Previous Article | Next Article 
The Journal of Neuroscience, April 15, 1998, 18(8):2856-2870
A Reluctant Gating Mode of Glycine Receptor Channels Determines
the Time Course of Inhibitory Miniature Synaptic Events in Zebrafish
Hindbrain Neurons
Pascal
Legendre
Institut des Neurosciences, Université Pierre et Marie Curie,
75252 Paris Cedex 05, France
 |
ABSTRACT |
Miniature IPSCs (mIPSCs) recorded in the Mauthner (M)-cell of
zebrafish larvae have a broad amplitude distribution that is attributable only partly to the functional heterogeneity of
postsynaptic glycine receptors (GlyRs). The role of the kinetic
properties of GlyRs in amplitude fluctuation was investigated using
fast-flow application techniques on outside-out patches. Short
applications of a saturating glycine concentration evoked outside-out
currents with a biphasic deactivation phase as observed for mIPSCs, and they were consistent with a rapid clearance of glycine from the synaptic cleft. Patch currents declined slowly during continuous applications of 3 mM glycine, but the biphasic deactivation
phase of mIPSCs cannot reflect a desensitization process because
paired-pulse desensitization was not observed. The maximum open
probability (Po) of GlyRs was close
to 0.9 with 3 mM glycine. Analyses of the onset of
outside-out currents evoked by 0.1 mM glycine are consistent with the presence of two equivalent binding sites with a
Kd of O.3-O.4 mM. Activation and
deactivation properties of GlyRs were better described with a kinetic
model, including two binding states, a doubly liganded open state, and
a reluctant gating mode leading to another open state. The 20-80%
rise time of mIPSCs was independent of their amplitude and is identical to that of outside-out currents evoked by the applications of a
saturating concentration of glycine (>1 mM). These results
support the hypothesis that GlyR kinetics determines the time course of synaptic events at M-cell inhibitory synapses and that large mIPSC amplitude fluctuations are mainly of postsynaptic origin.
Key words:
glycine receptors; reluctant gating mode; zebrafish
larva; miniature inhibitory synaptic currents; Mauthner cell; glycinergic synapses; channel kinetics
| |
INTRODUCTION |
At individual synapses, the duration
of postsynaptic events and their amplitude fluctuations can depend on
both post- and presynaptic factors, such as the number and the relative
proportions of receptors subtypes per synaptic bouton, their gating
properties, the neurotransmitter time course in the synaptic cleft, and
variations in the numbers of molecule release (Frerking and Wilson,
1996
).
Glycine receptors (GlyRs) mediate inhibitory synaptic transmission in
the vertebrate spinal cord and brain stem (Curtis and Johnston, 1974).
For the Mauthner cell (M-cell) in the hindbrain of the zebrafish
(Danio rerio) larvae, the amplitude distribution of
glycinergic miniature IPSCs (mIPSCs) cannot be resolved by a single
Gaussian curve (Legendre and Korn, 1994). This distribution is skewed
or even multi-modal, as often observed at central synapses (Bekkers et
al., 1990; Edwards et al., 1990; Manabe et al., 1992; Silver et al.,
1992; Tang et al., 1994). Two functionally different GlyRs have been
detected on M-cells (Legendre, 1997), and it has been proposed that
variations in their proportion from one synapse to another can partly
account for the broad amplitude distribution of mIPSCs (Legendre,
1997). However, this receptor heterogeneity is unlikely to be the sole
determinant of the variability of the mIPSCs amplitude (Legendre,
1997). Such a complex distribution can also reflect a variation in the
amount of neurotransmitter released (Frerking et al., 1995), intrinsic
kinetic properties of the postsynaptic receptor (Tang et al., 1994),
and a variable number of postsynaptic receptors from one synapse to
another (Hestrin, 1992). To resolve this issue, the kinetic properties
of GlyRs controlling the efficacy of inhibitory synapses first need to be characterized.
The analysis of the activation time course of ion channels in
outside-out patches using concentration-clamp techniques (Franke et
al.. 1987; Lester et al., 1990) has revealed important features of
intrinsic channel properties for the control of synaptic efficacy (Clements and Westbrook, 1991; Clements et al., 1992; Hestrin, 1992;
Lester and Jahr, 1992; Jones and Westbrook, 1996). It has been
demonstrated that central receptors can be saturated by a released
vesicle (Clements, 1996), whereas the time course of postsynaptic
events is determined by channel kinetics rather than by clearance of
neurotransmitters from the synaptic cleft (Clements, 1996). The
kinetics of glyRs has been studied previously using only steady-state
analysis, which is limited by uncertainties, including receptor
desensitization and possible cooperativity between binding sites
(Colquhoun, 1973).
The goal of this study was to characterize the kinetic properties of
GlyRs that control the time course of mIPSCs. The present report is
based on the analysis of the time course of mIPSCs recorded in the
zebrafish larva M-cell and during fast-flow application of glycine to
outside-out patches. Patches containing GlyRs with a single conductance
state of 40-46 pS were used because this GlyR subtype is dominant in
the M-cell (Legendre and Korn, 1994). The properties of these glyRs
were used to estimate the peak concentration of glycine at the synaptic
cleft and to determine the pre- and postsynaptic origin of the
amplitude fluctuations of mIPSCs.
| |
MATERIALS AND METHODS |
Isolated intact brain preparation. The isolated
intact zebrafish brain was prepared as described before (Legendre and
Korn, 1994). Briefly, the brains of newly hatched larvae were dissected out and glued to a coverslip using a plasma-thrombin embedding procedure (Gähwiler and Brown, 1985). Before the experiments were
started, brain preparations were stored for 15 min in an oxygenated
(95%O2, 5%CO2) bathing solution
containing (in mM): NaCl 145, KCl 1.5, CaCl2 2, MgCl2 1, NaHCO3 26, NaH2PO4 1.25, glucose 10, with the osmolarity
adjusted to 330 mOsm.
