See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/44614734

Neurones and Synapses for Systemic Models of Psychiatric Disorders

Article  in  Pharmacopsychiatry · May 2010

DOI: 10.1055/s-0030-1252025 · Source: PubMed

CITATIONS READS

15 825

3 authors:

Svetlana Postnova Epaminondas Rosa

The University of Sydney Illinois State University
66 PUBLICATIONS   664 CITATIONS    59 PUBLICATIONS   950 CITATIONS   

SEE PROFILE SEE PROFILE

Hans Albert Braun

Philipps University of Marburg
168 PUBLICATIONS   3,076 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Neurophysiologie View project

Computational Investigation of the Biophysical Mechanisms Underlying Thermoreception in the AFD neuron of Caenorhabditis elegans View project

All content following this page was uploaded by Hans Albert Braun on 22 May 2014.

The user has requested enhancement of the downloaded file.

S82 Original Paper
Neurones and Synapses for Systemic Models of
Psychiatric Disorders
Authors
Affiliations
Bibliography
DOI http://dx.doi.org/
10.1055/s-0030-1252025
Pharmacopsychiatry 2010;
43 (Suppl. 1): S82–S91
© Georg Thieme Verlag KG
Stuttgart · New York
ISSN 0176-3679
Correspondence
H. A. Braun
Institute of Physiology
Neurodynamics Group
University of Marburg
Deutschhausstraße 2
35037 Marburg
Germany
Tel.: + 49/6421/286 2307
Fax: + 49/6421/286 6967
braun@staff.uni-marburg.de
Postnova S et al. Neurones and Synapses for Systemic Models of Psychiatric … Pharmacopsychiatry 2010; 43 (Suppl. 1): S82–S91
S. Postnova1, 2, 3, E. Rosa Jr.4, H. A. Braun1
1
Institute of Physiology, University of Marburg, Germany
2
School of Physics, University of Sydney, Australia
3
Woolcock Institute of Medical Research, Sydney, Australia
4
Department of Physics, Illinois State University, Normal, USA
Abstract
▼ as impaired transmitter availability or reduced
We propose a mechanism-based modelling
approach which brings together the most rel-
evant features of neural dynamics and synaptic
transmission for clinically valuable simulations
of psychiatric disorders and their pharmaceu-
tical treatment. It is based on a minimal, but
physiologically justified concept, which allows to
account for a great diversity of neuronal dynam-
ics and synaptic mechanisms. It can simulate
ionotropic as well as metabotropic receptors
in addition to the effects of eventual co-trans-
mitters and external neuromodulators. The
proposed model can mimic the clinically most
Introduction
▼ simulations can demonstrate their viability. They
Psychiatric disorders are typically involving diverse
brain areas and functions. These are diseases of the
nervous system which are manifested in the
patients’ behaviour while pharmaceutical drug
treatments interfere mainly with cellular and sub-
cellular mechanisms. The drugs generally act at the
level of synaptic transmission, eventually introduc-
ing long term alterations of gene expression.
Accordingly, the understanding of psychiatric dis-
orders requires taking into account different func-
tional levels and time scales as emphasized, for
example, by Bender et al. [2], Carlsson [7], and
Braun et al. [6].
In such highly complex systems the critical
pathologies are not easy to extract and the effects
of drug treatment are hard to predict. Animal
models of psychiatric disorders are rare and have
their natural limitations. Chronic disease pro-
gression and long term treatment effects are, in
principle, difficult to examine experimentally.
For example, neuronal discharges are mostly and
ion currents are exclusively recorded in acute
experiments.
important aspects of synaptic disturbances, such
number of postsynaptic receptors, for example
due to their internalization as a function of trans-
mitter concentration. It also allows evaluation of
the effects of drugs with specific actions such as
receptor agonists and antagonists or reuptake
inhibitors. It is a major advantage of this physi-
ologically based approach that it can be adjusted
to different types of neurons and synapses, and
also can be extended to more elaborate physi-
ological situations, e. g. by including additional
receptors or ion channels, whenever this is indi-
cated by clinical or experimental data.
These are the typical situations when computer
can connect the actual knowledge of different
systems and physiological mechanisms with rea-
sonable assumptions about the missing links.
Different from theories, these connections have
to be made in a functionally rational way which
is both, physiologically appropriate and mathe-
matically consistent. The simulation data, in turn,
shall allow to make predictions for further, goal-
directed experiments and clinical studies to eval-
uate and refine the model’s assumptions. Thereby,
a successively better understanding of the dis-
ease relevant mechanisms should be achieved
and new treatment strategies may be elucidated.
However, this requires a modelling concept which
allows to understand the interdependencies
between the different functional levels, from
sub-cellular processes to brain dynamics and
behaviour.
The actual modelling strategies exhibit a great
diversity. Depending on the specific tasks, there
are simulation approaches of quite different com-
plexity. The probably most formal approaches for
use in large-scale models are based on circle

