Howard Hughes Medical Institute and Department of Neurobiology, Duke University School of Medicine, Durham, North Carolina 27710
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
McQuiston, A. Rory and Lawrence C. Katz. Electrophysiology of Interneurons in the Glomerular Layer of the Rat Olfactory Bulb. J. Neurophysiol. 86: 1899-1907, 2001. In the mammalian olfactory bulb, glomeruli are surrounded by a heterogeneous population of interneurons called juxtaglomerular neurons. As they receive direct input from olfactory receptor neurons and connect with mitral cells, they are involved in the initial stages of olfactory information processing, but little is known about their detailed physiological properties. Using whole cell patch-clamp techniques, we recorded from juxtaglomerular neurons in rat olfactory bulb slices. Based on their response to depolarizing pulses, juxtaglomerular neurons could be divided into two physiological classes: bursting and standard firing. When depolarized, the standard firing neurons exhibited a range of responses: accommodating, nonaccommodating, irregular firing, and delayed to firing patterns of action potentials. Although the firing pattern was not rigorously predictive of a particular neuronal morphology, most short axon cells fired accommodating trains of action potentials, while most delayed to firing cells were external tufted cells. In contrast to the standard firing neurons, bursting neurons produced a calcium-channel-dependent low-threshold spike when depolarized either by current injection or by spontaneous or evoked postsynaptic potentials. Bursting neurons also could oscillate spontaneously. Most bursting cells were either periglomerular cells or external tufted cells. Based on their mode of firing and placement in the bulb circuit, these bursting cells are well situated to drive synchronous oscillations in the olfactory bulb.
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The structure and
function of the olfactory system is conserved across many vertebrate
and invertebrate species, thus providing insights into general
principles of sensory processing (for reviews, see Hildebrand
and Shepherd 1997; Mori 1987
; Shipley and
Ennis 1996
). The first stage of olfactory information
processing occurs in the olfactory bulb (OB), where olfactory sensory
neurons make their first synapses in spherical structures called
glomeruli. Individual glomeruli consist of a shell of small neurons and
glial cells within which axons of sensory neurons synapse onto the
dendrites of both principal cells (mitral/tufted cells) and
interneurons surrounding the glomerulus (Pinching and Powell
1971a
,b
; White 1972
, 1973
), which have been
termed juxtaglomerular cells (Shipley and Ennis 1996
).
Synaptic interactions between the dendrites of mitral/tufted cells
(M/T) and juxtaglomerular cells also occur within glomeruli. Despite
the obviously important position of juxtaglomerular (JG) cells at this
first stage of olfactory processing, understanding of their functional
repertoire is incomplete.
JG cells include several different types of interneurons such as
periglomerular cells, short axon cells, and external tufted cells. In
the original descriptions (Pinching and Powell
1971a-c
), each class of cell was considered homogeneous.
Periglomerular cells were proposed to inhibit the mitral/tufted cells
via dendrodendritic synapses and lateral axonal collaterals, short axon
cells to exclusively inhibit other short axon and periglomerular cells
via axon collaterals, and external tufted cells to excite inhibitory
cells of glomeruli and possibly granule cells of the granule cell layer.
More recent experiments have demonstrated that JG cells are far more
heterogeneous (for review, see Kosaka et al. 1998). JG cells express a variety of different transmitters, including but not
limited to GABA (Ribak et al. 1977
), nitric oxide/NADPH
(Scott et al. 1987
), dopamine (Halasz et al.
1977
), and neuropeptide Y (Scott et al. 1987
).
Furthermore the dendrites of different periglomular cells target
specific functional domains of glomeruli. The dendrites of one
subclass, for example, avoid regions innervated by sensory axons and
only receive input from M/T cells, while other periglomerular cells
receive both sensory neuron and M/T cell input (Kosaka et al.
1998
).