Whole-cell and outside-out recordings. Standard whole-cell
and outside-out recordings (Hamill et al., 1981) were achieved under
direct visualization (Nikon Optiphot microscope) on the M-cell and on a
reticular neuron (MiM1) located in the fourth hindbrain rhombomere
(Metcalfe et al., 1986). The isolated brain preparation was perfused
continuously at room temperature (20°C) with the oxygenated bathing
solution (2 ml/min) in the recording chamber (0.5 ml). Patch-clamp
electrodes were pulled either from thin-wall (whole-cell recordings,
1-3 M
) or thick-wall (outside-out recordings, 10-15 M)
borosilicate glass. They were fire-polished and filled with (in
mM): CsCl 135, MgCl2 2, Na3ATP 4, EGTA 10, HEPES 10, pH 7.2. The osmolarity was adjusted to 290 mOsm.
Currents were recorded using an Axopatch 1D amplifier (Axon
instruments, Foster City, CA), filtered at 10 KHz, and stored using a
digital tape recorder (DAT DTR 1201, Sony, Tokyo, Japan). During
whole-cell recordings, the series-resistance (4-10 M) was monitored
by applying 2 mV hyperpolarizing pulses and compensated 50-70%. To
ensure cell dialysis, measurements were made on data obtained at least
3-5 min after the whole-cell configuration was established.
Drug delivery. During whole-cell experiments, drugs were
applied to the preparation via an array of four flowpipes of 400 µm
diameter positioned above the brain. TTX (1 µM) was
dissolved in a perfusate containing (in mM): NaCl 145, KCl
2, CaCl2 2, MgCl2 1, glucose 10, and HEPES 10, pH 7.2; osmolarity was 330 mOsm.
Outside-out single-channel currents were evoked using a fast-flow
operating system (Franke et al., 1987; Lester et al., 1990). Drugs were
dissolved in a control solution containing (in mM): NaCl
145, KCl 1.5, CaCl2 2, MgCl2 1, glucose 10, and
HEPES 10, pH 7.2; osmolarity was 330 mOsm. Control and drug solutions
were gravity-fed into the two channels of a thin-wall glass theta tube (2 mm outer diameter; Hilgenberg) pulled and broken to obtain a tip
diameter of 200 µm. The outside-out patch was positioned (45°
angle) 100 µm away from the theta tubing, to be close to the
interface formed between the flowing control and drug solutions. One
lumen of the theta tube was connected to a manifold with reservoirs filled with solutions containing different glycine concentrations. The
solution exchange was performed by rapidly moving the solution interface across the tip of the patch pipette, using a piezo-electric translator (model P245.30, Physic Instrument) (Clements and Westbrook, 1991). Concentration steps of glycine lasting 1-1000 msec were applied
every 5-10 sec. Exchange times of 20-80% (
0.04 msec) (see Fig. 1)
were determined after the seal was ruptured by monitoring the change in
the liquid junction evoked by the application of a 10% diluted control
solution on the open tip of the patch pipette (see Fig.
1B). The absolute exchange on the patch, however, may result partially from an unstirred layer around the patch. The theoretical limit to the speed of solution change was estimated using
the method published by Maconochie and Knight (1989). Assuming that the
patch has a spherical geometry with a radius r, the velocity profile can be evaluated as:
|
|
with v being the flow velocity and h the
distance to the patch. The time required for diffusion of the
neurotransmitter at a distance h from the patch surface
is:
where D is the diffusion coefficient of the
neurotransmitter. Assuming a patch diameter of 0.5 µm (pipette
resistance 10-15 M) and a diffusion coefficient for glycine close
to 10
5 cm2/sec (Faber et al., 1992),
the estimated absolute exchange time was found to be almost identical
to the experimental full exchange time (0.08 msec).
Analysis of whole-cell currents. Spontaneous synaptic
activity was digitized off-line with a Macintosh IICi computer at 24 kHz using MMII software (GW Instruments). The detection of synaptic events was automatically performed, as described previously (Ankri et
al., 1994; Legendre and Korn, 1994). Two extreme classes of mIPSCs
(50-100 pA and 250-400 pA, respectively) were selected according to
their histogram amplitudes (Legendre and Korn, 1994). Their time
courses were analyzed by averaging 25 isolated single events using
Axograph 3.0 (Axon Instruments) (filter cutoff frequency, 10 kHz), and
the first 100 msec of the decay phase was fitted with a sum of
exponential curves using Axograph 3.0 software. The presence of one or
two exponential components was tested by comparing the sum of squared
errors (SSEs) of the fits (Clements and Westbrook, 1991). The equation
with two exponential components always resulted in a significantly
better fit (SSE 1 exp > SSE 2 exp; paired t test;
p = 0.01; n = 10 cells).
Outside-out patch current analysis. Single-channel currents
were filtered at 10 kHz using an eight-pole Bessel filter (Frequency Devices), sampled at 62.5 kHz (Tl-1 interface, Axon Instruments) and
stored on an IBM AT compatible computer using Pclamp software 6.03 (Axon Instruments). Outside-out currents were analyzed off-line with
Axograph 3.0 software (Axon Instruments).
The open tip current (recorded after the seal was ruptured) used to
measure the exchange time duration usually preceded glycine-evoked currents (see Fig. 1A), as mentioned previously
(Jones and Westbrook, 1995). This caused difficulties in determining
the real initial onset of the outside-out current evoked by the
application of low glycine concentration. I therefore determined the
initial onset of slow responses evoked by intermediate glycine
concentrations (0.003-1.0 mM) from the onset of the
chloride currents evoked by the application of 3 mM
glycine.
The activation time constants of glycine-evoked currents were estimated
by fitting the onset of the responses with a sum of sigmoidal curves
(see Results) using Axograph 3.0 software. Decay time constants were
obtained as described above for the analysis of the mIPSCs decay phase.
For single-channel analysis, open time and closed time durations were
analyzed manually using Axograph 3.0. For display purposes, open time
and closed time histograms show the distributions in log intervals with
the ordinate on a square root scale (Sigworth and Sine, 1987).
Single-channel conductance was determined by constructing
point-by-point amplitude histograms of data segments.
To minimize incorrect assignments in the classification of openings as
short and long events, the method described by Jackson et al. (1983)
was used. A critical time duration t0 was
defined by the relation:
where Nf and
Ns are the relative area of the fast and slow
components, and 
0f and 0s are the mean
open times of short- and long-lived open states. An opening will be
classified as a short event when its duration is
<t0.
