Department of Pharmacology and Cancer Biology and Department of Neurobiology, Duke University Medical Center, Durham, North Carolina 27710
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Molnár, Péter and J. Victor Nadler. Mossy Fiber-Granule Cell Synapses in the Normal and Epileptic Rat Dentate Gyrus Studied With Minimal Laser Photostimulation. J. Neurophysiol. 82: 1883-1894, 1999. Dentate granule cells become synaptically interconnected in the hippocampus of persons with temporal lobe epilepsy, forming a recurrent mossy fiber pathway. This pathway may contribute to the development and propagation of seizures. The physiology of mossy fiber-granule cell synapses is difficult to characterize unambiguously, because electrical stimulation may activate other pathways and because there is a low probability of granule cell interconnection. These problems were addressed by the use of scanning laser photostimulation in slices of the caudal hippocampal formation. Glutamate was released from a caged precursor with highly focused ultraviolet light to evoke action potentials in a small population of granule cells. Excitatory synaptic currents were recorded in the presence of bicuculline. Minimal laser photostimulation evoked an apparently unitary excitatory postsynaptic current (EPSC) in 61% of granule cells from rats that had experienced pilocarpine-induced status epilepticus followed by recurrent mossy fiber growth. An EPSC was also evoked in 13-16% of granule cells from the control groups. EPSCs from status epilepticus and control groups had similar peak amplitudes (~30 pA), 20-80% rise times (~1.2 ms), decay time constants (~10 ms), and half-widths (~8 ms). The mean failure rate was high (~70%) in both groups, and in both groups activation of N-methyl-D-aspartate receptors contributed a small component to the EPSC. The strong similarity between responses from the status epilepticus and control groups suggests that they resulted from activation of a similar synaptic population. No EPSC was recorded when the laser beam was focused in the dentate hilus, suggesting that indirect activation of hilar mossy cells contributed little, if at all, to these results. Recurrent mossy fiber growth increases the density of mossy fiber-granule cell synapses in the caudal dentate gyrus by perhaps sixfold, but the new synapses appear to operate very similarly to preexisting mossy fiber-granule cell synapses.
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
A common feature of temporal lobe epilepsy
(Babb et al. 1991; Franck et al. 1995
;
Houser et al. 1990
; Represa et al.
1989
; Sutula et al. 1989
) and of animal
models of epilepsy (Buckmaster and Dudek 1997b
;
Mello et al. 1993
; Nadler et al. 1980
;
Stanfield 1989
; Sutula et al. 1988
) is
the development of a dense recurrent mossy fiber pathway in the
hippocampal formation. Recurrent mossy fiber growth creates
monosynaptic excitatory connections among dentate granule cells
(Okazaki et al. 1995
, 1999
; Wuarin
and Dudek 1996
). There is, at present, no anatomic evidence for
such connections in the normal brain. In nonepileptic animals, dentate
granule cells have been shown to resist the propagation of seizures
through the limbic circuit (Lothman et al. 1992
).
Although granule cell discharge can be synchronized by nonsynaptic
mechanisms (Schweitzer et al. 1992
), this occurs only
during very strong afferent bombardment (Lothman et al.
1992
). The recurrent mossy fiber pathway, if sufficiently powerful, could reduce the threshold for granule cell synchronization, thus enhancing the participation of these cells in seizures.
Considerable evidence appears to support this view (Buckmaster
and Dudek 1997a
; Cronin et al. 1992
;
Masukawa et al. 1992
; Patrylo and Dudek
1998
; Tauck and Nadler 1985
), although some
investigators have offered alternative interpretations (Kotti et
al. 1997
; Longo and Mello 1997
; Sloviter
1992
).
As one approach toward understanding the role of the recurrent
mossy fiber pathway in limbic seizures, we sought to characterize the
electrophysiological properties of mossy fiber-granule cell synapses.
In a previous study, we evoked excitatory postsynaptic currents (EPSCs)
by antidromic stimulation of the mossy fibers in hippocampal slices
from rats with recurrent mossy fiber growth (Okazaki et al.
1999). These responses appeared to be largely or entirely
monosynaptic, as indicated by our ability to record an
N-methyl-D-aspartate (NMDA)
receptor-mediated EPSC in the presence of an
-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)/kainate receptor antagonist. In addition, we recorded antidromically evoked EPSCs in some granule cells from control rats. These responses also
appeared to be largely monosynaptic, based again on the observation of
an NMDA receptor-mediated EPSC. The use of antidromic electrical stimulation raises at least the theoretical concern that excitatory pathways other than the mossy fibers might have contributed to the
synaptic response. To study mossy fiber-granule cell synapses in a way
that is not compromised by simultaneous activation of any other
pathway, the stimulus must be restricted to presynaptic granule cells.
Our electron microscopic (Okazaki et al. 1995
) and
electrophysiological (Okazaki et al. 1999
) data
suggested a low probability of connection between pairs of granule
cells, even after robust recurrent mossy fiber growth. Furthermore,
published anatomic studies provide no assistance in localizing coupled
cell pairs. Individual granule cells project not only to their close neighbors, but also to granule cells some distance away
(Buckmaster and Dudek 1999
; Okazaki et al.
1995
; Sutula et al. 1998
). These considerations
suggested that simultaneous recording from granule cell pairs would
have a low probability of success. Accordingly, we studied the
properties of mossy fiber-granule cell synapses with the use of
scanning laser photostimulation (Katz and Dalva 1994
).
By using highly focused ultraviolet (UV) light to release glutamate
from an inactive ("caged") precursor, extracellular glutamate
reached a concentration high enough to cause action potential firing by
only a small number of granule cells, and a large number of
photostimulation sites could be explored in a short period of time.
![]() |
METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Pilocarpine-induced status epilepticus
Recurrent mossy fiber growth was provoked by inducing status
epilepticus with pilocarpine. Male Sprague-Dawley rats (175-225 g;
Zivic-Miller Laboratories, Allison Park, PA) received a single injection of pilocarpine (330-360 mg/kg ip) preceded 30 min earlier with scopolamine methyl bromide and terbutaline hemisulfate (2 mg/kg
ip, each) to block peripheral side effects and maintain respiration.