JG neurons were initially described as bursting cells (Shepherd
1963), some of which were likely to be inhibitory
(Duchamp-Viret et al. 1993
; Getchell and
Shepherd 1975
). Others, however, have suggested that the JG
neurons may all be excitatory (Freeman 1974a
,b
) and that GABAergic synaptic transmission in the glomeruli may be
depolarizing (Siklos et al. 1995
). Although others have
observed the bursting behavior of JG neurons (specifically
periglomerular neurons) (Wellis and Scott 1990
), some
experiments using whole cell patch clamping in olfactory bulb slices
have failed to observe bursting neurons in the glomerular region. These
studies have described JG neurons [including anatomically identified
periglomerular (PG) neurons] as producing single action potentials or
trains of a few action potentials in response to depolarizing current injections (Bardoni et al. 1995
; Bufler
et al. 1992
; Puopolo and Belluzzi 1996
, 1998
).
In contrast, PG neurons of very young rabbit (P4) olfactory bulb slices
have been described as not spiking at all, where other JG neurons
produce a train of spikes in response to depolarizing current injection
(Bufler et al. 1992
). Thus there are a number of
discrepancies describing the physiological properties of JG neurons of
the olfactory bulb that need to be addressed for the physiological
nature of JG neurons to be fully understood.
Given their diversity, their importance at the first stage of sensory processing, and the paucity of coherent information regarding their physiological properties, we investigated the physiological properties of JG neurons in the rat olfactory bulb. In addition to this chemical and connectivity diversity, we found considerable heterogeneity of spiking behavior. Of particular interest was the substantial population of these interneurons that exhibited spontaneous and evoked bursting behavior. Synchronous firing of neurons in the bulb has been implicated in olfactory coding; these neurons may be important components of the intrinsic bulbar networks that generate rhythmic firing.
![]() |
METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Male Sprague Dawley rats (13-21 days old) were anesthetized with isoflurane and killed by decapitation. The olfactory bulbs (OB) were removed and placed in cold, oxygenated saline [which contained (in mM) 119 NaCl, 2.5 KCl, 1.0 CaCl2, 4 MgCl2, 1 NaHPO4, 26.2 NaHCO3, 11 glucose, and 1 kynurenic acid, pH 7.4 when bubbled with 95% O2-5% CO2]. The OBs were sectioned coronally or horizontally (300-350 µm thick) on a Vibratome (Lancer, Technical Products, St. Louis, MO), maintained submerged in saline (as in the preceding text but without kynurenic acid) supported on a nylon mesh. Slices were incubated at 30°C for the first 30 min and then at room temperature (~23°C); under these conditions slices remained healthy for 2-4 h. Although slices were usable for longer periods of time, clear deterioration of slice health (apoptotic cells) were obvious after several hours in the incubation chamber. Recordings lasted on average 30 min. Prolonged recordings and continued activation of the low-threshold spike (LTS) caused significant run down of the LTS making interpretation of the results difficult.
For whole cell patch-clamp recording, the tissue slice was submerged in
a recording chamber mounted on a custom made stage under a Zeiss
Axioscope FS microscope (Carl Zeiss, Thornwood, NY). The slice was
supported by a cover slip, which formed the bottom of the recording
chamber and was superfused with saline (mostly at room temperature,
some at 34°C) consisting of (in mM) 119 NaCl, 2.5 KCl, 2.5 CaCl2, 1.3 MgSO4, 1 NaHPO4, 26.2 NaHCO3, and 11 glucose, pH 7.4). Slices were visualized using a ×63 water-immersion objective trans-illuminated with infrared light. The image
was collected by a Hamamatsu C2400 CCD camera (Hamamatsu, Bridgewater, NJ) with contrast enhancement and displayed on a video monitor so that
glass patch pipettes could be visually advanced through the slice to
the surface of a cell (Dodt and Zieglgansberger 1990; MacVicar 1984
). Whole cell patch-clamp recordings were
made from visually identified JG cells in the glomerular layer of the OB.