To determine whether the length of successive openings was independent
(test of correlation), the runs test was used (Colquhoun and Sakmann,
1985). This test is calculated by converting the opening duration into
a series consisting of the digits 0 (short openings) and 1 (long
openings). The series consist of no 0 values and
n1 1 values (no + n1 = n). A run is defined as a
contiguous section of the series that consists entirely of one or more
0 values or 1 values, T being the number of the runs.
If the series is in random order then the mean of T will
be:
and its variance will be:
The test statistic z is defined as:
A positive correlation between the lengths of events will give
fewer observed runs than the predicted value meanT,
the value of z being smaller than 
2 (Colquhoun and
Sakmann, 1985).
Kinetic modeling programs. To obtain a kinetic model for
GlyR behavior, glycine-evoked currents were analyzed off-line using chemical kinetic modeling programs (Fastflow and Fitfastflow; gift from
J. D. Clements, University of Canberra) on a Power Macintosh (7600/132). This program first calculated the evolution of the number
of channels in each given state for given rate constants. This program
then varied systematically the rate constants to give the minimum sum
of squared errors between the experimental data and a given model
transient (Clements and Westbrook, 1991).
Patch currents and mIPSCs represent the average of 10 or more trials as
specified in the figure legends or the text. Results are presented as
mean values ± SD throughout, unless noted otherwise.
| |
RESULTS |
Two types of M-cell synaptically activated glycine-gated channels
were determined on the basis of their main conductance states and their
pharmacological properties (Legendre, 1997). Nonstationary kinetic
analysis was performed on the GlyR subtype characterized by a single
conductance state of 40-46 pS (Legendre and Korn, 1994; Legendre,
1997). Patches containing channels with multiple subconductance levels
in response to the application of low agonist concentrations were
omitted.
mIPSCs and glycine-evoked chloride currents have similar
time courses
Glycine receptor (GlyR) currents in outside-out patches of M-cell
(Fig. 1B) were evoked
by rapid switches between control solution and a glycine-containing
solution (see Materials and Methods). A rundown of the response was not
observed at the application frequency used (0.1-0.2 Hz). The absolute
solution exchange time was ~0.08 msec (see Materials and Methods)
(Fig. 1B) in all experiments. A series of trials
(10-15) separated by 5 sec or more were used to generate an ensemble
average trace. The decay phase of the outside-out currents evoked by
short steps (1-2 msec) into 3 mM glycine had two
components (Fig. 1C), with time constants of
fast = 4.9 ± 1.3 msec (66.3 ± 6.1% of total
amplitude; n = 10) and slow = 39.4 ± 12.4 msec (n = 10), respectively. These responses were completely blocked by the application of 0.1 µM
strychnine (data not shown).
View larger version (28K):
[in this window]
[in a new window]
|
Figure 1.
mIPSCs and glycine-evoked patch currents have a
closely similar decay. A, Example of a chloride current
evoked by a brief pulse (1.3 msec) of 3 mM glycine. Note
that the channels open in bursts at the end of the deactivation phase
(insert). B, Open tip current (top
trace) recorded during a 1.3 mM step into a 10%
diluted control NaCl solution. Bottom trace is rising
phase of averaged patch currents (n = 15) evoked by
3 mM glycine application (1.3 msec;
Vh = 50 mV). The onset of the open tip and
the patch currents were aligned to compare their time courses (see
Results). Note that the time to peak (ttp) of the open
tip current (full exchange time = 0.08 msec) was shorter than the
ttp of the patch current. C, The decay of the averaged
patch currents (n = 15) evoked by a step into 3 mM glycine (1.3 msec) was accurately fitted by a
biexponential curve [fast = 5.5 msec (80.4%);
slow = 29.5 msec]. D, Example of
averaged mIPSCs (50-150 pA; n = 25) with a
biphasic decay with time constants of fast = 5.3 msec
(76.6%) and slow = 42.6 msec.
|
|
mIPSCs also had two decay phase components (Fig. 1D).
In our previous studies (Legendre and Korn, 1994, 1995; Legendre,
1997), these two decay phase components for glycinergic mIPSCs were not detected because my analyses were limited by automatically fitting the
first 10 msec of the deactivation phase. Although the amplitude of
mIPSCs varied strongly (Legendre and Korn, 1994, 1995; Legendre, 1997)
(Fig. 2A), their decay
time constants were independent of the amplitude (Legendre and Korn,
1994, 1995; Legendre, 1997) (Fig. 2B,C). For mIPSCs
having an amplitude ranging from 50 to 150 pA, fast = 4.2 ± 0.9 msec (72.6 ± 8.99%) and slow = 29.3 ± 6.6 msec (n = 10 cells), whereas 250-400
pA mIPSCs had a fast and slow of 4.5 ± 1.1 msec (74.9 ± 12.7%) and 26.4 ± 6.5 msec (n = 10 cells), respectively. The small and large
mIPSCs decay time constants and the relative amplitudes of the two
exponential curves were not significantly different (unpaired
t test; p = 0.05; n = 10 cells). The mIPSC decay time constants and relative amplitudes of the
two exponential components were also not significantly different from
those obtained on outside-out current evoked by a brief pulse (1-2
msec) of 3 mM glycine (unpaired t test;
p = 0.05). These results are consistent with the
hypothesis that rebinding of glycine did not significantly account for
glycinergic mIPSC duration. Accordingly, the rate of clearance of free
neurotransmitter from the synaptic cleft is likely to be fast at
central synapses (Faber et al., 1992; Clements, 1996).
View larger version (21K):
[in this window]
[in a new window]
|
Figure 2.
Time course of mIPSCs is independent of their
amplitude. A, Amplitude histogram of mIPSCs recorded on
a Mauthner cell in the presence of 1 µM TTX
(n = 808; bin width = 8 pA;
Vh = 50 mV). B,
Superimposed averaged mIPSCs of 50-150 and 250-400 pA, respectively
(n = 25; Vh = 50 mV).