maps [36]. Already closer to physiology are the integrate-and-
fire neurons [9]. More complex spiking pattern can be achieved
with the widely used Izhikevich neurons [19] that are still suffi-
ciently simplified to allow large scale network simulations
[20, 24]. However, in most of these and related models, the syn-
aptic connections are often rather formally made via simple
delta or exponential functions [10, 23]. Other simulation
approaches, without considering individual spikes, are based on
Hopfield nets [14] where the neuronal elements are connected
via sigmoid output functions which are representing an inte-
grated activity. Sigmoid functions and relaxation terms are also
used in large scale models for the examination of dynamical
interactions between different brain areas [28].
Simulations of psychiatric disorders likewise make use of differ-
ent types of modelling approaches. An overview, with special
regard on schizophrenia, is given by Rolls et al. [35]. However, for
the examination of psychiatric disorders, the focus is mostly laid
on the imbalance of different transmitter systems [38], partly
with large scale modelling approaches [27] but mostly with
detailed simulations of transmitter metabolism and homeosta-
sis [34]. The question is, how the diverse modelling approaches
can be connected towards a better understanding of psychiatric
disorders which requires to understand the underlying distur-
bances of brain functions on the basis of cellular processes.
We propose that these connections can best be made with a
modelling approach which is built up at the level of neurones
and synapses. Moreover, the model parameters and variables
should represent physiological mechanisms which, in principle,
are experimentally or clinically accessible. The still most widely-
used method for mechanism-based modelling is the Hodgkin-
Huxley (HH) type approach [13]. It simulates neuronal dynamics
in terms of voltage- and time-dependent activation and inacti-
vation of ion channels. Such a conductance-based approach has
been realized in different forms for the simulation of neuronal
dynamics with a variety of simplifications as well as extensions
[3, 8, 12]. It can also be used for mechanism-based simulations of
synaptic transmission and plasticity with implementation of
transmitter-gated ion channels [1, 32, 37].
The HH-type models, however, have a fatal disadvantage. They
can easily become very complicated with many dimensions
which makes it difficult to extract the functionally relevant
dynamics. Accordingly, major efforts shall be made towards
physiologically justified model simplifications while the model’s
structure, nevertheless, should allow to examine specific func-
tions in more details whenever this is requested by specific
tasks.
Our approach is founded on previous studies of single neuron
dynamics [3, 5, 11, 15, 17] and more recent studies of neuronal
synchronization [29–31] including novel concepts for the mod-
elling of synaptic modulation [32]. This approach allows to con-
sider subcellular processes as well as neuronal network
dynamics. By its nature, the proposed model is flexible and can
easily be adjusted to fit particular conditions dictated by experi-
mental data and physiological insights. These particular capa-
bilities of Hodgkin-Huxley type modelling approaches have
most impressively been illustrated by Denis Noble [26] in
describing the development of a physiologically appropriate
model of heart excitability.
The following sections describe the physiological background
and the main concept of our conductance-based modelling
approach, specifically considering the simulation of voltage-
dependent and transmitter-gated ion channels. Throughout
Postnova S et al. Neurones and Synapses for Systemic Models of Psychiatric … Pharmacopsychiatry 2010; 43 (Suppl. 1): S82–S91
Original Paper S83
these sections, particular emphasis is laid on the description
and justification of the model simplifications and potential
extensions. The study is complemented by a few examples to
demonstrate potential applications. The paper concludes with a
discussion of the specific features of this modelling approach for
clinical applications.
Physiological Background
▼
Information processing in the nervous system has long been
considered to be based on neuronal networks constituted of
rather passive transducer elements, the so-called platonic nerve
cells [22]. These platonic nerve cells generate stereotypical out-
put signals in the form of action potentials in response to like-
wise stereotypical inputs form other neurons via electrical or
chemical synapses.
During recent decades, this picture has changed completely. It
has been demonstrated that neurons operate with a diversity of
ion channels, ion pumps, ion exchangers etc. which are respond-
ing to alterations of the membrane voltage and ionic concentra-
tions, intracellular signal substances and a variety of
neurotransmitters, neuromodulators and hormones (● ▶ Fig. 1).
In particular, synaptic transmission at electrical and, especially,
chemical synapses has turned out to be a complicated process.
In the following text by using the term synaptic transmission we
generally refer to chemical synapses unless otherwise indi-
cated.
Specific transmitter effects can be mediated via different recep-
tors which are located not only at the postsynaptic membrane
but also at the presynaptic terminals. The process of transmitter
release itself includes diverse steps which altogether result in
the fusion of synaptic vesicles with the presynaptic membrane
in response to the appearance of the presynaptic spike. Addi-
tionally, the release of transmitter to the cleft can be adjusted by
the action of autoreceptors located on the presynaptic mem-
brane. Elimination of the transmitters from the synaptic cleft
can happen by diffusion, frequently accompanied by active
reuptake processes which are major targets for pharmacological
treatment of mental disorders.
Activation of postsynaptic receptors can lead to the opening of
ion channels directly (ionotropic receptors) or through the acti-
vation of a diversity of intracellular signal cascades (metabo-
tropic receptors). The postsynaptic alterations of ionic
conductances and the thereby induced changes of the mem-
brane potential will necessarily interfere with the activation
state of voltage-dependent channels. The complex interdepend-
encies between transmitter- and voltage-gated ion channels
essentially determine the neuron’s dynamic state.
Further modifications of the neurons’ dynamics can appear due
to the action of a manifold of intra- and extracellular signal sub-
stances. Calcium ions are among the most ubiquitous ones, not
only controlling transmitter release and second messengers
pathways but also having direct effects on ion channels, e. g. Ca-
dependent K-channels. Many signal substances which are mod-
ulating the activation state of ion channels are related to
homeostatic mechanisms as, for example, ATP, adenosine or glu-
cose which are signalling/providing metabolic information.
Much of the homeostatic control is mediated by hormones
which either activate membrane receptors (like peptide hor-
mones) or permeate the double lipid layers (steroid hormones
S84 Original Paper
(9)
substance application
channel blockers
receptor agonists/antagonist
pump/re-uptake inhibitors
Ca2+
transmitter (6)
reuptake release
autoreceptors
ion pumps & leak voltage-gated
exchanger channels channels (8)
(1) (2) (3) G-protein
neuromodulators
postsynaptic neuron hormones
transmitter-gated channels
metabotropic
ionotropic (4)
nucleus gane expression
Fig. 1 Schematic representation of major components of neuronal
excitability, synaptic transmission and neuromodulation. There are (1)
ion pumps and ion exchangers to keep the ion concentrations in the
physiological appropriate range, (2) leak channels which set the neuron’s
potential to a typically negative value, (3) voltage gated channels which
allow the generation of action potentials as well as subthreshold changes
of membrane potential, (4) transmitter gated channels for synaptic
modulation through direct ionotropic or G-protein coupled metabotropic
receptors and (5) gap junctions providing diffusive exchange of ions
like cortisol etc.) thereby directly interfering with gene expres-
sion.
Gene expression can also be modified through synaptic trans-
mission which again is preferably mediated by calcium ions, e. g.
with activation of MAP-kinases and CREB via the cAMP system.
These mechanisms can up-regulate the synthesis of ion chan-
nels, receptors and pumps and their embedding in the neuronal
membrane, including protein-biosynthesis for synaptic sprout-
ing. Simultaneously, there is a continuous down-regulation of
the membrane proteins which are internalized and metabo-
lised.
This description can only give a rough overview on some major
mechanisms of neuronal activity and synaptic transmission
and ● ▶ Fig. 1 illustrates only parts of them without going into
details. This figure also cannot illustrate the complex functional
interdependencies with many nonlinear and time-delayed feed-
back loops. These can only be assessed by means of mathemati-
cal simulations. Even an understanding of model dynamics can
still be difficult, especially when the model tries to consider the
manifold of physiological processes in all details. This can easily
lead to mathematical systems of many dimensions with an
unmanageable number of control parameters.
Therefore, here we propose a minimal model with considerable
simplifications which nevertheless can simulate the functionally
most relevant dynamics of neuronal excitability and synaptic
transmission. Most importantly, its extraordinarily adaptable
and flexible structure should allow easy extensions whenever
specific mechanism have to be examined in more details.
Postnova S et al. Neurones and Synapses for Systemic Models of Psychiatric … Pharmacopsychiatry 2010; 43 (Suppl. 1): S82–S91
presynaptic terminal
(7) gap junction
coupled
neuron
gap junctions
(5)
2nd
messengers
metabotropic
second mess.
receptor/channel
internalization/embedding
with neighboured neurons. Synaptic modulation additionally includes
(6) presynaptic transmitter release and reuptake as well as (7) feedback
effects via autoreceptors. Additional modulation of the postsynaptic
neuron’s excitability can be introduced (8) by intra- and extracellular signal
substances or hormones either with direct effects on voltage- and/or
transmitter-gated ion channels or via alteration of gene expression. A
variety of effects can be introduced with (9) external application of diverse
neuroactive substances, e. g. ion channel blockers or ion pump inhibitors,
receptor agonists or antagonists.
A Conductance-Based Modelling Approach
▼
Our modelling approach is based on the physiological point of
view of an excitable membrane (● ▶ Fig. 2a), constituted by a
lipid bilayer separating the intracellular and the extracellular
space of different ion concentrations. The dynamically relevant
components are provided by the embedded membrane proteins
functioning as ion channels, receptors, or transporters, e. g. ion
pumps.
From the electrical point of view (see equivalent circuit
in ●
▶ Fig. 2b), the membrane can be considered as a capacitor (C)
which is charged and discharged by diverse ion currents (I) (see
membrane equation in ● ▶ Fig. 2). The currents strengths depend
on the conductance and driving force. The driving force is deter-
mined by the difference between the actual membrane poten-
tial and the equilibrium potentials of the diverse ions. The
equilibrium potentials, represented by batteries, are usually
considered constant.
The major control parameters are provided by variable conduct-
ances represented by potentiometers. Their factual conductance
can significantly change depending on the membrane potential
(voltage-gated ion channels) and on different types of intracel-
lular and extracellular signal substances, specifically on synaptic
transmitters (transmitter-gated ion channels). The reference
potential is set by diverse voltage and transmitter insensitive
leak conductances which are generally represented by a single
resistor.
Conductance changes are resulting from the opening and closing
of ion channels. For simplicity, ● ▶ Fig. 2a shows single channels
while the model equations principally refer to the complete set