Status epilepticus was defined as a continuous limbic motor seizure of
stage 2 or higher (Racine 1972). Status epilepticus was
terminated after 3-3.5 h with a single injection of phenobarbital sodium (50 mg/kg ip). Some pilocarpine-treated rats did not develop status epilepticus. These animals were used as drug-treated controls. Age-matched rats were used as untreated controls.
Preparation and incubation of hippocampal slices
Animals were studied 10-30 wk after the administration of
pilocarpine. The rat was decapitated under ether anesthesia, the brain
was removed, and 400-µm-thick transverse slices were cut from the
caudal third of the hippocampal formation with a vibratome. Slices used
for electrophysiological recording corresponded to horizontal plates
98-100 of Paxinos and Watson (1986). Slices reserved
for Timm histochemistry were taken from a level of the hippocampal
formation immediately rostral to this, corresponding to plates
101-103. For electrophysiological studies, the slices were transferred
to a beaker of artificial cerebrospinal fluid [ACSF, which contained
(in mM) 122 NaCl, 25 NaHCO3, 3.1 KCl, 1.8 CaCl2, 1.2 MgSO4, 0.4 KH2PO4, and 10 D-glucose, pH 7.4] and oxygenated at room temperature for
at least 1.5 h with 95% O2-5%
CO2. For Timm histochemistry, the slices were
immersed in 0.1% (wt/vol) Na2S, 0.1 M sodium
phosphate buffer, pH 7.3.
Whole cell patch-clamp recording
A slice was transferred to a glass-bottom Plexiglas
submersion-type recording chamber mounted on the stage of a Nikon
Optiphot-2 upright microscope (Nikon, Melville, NY) connected to a
Noran Odyssey confocal imaging system (Noran Instruments, Middleton, WI). The chamber was filled with ACSF that was recirculated at a rate
of 4 ml/min at room temperature (22-24°C). The total volume of
superfusion medium was 10 ml. Patch electrodes were pulled from
borosilicate glass (1.5 mm OD, 1.1 mm ID, Sutter Instruments, Novato,
CA) and had a tip resistance of 5-7 M. The tip of the electrode was
filled by vacuum with a solution that contained (in mM) 140 cesium
gluconate, 15 HEPES, 3.1 MgCl2, 1 CaCl2, 11 EGTA, pH 7.2 and 276 mosm. The
electrode was then backfilled with internal solution that contained (in
mM) 120 cesium gluconate, 10 HEPES, 2 MgATP, 1 EGTA, 5 creatine
phosphate, 20 units/ml creatine phosphokinase, 10 QX-314
(N-ethyl lidocaine) chloride, and 0.1 fluorescein-dextran
(10,000 MW), pH 7.2 and 276 mosm. Creatine phosphate, creatine
phosphokinase, and ATP, constituting an ATP-regenerating system, were
included to minimize the rundown of NMDA receptor-mediated currents
(Rosenmund and Westbrook 1993
). In current-clamp
experiments, potassium gluconate replaced cesium gluconate, and QX-314
was omitted. The superfusion medium contained 30 µM bicuculline
methiodide to block GABAA receptor-mediated
currents. Activation of postsynaptic GABAB
receptors was prevented by the use of a cesium-based internal solution
that included QX-314, but not GTP.
Recordings were made from granule cells located in the infrapyramidal
blade of the dentate gyrus near the apex of the granule cell arch,
because Timm histochemistry indicated that recurrent mossy fiber growth
is densest there (Okazaki et al. 1995). Gigaohm seals
were formed by the "blind" approach (Blanton et al.
1989
) on granule cell bodies located at least 30 µm below the
upper surface of the slice. Whole cell access was obtained in
current-clamp mode; only cells with Vm
more than
70 mV on break-in (after correction for a 10-mV liquid
junction potential) were accepted for study. Resting membrane
potentials for granule cells from the three treatment groups were as
follows: status epilepticus,
78 ± 4 mV (mean ± SD,
n = 79); treated controls,
77 ± 3 mV
(n = 71); untreated controls,
79 ± 2 mV
(n = 47). Input resistances for these cells after
intracellular dialysis with the cesium gluconate-based internal solution were as follows: status epilepticus, 440 ± 150 M
;
treated controls, 620 ± 160 M
; untreated controls, 660 ± 150 M
. In five granule cells from the treated control group, input
resistance after intracellular dialysis with the potassium
gluconate-based internal solution was 180 ± 30 M
. Granule
cell identity was confirmed by visualizing intracellular fluorescein
(excitation: 488 nm, 515 nm barrier filter) and observation of strong
spike-frequency adaptation during a suprathreshold depolarization.
Recordings were made with an Axon Instruments (Foster City, CA)
Axopatch ID patch-clamp amplifier beginning ~20 min after achieving
whole cell access. Series resistances ranged from 6 to 22 M and were
compensated ~50%. Signals were filtered at 2 kHz, digitized at 10 kHz, and stored to disk with use of a TL 1-125 digitizing board and
PClamp6 (Axon Instruments, Foster City, CA).
Scanning laser photostimulation
An 80-mW Coherent Enterprise 653 argon ion UV laser (Coherent Laser Group, Santa Clara, CA) was used for photostimulation. The laser was coupled to the epifluorescence input of the microscope by a fiber optic cable, and the output passed through an Olympus (Melville, NY) water-immersion, UV-corrected ×40 objective (NA, 0.7; working distance, 3.2 mm). The effective diameter of the laser beam within the plane of focus, uncorrected for light scattering within the slice, was 5.3 µm (Fig. 1). Shutter opening was controlled by PClamp6.
|
To locate a presynaptic granule cell, 200 µM -(CNB-caged)
L-glutamate (Molecular Probes, Eugene, OR) was added to the
superfusion medium. The laser beam was initially focused at a site in
the granule cell body layer ~75 µm from the recorded cell on the
side opposite the apex of the granule cell arch. Then five pulses of UV
light (4 ms, 50 mW) were applied at 10-s intervals. If none of the
stimuli evoked an EPSC, the laser beam was moved 25 µm farther from
the recorded cell (by moving the microscope stage), and another five
pulses were applied. Stimuli continued to be applied at 25-µm
intervals. If no EPSC had been recorded when the laser beam reached the
end of the granule cell body layer, uncaging sites were surveyed in a
similar manner on the opposite side of the recorded cell. The search
ended when laser photostimulation evoked an EPSC in the recorded cell.