Same cell as in A. C, Normalized averaged
mIPSCs shown in B. These two miniature events have
similar time courses. They were accurately fitted (not shown) with the
sum of two exponential curves with fast = 4.5 msec
(56.5%) and slow = 23 msec.
|
|
Steady-state concentration-response curve and maximum
open probability
To understand how mIPSCs are generated after exocytosis, we first
need to know the efficacy of glycine in activating its receptor channels. Figure 3A
illustrates chloride currents evoked by the application of glycine at
different concentrations. The duration of the applications was adjusted
to obtain steady-state responses. Three or four different glycine
concentrations were usually tested with each patch, and the response
amplitude was normalized to that obtained by the application of 3 mM glycine. Figure 3B shows the
concentration-response curve obtained from 11 experiments. The
normalized data were fitted using a single binding isotherm of the
form:
where I/Imax are the normalized
response amplitude, EC50 is the glycine concentration
[glycine] producing 50% of the maximal response and h is
the Hill coefficient. This fit produced an EC50 of 0.054 mM and a Hill coefficient of 1.56. This is closely similar to values obtained in other preparations using whole-cell recording techniques (Akaike and Kaneda, 1989; Zhang and Berg, 1995). A maximum
response was obtained for [glycine] >1 mM, suggesting that applications of millimolar glycine would lead to a near saturation of GlyRs.
View larger version (30K):
[in this window]
[in a new window]
|
Figure 3.
Concentration-response curve of glycine-evoked
currents. A, Responses of a patch to step application of
different concentrations of glycine. The duration of the application
was adjusted to obtain a steady-state amplitude of the responses. Each
trace represents the average of 10 responses. B,
Concentration-response plot of data obtained in 11 patches. Response
amplitudes were normalized to that obtained in the presence of 3 mM glycine. Each point is the average of 5-11
measurements. Data points were fitted with a single binding isotherm
(see Results). C, Superimposed averaged traces of the
first five and the last five responses from a set of 45 currents evoked
by identical step application of 3 mM glycine (2 msec;
Vh = 50 mV). D,
Variance-amplitude plot computed for 45 current transients (same patch
as in C). The amplitude and the variance were computed
for a period of 150 msec starting at the peak of the averaged response.
The curve represents the fitted model  2 = iI (I2/N) (see
Results) with i = 1.72 pA and n = 49.
|
|
At this concentration a maximum number of the available receptors would
be opened. To estimate the maximum open probability (Pomax), a series of responses
(20-45) to identical 2 msec steps of 3 mM glycine were
recorded (Fig. 3C,D). In each experiment the possibility of
rundown of the response amplitude and time-dependent changes in the
rate of deactivation were examined by comparing the averaged trace of
the first five responses to the averaged trace of the last five. In the
10 patches used for this analysis, little or no rundown of the response
amplitude or changes in the rate of deactivation was observed (Fig.
3C). Nonstationary variance analysis (Sigworth, 1980) was
used to estimate Pomax, assuming that the patch current is generated by superimposition of openings originating from independent GlyRs. The point-by-point relationship (bin interval = 0.016 msec; filter cutoff frequency = 2 kHz)
between variance (Vart) and current
(It) of the decay phase was fitted by:
where i is the elementary current,
It the macroscopic current, and N is
the total number of available receptor channels in the patch. The
relation between Vart and
It is clearly parabolic, justifying the fit
procedure (Fig. 3D). The number of estimated available glyRs
strongly varied from one patch to another and ranged from 19 to 160 channels. In the 10 patches tested, the maximum current was 0.91 ± 0.06 of the predicted maximal current to obtain
Vart = 0 (100% of the available GlyRs being
opened), indicating that Pomax of
GlyR at saturation is high and would be >0.9.
Desensitization of glycine-gated channels does not control the
duration of inhibitory synaptic response
Biphasic decay of GlyR responses after a brief application of
agonist might result from desensitization, as demonstrated for AMPA and
GABAA receptor channels (Hestrin, 1992; Jones and
Westbrook, 1995). In this case, a short agonist pulse is sufficient to
drive the channels into a desensitized state (Hestrin, 1992; Jones and Westbrook, 1995). It was proposed that multi-liganded receptors can
enter rapidly equilibrating desensitized states, which predicts a
biphasic deactivation phase near saturation of the receptors (Jones and
Westbrook, 1995).
To test the role of desensitization during the decay phase, long pulses
(1 sec) of glycine (3 mM) were first applied to evoke chloride currents. As shown in Figure
4A, this current
declined slowly. In the three patches tested, the extent of
desensitization at steady state ranged from 66.7 to 70% (mean = 68.1%) of the maximum current. GlyR desensitized with two exponential
components (Fig. 4A). fast ranged from
22 to 66 msec (mean = 38.7 msec; n = 3), and
slow ranged from 514 to 630 msec (mean = 569 msec; n = 3). Although these two components had time
constants closely similar to those for GABAA receptors
(Jones and Westbrook, 1995), the proportion of the fast component is
smaller. Effectively, fast represented 5.3-9.0% of the
total current (mean = 7.3%; n = 3) only (26% for
GABA) (Jones and Westbrook, 1995), whereas relative amplitude of
slow ranged from 58.6 to 62.4% (mean = 60.8%;
n = 3) of the total current. These results show that
33% of the glycine-evoked current does not desensitize. Slow
desensitization kinetics was not further analyzed.
View larger version (24K):
[in this window]
[in a new window]
|
Figure 4.
Desensitization of currents evoked by step
applications of glycine. A, Averaged traces of five
responses to long step applications of 3 mM glycine (0.63 and 1 sec). Desensitization to prolonged applications of glycine was
described by the sum of two exponential curves and a constant term for
steady-state current. Desensitization time constants were
fast = 67 msec and slow = 564 msec. Fast
desensitization was 7.6% of the total current amplitude; the slow
component represented 62.4% of the total current. B,
Superimposed averaged traces of responses (5 each) evoked by paired
pulses of 3 mM glycine (2 msec). C,
Interpulse intervals were 5, 15, 30, 50, 70, 100, 135, and 170 msec.
Note that paired pulses of glycine did not produce desensitization
(Vh = 50 mV).
|
|
These results are consistent with the hypothesis that a short glycine
pulse, mimicking synaptic glycine release, could not significantly
promote a rapidly entered and exited desensitized closed state.