a Fig. 2 Schematic illustration of the neuronal
extracellular
V mechanism-based, Hodgkin-Huxley type modelling
intracellular
b membrane capacitance c.
extracellular
lC lI lK
gI g*K
V C
VI VNa
intracellular
dV
C = –∑ Ij where j=l, K, Na, syn,...
dt
of ion channels of a specific type. Accordingly, the voltage
changes are calculated on the basis of compound ion currents.
Moreover, according to our rather generic approach, all values
are calculated for a unit membrane area of 1.0 cm2 which is the
conventionally used basis in such simulations. Our actual
approach also does not consider specific shapes of the neurons
or different compartments, e. g. of the cell body, the axon and
diverse dendrites. This normalization and the simplifications do
not affect the principal model dynamics but make the model
easier to manage. If requested, different compartments of differ-
ent size can be introduced with formally the same equations.
Also the different membrane elements can be implemented in
more details, e. g. by calculating the equilibrium potentials on
the basis of the Nernst- and/or Goldmann equations. For practi-
cal use, the ion permeabilities of the Goldmann equation should
appropriately be replaced by ion conductances [4, 21], primarily
because these values can be determined by experimental record-
ings.
It is the major advantage of this modelling approach that it
allows to implement the different functions, if required, in more
detail and also that it easily can be extended by additional ion
channels and/or receptors (● ▶ Fig. 2a). In practice, it simply
means to add the corresponding electrical pathways to the
equivalent circuit (●▶ Fig. 2b) and the appropriate current term
to the equations. The following sections will describe how this
can be realized for voltage- and transmitter-gated ion channels
in a mathematically minimized but physiologically justified
form.
Modelling Neurons: Voltage-Gated Ion Channels
▼ current curves on the right show the alterations of Na + - and K + -
Single neuron models of the HH-type mostly are composed,
apart from leak channels, by a set of voltage-gated ion channels.
In their simplest forms, they include only Na + - and K + -channels
for action potential generation. Implementation of additional
current terms can introduce much more complex dynamics, for
example when subthreshold currents lead to slow membrane
potential oscillations. Here, we focus only on voltage-gated ion
channels. Implementation of the physiologically relevant voltage
and time dependencies are illustrated in ● ▶ Fig. 3 (a and b)
Postnova S et al. Neurones and Synapses for Systemic Models of Psychiatric … Pharmacopsychiatry 2010; 43 (Suppl. 1): S82–S91
Original Paper S85
membrane with ion channels (a) and the
electrical equivalent circuit (b) according to
approaches. Alterations of membrane voltage over
time (dV/dt) are determined by the sum of ion
currents Ij which are charging and discharging the
lNa lsyn
g*Na g*syn
Vsyn
together with examples of dynamical current and voltage
changes during an externally induced action potential (● ▶ Fig. 3,
c and d).
These mechanisms are controlled by the voltage dependent
opening (activation) and closing (inactivation) of ion channels.
Activation and inactivation values (av and bv) are normalized to
1 thus representing not the absolute value but the relative
amount of open or inactivated ion channels, respectively. The
voltage dependencies (av∞ and bv∞) are implemented by sig-
moid curves (● ▶ Fig. 3a), i. e. by the same curves to which exper-
imental data are generally fitted. Sigmoid curves are fully
determined by only two parameters, the half-activation values
V0 and the slopes s, which means that they can be easily adjusted
to different conditions. The equation for bv is not specifically
shown. For voltage dependencies (● ▶ Fig. 3a) it is simply 1 minus
the given expression, for time dependencies (● ▶ Fig. 3b), the
equation is exactly the same.
The dynamically most important effects are introduced by the
time delays of activation and inactivation of the different ion
channels (● ▶ Fig. 3b). For example, an action potential can only
develop because of the much faster activation of Na-channels
(afast) then K-channels (aslow), and because Na-channel activation
is preceding its inactivation (bfast) for a sufficiently long time.
The time constants τv as well as the steady state values av∞ and
bv∞ can be determined in voltage/patch clamp experiments as
shown in ● ▶ Fig. 3b (right hand side). Here, first order time delays
are implemented with a single time constants τv.
The absolute values of the ion conductances appear only in the
current equations (● ▶ Fig. 3c) as the product of the dimension-
less activation variables av and the maximum conductances gv
achieved when all channels are in an open state, i. e. av = 1. The
currents during an action potential (● ▶ Fig. 3d) generated by
exactly these currents according to the activation kinetics given
by the equations in ● ▶ Fig. 3a, b. The intrinsic dynamics are initi-
ated by an external current stimulus. The particular forms of the
current curves, especially for INa, give an indication of the com-
plex dynamics which already can develop in such a minimal
model due to the mutual interdependencies between current
and voltage changes.
S86 Original Paper

