Calcium Dynamics Underlying Pacemaker-Like and Burst
Firing Oscillations in Midbrain Dopaminergic Neurons:
A Computational Study
B. AMINI,1 J. W. CLARK, JR.,1 AND C. C. CANAVIER2
1
Department of Electrical and Computer Engineering, Rice University, Houston, Texas 77005; and 2Department of
Psychology, University of New Orleans, New Orleans, Louisiana 70148
Amini, B., J. W. Clark, Jr., and C. C. Canavier. Calcium dynamics
underlying pacemaker-like and burst firing oscillations in midbrain
dopaminergic neurons: a computational study. J. Neurophysiol. 82:
2249 –2261, 1999. A mathematical model of midbrain dopamine
neurons has been developed to understand the mechanisms underlying
two types of calcium-dependent firing patterns that these cells exhibit
in vitro. The first is the regular, pacemaker-like firing exhibited in a
slice preparation, and the second is a burst firing pattern sometimes
exhibited in the presence of apamin. Because both types of oscilla-
tions are blocked by nifedipine, we have focused on the slow calcium
dynamics underlying these firing modes. The underlying oscillations
in membrane potential are best observed when action potentials are
blocked by the application of TTX. This converts the regular single-
spike firing mode to a slow oscillatory potential (SOP) and apamin-
induced bursting to a slow square-wave oscillation. We hypothesize
that the SOP results from the interplay between the L-type calcium
current (ICa,L) and the apamin-sensitive calcium-activated potassium
current (IK,Ca,SK). We further hypothesize that the square-wave oscil-
lation results from the alternating voltage activation and calcium
inactivation of ICa,L. Our model consists of two components: a
Hodgkin-Huxley-type membrane model and a fluid compartment
model. A material balance on Ca21 is provided in the cytosolic fluid
compartment, whereas calcium concentration is considered constant
in the extracellular compartment. Model parameters were determined
using both voltage-clamp and calcium-imaging data from the litera-
ture. In addition to modeling the SOP and square-wave oscillations in
dopaminergic neurons, the model provides reasonable mimicry of the
experimentally observed response of SOPs to TEA application and
elongation of the plateau duration of the square-wave oscillations in
response to calcium chelation.
INTRODUCTION
Midbrain dopaminergic neurons (DA) have been implicated
in several important functions in humans, including movement,
attention, learning, and reinforcement as well as in the etiology
of Parkinson’s disease and various thought disorders, such as
schizophrenia, attention deficit disorders, and depression (Carl-
son 1992). Therefore it is essential to understand how the firing
pattern in these neurons is generated and modulated. Whereas
in vivo DA neurons can exhibit either single-spike or burst-
firing activity as well as spontaneous shifts between modes
(Freeman et al. 1985; Grace and Bunney 1984), in a slice
preparation a regular, pacemaker-like firing pattern usually is
The costs of publication of this article were defrayed in part by the payment
of page charges. The article must therefore be hereby marked “advertisement”
in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
0022-3077/99 $5.00 Copyright © 1999 The American Physiological Society
observed, presumably due to the loss of synaptic afferents. In
Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017
this study, we will focus on the intrinsic oscillatory activity of
DA neurons in a slice preparation.
In addition to the endogenous regular single-spike firing
exhibited in a slice preparation, burst firing has been induced
by the application of apamin (Ping and Shepard 1996) or
N-methyl-D-aspartic acid (NMDA) (Johnson et al. 1992). The
firing pattern in these neurons is particularly significant be-
cause burst-firing in vivo is associated with behaviorally rele-
vant appetitive and sometimes novel stimuli and possibly with
movement (for a review, see Overton and Clark 1997). This
study will address the biophysical basis for the calcium-depen-
dent oscillations underlying regular single-spike firing and
apamin-induced burst firing.
Because midbrain DA neurons have an unusually depolar-
ized firing threshold (233 6 15 mV) (Grace and Onn 1989),
the membrane must be depolarized repeatedly from rest (260
mV) to this threshold to fire repetitively. Indeed, such an
underlying slow oscillatory potential (SOP) has been observed
when the sodium-mediated spikes are blocked by tetrodotoxin
(TTX) and is enhanced by the application of the potassium
channel blocker tetraethyl ammonium (TEA). The depolarizing
phase of the SOP has been variously termed the pacemaker-
like slow depolarization (PLSD) or simply the pacemaker-like
depolarization (PLD), which has been shown to be calcium-
dependent.
The SOP sometimes can be converted into a square-wave
oscillation by the application of apamin (Nedergaard et al.
1993; Ping and Shepard 1996), revealing a crucial role for the
apamin-sensitive current IK,Ca,SK in the repolarization of the
SOP. Just as the SOP is the underlying oscillation that sets the
rhythm for pacemaker-like single-spike firing, the square-wave
oscillation sets the timing for burst firing; in the absence of
TTX, a burst of spikes is generated on the plateau of the
square-wave and quiescence during the trough (Gu et al. 1992;
Ping and Shepard 1996; Shepard 1993; Shepard and Bunney
1988). In some cases, apamin induced irregular firing instead
of burst firing (Gu et al. 1992; Ping and Shepard 1996), and in
those cases, the subsequent application of TTX and TEA
revealed irregular high-threshold calcium spiking (Ping and
Shepard 1996). The generation of irregular firing patterns,
however, is not addressed in this study. Neither the SOP nor
the square-wave should be confused with the slow oscillation
in membrane potential revealed by TTX in the presence of
NMDA, which also can underlie burst firing, appears to be
2249
2250 B. AMINI, J. W. CLARK, AND C. C. CANAVIER

Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017

FIG. 1. Diagram of the dopaminergic (DA) neuron cell model. A: electrical equivalent circuit for the membrane; ICa is the
combined calcium current consisting of ICa,T, ICa,N, ICa,L, and ICa,HVA. IK,tot is the combined voltage-dependent potassium current
consisting of IK and IA.B: fluid compartmental model, including intracellular and extracellular spaces.

sodium rather than calcium dependent (Johnson et al. 1992)
and is also not addressed in this study.
The goal of this study was to test hypothesized mechanisms
for the SOP and the square wave. On the basis of data that
suggest that dihydropyridines abolish the SOP (see DISCUSSION),
we hypothesize that the SOP is generated by interplay between
the L-type calcium current ICa,L and the calcium-activated
IK,Ca,SK. We further hypothesize that in the presence of apamin,
the calcium-mediated inactivation of ICa,L is responsible for
repolarization of the square wave. We also hypothesize that the
mechanisms driving these phenomena are located in or near the
soma. Hence we constructed a quantitative somatic model
based on voltage-clamp, morphological, and calcium-imaging
data to confirm that the currents described under voltage clamp
are capable of generating the electrical activity observed under
current clamp. Minimum expectations for model performance
included production of SOPs similar to those of Ping and
Shepard (1996) in the presence of TTX and TEA, SOPs under
conditions where either the calcium current ICa,T or ICa,N were
blocked but not when ICa,L is blocked, and square waves
similar to those observed (Ping and Shepard 1996) in the
presence of apamin.
MODEL DEVELOPMENT
Our model of the DA neuron consists of a single compart-
ment Hodgkin-Huxley (HH)-type parallel conductance mem-
brane model (Fig. 1A) and a lumped fluid compartmental
model (Fig. 1B). The HH equivalent circuit is composed of a
somatic membrane capacitance of 15.8 pF (Kang and Kitai
1993b), shunted by resistive ion-selective channels, as well as
ion pumps, a sodium-calcium exchanger, and other ionic cur-
rents. The lumped fluid compartment model (Fig. 1B) consists
of an intracellular compartment containing constant concentra-
tions of Na1 and K1, and a calcium buffer (presumably cal-
modulin). A material balance for Ca21 describes the time rate
of change in cytosolic Ca21 concentration. The extracellular
space is assumed to have a relatively large volume, so that the
ionic concentrations of Ca21, Na1, and K1 there are assumed
to be constant.
MEMBRANE CURRENTS
Under space-clamp conditions, the differential equation de-
scribing the time-dependent changes in the membrane potential
(V) is
V̇ 5 2~I Ca,T 1 I Ca,L 1 I Ca,N 1 I Ca,HVA 1 I K 1 I A 1 I h 1 I Na,K
1 I Ca,pump 1 I Na,Ca !/C m (1)
where Cm is the whole cell membrane capacitance. Model
currents include a T-type calcium current (ICa,T), an L-type
calcium current (ICa,L), an N-type calcium current (ICa,N), a
residual high-voltage activated calcium current (ICa,HVA), a
delayed rectifier current (IK), a transient outward current (IA),
a small conductance calcium-dependent potassium current
(IK,Ca,SK), a hyperpolarization activated current (Ih), a sodium-
potassium pump current (INa,K), a calcium pump current
(ICa,pump), and a sodium-calcium exchanger (INa,Ca).
In the current descriptions that follow, the HH-type activa-
tion and inactivation gating variables are solutions of the
familiar first-order differential equations described as
 CALCIUM DYNAMICS IN PACEMAKER AND BURST FIRING 2251

Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017

FIG. 2. Model-generated approximations to voltage-clamp data for several membrane currents. Steady-state activation and
inactivation characteristics of the currents were characterized by fits (—) to experimental data (- - -). A: ICa,T, model was held at
288 mV and clamped at various voltages from 268 mV through 254 mV in increments of 2 mV following Kang and Kitai
(1993b). B: IA, model was held at 290 mV and clamped to various voltages from 260 to 75 mV in 15-mV increments after
experiments by Silva et al. (1990). C: Ih, model was held at 257 mV and clamped to various potentials from 2141 to 269 mV
in 12-mV increments after Mercuri et al. (1995). D: IK, model was held at held at 240 mV and clamped to 20 mV in 10-mV
increments after experiments by Silva et al. (1990). E: ICa,N, model was held at 284 mV and clamped to 258, 248, and 238 mV
following Kang and Kitai (1993b). Steady-state activation and inactivation curves for IA (p# and q# , respectively) and steady-state
activation curves for Ih (q# ) and IK (n# ) are displayed in F. Current parameters and equations are in Tables 1–3.

z# ~V! 2 z~V, t!
ż~V, t! 5
t z ~V!
where z#(V) is the steady-state value of the general gating
variable z at membrane voltage V. We have characterized the
steady-state gating variable z#(V) by a sigmoidal, or Boltzman-
type relationship, and the time constant tz(V) by a Gaussian
relationship.
Individual ionic membrane currents were characterized by
fits to published voltage clamp experiments where available.
The temperature at which the experiments were performed
ranged between 30 and 35°C, thus temperature adjustments
(e.g., Q10) were not needed during final integration of currents
into the model. To obtain fits to the currents, membrane po-
tential was held constant and the differential equations char-
acterizing the gating variables for each current were integrated
numerically and substituted into the appropriate equation for
ionic current. Examples of fits to the component currents of the
model are shown in Fig. 2.
Fits to voltage-clamp data are not the only criteria for
formulating ionic current descriptions, which may have to be
modified to fit whole cell transmembrane potential data. These
additional adjustments are justified considering that the volt-
age-clamp and the free-running SOP and square-wave re-
cordings were obtained from different cells and in different
laboratories and that the voltage-clamp experiments were
performed on a particular cell, and there is a considerable
variation in the waveshape of the ionic current response from
(2)
cell to cell.
Calcium currents—ICa
The mathematical equations used in the model for the dif-
ferent calcium currents are given in Table 1, and the voltage
dependence of the steady-state activation and inactivation char-
acteristics of these currents are shown graphically in Fig. 3.
The reversal potential for the calcium currents has been set to
a constant 50 mV by extrapolating current-voltage curves from
Kang and Kitai (1993b).
T-TYPE LVA CALCIUM CURRENT—ICA,T. The T-type calcium cur-
rent is based on data from Kang and Kitai (1993b). The
membrane potential was held at 288 mV and clamped at
various voltages from 268 mV through 254 mV in increments
of 2 mV. Their experiments show a two-stage inactivation: a
quick inactivation to a small current followed by a slow inac-
tivation that maintains the small current for the duration of the
voltage clamp (300 ms). Characterization for the long second-
ary component is based on data from neostriatal neurons
(Hoehn et al. 1993), whereas the activation and fast inactiva-
tion component was characterized by fitting data from Kang
and Kitai (1993b). The voltage-clamp fits to data are shown in
Fig. 2A. The current description fits the quick activation and
biphasic inactivation well at the expense of matching the peak
current values at some clamp potentials.
2252 B. AMINI, J. W. CLARK, AND C. C. CANAVIER

TABLE 1. Calcium currents—ICa because no calcium-independent inactivation was detected in

the presence of the calcium buffer EGTA and the addition of
ICa,T 5 g# Ca,TdT(fTf 1 0.04fTs)(V 2 ECa) the calcium buffer EGTA increased the inactivation time con-
stant. We were unable to find specific data regarding the
1.0 1.0
d# T 5 #fT 5 Ca21-mediated inactivation of ICa,N in DA neurons and hence
1.0 1 e2~V163.5!/1.5 1.0 1 e~V176.2!/3.0
employed a simple Michaelis-Menten relationship to modify
ddT d# T 2 dT 2
5 tdT 5 65.0e2((V 1 68.0)/6.0) 1 12.0 the channel conductance. The parameters of the gating variable
dt tdT
dfTf #fT 2 fTf
equations associated with the equations for ICa,N were deter-
2
5 tfTf 5 50.0e2((V172.0)/10.0) 110.0 mined generally by fitting voltage-clamp data (Fig. 2E) from
dt tfTf
dfTs #fT 2 fTs 2
Kang and Kitai (1993b) in a cell with N-type channels which
5 tfTs 5 400.0e2((V1100.0)/10.0) 1400.0 did not exhibit calcium-dependent inactivation. The cell was
dt tfTs
held at 284 mV and clamped to voltages of 258, 248, and
ICa,N 5 g# Ca,NdN fCa,N(V 2 ECa) 238 mV, respectively. Because the characterization of the
calcium-dependence of this current was not based on voltage-
1.0 0.00010 clamp data, the parameters associated with the steady-state
d# N 5 fCa,N 5
1.0 1 e2~V145.0!/7.0 0.00010 1 @Ca21#i calcium-dependent inactivation function ( fCa,N) were loosely

Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017

ddN d# N 2 dN 2 set in the fit to voltage-clamp data and determined finally in fits
5 tdN 5 18.0e2((V170.0)/25.0) 10.30
dt tdN of the complete model to SOP and square-wave data. The
specific equations for ICa,N are listed in Table 1.
ICa,L 5 g# Ca,LdL fCa,L(V 2 ECa)
L-TYPE HVA CALCIUM CURRENT—ICA,L. On the basis of experi-
1.0 0.00045 ments in other cells (Fox et al. 1987a,b; Johnston and Wu
d# L 5 fCa,L 5
1.0 1 e2~V150.0!/3.0 0.00045 1 @Ca21#i 1995), we have assumed that the nifedipine-sensitive, long-
ddL d# L 2 dL 2 lasting calcium current, ICa,L, in DA neurons has activation
5 tdL 5 18.0e2((V145.0)/20.0) 11.5
dt tdL characteristics that are similar to those of ICa,N. The steady-
state half-activation voltage (V0.5) associated with ICa,L was set
ICa,HVA 5 g# Ca,HVAdHVA fHVA(V 2 ECa)
to be more hyperpolarized compared with the V0.5 value of the
ICa,N activation characteristic. This is consistent with observa-
1.0 1.0
d# HVA 5 #fHVA 5 tions about the relationship between dihydropyridine- and
1.0 1 e2~V110.0!/10.0 1.0 1 e~V148.0!/5.0
v-conotoxin-sensitive currents (Kasai and Neher 1992; Regan
ddHVA d# HVA 2 dHVA 2
5 tdHVA 5 0.1e2((V162.0)/13.0) 10.05 1991) (see Fig. 3).
dt tdHVA
dfHVA #fHVA 2 fHVA 2
Recent experiments have shown that calcium chelation with
5 tfHVA 5 0.5e2((V155.6)/18.0) 10.5 bis-(o-aminophenoxy)-N,N,N9,N9-tetraacetic acid (BAPTA)
dt tfHVA
prolongs the plateau of the square-wave oscillation, which is
blocked by nifedipine (Ping and Shepard 1997). This is con-
N-TYPE HVA CALCIUM CURRENT—ICA,N. As modeled, ICa,N be- sistent with the hypothesis that the L-type current, like ICa,N, is
gins to activate at 260 mV and continues to activate at 245 inactivated by [Ca21]i. The inactivation of ICa,L has been
mV. The slow inactivation that appears in voltage-clamp modeled as calcium dependent (Johnston and Wu 1995). A
records (Kang and Kitai 1993b) is probably calcium-mediated, Michaelis-Menten relationship, ( fCa,L in Fig. 4) has been used
to characterize the calcium-dependent inactivation of this cur-
rent, and its parameters have been adjusted to give reasonable
fits to SOP and square-wave oscillation data. The equations for
ICa,L can be found in Table 1.