Accordingly, desensitization would not be apparent when chloride
currents were evoked by short glycine pulses. Paired-pulse experiments
confirmed this hypothesis (Fig. 4B,C). When paired glycine applications (3 mM; 2 msec) were given at variable
intervals (Fig. 4B), the second application did not
evoke a smaller peak current than the first application. Similar
results were obtained on three other patches.
Kinetic analysis performed on one glycine-gated channel
The analysis of opening and closing behaviors of a single receptor
channel activated by a saturating concentration of agonist gives
important information on the channel kinetics that determine the time
course of postsynaptic responses. In a patch pulled from a reticular
neuron (MiM1) (Metcalfe et al., 1986) located in the same rhombomere as
the M-cell, we successfully recorded one glycine-gated channel
activated by short applications (2 msec) of 3 mM glycine (Fig. 5A). This channel had a
single conductance state of 44 pS as observed for
hetero-oligomeric-like GlyRs of M-cell (Legendre, 1997).
View larger version (38K):
[in this window]
[in a new window]
|
Figure 5.
Patch with a single GlyR activated by short step
applications of 3 mM glycine. A, Example of
10 responses evoked by successive 2 msec glycine pulses (cutoff filter
frequency, 3 kHz; Vh = 50 mV). Note the
opening failure in epoch 5. Bottom trace: the averaged
current of 26 responses has a decay that is accurately fitted by a
bi-exponential curve with time constants of fast = 5.2 msec (65%) and slow = 27 msec. Open time
(B) and closed time (C)
duration histograms are shown as a function of a log intervals with the
ordinate on a square root scale.
|
|
As shown in Figure 5A, GlyR opens in bursts in response to a
short-step concentration of 3 mM glycine, suggesting that
this channel could reopen several times before the dissociation of the
agonist occurs. Only one failure was observed in 26 trials. Clusters of
short- and long-lived openings were also observed at the end of the
deactivation phase of outside-out currents evoked by short applications
of 3 mM glycine on patches containing several GlyRs (Fig.
1A).
Averaged traces revealed a deactivation phase that could be fitted with
the sum of two exponential curves, with time constants closely similar
to those calculated from glycine-evoked currents obtained on patches
pulled from the M-cell (Fig. 5A). Single openings and
closures were detected using a filter cutoff frequency of 4 kHz
(six-pole Bessel filter). The open time histograms could be fitted by
the sum of two exponential curves (Fig. 5B). The two mean
open times were o1 = 0.5 msec and o2 = 2.4 msec, which is similar to that observed for M-cell GlyR (Legendre
and Korn, 1994; Legendre, 1997), and their relative areas were 0.58 and 0.42, respectively. A single closed time was detected with
c1 = 0.42 msec (Fig. 5C). Longer closures
were also observed (4-5 msec), but they were too infrequent to be
fitted.
To determine whether long and short openings occurred randomly after a
concentration jump, each sweep was reexamined. Such analysis gives
information about the channel-gating mode (Colquhoun and Hawkes, 1987).
The first order of occurrence of a particular opening (long or short)
in response to a short step to a saturating concentration of agonist
was investigated first. To minimize the number of incorrect assignments
of openings as short- and long-lived events, we calculated a critical
time duration (t0 = 1.2 msec) as proposed by
Jackson et al. (1983) (see Materials and Methods). In 23 of the 25 sweeps analyzed, the first opening had a long duration (3.56 ± 1.73 msec). When a long opening occurred first, it could be followed by
one or two other long openings. These long openings were followed by
clusters of short openings (2-10 events) separated by rare long
events. These results suggest that the opening rate constant of the
long-lived open state is considerably faster than that of the
short-lived open state, or that the short-lived open state is not
directly linked to the fully liganded closed state. For example, these
two openings can reflect mono-liganded and di-liganded closed states,
as proposed previously (Bormann, 1990; Twyman and MacDonald, 1991;
Legendre and Korn, 1994). If the latter is true, open times within a
burst should be correlated (Colquhoun and Hawkes, 1987).
Only one closed time was detected, and therefore it was not possible to
determine whether the clusters of short-lived openings are true
individual bursts of short openings. This could be attributable to the
limitation of the record resolution if different classes of closures
have mean durations that are similar. In the absence of any evidence of
different mean closed times, correlation analysis of openings was
performed using the runs test (see Materials and Methods), considering
that reopenings of GlyR after a pulse of glycine represent a single
burst. The runs test with a critical time of 1.2 msec (see above) gave
a standard Gaussian deviate of z = 2.04. This is the
limit to affirm that runs of short and long openings exist.
Rise time of glycine-evoked currents
As shown in Figure
6A, the rise time of
glycine-evoked currents decreased when agonist concentration was
increased to reach a minimum at a glycine concentration of 3 mM. An increase in the glycine concentration to 10 mM did not change the activation phase (data not shown),
which suggests that the opening and closing rate constants became the
limiting factors for GlyR activation kinetics (Colquhoun and Hawkes,
1995). Because the initial onset of the response evoked by a step into
a glycine concentration >3 mM was likely to be distorted
by the agonist solution exchange rate (Fig. 1), the first 5% of the
onset was not used for the analysis. The 5-100% rise time of the 3 mM glycine-evoked responses was well fitted by a single
exponential curve, with a time constant (on)
close to 0.1 msec (0.121 ± 0.03 msec; n = 11). An
increase in the glycine concentration to 10 mM gave a
on value of 0.115 msec (n = 3).
View larger version (18K):
[in this window]
[in a new window]
|
Figure 6.
Activation time course of glycine-evoked
responses. A, Averaged traces of patch currents
(n = 10) showing the activation phase of the
responses evoked by the application of 0.03, 0.1, 0.3, 1.0, and 3.0 mM glycine. Traces were normalized to their maximum
amplitude. B, Normalized averaged current of 15 responses evoked by step applications of 0.1 mM glycine.