A test was considered unsuccessful if this procedure failed to evoke an
EPSC at any of 20-80 uncaging sites in that slice. We avoided exposing
the region within 75 µm of the recorded cell to UV irradiation to
minimize contamination of the recording by direct glutamate current and
to avoid UV-induced cytotoxicity. Exposure of granule cells to the
laser beam in the absence of caged glutamate evoked neither an inward
current nor cell firing.
When a presynaptic granule cell was located, the laser power was
reduced to the minimum required to evoke a visually identifiable EPSC.
The stimulus frequency was reduced to 1/15 s. In most experiments, EPSCs were recorded at alternating holding potentials of 80 and
20
mV. In some early experiments, recordings were initially made at
80
mV before switching to
20 mV. Between 15 and 30 stimuli were applied
at each holding potential.
In attempts to study mossy cell-granule cell pairs, the laser beam was
focused at sites within the dentate hilus. These sites were located at
least 50 µm from the hilar border of the granule cell body layer,
within 200 µm of the recorded granule cell and at least 25 µm from
the previous site. Scharfman (1995) reported that
granule cells in a hippocampal slice were most likely to be innervated
by mossy cells located no more than 200 µm away. An average of 27 uncaging sites was tested in each experiment.
Effect of -(CNB-caged) L-glutamate on NMDA
receptors
A bipolar stimulating electrode (25 µm diam insulated
nichrome wire, tip separation 0.3 mm) was inserted into the
perforant path where it crosses the subiculum. To block the
AMPA-kainate component of the EPSC, 5 µM
2,3-dihydroxy-6-nitro-7-sulfamyl-benzo(F)quinoxaline (NBQX) was added
to the superfusion medium. Rectangular current pulses of 100 µs
duration were applied with a Grass (W. Warwick, RI) stimulator and
stimulus isolator every 30 s. Stimulus strength was adjusted to
evoke a 150- to 200-pA inward synaptic current recorded at a holding
potential of
20 mV. The concentration of caged glutamate was the same
as that used for scanning laser photostimulation.
Visualization of synaptic connections
In some experiments, DiI was applied to the uncaging spot after
the electrophysiological recordings had been completed. DiI was
dissolved in hot cod liver oil (Hosokawa et al. 1995) at
a concentration of 1% (wt/vol) and pressure ejected through a glass micropipette (10 µm tip opening) with a Picospritzer (General Valve,
Fairfield, NJ). The diameter of the region labeled by DiI was ~50
µm. In these experiments, the recorded cell was filled with 0.6%
(wt/vol) Lucifer yellow instead of fluorescein-dextran. The slice was
then fixed in 4% (wt/vol) paraformaldehyde, 0.1 M sodium phosphate
buffer, pH 7.4, at 4°C for 7 days. It was then embedded in an
albumin-gelatin mixture (Okazaki et al. 1995
), and
sections of 50-µm thickness were cut with a vibratome. Fluorescent dyes were visualized under the confocal microscope by excitation at 488 nm and use of a 510-nm barrier filter. Images were captured and stored
with use of Image-1 software (Universal Imaging, West Chester, PA). The
sections were also viewed under an epifluorescence microscope equipped
with DiI and Lucifer yellow cubes.
Timm histochemistry
Slices remained in the Na2S solution for
90 min and were then stored in phosphate-buffered 10% Formalin at
4°C for 1-2 days. They were then embedded in albumin-gelatin, and
30-µm-thick sections were prepared with a vibratome. Slide-mounted
sections were stained for the presence of heavy metals as described by
Danscher (1981) and lightly counterstained with cresyl violet.
Data analysis
Waveforms were analyzed off-line with functions incorporated in
PClamp6. Traces without an observable stimulus-evoked inward current
were averaged, and the averaged trace was electronically subtracted
from traces that showed a stimulus-evoked current. Half-width and decay
time constant () were obtained with built-in PClamp routines,
whereas EPSC amplitude, latency to onset, and 20-80% rise time were
measured manually with cursors. These parameters were determined from
each recording, and the results were averaged to yield single values
for each synaptic connection. Spontaneous synaptic events were not
usually observed during the 100-ms period after laser photostimulation,
but occasionally a spontaneous event overlapped a portion of the evoked
response. In those instances, measurements were restricted to the
parameters that were uncontaminated. Multiple responses were defined as
two or more inward currents time locked to the stimulus whose onset
latency distributions did not overlap.
Materials
-(CNB-caged) L-glutamate, fluorescein dextran,
DMNB-caged fluorescein-dextran (10,000 MW), Lucifer yellow, and DiI
were purchased from Molecular Probes (Eugene, OR).
D-Gluconic acid lactone, HEPES, EGTA, creatine phosphate,
creatine phosphokinase, phenobarbital sodium, pilocarpine
hydrochloride, (
)scopolamine methyl bromide, and terbutaline
hemisulfate were obtained from Sigma Chemical (St. Louis, MO).
D-2-Amino-5-phosphonopentanoate (D-AP5) was
purchased from Tocris Cookson (Bristol, UK), bicuculline methiodide
from Research Biochemicals (Natick, MA), and cesium hydroxide (99.9%; 50 wt%) from Aldrich (Milwaukee, WI). QX-314 chloride was obtained from Astra USA (Westborough, MA) and Alomone Labs (Jerusalem, Israel).