1 Fig. 3 Voltage-gated ion currents. Upper

a bfast ∞ afast ∞ diagrams: equations and graphs of the voltage-
voltage-dependence
and time-dependencies of current activation,
1 represented by sigmoid steady state activation
av∞ = curves (a) and first order differential equations
1 + exp (–sv (V – V0v)) aslow ∞ (b), respectively. Lower diagrams: dynamic
0 alterations of ion currents (c) and the membrane
–100 30 voltage (d) during an action potential, which is
V, mV introduced by an external current pulse (lowest
1 trace).
b time-dependence bfast afast aslow

dav av∞ – av
=
dt Tv 0
V, mV 20
–100
0 time, ms 12

c IK
voltage-gated currents
I, µ A/cm2

0
Iv = gv av (V – Ev))
INa

0 15
time, ms

d membrane potential
0
V, mV

dV
C = –∑ Ij = – Il – INa – IK – ...
dt
I, µ A/cm2

2
0
0 time, ms 15

This modelling approach is significantly simplified compared to
the original HH-model and several others which have been
derived from it. First of all we do not refer to the transition rates
α and
of ion channel states to achieve the sigmoid voltage
dependencies. The state transitions are generally only indirectly
estimated from conductances that can be determined much eas-
ier experimentally. In the 1950’s, when Hodgkin, Huxley, Katz
and others made their recordings and calculation, the primary
aim was the understanding of the shape of an action potential
(AP). For nowadays approaches, especially in neuronal network
simulations, the frequency of action potentials and eventually
their temporal patterns are the most relevant parameters.
Accordingly, we do not specifically consider those mathematical
expressions which mainly are for a precise adjustment of the
action potential form, e. g. the voltage dependencies of the time
constants or the power functions of the activation and inactiva-
tion variables. For most applications, even inactivation, e. g. of
the spike-generating Na + -currents, can be neglected because
these channels will anyhow close due to the K + -induced repo-
larization. Moreover, as these channels open much faster than
any others, it also will not make a significant difference to model
them as instantaneously activating, i. e. without time delay.
With these simplifications we achieve a conductance-based
model with only two dimensions [32] that still shows the func-
tionally relevant tuning properties, quite similar to the four-
dimensional and much more complicated models of the original
HH-type that are conventionally used. This leads to an easily
understandable model structure. Especially, it allows simple
extensions for the simulation of more complex dynamics of
physiological relevance. Examples have been given in previous
studies with the development of most flexible neuronal pattern
generators with only two additional terms of slow subthreshold
currents. Both can be simulated as directly voltage dependent
currents [18], but also indirectly voltage gated currents, e. g.
Ca + + -dependent K + -currents, have been implemented in differ-
ent details [5, 16].
Indirect voltage gating via intracellular or extracellular sub-
stances shall not further be described here. Instead, in the next
section, we will focus on synaptic transmission. The mecha-
nisms of current gating are, anyhow, quite similar, independent
of a specific substance or transmitter. However, for the model-
ling of synaptic transmission we have additionally to consider
the presynaptic control of transmitter release.
Modelling Synapses: Transmitter-Gated Ion Channels
▼
Compared to the gating of voltage-dependent channels, synaptic
control of transmitter-gated ion channels is, on the one hand, a
much more complicated process that include many more steps.
On the other hand, it does not need to consider the particular
complexity arising from the mutual interdependencies between
voltage-dependent currents and current-dependent voltages.
For example, activation of postsynaptic receptors does not
depend on the postsynaptic voltage but on presynaptic trans-
mitter release. Synaptic transmission is organized in a more lin-

Postnova S et al. Neurones and Synapses for Systemic Models of Psychiatric … Pharmacopsychiatry 2010; 43 (Suppl. 1): S82–S91
 Original Paper S87

Fig. 4 Functionally relevant steps in synaptic

a transmission (equations and corresponding
presynaptic spike
functions). A presynaptic spike (a) induces
dVpre transmitter release (b) which, together with
C = –∑ Iv
dt transmitter elimination, determines transmitter
concentration in the synaptic cleft (c). The
b transmitters in the synaptic cleft activate both
transmitter release
ionotropic and metabotropic postsynaptic
cspike
crelease = receptors (d) which leads to the activation of
1 + exp (–ssyn (Vpre – Vspike)
transmitter gated currents with different time
delays for ionotropic (a) and metabotropic
c (b) receptors (e). Postsynaptic currents are
transmitter concentration
dccleft c c determined in the conventional way as the product
= release – cleft of the ionic conductance and the driving force (f)
dt taccum telim
and are added to the membrane equation (g).
d ionotropic (a) metabotropic (b)
receptors activation receptors activation p
rit ccleft r c ccleft
pit = pmt = mt cleft
cit + ccleft cmt + ccleft

e
current activation current activation
it mt
ait = Pit damt p a
= t mt – t mt
dt act inact

f postsynaptic current postsynaptic current

it mt
Ipostit = gitait (Vpost – Eit) Ipostmt = gmtamt (Vpost – Emt)