Patch pipettes were fabricated from borosilicate glass (KG33; 1.5 mm
OD, 1.0 mm ID; Garner Glass, Claremont, CA) pulled on a two-stage
vertical electrode puller (Narishige PP 83, East Meadow, NY). The
intracellular solution consisted of (in mM)120 K methylsulfate, 8 NaCl,
10 HEPES, 2 -methyl-
-phenylsuccinimide, paraformaldehyde (MgATP),
0.3 Na3GTP, and 0.1 BAPTAK4, biocytin,
0.2-0.5%, pH 7.25. Membrane potentials were monitored with an
Axoclamp 2B amplifier (Axon Instruments, Foster City, CA), and data
were collected onto a Pentium II personal computer using software
written in LabView (National Instruments; by Eric Schaible and Paul
Pavlidis, Stanford University). Data were analyzed using LabView and
Axum (Mathsoft, Cambridge, MA) software packages. Statistical
significance was determined by a two-tailed unpaired Student's
t-test for data of unequal variance. Values are reported as
means ± SE. Drugs were applied by bath superfusion or by flash
photolysis from a mercury arc lamp (Oriel, Stratford, CT) focused into
a 60-µm fiber optic (Ceram) placed above the slice and cell of
interest (Kandler et al. 1998
). Synaptic responses were
evoked by bipolar tungsten electrodes placed in the sensory nerve layer
(FHC, Bowdoinham, Maine).
Following physiological experiments, cells were processed for
morphological analysis as biocytin had been included in the patch
pipettes. Slices were fixed overnight in buffered formalin (0.1 M
phosphate buffer, 4% paraformaldehyde), and processed without further
sectioning. Slices were processed one of two waysby using either a
peroxidase conjugate to produce a dark reaction or a fluorescent
streptavidin conjugate to produce a fluorescent cell (Alexa Fluor-488,
Molecular Probes). To produce a dark reaction product, slices were
permeabilized (0.5% Triton X-100), treated with 0.3%
H2O2 to reduce background,
and incubated 1-3 days in avidin-biotin-peroxidase complex (Elite
Vectastain ABC kit; Vector Laboratories, Burlingame, CA). Sections were
then reacted with diaminobenzidine and intensified with nickel, mounted
on slides, cleared, and coverslipped. Cells were subsequently
reconstructed using a camera-lucida. For fluorescence, slices were
permeabilized with Triton X-100, incubated with goat serum to reduce
background, and incubated with streptavidin Alexa Fluor 488 for 1-2
days. Slices were then washed and mounted wet in Vectashield mounting medium (Vector Laboratories). Fluorescent cells were reconstructed with
a BioRad MRC 600 confocal microscope in 0.5- to 1-µm sections.
All chemicals were purchased from Fluka (Milwaukee, WI), except for the following: tetrodotoxin (TTX; Calbiochem, La Jolla, CA); MgATP (Sigma, St. Louis, MO); and potassium methylsulfate (ICN, Costa Mesa, CA).
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Whole cell patch-clamp recordings were made from visually
identified neurons in the periglomerular region of rat olfactory bulb
slices. Under infrared illumination glomeruli were easily recognizable
as lucent spherical objects surrounded by small cell bodies (5-15 µm
diameter). We recorded from a total of 179 cells with intermediate soma
diameters (8-12 µm), a range that includes neurons from the three
broad classes of JG neurons in this region: periglomerular, short axon
and external tufted cells (Pinching and Powell 1971a).
The membrane potential (Vm) of the JG
cells was maintained near their resting potential, 60-70 mV. The
internal pipette solution contained biocytin (0.2-0.5%) to allow
subsequent morphological analysis of a subset of recorded cells.
JG cells show diverse electrophysiological properties
JG cells exhibit a diverse array of firing properties in response to an artificial depolarization, which could be divided into two large groups (Figs. 1 and 2). One group (102 cells) fired standard trains of action potentials; the second group fired in bursts. The first group, which exhibited "standard" firing properties, consisted of cells that displayed one of four different firing patterns (Fig. 1). The second group (77 cells), which fired bursts of action potentials, was distinguished by the presence of a low-threshold calcium-dependent spike (Fig. 2).
|
|
In the simplest case, the standard group responded to brief current injection through the patch pipette with a regularly spaced train of action potentials (Fig. 1Ai, n = 31). Because the intervals between action potentials in the train remained similar throughout current injections, we termed this type of cell nonaccommodating. On hyperpolarization, 71% (22/31 cells, Fig. 1Aii) of nonaccommodating neurons displayed a small depolarizing sag in the membrane potential (Vm) during the hyperpolarizing pulse. In the remaining 29% (9/31 cells), the sag was absent.