NBQX was a gift from Novo Nordisk (Måløv, Denmark).
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Timm histochemistry
In slices from both pilocarpine-treated rats that did not develop
status epilepticus (treated controls) and age-matched, untreated rats
(untreated controls), scattered clusters of mossy fiber-like Timm
staining were present in the supragranular zone of the dentate molecular layer near the recording site (Fig.
2). Pilocarpine-induced status
epilepticus caused the appearance of dense Timm staining at this site
in every animal. Previous studies demonstrated that dense Timm staining
of the supragranular zone signifies the robust growth of recurrent
mossy fibers, at least some of which make synaptic contact with granule
cells (Frotscher and Zimmer 1983; Okazaki et al.
1995
; Sutula et al. 1989
).
|
Granule cell action potentials evoked by laser photostimulation
Preliminary studies determined the conditions necessary for
uncaged glutamate to depolarize a relatively small number of granule cells to threshold. In 15 experiments, the laser power was adjusted to
50 mW, and the duration of shutter opening was varied. The membrane
potential of the recorded cell was clamped at 80 mV, and flashes of
UV light were focused on the soma. A 2-ms light flash, the shortest
duration tested, evoked an inward current of ~100-pA peak amplitude.
A maximal current was produced with a pulse duration of 16 ms.
The granule cell's response to a 4-ms, 50-mW exposure to UV light,
which had produced a large, but submaximal, glutamate current, was
examined in current-clamp mode. The focus of laser stimulation was
moved to different locations, and the response to 10 light flashes
presented at 15-s intervals was recorded at each location. When
glutamate was uncaged by focusing the laser beam on the center of the
soma, the cell fired one to three action potentials (Fig. 3). At this site, the laser power could
be reduced to as low as 10-15 mW without loss of the response
(although the number of action potentials per stimulus was reduced to
one). The minimum peak glutamate current or injected electrical current
required to depolarize a granule cell to threshold was ~150 pA. When
the laser beam was focused on the perimeter of the soma, on the apical dendrite, within 12 µm from the center of the soma in the
xy plane or within 50 µm from the center of the soma in
the z-axis, a 4-ms, 50-mW stimulus evoked a single action
potential. Every light flash evoked an action potential during trial
periods that lasted as long as 25 min. At greater distances from the
recorded cell, laser photostimulation evoked an action potential, if at
all, only when the laser power or duration of shutter opening was
substantially increased. In each of seven experiments, there was an
abrupt transition between a location at which photostimulation always
evoked an action potential in the recorded cell and a location at which no action potentials could be evoked. We never found a site at which
the light flash evoked an action potential in some trials, but not in
others. Such variable responses, if they occurred at all, must have
been rare, because the maximal width of the transition zone between
sites at which laser photostimulation with 4-ms, 50-mW pulses always
evoked an action potential and sites at which it never evoked an action
potential was 2 µm (the limit of resolution of our stage drive;
shaded area between 2 and
3 in Fig. 3A). This result indicates
that, in subsequent experiments, the presynaptic granule cell fired an
action potential in response to every photostimulus.
|
The region of the cell body layer in which granule cells fired an
action potential would be expected to have had an hourglass shape
centered on the focal plane (Pettit et al. 1997). To
approximate the number of granule cells brought to threshold by the
4-ms, 50-mW flash of UV light used for scanning, one must estimate the volume of this oddly shaped region. However, it is sufficient for the
present purpose to compute maximal and minimal values based on the
volume of simpler shapes. Considering the region of action potential
firing as the frusta of a double cone provides an overestimate of the
volume, whereas considering the region as a cylinder provides an
underestimate. Assuming that the laser beam filled the objective, the
double frusta model yields an estimate of 68,161 µm3 [V = 2/3 · (
r12 +
r22 +
)
· h, where r1 (radius of activated
region within the focal plane = 12 µm) and
r2 (radius of activated region 50 µm
above or below the focal plane = 17.3 µm) are the radii of the
bases of each frustum and h (greatest distance above or
below the focal plane at which granule cells were activated = 50 µm) is the height of each frustum]. The cylinder model yields an
estimate of 45,239 µm3 [V =
· r2 · h, where
r (12 µm) is the radius of the cylinder and h
(100 µm) is the total height of the cylinder]. Granule cell density in aldehyde-fixed rat brain was determined to be 0.0008 cell/µm3 (Boss et al. 1985
).
Applying a correction of 30% for shrinkage during fixation
(Mott et al. 1997
) yields a granule cell density of
0.00056 cell/µm3 in a hippocampal slice. From
these values, we estimate that each exposure to UV light brought
between 25 and 38 granule cells to threshold. It should be noted that
in many experiments the minimal conditions for evoking an EPSC in the
recorded cell required less total UV light exposure per stimulus than
that provided by the standard pulse used for scanning. Thus the
stimulation parameters were reduced, sometimes by half or more. In
these instances, correspondingly fewer granule cells were activated.
EPSCs evoked by uncaging glutamate in the granule cell body layer
Minimal photostimulation in the granule cell body layer most
frequently evoked an EPSC in granule cells from rats that had developed
status epilepticus, with subsequent recurrent mossy fiber growth. We
were able to locate a presynaptic granule cell in 61% of these
experiments (Table 1). On average, seven
uncaging sites were tested before a connection was found (range:
1-20). Thus an average of between 175 and 266 granule cells had to be activated before a synaptic response was recorded. With use of the same
procedure, synaptic connections were also located in 16% of the
experiments with pilocarpine-treated control rats and in 13% of the
experiments with age-matched, untreated rats. On average, 11 uncaging
sites were tested in slices from treated controls and 9 in slices from
untreated controls before a synaptic connection was found (range:
3-22). The latency from the beginning of the light flash to the onset
of the EPSC averaged 30 ms (range: 9-94 ms). There was no significant
difference in onset latency among the treatment groups. In studies of
this type, onset latency is determined predominantly by the highly
variable and indeterminable interval between the photostimulus and cell
firing (Katz and Dalva 1994). Latency values cannot be
related to synaptic delay or axonal conduction.
|
Synaptic currents evoked by minimal photostimulation exhibited
properties consistent with their identification as unitary EPSCs. They
had a smooth rising phase (Fig.