g postsynaptic potential postsynaptic potential

it mt
dVpostit dVpostmt
C = –∑ Iv – Ipostit C = –∑ Iv – Ipostmt
dt dt

ear way from presynaptic action potentials to the generation of
postsynaptic potentials. The manifold of pre- and postsynaptic
processes which are involved in synaptic transmission can be
comprised in simplified but physiologically justified equations
as summarized in ● ▶ Fig. 4, with illustrations of corresponding
dynamics.
Synaptic transmission starts with the appearance of an action
The transmitter concentration in the synaptic cleft determines
how many of the postsynaptic receptors will be occupied and
activated. The absolute number additionally depends on the
availability of postsynaptic receptors. The same holds true for
presynaptic autoreceptors which, however, are not specifically
considered on this figure. Instead, we distinguish between iono-
tropic and metabotropic receptors (● ▶ Fig. 4d). Their activation,

potential at the presynaptic terminal (● ▶ Fig. 4a). Transmitter pi and pm, respectively, can be modelled in form of Michaelis-
release (● ▶ Fig. 4b) is triggered by voltage-dependent Ca-cur- Menten equation (see inset). In the equations of ● ▶ Fig. 4d (a),(b),

rents of high activation threshold. This is realized by an almost
step-like sigmoid activation curve (dashed curve in ● ▶ Fig. 4a,
eq. in ● ▶ Fig. 4b) which makes sure that transmitter release
essentially happens during the peak of the presynaptic spike
(●▶ Fig. 4b, c
release).
Presynaptic time delays, e. g. of Ca + + -accumulation and vesicle
fusion are comprised in the time delays of transmitter accumu-
lation in the synaptic cleft (● ▶ Fig. 4c, τ
accum). The functionally
more important time delays are those of transmitter elimination
τelim. Different processes, diffusion, degradation and/or reuptake,
may be involved.
The transmitter concentration in the synaptic cleft (ccleft, ● ▶ Fig. 4c)
is a central control variable of synaptic transmission, being also
cit and cmt are the transmitter concentrations at which half of the
receptors is activated. Accordingly, this parameter is reflecting
the transmitters’ affinity. The values of rit and rim represent the
availability of receptors determining the maximum activation
which can be reached when all receptors are occupied.
Receptor’ availability is another functionally relevant control
parameter which can significantly change, e. g. due to receptors’
internalization or re-embedding. Also alterations of synaptic
sprouting or the number of synaptic spines can be considered by
appropriate adjustment of the parameters rit and rmt. Especially,
application of receptor antagonists will decrease the receptors’
availability.
The next equations (● ▶ Fig. 4e) relate the activation of postsyn-

a major target of psychiatric drugs. Functionally relevant changes
can be introduced with alterations of transmitter release (cre-
lease), e. g. via autoreceptors, or the time constant of transmitter
elimination (τelim), e. g. via reuptake inhibitors.
aptic receptors to the activation of postsynaptic currents. These
are significantly different for ionotropic and metabotropic recep-
tors. In case of ionotropic receptors, where the binding site is
part of the ion channel, activation of the receptors pit leads
directly to the activation of the ion channels ait (direct-

Postnova S et al. Neurones and Synapses for Systemic Models of Psychiatric … Pharmacopsychiatry 2010; 43 (Suppl. 1): S82–S91
S88 Original Paper

receptor-antagonist Fig. 5 Simulations of synaptic transmission

(a) control (b) receptor-antagonist (c) + under control conditions (a) and with reduced
a reuptake inhibitor
20 availability of postsynaptic receptors, e. g.
Vpre, mV

by application of a receptor agonist (b),

compensated by a longer residence time of the
–90 neurotransmitters with application of reuptake
b 0.25
inhibitors (c). For simplicity, the model parameters
ccleft

have been adjusted in a way that, under control

0 conditions, a single presynaptic action potential
c 0.025
can generate a postsynaptic spike. The diagrams
show, from top to bottom: a presynaptic action
pi

0 potential (a, Vpre), transmitter concentration in the

d
Ipost, µA/cm2

0
synaptic cleft (b, ccleft), postsynaptic activation of
ionotropic receptors (c, pit), postsynaptic current
–2 (d, Ipost,it), and postsynaptic potential (e, Vpost).
e 20
Vpost, mV

–90
0 0.4 0 0.4 0 0.4
time, s time, s time, s

gating, ●▶ Fig. 4e (a)). In case of G-protein coupled metabotropic

a 0.025
receptors the ion channels are remote from the receptor sites
and their activation goes through second messenger cascades
with many intermediate steps (indirect-gating). The equation
pi

in ●▶ Fig. 4e (b) does not consider specific steps of these second

messenger cascades. The time constant τact accounts for the all-
0
over delays from receptor binding to ion channel activation. The b 20
additional time constant τinact represents inactivation or elimi-
nation of intermediate signal substances. The time constants are
Vpost, mV

the functionally relevant parameters, e. g. for the implementa-

tion of eventual disturbances in the second messengers cas-
cades.
The postsynaptic currents, for both types of receptors, are calcu-
lated according to the general current equation which is the –90
same as for voltage-dependent currents (see ● ▶ Fig. 3c), namely
0.1 0.3
time, s
as the product of the actual conductance and the driving force
(●▶ Fig. 4f). The actual conductance is given by the maximum
Fig. 6 Curves of postsynaptic receptor activation (a, pit), and
conductance (git, gmt) multiplied by the activation variable (ait, postsynaptic potentials (b, Vpost) from the three simulation runs as shown
amt) as determined in the preceding equations in ● ▶ Fig. 4. Finally, in ●
▶ Fig. 5, here plotted upon each other for easier comparison of

the synaptic currents can be added to the membrane equation amplitudes and time delays.
(●▶ Fig. 4g) where they introduce postsynaptic potentials (PSPs)

with additional time-delays due to the membrane capacitance.