Another subset of cells in the standard group fired trains of action potentials that decreased in frequency during depolarization (Fig. 1Bi, n = 14) and were therefore termed accommodating cells. Most cells of this type produced a depolarizing sag during a hyperpolarizing current pulse (Fig. 1Bii, 11/14). A related group of cells, also termed accommodating, produced only one action potential during the depolarization regardless of the magnitude of the depolarizing current pulse (Fig. 1Ci, n = 21). The majority of the cells, (67%, 14/21), that fired single action potentials produced a depolarizing sag during hyperpolarizing steps (Fig. 1Cii).
Infrequently we encountered neurons that fired sporadically with no clear pattern when depolarized. Most of these irregular firing cells also produced depolarizing sags during hyperpolarizing steps (Fig. 1Dii, 4/5).
The last group of cells in the standard category fired action
potentials with a variable delay, during which the
Vm slowly ramped toward the action
potential firing threshold during depolarization. These cells also
produced subthreshold oscillations during the ramp period before the
initial action potential (Fig. 1Ei, n = 10).
In contrast to the other standard cells, 50% of these delay cells did
not produce a depolarizing sag during a hyperpolarization (Fig.
1Eii, 5 of 10). The properties of this cell type resembled those observed in mitral cells (Chen and Shepherd 1997).
Based on their electrophysiological properties, we classified these
four types of neurons as one group of JG cells. All had very large
input resistances (1,027 ± 123 M, n = 46).
JG cells with LTSs
The second large group of JG cells produced a LTS evoked by
depolarization from resting Vm or from
an anode break response following a hyperpolarizing current pulse (Fig.
2, A and B). All cells with an LTS also had a
depolarizing sag during a hyperpolarization (Fig. 2, Aii and
Bii). The presence of a depolarizing sag in all cells that
produce an LTS is significant since other neuronssuch as
thalamocortical cells
that produce behaviorally important slow oscillations (during sleep), critically depend on the interplay between
the LTS and a depolarizing sag to produce slow (~4 Hz) oscillations.
Termination of the hyperpolarizing pulse produced an overshoot of the
resting Vm and subsequently induced an
LTS. Generally, fast action potentials accompanied the LTS (Fig. 2, Aii and Bii, bottom). A variable number of fast
action potentials occurred on the crest of the LTS (0-10; mean
2.4 ± 0.2, n = 77; Fig. 2, Aii and
Bii). The frequency of these fast action potentials was in
the gamma range (67 ± 4 Hz, n = 38). The duration
of the LTS from the point of inflection to the corresponding
Vm on the falling phase varied between
75 and 960 ms (mean 298 ± 2 ms, n = 73).
Most experiments were performed at room temperature; however the presence of the LTS was sensitive to prolonged recording and particularly to elevated temperatures. At physiological temperatures (35-37°C), the LTS duration was significantly decreased by roughly half (141 ± 28 ms, n = 5, P < 0.002, range 70-215 ms), but physiological temperature did not significantly affect the number of action potentials (3.6 ± 0.9, n = 5, P > 0.25). Although the frequency of fast action potentials appeared to increase at physiological temperature, the difference did not reach significance, probably because of the low number of cells recorded (163 ± 34 Hz, n = 4, P > 0.05).
At room temperature a few cells (n = 3) produced plateau potentials (Fig. 2C, i and ii). Similar to cells with LTSs, plateau potentials could be produced by depolarizing the cell from resting Vm (Fig. 2Ci) or from an anode break response following a hyperpolarizing current injection (Fig. 2Cii). Plateau potentials produced fast action potentials (1-7 ms; mean, 3 ± 2) and plateau durations lasted longer than a second (1,450-7,000 ms; mean, 4,817 ± 1,708). However, given their infrequency, they were not investigated further.
We collectively termed the cells having an LTS and those producing
plateau potentials as bursting cells. Both the "standard" firing
cells and the bursting cells had indistinguishable, very high-input
resistances [1,027 ± 123, n = 46 (bursting) vs.