4A), appeared at a constant
latency (±5 ms) with respect to the onset of the light flash and
exhibited all-or-none behavior when the laser power was varied (Fig.
5). In a few instances, the rising phase
of the synaptic current exhibited a slightly jagged or inflected
appearance (e.g., Fig. 4B). This irregularity may have
resulted from a slightly asynchronous release of transmitter from
different sites in the same bouton, as described for some unitary mossy
fiber EPSCs recorded in CA3 pyramidal cells (Jonas et al.
1993). Responses were abolished by addition of 2 mM
Cd2+ to the superfusion medium.
|
|
Multiple responses were commonly observed with the present experimental approach. These consisted of two to four EPSCs that appeared at distinctly different times after the photostimulus (Fig. 4B). Of the 48 granule cells from the status epilepticus group in which EPSCs were recorded, multiple responses were recorded in 18. Multiple responses were also recorded in 7 of the 11 successful experiments from the treated control group. No multiple responses were recorded in granule cells from the untreated control group. Reducing either the laser power or duration of the light flash usually did not reduce the number of temporally distinct responses. In some experiments, the individual EPSCs exhibited very similar peak amplitudes and response kinetics. They may therefore have arisen from repetitive firing of a single presynaptic cell. In other experiments, the individual EPSCs exhibited distinctly different peak amplitudes and/or response kinetics, suggesting that they arose from the glutamate-evoked firing of more than one presynaptic cell. Because we could not distinguish between these two mechanisms with certainty, quantitative data were compiled in two ways: by assuming that each temporally distinct EPSC arose from the activation of a different presynaptic cell and by assuming that multiple responses always arose from repetitive activity in a single presynaptic cell. However, the two calculations produced nearly indistinguishable results.
Properties of the EPSC evoked by minimal photostimulation did not vary
significantly among the treatment groups. The EPSC had a mean peak
amplitude of ~30 pA, 20-80% rise time of ~1.2 ms, decay time
constant () of ~10 ms and half-width of ~8 ms (Table 1). Peak
amplitudes varied over a 5.5-fold range (12.8-70.5 pA). Response
kinetics for EPSCs in granule cells from untreated controls were
generally slightly faster than for EPSCs in the other treatment groups,
but the differences were not statistically significant.
The mean failure rate was ~70%. That is, only ~3 of 10 photostimuli evoked an EPSC in the recorded cell (Fig. 4A). Few, if any, failures could have resulted from our inability to distinguish a synaptic response from noise. The smallest EPSC recorded in this study had a peak amplitude of 12 pA, about 3 times the peak-to-peak noise in our recordings.
In some granule cells, laser photostimulation evoked a large, complex
inward current (Fig. 4C). The peak amplitude of these responses ranged from 100 to 500 pA. They appeared at varying latencies
and were evoked in some trials and not others. In some experiments,
they were evoked by the same stimuli that also evoked one or more
minimal EPSCs in the recorded cell. However, 70% of the large, complex
inward currents occurred in the absence of a minimal EPSC. These
responses were most commonly observed in granule cells from the status
epilepticus group (16/79, 20%), but also appeared in 7 of 71 granule
cells (10%) from the treated control group and in 1 of 47 granule
cells (2%) from the untreated control group. They closely resembled
delayed inward currents evoked by antidromic stimulation of the mossy
fiber pathway in the presence of a GABAA receptor
antagonist (Okazaki et al. 1999). The antidromically
evoked delayed inward currents were abolished by AMPA/kainate receptor
antagonists. Thus they were considered to be polysynaptic EPSCs that
resulted from reverberating excitation among granule cells.
Unfortunately, we could not use AMPA/kainate receptor antagonists to
test whether the large, complex inward currents observed in the present
study were also polysynaptic EPSCs, because AMPA receptors must be
activated in order for uncaged glutamate to fire granule cells.
NMDA receptors contribute to the EPSC
To determine whether the EPSC evoked by minimal photostimulation
has an NMDA receptor-mediated component, we compared evoked responses
recorded at alternating holding potentials of 80 and
20 mV (Fig.
6). A late component of the EPSC with
slow decay kinetics appeared at a holding potential of
20 mV, but was
not apparent at a holding potential of
80 mV (Fig. 6, A
and B). The NMDA receptor antagonist D-AP5 (50 µM) was then added to the superfusion medium. Addition of
D-AP5 usually abolished the evoked response, but the
response could always be recovered simply by increasing the duration of
shutter opening. This action of D-AP5 presumably resulted
from block of NMDA receptors whose activation by uncaged glutamate
contributed to depolarization of the presynaptic cell. After the EPSC
was recovered, it could be seen that D-AP5 barely affected the response recorded at
80 mV, but it abolished the late,
slow component of the response recorded at
20 mV (Fig. 6C). Electronic subtraction of traces recorded at
20
mV before and after superfusion with D-AP5 revealed a small
NMDA component (Fig. 6D). Similar results were obtained
in nine granule cells from the status epilepticus group and six granule
cells from the treated control group. An NMDA/AMPA ratio was computed
by dividing the peak amplitude of the NMDA receptor-mediated EPSC
recorded at
20 mV by the peak amplitude of the AMPA/kainate
receptor-mediated EPSC recorded at
80 mV. There was no significant
between-group difference. Mean values (±SD) were as follows: status
epilepticus group, 0.19 ± 0.06; treated control group, 0.16 ± 0.05.
|
-(CNB-caged) L-glutamate did not desensitize
NMDA receptors
High concentrations of -(CNB-caged) L-glutamate
have been found to activate NMDA receptors (Kandler et al.