FIG. 3. Steady-state characteristics of calcium currents. d and f represent
the activation and inactivation gating variables, respectively, for calcium
currents. Steady-state activation (d# T) and inactivation (f#T) characteristics of
ICa,T were obtained from voltage-clamp studies of Kang and Kitai (1993b).
Steady-state activation of ICa,N (d# N) was obtained by fitting voltage-clamp data
from the same source. Steady-state activation of ICa,L (d# L) is more hyperpo-
larized than that of ICa,N. Steady-state activation (d# HVA) and inactivation (f#HVA)
of ICa,HVA, are adapted from Cardozo and Bean (1995) and Kang and Kitai
(1993b), respectively.
FIG. 4. Steady-state calcium-dependent characteristics associated with var-
ious ionic currents. A: summary of calcium-dependent ionic currents. fCa,L,
fCa,N, and pK,Ca are indicated. B: activation of IK,Ca,SK by calcium during slow
oscillation potential (SOP). IK,Ca activation variable (pK,Ca) is plotted against
the internal calcium concentration, [Ca21]i. Extent of [Ca21]i excursion during
SOP is indicated by the shaded region. C: inactivation of ICa,L and ICa,N by the
calcium-dependent fCa,L and fCa,N, respectively, during square-wave oscilla-
tions. Extent of [Ca21]i excursion during the square-wave oscillations is
indicated by the shaded region.
 CALCIUM DYNAMICS IN PACEMAKER AND BURST FIRING 2253

TABLE 2. Potassium currents TRANSIENT OUTWARD CURRENT—IA. The 4-aminopyridine (4-

AP)-sensitive current, IA, has been observed in DA neurons
1.0 and plays a role in the regulation of action-potential frequency
IK,Ca,SK 5 g# K,Ca,SK pK,Ca~V 2 EK! pK,Ca 5
1.0 1 e~@Ca
21#i2KM,K,Ca!/20.000004
by slowing the recovery of membrane potential to baseline
IA 5 g# Ap3(0.70q1 1 0.30q2)(V 2 EK) levels (Silva et al. 1990). The steady-state activation (p) and
inactivation (q) characteristics of IA (Fig. 2F) were determined
1.0 1.0 by fitting published voltage-clamp data obtained from DA
p# 5 q# 5
1.0 1 e2~V139.0!/27.0 1.0 1 e~V164.0!/5.5 neurons of the substantia nigra pars compacta (SNc) of the rat
dp p# 2 p (Silva et al. 1990). The cell was held at 290 mV and clamped
5 tp 5 5.5e2~~V142.0!/10.0! 1 4.0
2

dt tp to various voltages in the range 260 to 75 mV in 15-mV

dq1 q# 2 q1 70.0 30.0 increments. Equations describing this current are given in
5 tq1 5 1 1 45.0
dt tq1 1.0 1 e~V18.0!/12.0 1.0 1 e~V260.0!/215.0 Table 2, and Fig. 2B shows the model-generated fits to voltage-
dq2 q# 2 q2 1400.0 clamp data.
5 tq2 5 1 450.0
dt tq2 1.0 1 e~V275.0! /213.0
HYPERPOLARIZATION-ACTIVATED CATION CURRENT—IH. The hy-

IK 5 g# Kn3(V 2 EK)
perpolarization-activated cation current (Ih) has been shown to
be important for pace-making activity in thalamic and cortical
1.0 neurons (McCormick and Pape 1990). In DA neurons, how-

Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017

n# 5 tn 5 5.0
1.0 1 e~V 1 39.0!/ 2 12.44
dn n# 2 n
5
dt tn
COMBINED HVA CALCIUM CURRENT—ICA,HVA. The remainder of
the HVA current has been lumped together as ICa,HVA. This
includes the v-agatoxin-sensitive P-type current and the R-type
current, which is not selectively blocked by any known phar-
macological agents. The parameters used for voltage-activation
(V0.5,act 5 210.0 mV, k 5 10 mV) are similar to those obtained
by Cardozo and Bean (1995), which were 25.0 mV and 13.8,
respectively. The parameters for voltage-inactivation are sim-
ilar to those obtained by Kang and Kitai (1993b), (V0.5,inact 5
248 mV, k 5 5.0 mV), which were 248 mV and 6.0, respec-
tively.
CALCIUM-ACTIVATED POTASSIUM CURRENT—IK,CA. Dopaminer-
gic neurons are known to contain at least two types of calcium-
activated potassium currents, the BK and SK currents (Shepard
and Bunney 1988; Silva et al. 1990). The BK, or maxi type
channel, has a high channel conductance and voltage- and
calcium-dependent activation and is known to be blocked
selectively by iberiotoxin and nonselectively blocked by TEA.
Specifically, Silva et al. have shown that IK,Ca,BK is blocked by
TEA in DA neurons (Silva et al. 1990). The smaller conduc-
tance SK channel has a purely calcium-dependent activation
and obtains its voltage dependence indirectly via the voltage
dependence of calcium entry mechanisms (e.g., calcium chan-
nels) (Hille 1992). It is blocked selectively by apamin and is
insensitive to TEA. For the purposes of simulation, we have
assumed that apamin partially blocks IK,Ca,SK, leaving it at
;20% of its original strength. Further, we have assumed that
IK,Ca,SK is activated by intracellular Ca21 through the Boltz-
man-type function pK,Ca with a half-activation calcium concen-
tration of KM,K,Ca given in Table 2 and seen in Fig. 4. The
calcium half-activation concentration value KM,K,Ca has been
set to 190 nM. This is reasonable in view of experimental
observations that the maximum calcium excursion of DA neu-
rons during SOP is ;200 nM (Callaway and Wilson 1997).
Because IK,Ca,BK is inactivated by TEA application as men-
tioned earlier, it cannot be essential for the slow underlying
oscillations (SOP). In the present modeling study, we have
excluded this active spiking mode of oscillation, and therefore
we have not included IK,Ca,BK in the model.
ever, Ih has been shown to have a negligible effect on sponta-
neous firing, resting membrane potential, and normal resting
conductance (Mercuri et al. 1995). Our simulations confirm
that Ih is not essential for the generation of the SOPs or for
setting the normal resting input impedance of the model (see
Table 3). However, Ih is the major component of the input
impedance of the membrane at more hyperpolarized potentials
(about 2100 mV).
Ih activates with hyperpolarization beyond approximately
260 mV and exhibits no inactivation. Its reversal potential Eh
is 235 6 4 mV (Mercuri et al. 1995), which lies between the
reversal potentials for potassium and sodium currents (about
270 and 60 mV, respectively). Ih therefore is modeled as a
mixed cation current predominately permeable to sodium and
potassium using a previously developed characterization
(Demir et al. 1994). Model-generated fits to voltage-clamp data
from rat by Mercuri et al. (1995) are shown in Fig. 2C. The cell
was held at 257 mV and clamped to various potentials from
2141 to 269 mV in 12-mV increments.
DELAYED-RECTIFIER CURRENT—IK. The delayed-rectifier cur-
rent, IK, was characterized by fitting published voltage-clamp
data from rat DA neurons (Silva et al. 1990) in which the cell
was held at 240 mV and clamped to 20 mV in 10-mV
increments (Fig. 2D). The equations characterizing IK are
given in Table 2. As mentioned in the introduction, SOPs are
observed after TTX application in the presence IK and en-
hanced after its blockade by TEA. The significance of this
observation is that IK is not necessary for SOP and, unless
otherwise noted, is not present in either SOP or square-wave
oscillations.
BACKGROUND CURRENT—IB. Although important, there is little
quantitative experimental data upon which to base a mathe-
TABLE 3. Hyperpolarization-activated cation current—Ih
Ih 5 In,Na 1 Ih,K
1.0 3136.0
y# 5 ty 5 1 26.21
1.0 1 e2~V177.62!/17.32 1.0 1 e~V229.60!/22.69
dy y# 2 y dy y# 2 y
5 5
dt ty dt ty
gh,NA 5 0.35gh gh,K 5 0.65gh
Ih,Na 5 gh,N,A y2(V2ENa) Ih,K 5 gh,K y2(V 2 EK)
2254 B. AMINI, J. W. CLARK, AND C. C. CANAVIER