Note that the activation phase has two components (see Results) with
time constants of on1 = 1.9 msec (71%) and
on2 = 8.4 msec. Only every 25th data point is plotted
for clarity. C, First 5 msec onset of an ensemble
average (n = 15 traces) after a step application of
0.1 mM glycine ( ) is plotted with logistic equations for
one, two, and three binding sites (see Results). Only every fifth data
point is plotted for clarity. D, The sum of squared
errors between the ensemble average traces and each of the three
equations was calculated over the first 5 msec of the onset. For eight
patches, data were better fitted with a two binding sites
equation.
|
|
The analysis of the onset of the current evoked by the application of a
low concentration of agonist can give information concerning the number
of binding sites (Clements and Westbrook, 1991). Steps into a
concentration of agonist close to the EC50 (as determined
in Fig. 3B) will evoke responses for which the rising phase
reflects the rate of agonist binding. The number of glycine binding
sites on GlyRs was therefore determined by analyzing the onset of the
responses evoked by the application of 0.1 mM glycine.
Because the exchange solution was delayed from the application artifact
(see above), the origin of the onset was determined from that of the
current induced by the application of 3 mM glycine. A step
into 0.1 mM glycine evoked responses with a clear sigmoidal
onset (Fig. 6B), which is consistent with the presence of more than one binding site per GlyR (Clements and Westbrook, 1991). Although at an intermediate concentration (0.1 mM) the activation was clearly biphasic, its slow component
did not significantly disturb the initial onset (Fig.
6B). To determine further the number of binding sites
we analyzed the degree of sigmoidicity of the onset by fitting the
first 5 msec of the rising phase (bin interval = 16 µsec; filter
cutoff frequency = 10 kHz) using the following equation:
|
|
where a is the maximum amplitude of the response,
act its activation time constant, and b the
degree of sigmoidicity. Three models were tested with a degree of
sigmoidicity of 1, 2, or 3 (Fig. 6C) by comparing SSEs
(Clements and Westbrook, 1991). A total of eight glycine responses
obtained on each of eight different patches were analyzed. SSEs
calculated over the first 5 msec based on the degree of sigmoidicity of
the curves are shown in Figure 6D. The equation with
a degree of sigmoidicity of 2 always resulted in a significantly better
fit (SSEs 1 site or SSEs 3 sites > SSEs 2 sites; paired
t test; p = 0.01). These results provide
evidence for the presence of two glycine binding sites per GlyR, as
proposed previously on spinal cord neurons (Sakmann et al., 1983;
Bormann, 1990; Takahashi and Momiyama, 1991; Walstrom and Hess, 1994). Although my experimental data were well fit by this model, it did not
take into account a possible allosteric interaction between the two
binding sites. This cannot be excluded completely, but such a mechanism
would result in a deviation from the predicted sigmoidal onset of the
responses (Clements and Westbrook, 1991), which was not observed in
these experiments.
As mentioned above, the responses evoked by the application of
intermediate concentrations of glycine had a biphasic rising phase, and
it was better fitted (Fig. 6B) with a sum of two
sigmoidal functions of the form:
where a and b are the relative
amplitudes of the two components and fast and
slow are the corresponding time constants. An increase
in the degree of sigmoidicity of the second component to 3 did not
increase the goodness of the fit. fast and
slow were strongly dependent on the concentration (Fig.
7A), suggesting that
agonist binding is the rate-limiting step for [glycine] <3 mM. The relative areas of these two components were also
dependent on the concentration that was applied (Fig.
7B). Slow component appeared for [glycine] <3
mM, and its relative area increased as the concentration
decreased. The responses evoked by a step into 0.01 mM
[glycine] had a fast = 12.5 ± 3.75 msec and a
slow = 116.4 ± 29.4 msec, with relative areas of
55 ± 10% and 45 ± 10%, respectively
(n = 6).
View larger version (15K):
[in this window]
[in a new window]
|
Figure 7.
A, Plots of the current
rising rates (1/on) versus glycine concentration
for the two rising phase components were fitted to the equation
1/on =  +  ([glycine]2/[glycine]2 + EC502) (see Results). Each point
represents the averaged data of 3-11 experiments. was 316.3 sec1 and was 8938 sec1 for
on1, and they were 41.9 and 2299 sec1 for on2. Note that EC50
of the two components has similar values: 0.96 and 0.72 mM,
respectively. B, Plot of the relative proportions of the
fast () and the slow (O) rising phase components versus
concentration. The relative proportion of the slow component decreased
when the concentration of glycine was increased. Each point is the
average of 3-11 measurements.
|
|
The plot of patch-current rise rate for the fast
(1/fast) and slow component
(1/slow) versus concentration (Fig.
7A) was fitted to the following equation:
where is an approximation of the closing rate constant, is
an approximation of the opening rate constant, [glycine] is the
glycine concentration, n is the number of binding sites
(two, as described above), and EC50 is the concentration of
glycine that gave 50% of the theoretical opening rate constant
(Colquhoun and Hawkes, 1995). Plots of fast and slow rise rates gave a
closing rate constant of 316 and 41.9 sec1 and an opening
rate constant of 8938 and 2299 sec1, respectively. These
two plots gave closely similar EC50 values (0.96 mM and 0.72 mM, respectively), which make it
unlikely that these two activation phases resulted from the presence of
binding sites with different affinities.
These results are inconsistent with the hypothesis that mono-liganded
opening can participate significantly in the activation phase of GlyR
responses. Mono-liganded openings should add a linear component to the
onset of GlyR responses, which should begin to rise instantaneously
(Clements and Westbrook, 1991), and its relative proportion should have
increased when the glycine concentration applied was decreased. This
was not apparent in my experiments, which confirms that at glycine
concentration >0.003 mM, mono-liganded openings will not
be frequently observed (Legendre and Korn, 1994).
A kinetic model for glycine-gated channels
To understand how a biphasic deactivation of patch current and
mIPSCs can occur without accumulation of channels in a fast desensitized state, we tested two different models of GlyR channel gating by fitting experimental traces using chemical kinetic modeling programs (Clements and Westbrook, 1991) (also see Materials and Methods).
The better model must predict a biphasic deactivation, a biphasic
rising phase of the responses evoked by the application of low glycine
concentrations, and no paired-pulses desensitization. Finally it must
be consistent with the fit of the amplitude versus concentration and
the measured Pomax.