1998
). One possible reason for the relatively small amplitude
of the NMDA receptor-mediated EPSC was that caged glutamate
desensitized NMDA receptors. To evaluate this possibility, we studied
the effect of caged glutamate on the pharmacologically isolated NMDA
receptor-mediated EPSC evoked by perforant path stimulation. In five
experiments, addition of 200 µM caged glutamate to the superfusion
medium altered neither the peak amplitude of the response nor the clamp
current. This result indicates that the compound did not activate or
desensitize NMDA receptors under the present experimental conditions.
Laser photostimulation in the dentate hilus did not evoke an EPSC
The only other known excitatory projection to granule cells that
could have been activated by laser photostimulation is the dentate
associational pathway. This pathway originates from mossy cells within
the dentate hilus (Buckmaster et al. 1992; Ribak et al. 1985
). We attempted to activate associational
connections by laser photostimulation directed at mossy cells. We were
mainly concerned with activation of associational synapses in slices from control rats, because most mossy cells are believed to die during
status epilepticus (Scharfman and Schwartzkroin 1990
;
Sloviter 1987
). Therefore age-matched, untreated control
rats were used in this study. The laser beam was focused within the
region of the dentate hilus in which presynaptic mossy cells were most
likely to be found.
We surveyed a total of 601 uncaging sites in 22 slices from 15 rats. Only one granule cell was studied in each slice. The laser parameters (4 ms, 50 mW) were the same as those used to locate presynaptic granule cells. In none of these experiments did laser photostimulation evoke an EPSC.
Visualizing potential mossy fiber-granule cell connections
To visualize presynaptic mossy fibers, DiI was pressure ejected at
the site of glutamate uncaging. In 12 experiments where minimal
photostimulation evoked an EPSC (10 from the status epilepticus group
and 1 each from the treated control and untreated control groups), a
single fluorescent axon was observed to traverse the molecular layer
approximately perpendicular to the dendrites of the recorded granule
cell (Fig. 7). In three additional
experiments on slices from rats that had developed status epilepticus,
DiI ejection labeled two axons with a similar trajectory that coursed parallel to each other at different depths within the molecular layer.
The labeled axons were studded with brightly fluorescent varicosities,
the largest of which were ~2 µm diam. Both confocal and
fluorescence microscopy showed that one of these varicosities was in
close apposition to a dendritic branch of the recorded cell. These
contact points were located as close as 22 µm and as distant as 130 µm from the center of the soma. Two-thirds of them were located
between 50 and 85 µm away. The size of the labeled boutons and the
location of the contact sites matches the description of recurrent
mossy fibers provided by retrograde labeling with biocytin
(Okazaki et al. 1995). There was no correlation between the number of temporally distinct EPSCs evoked by photostimulation and
the number of axons labeled in the same experiment. Only a single EPSC
was evoked in each of the three experiments in which two putatively
presynaptic axons were labeled and only a single axon was labeled in
two experiments in which multiple responses were observed.
|
In six slices (4 from the status epilepticus group and 2 from the treated control group), DiI was pressure ejected at a site where photostimulation failed to evoke an EPSC. In none of these experiments was a fluorescent axon observed to cross the dendritic tree of the recorded cell.
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Our results reinforce and extend the existing electrophysiological
evidence for synaptic connections between dentate granule cells
(Okazaki et al. 1999; Wuarin and Dudek
1996
). They indicate that some mossy fiber-granule cell
synapses are normally present in the rat dentate gyrus, that
pilocarpine-induced status epilepticus substantially increases the
number of these synapses, that EPSCs evoked by activating preexisting
and newly formed synapses have very similar properties, and that both
AMPA/kainate and NMDA receptors mediate transmission at these sites.
We evaluated the possibility that some of the EPSCs evoked by minimal
photostimulation arose from activation of an associational (granule
cell-mossy cell-granule cell) circuit within the slice. Hilar mossy
cells have been shown to make monosynaptic excitatory synapses on
granule cells (Scharfman 1995). They are highly
excitable, because their normal resting membrane potential is only
~7-8 mV above their action potential threshold (compared with ~30
mV for granule cells; Staley et al. 1992
), inhibitory
postsynaptic potentials (IPSPs) are small and their input resistance is
relatively high (Scharfman and Schwartzkroin 1988
). Thus
mossy cells are readily activated by excitatory input from mossy fibers
(Scharfman et al. 1990
). In the present study, the
glutamate-evoked firing of as many as 25-38 granule cells per UV light
flash could potentially have activated mossy cells. However, we did not
activate a single mossy cell-granule cell connection by laser
photostimulation of the dentate hilus in 601 attempts; the probability
of evoking an EPSC by focusing the laser beam in the granule cell body
layer with use of the same power and shutter open time was much higher. Similarly, Wuarin and Dudek (1996)
observed no change in
the spontaneous excitatory postsynaptic potential (EPSP) frequency when
they applied glutamate to the dentate hilus of hippocampal slices from
kainate-treated rats. In contrast, spontaneous EPSP frequency was
markedly increased by similar glutamate applications to the granule
cell body layer. In addition, the properties of minimal EPSCs from the
status epilepticus and control groups were indistinguishable; the major
difference was that these responses were much more frequently evoked in
granule cells from the status epilepticus group. This result was
opposite to what would be expected if minimal photostimulation had
evoked associational synaptic responses, because mossy cells are more numerous in slices from control rats. We therefore conclude that minimal photostimulation in the granule cell body layer rarely, if
ever, activated associational synapses on granule cells. Perhaps essentially all associational circuitry was cut during preparation of
our slices. More likely, mossy cells, being extremely vulnerable to
insults that raise intracellular calcium (Scharfman and
Schwartzkroin 1989
), simply fail to survive slice preparation
and incubation in older rats.