TABLE 4. Background current—IB TABLE 6. Parameter values

IB 5 IB,Na 1 IB,K 1 IB,Ca g# Ca,T 5 0.0240 mS KNa,Ca 5 0.000200 mM24

IB,Na 5 g# B,Na(V 2 ENa) g# Ca,N 5 0.00039 mS DNa,Ca 5 0.05 mM24
IB,Ca 5 g# B,Ca(V 2 ECa) g# Ca,L 5 0.00050 mS g 5 0.5
IB,K 5 g# B,K(V 2 EK) g# Ca,HVA 5 0.00089 mS [B]i 5 0.050 mM
g# K,Ca,SK 5 0.0008 mS kU 5 200.0 mM21 ms21
g# A 5 0.0126 mS kR 5 0.238 ms21
matical description of IB. The linear background current is g# K 5 0.005 mS n 5 2.0
modeled as a sum of three separate components: a sodium g# h 5 0.003 mS [Na1]i 5 5.0 mM
current (IB,Na) a calcium current (IB,Ca), and a potassium com- g# B,Na 5 0.000150 mS [Na1]o 5 154.0 mM
g# B,Ca 5 0.000090 mS [K1]i 5 155.0 mM
ponent (IB,K) The equations for these linear leak conductances g# B,K 5 0.000090 mS [K1]o 5 5.9 mM
are given in Table 4. The input impedance of the model was KM,K,Ca 5 0.00019 mM [Ca21]o 5 2.4 mM
obtained by clamping the membrane to 260 mV from a hold- ICa,pump 5 0.0130 nA Voli 5 0.0141 nL
ing potential of 250 mV and observing the evoked transient KM,CaP 5 0.000500 mM F 5 96500 C/mol
current. A DV of 10 mV was divided by the resultant DI to I#Na,K 5 0.0915 nA R 5 8314 J/(KgmolK)
T 5 300 K
obtain an input impedance of 365 MV, which is within the
experimentally reported range for these neurons (Grace and
binding to an intracellular protein such as calmodulin. The

Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017

Onn 1989; Johnson and North 1992; Yung et al. 1991).
PUMP AND EXCHANGER CURRENTS—ICA,P, INA,CA, AND INA,K. The
equations (see Table 5) for the sodium-calcium exchanger and
the calcium pump are adapted from Canavier et al. (1991) and
the weakly voltage-dependent sodium-potassium pump from
Lindblad (1996). The value for the half-activation calcium
concentration for ICa,pump (KM,CaP) was set to 0.50 mM, which
is within the range of experimental observations for other
preparations (Lichtman et al. 1981; Michaelis et al. 1983).
Fluid compartment model
Calcium dynamics are an important part of DA neuron
activity because Ca21-mediated processes are hypothesized to
be involved in the generation of the subthreshold oscillations,
and, as a result, the mechanisms of rhythmic firing. The ma-
terial balance on calcium depends on three processes: entry,
extrusion, and buffering of calcium. The differential equation
representing the rate of change of cytosolic Ca21 is
z ~I Na,Ca 2 I Ca 2 I B,Ca 2 I Ca,pump !
@ Ca 21 # i 5
2Vol i F
where Voli is the cytosolic volume in nanoliters, F is Faraday’s
constant in C/mol, and [B]i is the concentration of the internal
calcium buffer in millimolar with n binding sites for Ca21. nBi
gives the total concentration of binding sites for Ca21. The
volume is calculated based on a spherical soma of radius 15
mm for a total intracellular volume of 0.0141 nl.
The buffering of intracellular calcium has been modeled as
formula for the buffer (Robertson et al. 1981) is given by
Ȯ C 5 k U @Ca 21 # i ~1 2 O C ! 2 k R O C (4)
where OC is the buffer occupancy or the fraction of sites that
already are occupied by Ca21 and therefore unavailable for
binding. All binding sites are considered as a single population
and assumed to be independent. This assumption ignores the
nonlinearities that occur as a result of multiple cooperative
binding sites.
COMPUTATIONAL ASPECTS
The complete system consists of 15 state variables: 3 dif-
ferential equations describing the time rate of change in mem-
brane voltage (V), internal calcium concentration ([Ca21]i),
and fractional occupancy of the calcium buffer (OC) and 12
other differential equations characterizing voltage-dependent
gating variables. Equations 1, 3, and 4, along with the tables
for individual membrane currents (Tables 1, 2, 3, 4, and 5);
parameter values (Table 6); and initial conditions (Table 7),
2 n@B# i Ȯ C (3)
contain all the information necessary to carry out the simula-
tions presented in this paper. The units used in the model are
time in milliseconds (ms), voltage in millivolts (mV), concen-
tration in millimoles/liter (mM), current in nanoamperes (nA),
conductance in microsiemens (mS), capacitance in nanofarads
(nF), volume in nanoliters (nl), and temperature in degrees
Kelvin (°K). All simulations were performed on a Sun Ultra I
TABLE 7. Initial conditions

TABLE 5. Pumps and exchangers V 5 241.7336 V 5 248.884

[Ca21]i 5 0.00018411 [Ca21]i 5 0.00034369
OC 5 0.133842 OC 5 0.224099
@Ca2 1 #i n 5 0.436929 n 5 0.312783
ICa,pump 5 I#Ca,pump 21 y 5 0.183531 y 5 0.137947
@Ca #i 1 KM,CaP
m 5 0.22642 m 5 0.071887
DFin 2 DFout h 5 0.523627 h 5 0.685456
INa,Ca 5 KNa,Ca dT 5 0.999997 dT 5 0.999949
S
fT 5 1.58247e205 fT 5 0.000104658
DFin 5 [Na1]3i [Ca21]oe(2gVF/RT)
fT2 5 0.000594264 fT2 5 5.1266e205
DFout 5 [Na1]3o[Ca21]ie(2(g21)VF/RT) dL 5 0.864897 dL 5 0.617111
dN 5 0.598077 dN 5 0.370159
S 5 1 1 DNa,Ca([Na1]3o[Ca21]i 1 [Na1]3i [Ca21]o) dHVA 5 0.0401651 dHVA 5 0.0200705
fHVA 5 0.224057 fHVA 5 0.543193
@Na 1 #1.5 @K 1 #o ~V 1 150.0! p 5 0.473764 p 5 0.409666
INa,K 5 #INa,K
i
1 1.5 1 q1 5 0.0902836 q1 5 0.0487723
@Na # 1 KM,Na @K #o 1 KM,K ~V 1 200.0!
i
1.5
q2 5 0.109643 q2 5 0.0378656
 CALCIUM DYNAMICS IN PACEMAKER AND BURST FIRING 2255

by forward integration of the coupled system of differential

equations using an implicit fifth-order Runge-Kutta method
with variable step size designed for stiff systems of differential
equations (Hairer and Warner 1990).

RESULTS

In addition to providing a quantitative test that the current

descriptions derived from voltage-clamp data are sufficient to
reproduce experimental data generated under current-clamp
conditions, simulations of the dopaminergic neurons in this
study enable us to gain insight into the ionic mechanisms
responsible for rhythmic firing as well as to utilize this insight
to make experimentally testable predictions. Figure 5 summa-
rizes the model results for the two oscillations of interest and
compares model-generated and experimentally observed be-
havior under different simulated experimental conditions. Be-

Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017

cause TTX must be present to reveal slow sinusoidal oscilla-
tions, INa has not been modeled. TEA application is simulated
by letting g# K 5 0. which results in slow oscillatory potentials
(SOPs, Fig. 5A). Simulation of the application of apamin
blocks most of the outward current IK,Ca,SK and results in
square-wave oscillations (Fig. 5B). In either case, the model
provides SOPs or square-wave oscillations that agree qualita-
tively with data from Ping and Shepard (1996).