The examples are given for transmitter dependent opening
(activation) of ion channels which are inducing depolarizing
inward currents (negative deflections in ● ▶ Fig. 4f) and,
accordingly, lead to excitatory postsynaptic potentials
(EPSPs, ●▶ Fig. 4g). Inhibitory postsynaptic potentials (IPSPs) are
not specifically illustrated but can be achieved with exactly the
same equations. Whether an EPSP or IPSP will occur only
depends on the driving force, i. e. on the equilibrium potential of
the transmitter-gated ion channels. Also transmitter dependent
closing of ion channels can easily be implemented, simply
replacing g*b by g*(1-b) in the current equation, similar to the
inactivation of voltage-gated channels.
Altogether, despite many simplifications, the here presented
equations can account for the physiologically most relevant
mechanisms of synaptic transmission and allow to consider
clinically important disturbances and major effects of pharma-
cological treatment. Examples are given in the next chapter.
Examples of Synaptic Modulation
▼
Synaptic transmission is far away from being an invariant proc-
ess. It can significantly change in response to physiological
mechanisms as well as on drug application. In the following we
give a brief overview with a representative example illustrating
how pathophysiological changes and compensatory drug effects
can be mathematically implemented.
Major physiological, pathophysiological and pharmacological
variants of synaptic transmission are related to 1) the amount of
transmitter release, 2) the speed of transmitter elimination from
the synaptic cleft, 3) the availability of postsynaptic receptors, 4)
the current through the postsynaptic ion channels. All these
physiological mechanisms are represented by appropriate model
variables and parameters as described above. For example, alter-
ations of transmitter release can be simulated by corresponding
changes of the parameter cspike which scales the variable crelease,
and the postsynaptic current amplitude Ipost can physiologically

Postnova S et al. Neurones and Synapses for Systemic Models of Psychiatric … Pharmacopsychiatry 2010; 43 (Suppl. 1): S82–S91