Properties of the mossy fiber-granule cell EPSC evoked by minimal photostimulation
The mean peak amplitude (~30 pA) of the EPSC evoked by minimal
photostimulation fell between reported values obtained under similar
recording conditions for mossy fiber synapses on CA3 pyramidal cells
(~70 pA) (Jonas et al. 1993) and for hippocampal
synapses made by small terminals with one or at most two release sites (
15 pA) (Allen and Stevens 1994
; Kneisler and
Dingledine 1995
; Raastad et al. 1992
). This
result appears intuitively reasonable for recurrent mossy fiber
synapses, assuming that each fiber usually makes only a single contact
with each granule cell. All other factors being equal, peak amplitude
would be expected to vary with the number of release sites per bouton.
This appears to be the case for mossy fiber synapses on CA3 pyramidal
cells: the quantal content of different EPSCs (Jonas et al.
1993
) varies over the same range as the number of synaptic
contacts per bouton (Chicurel and Harris 1992
). Mossy
fiber boutons that contact dentate granule cells usually have multiple
release sites, but are, on average, only about one-third to one-half
the size of mossy fiber boutons that contact CA3 pyramidal cells
(Okazaki et al. 1995
). The largest of these boutons in
the inner third of the dentate molecular layer is ~2 µm diam as
determined by retrograde labeling with biocytin (Okazaki et al.
1995
) and by DiI labeling (present study). However, many
recurrent mossy fiber boutons are <2 µm and may contact only a
single spine. Thus the relatively large size of and the 5.5-fold
variability in the minimal EPSC amplitude appear anatomically reasonable.
Response kinetics were relatively slow compared with values reported
for mossy fiber synapses on other postsynaptic cells. DiI labeling
suggests that this difference probably cannot be explained by greater
attenuation and filtering in the dendritic tree. Although we cannot be
certain that the labeled contact site represented the synapse that was
studied electrophysiologically, the distribution of potential contact
sites provides information on the range of synaptic placements relative
to the recording site at the soma. The synaptic currents we studied
were probably generated at roughly the same distance from the somatic
voltage clamp as in studies of CA3 pyramidal cells (Jonas et al.
1993), and errors due to imperfect space clamp are likely to
have been small (Carnevale et al. 1997
; Spruston
et al. 1993
). Our use of higher resistance patch electrodes,
which we found necessary in slices from older rats, may have
exaggerated the effect of any series resistance errors. However, the
slower response kinetics may also reflect differences among
postsynaptic cells in their receptor mechanisms and in the extent to
which postsynaptic receptors are saturated with released glutamate. In
addition, much younger rats were used to study mossy fiber synapses on
other cell types; response kinetics of mossy fiber EPSCs may change
with age. Finally, response kinetics of the mossy fiber-granule cell
EPSC may be influenced to some degree by the nonuniform electrotonic
structures of granule cells (Carnevale et al. 1997
).
The consistently high failure rate (~70%) observed in this study did
not result from variable activation of the presynaptic granule cell by
uncaged glutamate. Rather, transmission at mossy fiber-granule cell
synapses may be relatively unreliable. A similarly high failure rate
was reported for Schaffer collateral-commissural synapses (Allen
and Stevens 1994). The percentage of failures at mossy
fiber-CA3 pyramidal cell synapses depends on the stimulation rate, the
greatest incidence of failures occurring during stimulation at <0.5
and >2 Hz (Jonas et al. 1993
). During stimulation at 1 Hz, the failure rate was ~11%. An average failure rate of 33% was
reported for the mossy fiber-dentate basket cell synapse during stimulation at 0.2-0.33 Hz (Kneisler and Dingledine
1995
). Thus the lower failure rate observed in previous studies
of mossy fiber synapses compared with the present study may reflect the
use of higher stimulus frequencies. Mossy fiber synapses on CA3
pyramidal cells exhibit dramatic frequency facilitation when activated
at rates of 0.1-1 Hz (Salin et al. 1996
). A stimulus
frequency of 1/15 s, as employed in the present study, may be a worst
case for transmission failure. The high failure rate at Schaffer
collateral-commissural synapses results mainly from the probabilistic
nature of transmitter release at sites of low release probability
(Allen and Stevens 1994
). For the mossy fiber pathway,
the possibility of conduction failure should also be considered, due to
the impedance mismatch between the thin axon and large bouton
(Lüscher and Shiner 1990
). It should be noted,
however, that some of the minimal EPSCs we recorded, especially in the
status epilepticus group, may not have been monosynaptic. They could
have been the end product of reverberating excitation among granule
cells; indeed we found evidence of such polysynaptic activity in some
experiments. We cannot rule out the possibility that the higher failure
rate of polysynaptic transmission influenced our results.
NMDA receptors contribute to the mossy fiber-granule cell EPSC,
as they do to mossy fiber EPSCs recorded in CA3 pyramidal cells
(Weisskopf and Nicoll 1995) and dentate basket cells
(Kneisler and Dingledine 1995
). We found no difference
between status epilepticus and control groups in the relative size of
the NMDA component; the peak amplitude of the NMDA component recorded
at
20 mV was a small fraction (~20%) of the AMPA/kainate component
recorded at
80 mV in each case. In contrast, a preceding status
epilepticus usually increased the NMDA/AMPA ratio to a value of
1.4-2.2 when mossy fibers were activated by antidromic electrical
stimulation (Okazaki et al. 1999
). In light of the
present results, this effect cannot easily be explained by a change in
the NMDA receptor itself. One possibility is that the size of the NMDA
component depends, in part, on spillover of glutamate from mossy fiber
synapses on nearby granule cells. Due to the much higher affinity of
NMDA receptors for glutamate compared with AMPA/kainate receptors, glutamate diffusing from the synaptic cleft predominantly activates NMDA receptors (Kullmann and Asztely 1998
). The region
of the dentate molecular layer into which recurrent mossy fibers grow normally has a high density of NMDA receptors (Monaghan and
Cotman 1985
). The effect of spillover is limited by active
transport and is thus particularly prominent at a less than
physiological recording temperature (Asztely et al.