Slow oscillatory potentials

Figure 6 shows the ionic currents underlying the SOP.
Figure 6A shows the SOP (same as Fig. 5A), whereas the
intracellular calcium concentration is shown in B. The major
currents involved in SOPs are ICa,L and IK,Ca, whereas ICa,N is
small relative to ICa,L (Fig. 6C). Note that ICa,T and ICa,HVA (E)
are not fully activated because the SOPs occur in a potential
range that is outside their range of activity. On the other hand,
the calcium inactivation characteristic of ICa,N (Fig. 4A) keeps
the magnitude of this current below that of ICa,L in the calcium
range of oscillation (nominally, 180 –200 nM). The other
model currents consist of background and pump currents as
seen in Fig. 6D as well as the transient outward current, IA (E).
These currents are important in biasing the cell in the appro-
priate range of membrane potential and cytosolic Ca21 con-
centration. The hyperpolarization-activated mixed cation cur-
FIG. 5. Model-generated and experimentally observed oscillations of DA
neurons under influence of blocking agents. A: model and data comparison of
the SOP. IK shut off to simulate TEA application. B: model and data behavior
during oscillations produced by partial blockage of IK,Ca,SK to simulate apamin
application. Data from rat scanned and digitized from Fig. 2C of Ping and
Shepard (1996).
FIG. 6. Underlying currents of SOP. A: membrane potential during SOP. B:
[Ca21]i oscillations conform well to data from Callaway and Wilson (1997). C:
ICa,L and IK,Ca are the main ionic currents during the SOP with minor contri-
bution from ICa,N. D: background currents IB, sodium-dependent currents INa,K
and INaCa, and the calcium pump, ICa,pump. E: remaining calcium currents ICa,T
and ICa,HVA and the transient outward current IA.
rent, Ih, can be eliminated with little impact on the shape and
frequency of the SOPs (not shown).
Mechanisms underlying SOP
The mechanism underlying the SOP (Fig. 6A) can be under-
stood by considering the calcium- and voltage-dependent pro-
cesses in the model. Broadly, activation of ICa,L depolarizes the
membrane and leads to increased [Ca21]i (Fig. 6C), which
activates IK,Ca,SK (Fig. 4B). Activation of IK,Ca,SK, in turn,
hyperpolarizes the membrane, decreases ICa,L, and reduces
[Ca21]i.
Clearly, the calcium transient (Fig. 7C) lags behind the
calcium current ICa,L, whereas the potassium current (IK,Ca,SK)
is in phase. As IK,Ca,SK increases it limits the membrane depo-
larization due to ICa,L and institutes a first stage of repolariza-
tion (slope L1 in Fig. 7A). The peak of ICa,L is long lasting, and
it is only after it begins to decline and concomitantly, IK,Ca,SK
peaks, that a second, stronger phase of repolarization is brought
about (slope L2 in Fig. 7A).
The model-generated calcium transient during SOP con-
forms well to experimental observations (Callaway and Wilson
1997). Calcium levels vary within the reported range of 160
and 220 nM, and the calcium oscillations lag the voltage
oscillations by ;90°. Nedergaard et al. (1993) have shown that
2256 B. AMINI, J. W. CLARK, AND C. C. CANAVIER
FIG. 7. Important currents underlying sinusoidal oscillations. Membrane
potential during SOP (A) responds weakly (with slope of line L1) at first to the
activation of IK,Ca (C). ICa,L (B) remains active, increasing [Ca21]i (D). Finally,
ICa,L regeneratively deactivates, hyperpolarizing the membrane (with slope of
line L2).
neither Ni21 nor v-conotoxin can block the SOP in guinea pig
SNc DA neurons, whereas nifedipine application alone is ef-
fective. Figure 8 shows that blockade of ICa,T, ICa,N, and ICa,L
in the model for the purposes of simulating Ni21, v-conotoxin,
and nifedipine application, respectively, conforms to these
experimental results. The increase in amplitude after blockade
of ICa,T is due to the removal of a depolarizing drive and
reduction in calcium entry, which allows ICa,L to remain active
longer, raising the peak potential.
Square-wave oscillations
The component currents underlying the square-wave oscil-
lations are shown in Fig. 9 in the same order as those for the
sinusoidal case (Fig. 6). The major depolarizing currents are
again ICa,L and ICa,N; however, the outward K1 current IK,Ca,SK
has been blocked. Our model predicts that the mechanism for
Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017
FIG. 9. Underlying currents of square-wave oscillations. A: membrane po-
tential during square-wave oscillations. B: [Ca21]i variation during square-
wave oscillations is greater than that seen during SOPs. C: ICa,L is the main
ionic current. The residual IK,Ca,SK current and ICa,N are minor contributers in
this mode of oscillation. D: calcium and sodium background currents, IB,Ca and
IB,Na. E: remaining calcium currents ICa,T and ICa,HVA. F: sodium-dependent
pump currents, INa,K and INa,Ca. G: transient outward current (IA), potassium
background current (IB,K), and the calcium pump (ICa,P).
the square-wave oscillations need not depend on the presence
of a distinct hyperpolarization current in this mode of oscilla-
tion. Without the activation of IK,Ca,SK by calcium, internal
calcium concentration continues to rise and achieves a higher
peak level than in sinusoidal oscillations (Fig. 9B), and the
frequency of oscillation decreases. ICa,L is inactivated more
strongly at the elevated levels of [Ca21]i. As ICa,L declines due
to inactivation, the remaining “residual” currents (IRes, Fig. 10)
sum to repolarize the membrane.
FIG. 8. Simulation of Ni , v-conotoxin, and nifed-
21
ipine application. A: control. Application of Ni21 (B)
and v-conotoxin (C) produce small changes in the
character of the SOP, whereas nifedipine application
(D) eliminates the SOP.
 CALCIUM DYNAMICS IN PACEMAKER AND BURST FIRING
FIG. 10. Time-expanded view of square-wave oscillations. A and B: plotted
as before and show the membrane potential and cytosolic Ca21 during square-
wave oscillations. C: ICa,L and the sum of the remaining currents (IRes 5 Itot 2
ICa,L).
Functional tests
Certain functional tests reveal the robustness of the model
and provide explanations of experimental observations. It has
been shown that application of TTX alone will result in lower-
amplitude and higher-frequency sinusoidal oscillations than
those resulting from the coapplication of TTX and TEA (Ne-
dergaard and Greenfield 1992; Kang and Kitai 1993a). Figure
11B is a simulation of membrane potential under TTX appli-
cation alone. The resulting oscillations in this case are due to
interactions between ICa,L and the two outward currents IK, and
IK,Ca,SK. Recall from Fig. 7, that due to its calcium-dependent
activation, IK,Ca,SK reaches its maximum strength ;100 ms
after membrane potential peaks. A voltage-dependent current
with relatively fast activation, on the other hand, would lead to
a hyperpolarizing drive that coincides with the peak voltage.
This drive leads to an earlier onset of hyperpolarization, lead-
ing to a lower amplitude and a shorter period.
The mechanisms of the square-wave oscillations also were
tested via model simulation and functional tests. Ping and
Shepard (1997) have shown that chelation of DA neurons
during square-wave oscillation by intracellular injection of the
calcium buffer BAPTA prolongs the plateau phase of the
oscillation. Because BAPTA is a calcium buffer, we assume
that the effect of adding more buffer to that already present in
the cytosol is similar to increasing the concentration of the
internal calcium buffer [B]i. The value of [B]i used nominally
in the model (0.050 mM) was increased to 0.080 mM to
simulate calcium chelation. Figure 12 shows the model-gener-
ated response to this increase in [B]i. At first, the results seem
counterintuitive. An increase in buffer concentration should
lead to a decrease in the rate of change of [Ca21]i, resulting in
a smaller excursion in calcium concentration. However, on
closer inspection of the figure, we see that the increase in [B]i
FIG.11. Effect of TEA on the model. A: control SOP in the presence of
TTX and TEA. B: SOP with bath application of TTX alone results in lower
amplitude oscillations.
2257
Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017
FIG. 12. Increased buffering and calcium-mediated currents. A: simulation
of chelation by increasing the concentration of the internal calcium buffer
results in increased plateau duration as compared with the control case. B:
increased buffering delays the rise in [Ca21]i which allows ICa,L (C) to remain
active longer. There is little change in the strength of the other calcium-
dependent currents, INa,Ca and ICa,P.
also leads to a decrease in the rate of ICa,L inactivation, which
causes a larger calcium influx and a greater excursion in
[Ca21]i.
In addition to changes in buffer concentration, the calcium
availability can be changed by adjusting other calcium-depen-
dent processes in the model. Two additional currents that have
the potential for altering calcium availability in the cytosol are
