode of
length , which injects a total current Ie. We'll have

 I e se x   / 2
2 a  im ( x, t )  
 0 se x   / 2

Going to the limit for   0 , and remembering the properties of the Dirac delta function, we can
write.

 Ie
lim se x   / 2 Ie
im ( x, t )   0 2 a    ( x)
 0 se x   / 2 2 a

Let's consider the solution to equilibrium, when the transient is over. Then, for every x  0, eq. (19)
becomes:

d 2 v ( x, t )
v( x )  2 x0 (20)
dx2

whose solution has the following form

v ( x )  A exp x    B exp x   x0 (21)

We need to find the two constants A and B, imposing the boundary conditions. Since the cable is
assumed to be of infinite length, we require that the solution not diverge. Then A = 0 for x < 0, B =
0 for x > 0. We get

 A exp x   x0
v( x)  
 B exp x   x0

Furthermore, for symmetry, it must be A = B. From the knowledge of the current Ie, by imposing the
derivative of the voltage V in the origin, it is finally possible to derive the constant A, which will
naturally be proportional to the current. The proof is omitted (see, for example, Dayan and Abbott,
I r
Theoretical Neuroscience). It turns out A  e m . The general form of the solution is then the
2 2a
following one:

v( x )  A exp x    exp x  
I e rm
(22)
2 2a

16
Finally, let us consider the case in which a current pulse of limited duration is injected. That is, now
we have:

I e m
im ( x, t )   ( x ) ( )
2 a

This means that the total charge injected (current multiplied the duration) is equal to I e m .

We provide the solution, omitting the proof (see, for example, Tucker, 1988). Of course now, being
in a non-stationary case, the solution will depend on both x and t. It turns out

I e rm   x2   t 
v ( x, t )  exp  m 2  exp   (23)
2a  4 t   m 

Eq. (23) tells us that the solution is provided by a Gaussian whose variance increases with t, but
whose amplitude is exponentially attenuated with time constant m.

An example of solutions (22) (constant current) and (23) (impulse current) is presented in the figure
below.

Trend of the membrane potential propagated along a cylindrical axon, assumed to be of infinite
length (from Dayan and Abbott, “Theoretical Neuroscience”, page 209). A) The figure on the left
refers to the stationary case, i.e. the injection of constant current at the coordinate x = 0 (Eq. (20).
The potential decays exponentially along both directions. B) The figure on the right refers to the
dynamic case, characterized by the injection of an impulsive and point-like current at the instant t =
0 and at the position x = 0 (Eq. (21)). The potential is described by a Gaussian, whose standard
deviation increases over time, but whose amplitude gradually decays.

17
Finally, note that an expression similar to (20), but more general, can be reached by transforming
Eq. (19) according to Laplace (and taking advantage of the fact that the model is linear). It is
obtained (assuming zero initial conditions everywhere)
2
 m s  1v( x, s)  rmie ( x, s)  2 d v ( x, s )
(24)
x 2

If we now suppose an impulsive external current, and therefore the external current is null
everywhere for x  0, we obtain an expression similar to (20) but in the Laplace domain:

2 d 2 v ( x, s )
v ( x, s )  x0 (25)
 m s  1 dx 2

whose solution takes the form:

 x  ms  1   x  ms  1 
v ( x )  A exp    B exp  x0 (26)
     
   

The first term of (26) represents a direct wave (which propagates in the direction of the positive x)
while the second term is an inverse wave. The constants A and B must be assigned on the basis of
the boundary conditions (shape of the injected current, any load impedances of the line or any
bifurcations, etc ...).

18

You might also like