1997
), such as that used in both our previous (Okazaki
et al. 1999
) and present studies. The spillover effect is also
limited by the number of activated glutamate synapses in close
proximity to the synapses under study. With respect to the recurrent
mossy fiber pathway, antidromic electrical stimulation activates many
more synapses than minimal photostimulation, and seizure-induced mossy
fiber growth increases the likelihood that antidromic stimulation will
release glutamate from a large number of nearby terminals. According to
this view, the enhanced contribution of NMDA receptors to the mossy
fiber-granule cell EPSC evoked by antidromic stimulation is simply the
inevitable consequence of increasing the synaptic density of the
recurrent projection. Further studies are needed to evaluate this hypothesis.
Mossy fiber growth increases the number of recurrent synapses on dentate granule cells
Our results reinforce the suggestion from our previous work
(Okazaki et al. 1999) that mossy fibers make recurrent
synapses with granule cells in the normal brain. Although there is no
published anatomic evidence for the existence of these synapses, mossy
fiber-like Timm stain is present in the supragranular zone at some
locations, especially near the caudal end of the dentate gyrus
(Gaarskjaer 1978
; Haug 1974
). It has
remained unclear to what extent these presumptive recurrent mossy
fibers innervate granule cells, as opposed to interneurons.
Wuarin and Dudek (1996)
reported that application of
glutamate to the granule cell body layer failed to evoke an excitatory
response in granule cells from normal rats. However, the likelihood of
activating connections between granule cells probably depends on the
site in the dentate gyrus at which recordings are made. In slices
immediately rostral to those we used for electrophysiological
recording, scattered clusters of Timm granules were present in the
inner third of the molecular layer near the recording site. This is the
same location in which DiI-labeled axons were found. Our results
suggest that these Timm granules do, in fact, represent mossy fiber
boutons and that at least some of these boutons engage in synaptic
contact with granule cells. Recurrent mossy fiber synapses may
synchronize granule cell discharge in regions of the dentate gyrus
where their numbers are significant.
Laser photostimulation much more readily evoked an EPSC in slices from rats that had developed status epilepticus, but the EPSC was virtually identical to control with respect to its mean peak amplitude, duration, kinetics, onset latency, failure rate, and contribution from NMDA receptors. Thus the new mossy fiber synapses appear to operate very similarly to preexisting synapses, but there are many more of them. We were able to evoke an EPSC in about four times as many slices from the status epilepticus group and, in successful experiments, a connection was found after ~30% fewer uncaging trials. Therefore reactive growth after pilocarpine-induced status epilepticus may increase the density of mossy fiber-granule cell synapses in the caudal dentate gyrus about sixfold.
It should be noted that our results provide only a rough estimate of granule cell interconnectivity. The finding that 175-266 granule cells had to be activated in slices from the status epilepticus group before a synaptic connection was located does not necessarily imply a 0.5% probability of interconnection in vivo. Some recurrent circuitry may have been cut when the slices were prepared. Furthermore, this study was designed to identify and characterize mossy fiber-granule cell synapses, not to estimate their absolute numbers; the search for a synaptic connection ended whenever a connection was found. Additional studies specifically directed at the quantitation of granule cell interconnectivity are needed.
Functional implications of the recurrent mossy fiber projection
All rats that develop status epilepticus after administration of
pilocarpine invariably exhibit chronic limbic motor seizures after a
latent period of 1-3 wk (Mello et al. 1993;
Lemos and Cavalheiro 1996
). Synaptic connections
among pyramidal cells serve as the anatomic substrate for epileptiform
burst firing in area CA3 (Miles et al. 1984
). Recurrent
mossy fiber growth creates circuitry in the dentate gyrus similar to
that normally present in the CA3 area. Given that dentate granule cells
are difficult to recruit into epileptiform discharge and thus present a
barrier to seizure propagation (Lothman et al. 1992
),
the new mossy fiber-granule cell connections may contribute to
seizures in pilocarpine-treated animals. The recurrent mossy fiber
circuit may also be contributory in persons with epilepsy whose
hippocampi have undergone this same synaptic rearrangement.
The circumstances under which mossy fiber-granule cell synapses
contribute to epileptogenesis remain to be defined. Our results suggest
that the synchronous activation of about five such synapses may be
sufficient to bring a granule cell to threshold under our experimental
conditions (~150 pA inward current required ~30 pA unitary
synaptic current). However, transmission at these synapses appears to
be relatively unreliable, granule cells normally fire action potentials
in vivo at low rates (Jung and McNaughton 1993
), and
these cells are subject to strong GABA inhibition (Freund and
Buzsáki 1996
; Otis et al. 1991
). These
considerations suggest that the synchronous activation of recurrent
mossy fibers might fail to bring more than a small percentage of
granule cells to threshold. Conversely, other factors might enhance the
ability of recurrent mossy fibers to promote seizures. Both the higher temperature and the background depolarization due to perforant path
input present in vivo will increase granule cell excitability. Furthermore, if mossy fiber-granule cell synapses exhibit marked activity-dependent facilitation similar to that of mossy fiber synapses
on CA3 pyramidal cells (Salin et al. 1996
), then the recurrent mossy fiber pathway might preferentially support synchronous granule cell discharge when it is driven at particular frequencies or
in particular firing patterns that maximize facilitation. Activation of
the recurrent pathway at frequencies near 5 Hz may also enhance its
synchronizing capability by depressing GABA inhibition (Mott et
al. 1993
). Finally, granule cells may interact through
electrotonic, as well as synaptic, coupling (MacVicar and Dudek
1982
).
![]() |
ACKNOWLEDGMENTS |
---|
We thank Drs. G. J. Augustine, R. J. Dingledine, G. Einstein, A. Fine, L. C. Katz, and M. M. Okazaki for helpful discussions and advice and K. Gorham for secretarial assistance.
This research was supported by National Institute of Neurological Disorders and Stroke Grant NS-17771 and by National Science Foundation Grant BIR-9318101.
![]() |
FOOTNOTES |
---|
Address for reprint requests: J. V. Nadler, Dept. of Pharmacology and Cancer Biology, Duke University Medical Center, Durham, NC 27710.
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.
Received 12 April 1999; accepted in final form 8 June 1999.
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|