the calcium pump, ICa,pump, and the sodium-calcium ex-
changer, INa,Ca. Increasing the strength of either calcium ex-
trusion mechanism should decrease cytosolic Ca21 availability
and yield results similar to increasing [B]i. In addition to
increasing calcium extrusion, calcium availability also can be
restricted. Because we have considered ICa,N nonessential for
SOPs, it can be reduced without altering the self-excitability of
the model. Figure 13 summarizes the results of these adjust-
ments. The maximal conductances for INa,Ca, ICa,pump, and ICa,N
(KNa,Ca, I#Ca,pump, and g# Ca,N, respectively) were each changed
by 5% in these experiments. We found that INa,Ca alone was
able to affect duty cycle, i.e., plateau duration could be
changed without a significant change in the period of the
oscillation. Figure 13A shows the control square-wave oscilla-
tions followed by three different schemes for increasing pla-
teau duration by reducing cytosolic Ca21. Of the three adjust-
ments, only increasing INa,Ca (Fig. 13D) has the desired effect.
Decreasing ICa,N (B) and increasing ICa,pump (C) in fact have
very little effect.
DISCUSSION
We have developed a semiquantitative model that can mimic
the behavior of DA neurons under a wide range of experimen-
2258 B. AMINI, J. W. CLARK, AND C. C. CANAVIER
FIG. 13. Altering plateau duration. Control square-wave oscillations are
shown (A) followed by attempts at increasing plateau duration by decreasing
ICa,N (B), increasing ICa,pump (C), and increasing INa,Ca (D). Lightly and darkly
shaded regions divide the control period at the minimum, and the dashed line
marks the control period from that minimum (P). As the lightly shaded regions
show, only increasing INa,Ca had the desired effect of increasing plateau
duration. Additionally, INa,Ca adjustment allows for control of plateau duration
without significant change in oscillation frequency.
tal conditions, including: SOPs in the presence of TTX and
TEA; the changes in the period and amplitude of sinusoidal
oscillations in response to TEA application; square-wave os-
cillations in the presence of TTX, TEA and apamin; and the
increase in the duration of the plateau phase of square-wave
oscillations observed in response to calcium chelation. The key
feature of the model is that it can be used to explore the
putative mechanisms underlying SOP and square-wave oscil-
lations.
Somatic origin of calcium-dependent oscillations
This model is a representation of the soma only and there-
fore is a first approximation of the full dynamics in a dopamine
neuron, which has a dendritic tree whose branches frequently
extend .500 mm from the soma. The data on which the current
descriptions are drawn are characteristic of the soma and
proximal dendrites because the studies on which the descrip-
tions are based (Cardozo and Bean 1995; Kang and Kitai
1993b; Silva et al. 1990) were conducted in acutely dissociated
or thin (150 –200 mm) coronal slice preparations. In both of
these types of preparations, the dendritic arbor can be trun-
cated. In particular, ICa,T may be localized preferentially on
distal dendrites because only 11 of 15 cells in thin coronal slice
(Kang and Kitai 1993b) and none in the dissociated preparation
(Cardozo and Bean 1995) exhibited this current.
On the other hand, the presence of large calcium currents on
the distal dendrites has been demonstrated in thick (400 mM)
coronal sections (Dunia et al. 1996). Nonetheless, the assump-
tion that the somatic currents and calcium dynamics drive the
SOP is justified by several experimental observations. First,
during the SOP the calcium transients in the soma are a
substantial fraction of those observed in the dendrites (Calla-
way and Wilson 1997) despite the large disparity in surface
area to volume ratio. Hence there is a significant amount of
calcium entry into the soma. Second, the PLD can be evoked
by brief depolarizations of the soma, and its rate of activation
is strongly dependent on the membrane potential in the soma
(Grace and Onn 1989). Finally, one study found that sectioning
the distal dendrites of DA neurons had no significant effect on
the pacemaker firing frequency (Nedergaard and Greenfield
1992). Assuming the firing frequency is strongly influenced by
the SOP, the distal dendrites do not contribute significantly to
the SOP as recorded at the soma.
Predictive value of model can aid in experimental design
Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017
Although the modeled voltage- and calcium-dependence of
ICa,L is consistent with available data, the precise form of the
voltage dependence and even the presence of calcium inacti-
vation of this current have not been directly demonstrated in rat
dopamine neurons. Hence the model description of this current
can be considered a prediction (see following text). In addition,
model predictions that INa,Ca controls the duty cycle (Fig. 13D)
could be tested if a specific INa,Ca blocker could be applied to
the bath.
Homogeneity of DA neurons within and across species
Homogeneity of midbrain DA neurons across species fre-
quently is assumed (Cardozo 1993; Lacey et al. 1989; Sang-
hera and German 1984; Tepper et al. 1987; Yung et al. 1991)
because the PLD, depolarized spike threshold, long-duration
action potentials, inward rectification, and other characteristics
are observed across species. It has been argued (Kotter and
Feizelmeier 1998) that scaling alone does not ensure constant
activity patterns between species, but additional compensatory
mechanisms (LeMasson et al. 1993; Turrigiano et al. 1994)
likely are involved. The model in the current study applies
specifically to rat mesencephalic neurons because it is based on
data from rat (Cardozo and Bean 1995; Kang and Kitai
1993a,b; Ping and Shepard 1996; Silva et al. 1990).
Within a species, these neurons also exhibit a degree of
electrophysiological homogeneity (Cardozo and Bean 1995;
Johnson and North 1992; Yung et al. 1991) and similar re-
sponses to stimuli (Schultz et al. 1995), although they are
located in three adjacent regions: the ventral tegmental area
(VTA or A10), the substantia nigra pars compacta (A9), and
the retrorubral area (A8). Although we have treated the DA
neurons as a homogenous population, some have suggested
that there are subpopulations of DA neurons (Nedergaard and
Greenfield 1992; Shepard and Bunney 1988), and there is
evidence of some heterogeneity in their neurochemistry, phar-
macology, and electrophysiology (Roth and Elsworth 1995). In
addition, the mix of calcium currents in these cells can vary
with age (DeFazio and Walsh 1995).
Does the dihydropiridine-sensitive ICa,L mediate the SOP in
all cases?
We have assumed that dihydropyridines block the SOP,
based on the following data. Mercuri et al. (1994) reported that
 CALCIUM DYNAMICS IN PACEMAKER AND BURST FIRING
30 mM nifedipine blocked spontaneous pacemaker activity in
rats, and by inference, the SOP as well. This is consistent with
findings in rat that the calcium conductance mediating the SOP
is dihydropyridine sensitive (Ping and Shepard 1997) and with
those of Kang and Kitai (1993a,b), who found that the calcium
channel blocker Cd abolished the SOP in rat. These findings
are also consistent with the results of Nedergaard et al. (1993),
who reported that 0.5–20 mM nifedipine abolished the SOP in
guinea pig. On the other hand, Fujimura and Matsuda (1989)
reported that although the calcium channel blockers Cd21 and
Co21 as well as Ca-free saline blocked the SOP in guinea pig,
100 mM nifedipine did not. Furthermore Yung et al. (1991)
reported that 500 mM Ni21 abolished the SOP in guinea pig,
whereas Nedergaard et al. (1993) reported that 500 mM Ni21
only slightly attenuated it.
Thus, in guinea pig at least, there are conflicting data.
However, the evidence argues for a dominant role for ICa,L. For
example, Kang and Kitai hypothesized that ICa,N was respon-
sible for the PLD; however, they did not show that v-cono-
toxin abolishes the SOP. Nedergaard et al. (1993) did perform
this test in guinea pig neurons and found that 1–10 mM
v-conotoxin did not block the SOP; this casts doubt on the
importance of ICa,N in generating these oscillations. The data of
Yung et al. (1991) suggest a role for ICa,T because Ni21
selectively blocks this current (but see Ellinor 1993). Our
simulations have shown that it is very unlikely that ICa,T, as
characterized by the voltage-clamp data of Kang and Kitai,
contributes significantly in the voltage range of the PLD.
Similarly, although Kang and Kitai have shown that ICa,N
exists in the correct voltage range for the PLD, the simulations
show that to generate both SOP and square-wave oscillations
under the appropriate conditions, a calcium current with dif-
ferent activation characteristics from ICa,N is necessary. We
predict that ICa,L, as we have characterized it, is this missing
current. The voltage range of activation of this current is more
negative than that typically associated with L-type calcium