1,005 ± 30 M, n = 61 (standard),
P > 0.88].
Ionic basis of LTSs
We next investigated the ionic basis of the LTS. To determine whether Na+ influx was required, we first replaced extracellular Na+ with choline (Fig. 3A). Choline eliminated fast action potentials, but the LTS persisted (Fig. 3A, middle, n = 3). In a second manipulation, we included QX314 (5 mM), an inhibitor of voltage-dependent sodium channels, in the intracellular solution (Fig. 3B). QX314 inhibited the generation of fast action potentials but did not prevent the activation of an LTS (n = 6). Therefore sodium influx does not appear to be required for the production of an LTS in JG neurons.
|
LTSs in neurons and nonneuronal cells often result from activation of
low-voltage-activated calcium currents (LVA) (Huguenard 1996). We tested this possibility by examining the effect of
antagonists to these channels on the LTSs. High concentrations of
nickel (1 mM) completely and reversibly inhibited the LTSs in JG
neurons (Fig. 4A,
n = 13). Lower concentrations of nickel did not
completely inhibit the LTSs. Similarly, high concentrations (5 mM) of
alpha-methyl-alpha-phenylsuccinimide (MPS), which are known to inhibit
LVA calcium currents (Huguenard 1996
), completely
blocked the LTSs; lower concentrations (1 mM) of MPS suppressed but did
not completely inhibit the LTSs (not shown) (Fig. 4B,
n = 5). Therefore the LTSs in JG neurons appears to
arise from the activation of LVA calcium channels.
|
Physiological stimuli activate LTSs in JG neurons
Although an LTS could be evoked by nonphysiological stimuli such
as intracellular current injection, if they are involved in bulbar
information processing, more physiologically relevant stimuli should
activate them as well. Excitatory synaptic inputs to JG cells arise
from the sensory neuron axons and mitral/tufted cell dendrites. Both
use glutamate as the excitatory transmitter (Bardoni et al.
1996; Berkowicz et al. 1994
; Chen and
Shepherd 1997
; Ennis et al. 1996
;
Trombley and Shepherd 1992
). Given that an LTS could be
evoked by depolarization from the resting
Vm (Fig. 2, Ai and
Bi), we tested whether an LTS could be evoked by excitatory synaptic input to the JG cells or by photolysis of caged glutamate onto
the dendritic region of the cell (Kandler et al. 1998
).
With a stimulating electrode in the sensory nerve layer, brief stimuli
evoked excitatory synaptic input onto JG cells, which in turn elicited
an LTS (Fig. 5A,
n = 25). Paired synaptic stimulation with intervals
varying between 100 ms and 1 s produced paired-pulse inhibition
sufficient to prevent the occurrence of a second LTS (Fig.
5A). An LTS could also be evoked in JG neurons by flash photolysis of caged glutamate onto JG cell dendrites (Fig.
5B, , n = 7). Interestingly, many JG
cells displayed considerable spontaneous synaptic activity (Fig.
5D). In many JG cells maintained at their resting membrane
potential, these spontaneous excitatory postsynaptic potentials (EPSPs)
were sufficient to produce an individual spontaneous LTSs or a series
of spontaneous LTSs (Fig. 5D, n = 25 cells).
These results demonstrate that an LTS can be evoked in JG cells by
excitatory synaptic input and activation of glutamate receptors.
|
Given that the LTSs could be produced following a hyperpolarization, we
tested whether an LTS in JG cells could also be produced by an
inhibitory PSP (IPSP). By placing a stimulating electrode in the
glomerular region, we could periodically evoke IPSPs. On occasion this
IPSP could produce an LTS (Fig. 5C, n = 3).
Therefore as in other regions of the CNS (McCormick and Bal
1997), an IPSP in JG cells can elicit an LTS following the
return of the membrane potential to rest. In addition to these various
methods of evoking an LTS, we found that JG cells could also exhibit
spontaneous rhythmic bursting (Fig. 5E, n = 11). The oscillations in JG cells varied in frequency but fell in the
delta range (1.7 ± 0.2 Hz) as seen in the thalamus
(McCormick and Bal 1997
).