correctly be adjusted by the maximum conductance git,mt in the
current equations.
The simulation examples in ● ▶ Fig. 5 illustrate the impairment of
synaptic transmission by reduced availability of postsynaptic
receptors as caused, for example, by application of receptor
antagonists or by some pathophysiological processes, like
impaired gene expression. Furthermore, it demonstrates how
such situations can pharmacologically be improved by decelera-
tion of transmitter elimination, e. g. with application of trans-
mitter reuptake inhibitors. For simplicity, we are only considering
ionotropic receptors and have adjusted the model parameters so
that, under control conditions (● ▶ Fig. 5 (a)), a single presynaptic
spike (● ▶ Fig. 5a (a)) can trigger a postsynaptic action potential
(●▶ Fig. 5e (a)). The mid diagrams in ● ▶ Fig. 5 show the transmit-
ter concentration in the synaptic cleft (● ▶ Fig. 5b), the activation
of postsynaptic receptors (● Fig. 5c) which, for ionotropic trans-
▶ the presynaptic neuron does not only generate single action
mission, changes in parallel with the transmitter concentration
(not shown), and the postsynaptic current (● ▶ Fig. 5d).
Application of a receptor antagonist (● ▶ Fig. 5 (b)) is imple-
mented by a reduction of rit (here to 80 % of the control value)
according to a reduced number of postsynaptic receptors to
which the transmitters can bind. Although the transmitter con-
centration (ccleft, ●
▶ Fig. 5b (b)) is the same as under control con-
ditions, the probability of receptor activation (pi, ● ▶ Fig. 5c (b))
decreases. Thereby, also the postsynaptic current (Ipost, ● ▶ Fig. 5d
(b)) is reduced and, in this example, is too small for the initiation
of a postsynaptic spike but only induces a gradually changing
EPSP (Vpost, ●▶ Fig. 5e (b)).
When additionally a transmitter reuptake inhibitor is applied
(●▶ Fig. 5 (c)), induction of a postsynaptic spike can be re-
installed. Although the model, in this minimal form, does not
explicitly consider reuptake processes, their inhibition can phys-
iologically consistently be simulated by the lengthening of the
time constant τelim (equation of ● ▶ Fig. 4c). In ●▶ Fig. 5 (c), the
time constant of transmitter elimination is three times longer
than under control conditions. This extends the residence time
of the transmitter in the synaptic cleft (ccleft, ● ▶ Fig. 5b (c)).
Accordingly, the postsynaptic receptors remain activated for a
longer time (● ▶ Fig. 5c (c)) which induces a correspondingly
longer current flow (● ▶ Fig. 5d (c)) for charging of the postsynap-
tic membrane capacitor. This can be sufficient for the induction
of a postsynaptic action potential (● ▶ Fig. 5e (c)) although the
current amplitude is not increased. With regard to postsynaptic
spike-generation, the reduction of the current amplitude by
receptor antagonists have been compensated by a longer current
flow due to reuptake inhibition.
The transient deflections in the current curves of ● ▶ Fig. 5d (a)
and ● ▶ Fig. 5d (c) do not appear in the subthreshold current
curve of ● ▶ Fig. 5d (b). They are caused by the strong changes of
the driving force during the postsynaptic action potential, simi-
lar to the deflections of the voltage gated Na + -current during an
action potential (for comparison see ● ▶ Fig. 3c).
More details can be recognized in ● ▶ Fig. 6, where the receptor
activation curves (pit) and postsynaptic voltage traces (Vpost) of
the above simulations have been plotted upon each other. These
curves show the reduction of the maximum of receptor activa-
tion after application of the receptor blockers in direct compari-
son with the control curve (● ▶ Fig. 6a, red vs. black curve). They
also demonstrate that the reuptake inhibitor does have an effect
on the maximum amplitude. The reduced amplitude is compen-
sated by slower inactivation (● ▶ Fig. 6a, blue curve). This is suf-
ficient for the generation of an action potential, although it
Postnova S et al. Neurones and Synapses for Systemic Models of Psychiatric … Pharmacopsychiatry 2010; 43 (Suppl. 1): S82–S91
Original Paper S89
occurs with clear time delay compared to control conditions
(●▶ Fig. 6b, blue vs. black curve). Simultaneously, the spike
amplitude is reduced – as a consequence of the longer delay.
There is more time for Na + -channel inactivation as well as for
the opening of K + -channels, both counteracting the depolariza-
tion by Na + -channel activation.
In any case, the generation of spikes, although of reduced ampli-
tude, will be facilitated by reuptake inhibitors. With higher con-
centrations, i. e. further lengthening of the elimination time
constant τelim, a single presynaptic spike can be sufficient to gen-
erate a series of postsynaptic action potentials. Also in this
example, it can be recognized that the EPSP continues after
action potential generation (● ▶ Fig. 6b, blue curve) when other-
wise the membrane potential has approached its steady state
(●▶ Fig. 6, black and red curves). In the more realistic case when
potentials but trains of spikes, the lengthening of postsynaptic
potentials can significantly facilitate their superposition.
The here presented example is referring to a rather simple situa-
tion of synaptic transmission illustrating basic effects of synap-
tic impairment and possible drug treatment. Nevertheless, it
elucidates physiologically relevant dynamics, essentially arising
from the intimate interdependencies between synaptic effects
and the neuron’s intrinsic properties. Noteworthy, these effects
can only be seen in a mechanism-based model.
Discussion
▼
We have introduced a conductance-based modelling approach
for the simulation of synaptic modulation of neuronal excitabil-
ity. Emphasis is laid on direct relations of the model parameters
and variables to those of the physiological processes. The idea is
to bring together neuronal dynamics and synaptic transmission
on a common, physiologically motivated model platform.
Because of the manifold and diversity of neural and synaptic
mechanisms, such an integrative modelling approach requires
simplifications (dimension reduction) for reasonable applicabil-
ity. At the same time, the model has to allow the implementa-
tion of specific mechanisms in more details if required for the
examination of specific functions. Altogether, a most flexible
and adaptable model structure is needed to allow simplifica-
tions – where appropriate – as well as extensions – where
requested. Especially, a model of clinical relevance has to use
variables and parameters that have direct physiological corre-
lates. This is the goal towards which the here presented model-
ling concept has been developed.
The neuronal part of the model is designed on the basis of previ-
ous simulations that use a HH-type approach which, on the one
hand, is considerably simplified but, on the other hand, can eas-
ily be extended by additional terms [5, 16, 17]. We have used this
conductance-based simulation strategy also for the implemen-
tation of synaptic transmission. The connections are made via
the membrane equation that includes voltage- as well as trans-
mitter-gated currents with mutual interdependencies due to
their voltage dependencies. In this way, the neuronal and synap-
tic model parts become an entity.
The synapse part is made up on recent modelling studies of
activity dependent synaptic plasticity in context with homeo-
static sleep-wake regulation [32]. This introduced further feed-
back loops because the neurons’ activity, in turn, depends on the
synaptic efficacy. Moreover, this study [32] and a following one
S90 Original Paper
[33] could demonstrate that the results of such conductance-
based simulations can even be connected to alterations of behav-
ioural states. The correlation of neuronal activity patterns and
behavioural states furthermore allows to connect quite different
time scales of physiological functions, from the millisecond scale
of neuronal impulse generation to the scales of hours and days of
synaptic plasticity and circadian cycles. These properties can
become of particular importance for the simulation of psychiat-
ric disorders, related to disturbances of neuronal dynamics but
with chronic progression and partly long term treatment
effects.
In contrast to the above mentioned applications, the actual study
specifically emphasizes the modelling principles and aims to
provide detailed descriptions of the physiological background of
the model equations, variables and parameters. The idea is to
propose a model concept which can account for a manifold of
physiological processes of synaptic transmission, their distur-
bances and targets for pharmacological treatment.
Already transmitter release, for example, can be modified by
diverse mechanisms, e. g. by presynaptic inhibition or via autore-
ceptors, by transmitter depletion or increased availability, by
long term potentiation or depression (LTP or LTD). All these
effects can be considered with appropriate adjustment of the
parameter cspike, which scales crelease. Most importantly, depend-
ing on the underlying mechanisms, the alterations of these
parameters can be modelled dynamically as a function of other
model parameters and variables. This can introduce additional,
functionally relevant feedback loops. For example, the activation
of autoreceptors and the thereby induced inhibition of transmit-
ter release depends on transmitter concentration in the synaptic
cleft which, in turn, is essentially determined by transmitter
release. Accordingly, for physiologically correct simulations of
such autoreceptor actions, the corresponding alterations of the
parameter cspike have to be modelled as a function of the param-
eter ccleft that, itself, depends on cspike.
We have given a single example to illustrate potential applica-
tions for the modelling of synaptic disturbances and pharmaco-
logical treatment (● ▶ Figs. 5 and 6). It simulates the effects of
reuptake inhibitors after impairment of the synaptic efficacy
due to reduced availability of postsynaptic receptors. This points
to a clinically important aspect, namely that drug treatment
often is not a causal treatment. In the given example, the original
functioning is not reinstalled but the deficiency is compensated
by manipulation of another mechanisms.
An alternative treatment of psychiatric disorders, especially
major depression, would be the application of monoamine oxi-
dase inhibitors (MAOIs). These substances increase the trans-
mitter availability and thereby also can improve the synaptic
efficacy. However, neither SSRIs not MAOIs are directly increas-
ing an eventually reduced number of postsynaptic receptors or
dendritic spines, although such changes may be introduced as
long term effects [25].
Hence, while the regular functioning may appear to be rein-
stalled by a certain drug, the system dynamics is not necessarily
fully identical to that in the healthy state. This can be completely
uncritical in most situations, but there may exist conditions
under which the treated system reacts different from the origi-
nal, healthy system, thereby eventually going back into a patho-
logical state. The estimation of such bifurcation points can
become a major goal of mechanism-based computer simula-
tions, e. g. for the elucidation of biological markers which can
Postnova S et al. Neurones and Synapses for Systemic Models of Psychiatric … Pharmacopsychiatry 2010; 43 (Suppl. 1): S82–S91
indicate when a patient approaches such critical states and in
search for appropriate strategies for its prevention.
Acknowledgement
▼
This work was supported by the European Union through the
Network of Excellence BioSim contract No LSHB-CT-2004-
005137. E. Rosa acknowledges financial support from Research
Corporation.
Disclosure
▼
The authors declare that there are no financial interests.
References
1 Abbott LF, Regehr WG. Synaptic computation. Nature 2004; 431 (2010):
796–803
2 Bender W, Albus M, Moller HJ et al. Towards systemic theories in bio-
logical psychiatry. Pharmacopsychiatry 2006; 39 (Suppl. 1): S4–S9
3 Braun HA, Huber M, Dewald M et al. Computer Simulations of neuro-
nal signal transduction: the role of nonlinear dynamics and noise. Int
J Bifurcation & Chaos 1998; 8: 881–889
4 Braun HA, Schneider H, Wollweber B et al. cLabs-Neuron (http://www.
clabs.de/): Membrane Properties – Conductances. Marburg: BM&T;
2002
5 Braun HA, Voigt K, Huber MT. Oscillations, resonances and noise: basis
of flexible neuronal pattern generation. Biosystems 2003; 71 (1–2):
39–50
6 Braun HA, Postnova S, Wollweber BT et al. Biological Rhythms in Men-
tal Disorders. In: Bertau M, Mosekilde E, Westerhoff H, editors.
Biosimulation in Drug Development: Wiley –VCH; 2007; 197–231
7 Carlsson A. The neurochemical circuitry of schizophrenia. Pharma-
copsychiatry 2006; 39 (Suppl 1): S10–S14
8 Chay TR, Fan YS, Lee YS. Bursting,spiking,chaos,fractals and univer-
sality in biological rhythms. Int J Bifurcation and Chaos 1995; 5:
595–635
9 Christodoulou C, Bugmann G, Clarkson TG. A spiking neuron model:
applications and learning. Neural Netw 2002; 15 (7): 891–908
10 Destexhe A, Bal T, McCormick DA et al. Ionic mechanisms underlying
synchronized oscillations and propagating waves in a model of ferret
thalamic slices. J Neurophysiol 1996; 76 (3): 2049–2070
11 Finke C, Vollmer J, Postnova S et al. Propagation effects of current and
conductance noise in a model neuron with subthreshold oscillations.
Math Biosci 2008; 214 (1–2): 109–121
12 Fitzhugh R. Impulses and Physiological States in Theoretical Models of
Nerve Membrane. Biophys J 1961; 1 (6): 445–466
13 Hodgkin AL, Huxley AF. A quantitative description of membrane cur-
rent and its application to conduction and excitation in nerve. J Phys-
iol 1952; 117 (4): 500–544
14 Hopfield JJ. Neural networks and physical systems with emergent col-
lective computational abilities. Proc Natl Acad Sci USA 1982; 79 (8):
2554–2558
15 Huber MT, Braun HA, Voigt K et al. Computational properties of intrin-
sic subthreshold oscillations. In: Bower J, editor. Computational Neu-
roscience: Trends in Research. New York: Plenum Press; 1998;
197–202
16 Huber MT, Krieg JC, Dewald M et al. Stochastic encoding in sensory
neurons: impulse patterns of mammalian cold receptors. Chaos,
Solitons and Fractals 2000; 11 (12): 1895–1903
17 Huber MT, Braun HA. Stimulus – response curves of a neuronal model
for noisy subthreshold oscillations and related spike generation. Phys-
ical Reviews E 2006; 73, 041929: 1–10
18 Huber MT, Braun HA. Conductance versus current noise in a neuronal
model for noisy subthreshold oscillations and spike generation. Bio-
systems 2006, doi: 10.1016/j.biosystems.2006.1005.1009
19 Izhikevich EM. Simple model of spiking neurons. IEEE Trans Neural
Netw 2003; 14 (6): 1569–1572
20 Izhikevich EM, Edelman GM. Large-scale model of mammalian tha-
lamocortical systems. Proc Natl Acad Sci USA 2008; 105 (9): 3593–
3598