As a basic kinetic scheme, the model must possess two sequential
binding steps (yielding A + AC and A2C) according to the shape of the
rising phase of the outside-out currents (Fig. 6). The slow
desensitization process was not taken into account because it is
improbable that it affects the decay of responses evoked by short
glycine applications. Two open states (O1 and O2) also need to be
incorporated, because two mean open times were detected during
single-channel recording (Fig. 5). These two mean open times are
closely similar to those published previously for the postulated
mono-liganded and doubly liganded open states (Legendre and Korn,
1994).
Although short openings appear to be grouped in clusters and occur
mainly during the deactivation phase of outside-out currents, the runs
test did not permit rejection of the hypothesis that successive
openings are not correlated (see above). Two types of models can
therefore be envisioned (Colquhoun and Hawkes, 1987) that can give
outside-out current with a biphasic rising phase in response to the
application of intermediate concentrations of glycine. The first model
does not give correlation between openings within a burst, and the
other one allows interconversion between the two opening modes (short
and long openings).
The first model used is characterized by two open states linked
independently to the doubly liganded closed state:
|
|
A model with two sequential opening states linked to the doubly
liganded closed state (A2C
O1O2) is unlikely because it will never
give a response with two exponential decay components.
The second model is characterized by two open states linked to two
closed states: the doubly liganded closed state (A2C) and a rapidly
equilibrating closed state (A2C*) linked to A2C (Fig. 8A). This model is
very similar to the reluctant gating mode model proposed for the N-type
calcium channel in bullfrog sympathetic neurons (Bean, 1989; Elmslie et
al., 1990; Boland and Bean, 1993; Elmslie and Jones, 1994). A model
with a mono-liganded opening state as proposed for the nicotinic
acetylcholine receptors (Colquhoun and Sakmann, 1985) is unlikely for
GlyRs because this mono-liganded opening needs to have a fast opening
rate to allow long bursts of short openings during deactivation (with
respect to the off rate: koff). Such
an opening rate will give a linear component to the onset of GlyR
responses below EC50 (Clements and Westbrook, 1991), which
was not observed for [glycine] >0.003 mM.
View larger version (19K):
[in this window]
[in a new window]
|
Figure 8.
A, A Markov model reproducing the
gating properties of the glycine-activated channel of the zebrafish
M-cell. This model possesses two sequential equivalent agonist binding
steps, the doubly liganded closed state A2C providing access to a
reluctant closed state (A2C*). These two closed states also provide
access to two doubly liganded independent open states. B,
C, Example of the activation and the decay phases of an
averaged trace of 15 responses to 1.3 msec step applications of 3 mM glycine fitted by a kinetic model with two open states
linked to the di-liganded closed state (model 1) and by
the the kinetic model shown in A (model
2). Note that model 1 cannot properly fit experimental data
(see Results). With model 2, a good fit was obtained
with kon = 8 µM1 · sec1,
koff = 2400 sec1,
1 = 738 sec1, 1 = 8938 sec1, 2 = 1300 sec1,
2 = 2610 sec1, d = 990 sec1, and r = 180 sec1.
|
|
The ability of the models to fit experimental data obtained by 1-2
msec steps into 3 mM glycine (including an absolute
solution exchange time of 0.08 msec) was first tested. Because long
openings occurred first after a concentration jump, it is likely that
their opening rate constant (1) determines the
rising phase of the outside-out current. The opening rate constant
(1 = 8938 sec1) was fixed based on the
measurement obtained from the plot of the rising phase versus
concentration (Fig. 7). The closing rates (1 and
2) were determined from the measured open dwell
time constants published previously for M-cell GlyRs (Legendre and Korn, 1994; Legendre, 1997). 2 was set to 1200-1400
sec1. An opening rate of 8000-9000 sec1
predicts very short closures within a burst that cannot be clearly detected using a filter cutoff frequency of 2-4 kHz, as used for stationary single-channel analysis (Bormann, 1990; Twyman and MacDonald, 1991; Legendre and Korn, 1994) (Fig. 5). The long mean open time that occurs first after a jump in a high glycine
concentration will therefore be overestimated (Colquhoun and Hawkes,
1995). As a first approximation, we first fixed the closing rate
1 to 500-700 sec1. The dissociation rate
(koff), the opening rate
2, the number of channels and in some cases
1, and for the second model, the rate
(d) and the rate (r)
between A2R and A2R* were free parameters. The association rate
constant (kon) was first arbitrarily
fixed to 5-10 106
M1 · sec1, a value that is
very similar to that described for NMDA and GABAA receptors
(Clements and Westbrook, 1991; Jones and Westbrook, 1995). When
kon was changed from 1 to 15 106
M1 · sec1 it did not
significantly modify the optimal calculated values of the free
parameters for a glycine concentration of 3 mM. The fitting
procedure systematically varied the free parameters until it reached
the minimum SSEs between the simulated and the experimental responses
over the 80 msec of the current transients (Clements and Westbrook,
1991).
The first model never gave a good fit of the experimental data (Fig.
8C), even with 1 and 2 as
free parameters. A significantly better fit was always obtained using
the reluctant gating mode model (SSEs of model 1 > SSEs of model
2; paired t test; p = 0.001; n = 10), with averaged rate constants of
koff = 1452.6 ± 252.7 sec1, d = 536.5 ± 95.7 sec1, r = 136.1 ± 65.3 sec1, 2 = 2889 ± 595.7 sec1 parameter (n = 10; mean ± SEM), and 1 = 680 ± 105.4 sec1 when
it was set as a free parameter (Fig. 8B,C). The
association rate constant (kon) was
then adjusted by fitting the reluctant gating mode model to current
transients evoked by steps into 0.1 mM glycine
concentration. The rate kon,
koff, d, and
r were set as free parameters. A good fit was obtained
with averaged optimum rate constants of kon = 4.7 ± 0.5 106
M1 · sec1,
koff = 2005.2 ± 453.8 sec1, d = 329.8 ± 124.9 sec1, and r = 122.3 ± 37.9 sec1 (n = 6; mean ± SEM).