Therefore an LTS can be activated by a range of synaptic mechanisms, including spontaneous and evoked EPSPs and IPSPS; these LTSs also endow JG cells with the ability to oscillate spontaneously.
Morphology of JG neurons exhibiting LTSs
Based on morphological criteria, neurons in the JG region comprise
several different classes (Shipley and Ennis 1996), and each class is heterogeneous (Kosaka et al. 1998
). Given
that not all neurons in the glomerular region produced an LTS (Fig. 1), we attempted to correlate the morphological classes of the JG neurons
with their observed firing patterns.
JG neurons that produced an LTS fell into two morphological categories
(Fig. 6). One group had morphological
characteristics typical of external tufted (ET) cells. For the most
part, the dendrites were devoid of spine-like appendages (Fig.
6C, i and ii), although some dendrites
showed sparse spines (Fig. 6Aii, arrow). These cells often
had axonal projections emerging from the glomerular region and
penetrating into the internal plexiform layer (IPL) and/or granule cell
layer (Fig. 6C). All these characteristics are well-defined
features of ET cells (Pinching and Powell 1971a).
|
However, not all JG neurons that produced an LTS displayed
morphological features of external tufted cells. The dendrites of some
cells did not completely fill the glomerulus, their dendrites had many
large spine-like appendages (gemmules), and their axons were confined
to the glomerular region (Fig. 6B, i and ii).
Based on previous morphological studies, these cells have the
characteristics of inhibitory periglomerular cells (Pinching and
Powell 1971a). Therefore at least two morphological classes of
JG cells can produce an LTS.
JG cells that did not produce an LTS had morphological features that overlapped with cells that did produce an LTS (Fig. 7). However, short axon cells rarely produced an LTS (Fig. 7A). Short axon cells had sparse dendrites and axons whose arbors avoided glomeruli and were confined to the periglomerular region (Fig. 7A). Some external tufted cells, characterized by their lack of spines, beaded dendrites (Fig. 7B), and/or presence of secondary dendrites along the bottom of the glomerular layer (Fig. 7C) lacked an LTS (Fig. 7, B and C). Some periglomerular neurons also did not produce an LTS (Fig. 7D). These cells were usually smaller than those that did exhibit an LTS but otherwise had dendritic gemmules and axons confined to the glomerular layer characteristic of cells of this class.
|
While considerable heterogeneity in both morphology and physiology was
observed, some trends were clear (Tables
1 and 2). Roughly two-thirds of external
tufted and periglomerular cells produced an LTS, while two-thirds of
short axon cells did not. Table 2
summarizes the firing properties of 25 morphological neurons that did
not produce an LTS. There was no clear single type of firing property
for periglomerular or external tufted cells; however, most short axon
neurons produced accommodating trains of action potentials. Two of
these neurons that fired with a delay were external tufted cells, which
is not surprising given that the delay to firing pattern is similar to
that seen in the other excitatory neurons in the olfactory bulb, the
mitral cells (Chen and Shepherd 1997).
|
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Based on modes of firing, we have found that JG cells fall into
two different physiological classes: standard nonbursting cells and
bursting neurons that exhibit a low-threshold calcium-dependent spike. Nonbursting neurons exhibited a range of firing properties, from
regular trains of action potentials to highly irregular patterns of
spikes. Firing patterns did not strongly correlate with the morphological class of JG cell from which we recorded. A similar lack
of correspondence between interneuron morphology and physiology has
been observed in other cortical structures such as the hippocampus, where interneurons of similar morphology have striking differences in
their passive electrical properties and action potential firing patterns (Cauli et al. 1997; Mott et al.
1997
; Parra et al. 1998
). This heterogeneity
could be explained by numerous factors, including the distribution of
ion channels; indeed there is evidence for considerable heterogeneity
in the distribution of potassium channels in JG cells, which would be
expected to alter firing patterns (Puopolo and Belluzzi
1998
).