21 Kandel E, Schwartz J, Jessel T. Principles of Neural Sciences. New York:
Mac Craw Hill; 1991
22 Llinas RR. The intrinsic electrophysiological properties of mammalian
neurons: insights into central nervous system function. Science 1988;
242 (4886): 1654–1664
23 Lumer ED, Edelman GM, Tononi G. Neural dynamics in a model of the
thalamocortical system. I. Layers, loops and the emergence of fast
synchronous rhythms. Cereb Cortex 1997; 7 (3): 207–227
24 Modolo J, Henry J, Beuter A. Dynamics of the Subthalamo-pallidal Com-
plex in Parkinson’s Disease During Deep Brain Stimulation. J Biol Phys
2008; 34 (3–4): 251–266
25 Nestler EJ, Barrot M, DiLeone RJ et al. Neurobiology of depression. Neu-
ron 2002; 34 (1): 13–25
26 Noble D. From the Hodgkin-Huxley axon to the virtual heart. J Physiol
2007; 580 (Pt 1): 15–22
27 Noori HR, Jager W. Neurochemical oscillations in the basal ganglia.
Bull Math Biol 2010; 72 (1): 133–147
28 Phillips AJ, Robinson PA. A quantitative model of sleep-wake dynamics
based on the physiology of the brainstem ascending arousal system.
J Biol Rhythms 2007; 22 (2): 167–179
29 Postnova S, Voigt K, Braun HA. Neural Synchronization at Tonic-to-
Bursting Transitions. J Biol Phys 2007; 33 (2): 129–143
30 Postnova S, Wollweber B, Voigt K et al. Impulse pattern in bi-direction-
ally coupled model neurons of different dynamics. Biosystems 2007;
89 (1–3): 135–142
Postnova S et al. Neurones and Synapses for Systemic Models of Psychiatric … Pharmacopsychiatry 2010; 43 (Suppl. 1): S82–S91
View publication stats
Original Paper S91
31 Postnova S, Finke C, Jin W et al. A computational study of the interde-
pendencies between neuronal impulse pattern, noise effects and syn-
chronization. J Physiol Paris 2009: doi: 10.1016/j.jphysparis.2009.11.
022
32 Postnova S, Voigt K, Braun HA. A mathematical model of homeostatic
regulation of sleep-wake cycles by hypocretin/orexin. J Biol Rhythms
2009; 24 (6): 523–535
33 Postnova S, Voigt K, Braun HA. Modelling the Hypothalamic Control of
Thalamic Synchronization along the Sleep-Wake Cycles. Advances in
Cognitive Neurodynamics 2010; 2, in press
34 Qi Z, Miller GW, Voit EO. A mathematical model of presynaptic
dopamine homeostasis: implications for schizophrenia. Pharmacops-
ychiatry 2008; 41 (Suppl 1): S89–S98
35 Rolls ET, Loh M, Deco G et al. Computational models of schizophrenia
and dopamine modulation in the prefrontal cortex. Nat Rev Neurosci
2008; 9 (9): 696–709
36 Rulkov NF, Timofeev I, Bazhenov M. Oscillations in large-scale corti-
cal networks: map-based model. J Comput Neurosci 2004; 17 (2):
203–223
37 Soto-Trevino C, Thoroughman KA, Marder E et al. Activity-dependent
modification of inhibitory synapses in models of rhythmic neural net-
works. Nat Neurosci 2001; 4 (3): 297–303
38 Tretter F, Gebicke-Haerter PJ, Albus M et al, Schwegler H. Systems
biology and addiction. Pharmacopsychiatry 2009; 42 (Suppl 1):
S11–S31