currents; however, similar L-type Ca21 channels have been
found in several locations, including guinea pig motoneurons
(Hsiao et al. 1998), turtle motoneurons (Russo and Hounsgaard
1996), and rat supraoptic neurons (Fisher and Bourque 1996).
Further experimental work is required to verify the validity of
this prediction and determine the activation characteristics of
ICa,L in DA neurons.
It is widely accepted that a persistent, TTX-sensitive sodium
current also contributes to the SOP (Grace and Onn 1989;
Nedergaard et al. 1993; Ping and Shepard 1996). In the dopa-
mine neurons of the retina, which exhibit similar regular pace-
maker activity in a slice preparation (Feigenspan et al. 1998),
the sodium current, and not a calcium current, is the dominant
pacemaking current. Hence under different circumstances, dif-
ferent currents may contribute to the SOP in differing propor-
tions, perhaps due to self-regulatory compensatory mecha-
nisms that dictate that dopamine cells are intrinsic pacemakers.
Comparison with previous models
Li et al. (1996) developed a model that replicated both
NMDA- and apamin-induced burst firing but ignored the SOP
and regular pacemaker firing. Their model was conceptual in
nature with arbitrary parameters not based on the voltage-
clamp data of the particular currents in DA cells. The mecha-
2259
nism that they proposed for NMDA-induced bursting is that
same as that proposed by a later, more physiologically based
model by Canavier (1999), namely the interaction of the
NMDA-induced current (INMDA) and the sodium pump. Re-
generative voltage activation of INMDA is postulated to be
responsible for the depolarization during a burst, and sodium
accumulation due to entry via INMDA is hypothesized to acti-
vate the electrogenic sodium pump, which is a net outward
current, until a regenerative hyperpolarization due to the volt-
age-dependent closing of INMDA channels. Sodium is removed
during the hyperpolarizing phase, allowing the cycle to begin
again. On the other hand, the mechanism proposed by Li et al.
(1996) for apamin-induced, calcium-dependent bursting as-
signs a crucial role to ICa,T rather than ICa,L, which is unlikely
in view of the more physiologically based simulations in the
current study. In addition to the models described above,
Kotter and Feizelmeier (1998) also have modeled DA neurons,
Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017
but it is difficult to compare the models because that study did
not discuss underlying ionic mechanisms, but merely showed
that different activity patterns, such as single-spike firing and
burst-firing, could arise from scaling parameters to account for
different neuronal sizes.
Whereas the model of Canavier (1999) focused solely on
sodium dynamics and the current study focuses solely on
calcium dynamics, in reality these systems operate in tandem
and interact with each other. Further work is required to
elucidate this interaction, as apamin is known to facilitate
NMDA-induced burst firing (Seutin et al. 1993). As far as the
relative contributions of these types of burst-firing to the situ-
ation in vivo, it is well known that NMDA promotes burst
firing in vivo (Overton and Clark 1992) and that DA neurons
receive excitatory amino acid input (Carter 1982; Robledo and
Feger 1990; Scarnati et al. 1986). To date, an endogenous
substance that modulates IK,Ca,SK in vivo has not been identi-
fied. Some have argued that apamin-induced burst firing re-
sembles in vivo firing patterns more closely than burst firing
induced by NMDA (Overton and Clark 1997), so a role for
IK,Ca,SK may yet be discovered in addition to the well-estab-
lished role for INMDA.
APPENDIX A: NULLCLINE ANALYSIS
Nullcline analysis was an invaluable tool in the development of the
model. Repeated, time-consuming simulations were avoided, and
considerable insight into the model was gained through nullcline
analysis, which is a subset of the more general phase space analysis of
nonlinear systems (see Rinzel and Ermentrout 1998).
Analysis of multidimensional systems—application to DA
model
A multidimensional system can be analyzed with two-dimensional
nullcline methods if it can be reduced to a two-dimensional system.
This is done by eliminating the kinetics of all but two state variables
and assuming that they can satisfactorily characterize the system. As
we have seen, the slow cyclic variation in [Ca21]i along with mem-
brane potential are essential contributors to oscillation in DA neurons.
Therefore we reduce the system to a two-dimensional one and set all
state variables to their steady-state values. We obtain the nullclines for
the reduced system by setting all state variables to their steady state
values and numerically solving the following equations:
2260 B. AMINI, J. W. CLARK, AND C. C. CANAVIER
FIG. A1. DA neuron nullclines, associated trajectories, and kinetics of the
reduced model. Oscillation trajectory in the V-[Ca21]i plane for SOPs and
square-wave oscillations (A). Time course of the potential and calcium oscil-
lations is shown in B. Reduced model (- - -) has a smaller calcium excursion
and a higher frequency compared with the complete model (—).
dV
50
dt
d@Ca 21 # i
50 (A1)
dt
Additionally, we assumed that a rapid equilibrium is reached between
the free calcium ([Ca21]i) and the buffered calcium (n[B]OC). We
then used the ratio of the free calcium to the total calcium to scale the
derivative of [Ca21]i (Chay and Keizer 1983). If we assume that the
[Ca21]i kinetics are slow compared with those of V, we can use the
slope of the potential nullcline at the equilibrium point as an indica-
tion of the stability of the system. If the slope is negative, we have a
stable equilibrium as any perturbations return the trajectory to the
equilibrium point. A positive slope indicates an unstable equilibrium
point as it repels all points in its vicinity to a limit cycle. The DA
neuron nullclines for the SOP and square-wave oscillations along with
their trajectories are shown in Fig. A1A. The waveforms generated by
the reduced (two dimensional) model (Fig. A1B, - - -) in are compared
with those of the complete model (—). In both SOP and square-wave
oscillations, the reduced model has a smaller calcium excursion, and
as a result, a higher frequency.
Rough tunings of the model were performed through analysis of the
nullclines of the reduced system. First, the slope of the potential
nullcline at the equilibrium point of the reduced system was a good
predictor of the stability of the complete system. Using this rule, we
were able to make large changes in the system equations while
maintaining the model in the correct range for oscillation. Also,
because [Ca21]i varies slowly relative to the potential, the trajectory
in phase plane remains more or less on the potential nullcline, jumping
between stable branches when an unstable region is encountered. This
was very helpful in estimating from the nullclines, the extent of
[Ca21]i and V excursion without running the full simulations.
APPENDIX B: MODEL EQUATIONS AND
PARAMETERS
The equations describing the dopaminergic neuron model are con-
tained within this appendix. Expressions for the transmembrane cur-
rents are given in Tables 1– 4. The pump and exchanger currents are
given in Table 5. The gating variables in the model are n, dT, fTf, fTs,
dN, dL, dHVA, fHVA, p, q1, and q2. The equations for z#(V) and tz(V) can
be found in Tables 1–3. A listing of model parameter values is given
in Table 6, and the initial conditions needed to run the model in both
modes of oscillation are given in Table 7 with SOPs and square-wave
oscillations in the left and right columns, respectively.
This work was funded by a Biomedical Engineering Research grant from the
Whitaker Foundation and National Institute of Neurological Disorders and
Stroke Grant NS-37963.
Address for reprint requests: C. Canavier, Dept. of Psychology, University
of New Orleans, New Orleans, LA 70148.
Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017
Received 9 February 1999; accepted in final form 29 June 1999.
REFERENCES
CALLAWAY, J. AND WILSON, C. Imaging voltage-dependent calcium oscillations
in the dopaminergic neurons of the substantia nigra, pars compacta. Soc.
Neurosci. Abstr. 23: 1278, 1997.
CANAVIER, C. C. Sodium dynamics underlying burst firing and putative mech-
anisms for the regulation of the firing pattern in midbrain dopamine neurons:
a computational approach. J. Comput. Neurosci. 6: 49 – 69, 1999.
CANAVIER, C. C., CLARK, J. W., AND BYRNE, J. H. Simulation of the bursting
activity of neuron RIS in Aplysia: role of ionic currents, calcium balance,
and modulatory transmitters. J. Neurophysiol. 66: 2107–2124, 1991.
CARDOZO, D. L. Midbrain dopaminergic neurons from postnatal rat in long
term primary culture. Neuroscience 56: 409 – 421, 1993.
CARDOZO, D. L. AND BEAN, B. P. Voltage-dependent calcium channels in rat
midbrain dopamine neurons: modulation by dopamine and GABAB recep-
tors. J. Neurophysiol. 743: 1137–1148, 1995.