Calculation of the microscopic dissociation constant
(Kd) gives values ranging from 0.3 to 0.4 mM, which is consistent with previously reported kinetic
analysis of hetero-oligomeric-like GlyRs from spinal cord neurons
(Walstrom and Hess, 1994). In this study, the microscopic
Kd was found to be 0.380 mM (Walstrom and Hess,
1994).
As shown in Figure 8B,C, the reluctant gating
mode model is able to account for the observed experimental data. It
predicts a bi-exponential decay of current transients evoked by steps
into 3 mM glycine, and it gives transients with two rising
phase components when the responses were evoked by <1 mM
glycine applications (Fig. 9A). This model does not
generate paired-pulse desensitization (Fig. 9B), but it
predicts a fast desensitized component of low amplitude (Fig.
9C) and gives responses with concentration-independent decay times (Fig. 9D). The theoretical response
has a Pomax close to 0.9 (Fig.
9E) and a concentration-response curve (Fig.
9F) with an EC50 value of 0.089 mM and a Hill coefficient of 1.47, very similar to the
experimental data (Fig. 3B).
View larger version (27K):
[in this window]
[in a new window]
|
Figure 9.
A, This model predicts two
components of the activation phase of responses evoked by step into <1
mM glycine. In this example the optimum fit of the averaged
trace of 15 responses to long step application of 0.1 mM
glycine was obtained with kon = 7 µM1 · sec1,
koff = 3600 sec1,
1 = 680 sec1, 1 = 8938 sec1, 2 = 1300 sec1,
2 = 3180 sec1, d = 814 sec1, and r = 270 sec1.
The experimental trace had two rising phase components with
on1 = 1.6 msec (64%) and on2 = 7.5 msec
(see Results for the fit procedure and the equations used). In this
example, the model predicts rising time constants of on1 = 1.55 msec (62.5%) and on2 = 7.2 msec.
B, Simulated paired-pulse responses generated by the
(Figure legend continued) kinetic model. Rate parameters used were
kon = 5 µM1 · sec1,
koff = 1500 sec1,
1 = 680 sec1, 1 = 8938 sec1, 2 = 1300 sec1,
2 = 3180 sec1, d = 540 sec1, and r = 140 sec1). Note that desensitization does not occur.
C, This model predicts a small desensitization component
during steady-state application of glycine. D, It also
generates responses with a concentration-independent decay time.
E, Maximum open probability of simulated responses
([glycine] = 3 mM; N = 44, ichannel = 2.2 pA). The variance-amplitude
plot was computed for 15 generated transient currents. The fit of this
plot gave a ichannel of 1.72 pA, an
N of 49, and a Pomax
of 0.93. F, Concentration-response curve of simulated
glycine-evoked currents. Fit of the theoretical data points using a
single binding isotherm gave an EC50 value (0.089 mM) and a Hill coefficient (1.47) in good agreement with
the experimental results.
|
|
Using simulated pulses of 1 msec (solution exchange time constant = 0.08 msec) of 3 mM glycine, I examined the consequences of varying the dissociation rate
(koff) and the on
(d) and off (r) rates
linking the doubly liganded closed states (A2C and A2C*). The
consequences of varying d and r rate
constants, the d/r ratio being constant,
was also analyzed.
Glycine current decay can be shaped by a precise balance between these
rates. A reduction in the unbinding rate constant from 18,000 to 500 sec1 increased both decay times and increased the
relative amplitude of the slow component (Fig.
10A). This is
consistent with the idea that slowing the unbinding rate will promote
reopening of the channel and will increase the probability of falling
into the second opening mode via the closed state A2C*. Increasing the relaxation rate d from 140 to 2040 sec1
increases the probability for the channel to open into the second mode
(A2C*O2), which results in an increase of the slow time constant and
an increase of its relative amplitude (Fig. 10B). In this case, the fast decay time constant decreases to a smaller extent, which reflects a decreased probability for the channel to
reopen directly from the doubly liganded closed state.
View larger version (20K):
[in this window]
[in a new window]
|
Figure 10.
The model parameters were used (see Fig.
9B) to examine the effect of varying the unbinding rate
koff, the rate d and
the rate r on the deactivation phase, and the amplitude
of glycine-evoked responses. A, Reducing
koff dramatically decreased the response
duration. B, C, Reducing the rate
constant d or r primarily affects the
slow decay phase component. D, Increasing rates constant
d and r, the
d/r ratio being constant, induces a
progressive lost of the biphasic shape of the deactivation phase.
Varying these binding rates, however, had little or no effect on the
response amplitudes.
|
|
The reluctant gating mode model also predicts that increasing the dwell
time (r) in A2C* will promote bursts of openings
from O2 and prolong the slow decay component. As shown in Figure
10C, varying the recovery rate r from 20 to 340 sec1 gives little variation of the fast decay
time, whereas the relative amplitude of the slow component is increased
with its decay time being dramatically decreased. The decay phase
becomes mono-exponential at r 
340 sec1 with a decay time constant of
toff = 7.9 msec. In this case, the burst of
openings occurring from A2O and A2O* will have similar durations.
Although rate constants d and r can vary
from patch to patch (1- to 10-fold), their ratio remained relatively
more constant and ranged from 1.5 to 5.5 (3.74 ± 1.35 SD;
n = 16). An increase in d (from 135 to 2160 sec1) and r (from 35 to 560 sec1) with d/r constant
decreases the two decay times and increases the relative area of
the slow component (Fig. 10D). An increase in
these rates to d = 2160 sec1 and
r = 560 sec
Top
Abstract
Introduction
![I/I<SUB><UP>max</UP></SUB>=1/(1+(<UP>EC</UP><SUB>50</SUB>/[<UP>glycine</UP>])<SUP><UP>h</UP></SUP>),](/content/vol18/issue8/fulltext/2856/img007.gif)
![1/&tgr;<SUB><UP>on</UP></SUB>=&agr;+&bgr;×([<UP>glycine</UP>]<SUP><UP>n</UP></SUP>/[<UP>glycine</UP>]<SUP><UP>n</UP></SUP>+<UP>EC</UP><SUB>50</SUB><SUP><UP>n</UP></SUP>),](/content/vol18/issue8/fulltext/2856/img011.gif)