CARLSON, N. R. Fundamentals of Physiological Psychology (2nd ed.). Need-
ham Heights, MA: Allyn and Bacon, 1992.
CARTER, C. J. Topographical distribution of possible glutaminergic pathways
from the frontal cortex to the stratum and substantia nigra in rats. Neuro-
pharmacology 21: 379 –383, 1982.
CHAY, T. R. AND KEIZER, J. Minimal model for membrane oscillations in the
pancreatic beta cell. Biophys. J. 42: 181–190, 1983.
DEFAZIO, T. AND WALSH, J. P. Differential ontogeny of calcium currents in the
rat nigrostriatal system. Soc. Neurosci. Abstr. 21: 1577, 1995.
DEMIR, S. S., CLARK, J. W., MURPHY, C. R., AND GILES, W. R. A mathematical
model of a rabbit sinoatrial node cell. Am. J. Physiol. 35: 832– 852, 1994.
DUNIA, R., BUCKWALTER, G., DEFAZIO, T., VILLAR, F. D., MCNEIL, T., AND
WALSH, J. Decreased duration of Ca21-mediated plateau potentials in striatal
neuron from aged rats. J. Neurophysiol. 764: 2353–2363, 1996.
ELLINOR, P. T., ZHANG, J. F., RANDALL, M., ZHOU, A. D., SCHWARZ, T. L.,
TSIEN, R. W., AND HORNE, W. A. Functional expression of a rapidly
inactivating neuronal calcium channel. Nature 363: 455– 458, 1993.
FEIGENSPAN, A., GUSTINCICH, S., BEAN, B. P., AND RAVIOLA, E. Spontaneous
activity of solitary dopaminergic neurons of the retina. J. Neurosci. 18:
6776 – 6789, 1998.
FISHER, T. E. AND BOURQUE, C. W. Calcium-channel subtypes in the somata
and axon terminals of magnocellular neurosecretory cells. Trends Neurosci.
19: 440 – 444, 1996.
FOX, A. P., NOWYCKY, M. C., AND TSIEN, R. W. Kinetic and pharmacological
properties distinguishing three types of calcium currents in chick sensory
neurones. J. Physiol. (Lond.) 394: 149 –172, 1987a.
FOX, A. P., NOWYCKY, M. C., AND TSIEN, R. W. Single-channel recordings of
three types of calcium channels in chick sensory neurones. J. Physiol.
(Lond.) 394: 173–200, 1987b.
FREEMAN, A., MELTZER, L., AND BUNNEY, B. Firing properties of substantia
nigra dopaminergic neurons in freely moving rats. Life Sci. 3620: 1983–
1994, 1985.
FUJIMURA, K. AND MATSUDA, Y. Autogenous oscillatory potentials in the
neurons of the guinea pig substantia nigra pars compacta in vitro. Neurosci.
Lett. 104: 53–57, 1989.
 CALCIUM DYNAMICS IN PACEMAKER AND BURST FIRING
GRACE, A. AND BUNNEY, B. The control of firing pattern in nigral dopamine
neurons: burst firing. J. Neurosci. 4: 2877–2890, 1984.
GRACE, A. A. AND ONN, S.-P. Morphology and electrophysiological properties
of immunochemically identified rat dopamine neurons recorded in vitro.
J. Neurosci. 910: 3463–3481, 1989.
GU, X., BLATZ, A., AND GERMAN, D. Subtypes of substantia nigra dopaminer-
gic neurons revealed by apamin: autoradiographic and electrophysiological
studies. Brain Res. Bull. 28: 435– 440, 1992.
HAIRER, E. AND WARNER, G. Solving Ordinary Differential Equations. II. Stiff
and Differential-Algebraic Problems. Springer Series in Computational
Mathematics. New York: Springer-Verlag, 1990.
HILLE, B. Ionic Channels of Excitable Membranes (2nd ed.). Sunderland: MA:
Sinauer Associates Sunderland, 1992.
HOEHN, K., WATSON, T. W., AND MACVICAR, B. Multiple types of calcium
channels in acutely isolated rat neostriatal neurons. J. Neurosci. 133: 1244 –
1257, 1993.
HSIAO, C. F., DEL NEGRO, C. A., TRUEBLOOD, P. R., AND CHANDLER, S. H. Ionic
basis for serotonin-induced bistable membrane properties in guinea pig
trigeminal motoneurons. J. Neurophysiol. 796: 2847–2856, 1998.
JOHNSON, S. W. AND NORTH, R. A. Two types of neuron in the rat ventral
tegmental area and their synaptics inputs. J. Physiol. (Lond.) 450: 455– 468,
1992.
JOHNSON, S. W., SEUTIN, V., AND NORTH, R. A. Burst-firing in dopamine
neurons induced by N-methyl-D-aspartate: role of electrogenic sodium
pump. Science 258: 665– 667, 1992.
JOHNSTON, D. AND WU, S. Foundations of Cellular Neurophysiology. Cam-
bridge, MA: MIT Press, 1995.
KANG, Y. AND KITAI, S. Calcium spike underlying rhythmic firing in dopami-
nergic neurons of the rat substantia nigra. Neurosci. Res. 18: 195–207,
1993a.
KANG, Y. AND KITAI, S. A whole cell patch-clamp study on the pacemaker
potential in dopaminergic neurons of rat substantia nigra compacta. Neuro-
sci. Res. 18: 209 –221, 1993b.
KASAI, H. AND NEHER, E. Dihydropyridine-sensitive and v-conotoxin-sensitive
calcium channels in a mammalian neuroblastoma glioma cell line.
J. Physiol. (Lond.) 448: 161–188, 1992.
KOTTER, R. AND FEIZELMEIER, M. Species-dependence and relationship of
morphological and electrophysiological properties in nigral compacta neu-
rons. Prog. Neurobiol. 545: 619 – 632, 1998.
LACEY, M., MERCURI, N., AND NORTH, R. Two cell types in rat substantia nigra
zona compacta distinguished by membrane properties and the actions of
dopamine and opioids. J. Neurosci. 9: 1233–1241, 1989.
LEMASSON, G., MARDER, E., AND ABBOTT, L. F. Activity-dependent regulation
of conductances in model neurons. Science 259: 1915–1917, 1993.
LI, Y., BERTRAM, R., AND RINZEL, J. Modeling N-methyl-D-aspartate-induced
bursting in dopamine neurons. Neuroscience 712: 397– 410, 1996.
LICHTMAN, A., SEGEL, G., AND LICHTMAN, M. Calcium transport and calcium-
ATPase activity in human lymphocyte plasma membrane vesicles. J. Biol.
Chem. 12256: 6148 –54, 1981.
LINDBLAD, D. S., MURPHY, C. R., CLARK, J. W., AND GILES, W. R. A model of
the action potential and underlying membrane currents in a rabbit atrial cell.
Am. J. Physiol. 271: 1666 –1696, 1996.
OVERTON, P. AND CLARK, D. Iontophoretically administered drugs acting at the
N-methyl-D-aspartate receptor to modulate burst firing in A9 dopamine
neurons in the rat. Synapse 10: 131–140, 1992.
OVERTON, P. AND CLARK, D. Burst firing in midbrain dopaminergic neurons.
Brain Res. Rev. 25: 312–334, 1997.
2261
PING, H. AND SHEPARD, P. Apamin-sensitive Ca21 activated K1 channels
regulate pacemaker activity in nigral dopamine neurons. Neuroreport. 73:
809 – 814, 1996.
PING, H. AND SHEPARD, P. Intracellular Ca21 chelation evokes bursting plateau
potential in nigral dopamine neurons in vitro. Soc. Neurosci. Abstr. 23:
1211, 1997.
REGAN, L. J. Voltage-dependent calcium currents in purkinje cells from rat
cerebellar vermis. J. Neurosci. 11: 2259 –2269, 1991.
RINZEL, J. AND ERMENTROUT, G. B. Analysis of neuronal excitability and
oscillations. In: Methods in Neuronal Modeling: From Ions to Networks,
edited by C. Koch and I. Segev. Cambridge, MA: MIT Press, 1998, p.
251–292.
ROBERTSON, S., JOHNSON, J., AND POTTER, J. The time course of Ca21 exchange
with calmodulin, troponin, and myosin in response to transient increases in
Ca. 21. Biophys. J.34: 559 –569, 1981.
ROBLEDO, P. AND FEGER, J. Excitatory influence of rat subthalamic nucleus to
substantia nigra pars reticulata and the pallidal complex: electrophysiolog-
ical data. Brain Res. 518: 47–54, 1990.
ROTH, R. H. AND ELSWORTH, J. D. Psychopharmacology: The Fourth Gener-
ation of Progress (4th ed.). New York: Raven Press 1995, vol. 1, chapt. 21,
Downloaded from http://jn.physiology.org/ by 10.220.33.4 on November 5, 2017
p. 227–243.
RUSSO, R. E. AND HOUNSGAARD, J. Plateau-generating neurones in the dorsal
horn in an in vitro preparation of the turtle spinal cord. J. Physiol. (Lond.)
493: 39 –54, 1996.
SANGHERA, M. AND GERMAN, M.T.D. Electrophysiological properties of mouse
dopamine neurons: in vivo and in vitro studies. Neuroscience 123: 793– 801,
1984.
SCARNATI, E., PROIA, A., CAMPANA, E., AND PACITTI, C. A microiontophoretic
study on the nature of the putative synaptic neurotransmitter involved in the
pedunculopontine-substantia nigra pars compacta excitatory pathway of the
rat. Exper. Brain Res. 46: 470 – 478, 1986.
SCHULTZ, W., ROMO, R., LJUNGBERG, T., MIRENOWICZ, J., HOLLERMAN, J. R.,
AND DICKINSON, A. Reward-related signals carried by dopamine neurons. In:
Models of Information Processing in the Basal Ganglia, edited by J. C.
Houk, J. L. Davis, and D. G. Beiser. Cambridge, MA: MIT Press, 1995,
chapt. 12, p. 233–248.
SEUTIN, V., JOHNSON, S. W., AND NORTH, R. A. Apamin increases NMDA-
induced burst-firing of rat mesencephalic dopamine neurons. Brain Res.
630: 341–344, 1993.
SHEPARD, P. The cellular basis of conditional bursting in mesencephalic
dopamine-containing neurons. Schizophr. Res. 9: 167, 1993.
SHEPARD, P. AND BUNNEY, B. Effects of apamin on the membrane properties of
putative dopamine-containing neurons in vitro. Soci. Neurosci. Abstr. 14:
932, 1988.
SILVA, N., PECHURA, C., AND BARKER, J. Postnatal rat nigrostriatal dopaminer-
gic neurons exhibit five types of potassium conductances. J. Neurophysiol.
641: 262–272, 1990.
TEPPER, J., SAWYER, S., AND GROVES, P. Electrophysiologically identified
nigral dopaminergic neurons intracellularly labeled with hrp: light-micro-
scopic analysis. J. Neurosci. 97: 2794 –2806, 1987.
TURRIGIANO, G., ABBOT, L. F., AND MARDER, E. Activity-dependent
changes in the intrinsic properties of cultured neurons. Science 264:
974 –977, 1994.
YUNG, W., HAUSSER, M., AND JACK, J. Electrophysiology of dopaminergic and
nondopaminergic neurons of the guinea-pig substantia nigra pars compacta
in vitro. J. Physiol. (Lond.) 436: 643– 667, 1991.