However, some consistencies were observed. The majority of short axon
cells fired accommodating trains of action potentials, and two of three
cells that showed delay to firing when depolarized were external tufted
cells. This is consistent with previous results showing that the delay
to firing pattern is characteristic of mitral and deep tufted cells
(Chen and Shepherd 1997).
Our studies show additional similarities to previous investigations.
Studies in vivo using both extracellular and intracellular techniques
have demonstrated a class of neuron in the glomerular layer that fire
bursts of action potentials similar to our bursting neurons
(Freeman 1974a; Shepherd 1963
;
Wellis and Scott 1990
). There are some differences in
the bursting we observed and that of previous studies. Most obvious is
that the duration of our bursts is 100-200 ms longer than those
reported in vivo. There are a number of possible reasons for this
discrepancy. First, most of our experiments were conducted at room
temperature. Increasing the temperature of the slice did reduce the
duration of our bursts, albeit not down to the durations observed in
vivo. We produced and measured our bursts by directly changing the
membrane potential of neuron. The in vivo studies used synaptic
activation that would activate additional neurons that could inhibit
and reduce the duration of the burst. The extracellular studies
measured the duration of spiking (Freeman 1974a
;
Shepherd 1963
), which is considerably shorter than the
entire envelope of the burst. And finally, there could be a difference
created by the very difficult experiments obtained with leaky
intracellular recordings (Wellis and Scott 1990
) and
those obtained with tight whole-seal recordings.
In contrast, previous studies in vitro have not observed bursting cells
in the glomerular layer. Whole cell recordings in olfactory bulb slices
have claimed that all JG neurons fire single or trains of spikes
similar to our standard firing cells (Bardoni et al.
1995; Puopolo and Belluzzi 1996
, 1998
). A
different study showed that PG neurons do not fire spikes at all
(Bufler et al. 1992
). A possible reason previous studies
did not observe the bursting behavior in vitro may be due to the whole
cell patch-clamping technique; we observed that the bursting phenomenon
was sensitive to prolonged recording and activation. Prolonged
recording and excessive activation resulted in a hyperpolarization of
the bursting JG cell resting potential, a reduction in the input
resistance of the cell, and the ability to fire only one fast action
potential. The previous observation that PG cells do not fire any
spikes at all may differ from our observations because of species and developmental differences. Previous studies used P4 rabbits, while we
used P13-21 rats.
The neurons generating bursts based on LTSs are of particular interest.
The behavior of these cellsespecially the circumstances that elicit
LTSs
resembles that of the reticular nucleus of the thalamus and
thalamocortical neurons (Andersen et al. 1968
;
Deschenes et al. 1984
; Huguenard and Prince
1992
; Roy et al. 1984
). LTSs in both of these
types of thalamic neurons results from activation of a LVA calcium
current (Huguenard and Prince 1992
; Jahnsen and Llinas 1984a
,b
). LVA calcium channels also appear to underlie the LTSs in JG neurons, as antagonists of LVA channels could completely inhibit the LTSs. The concentrations of antagonists required to inhibit
the LTSs in JG cells were high but within the widely variable range of
concentrations observed to inhibit LVA currents in different cell types
(for review, see Huguenard 1996
; Perez-Reyes
1998
). Furthermore, inhibition of voltage-dependent sodium
channels by the antagonist QX314, and inhibition of sodium influx by
the removal of extracellular sodium, did not inhibit the LTSs.
Consistent with our findings, a LVA calcium channel has been cloned
from mouse (Klugbauer et al. 1999
) and is expressed
widely throughout the olfactory bulb. However, two potential
alternative explanations for the mechanism producing the LTSs might be:
1) the release of glutamate from the recorded JG neuron acts
on glutamatergic autoreceptors to produce the LTSs and 2) a
network mechanism in which glutamate is released from the recorded JG
neuron, resulting in excitatory feedback from other neurons to produce
the LTSs. However, these alternatives are inconsistent with our data.
First, LTSs could be produced in the presence of glutamatergic receptor antagonists that should inhibit both glutamatergic autoreceptors and
network activity. Second, LTSs were observed in the presence of sodium
channel blockers that should inhibit synaptic transmission. Finally,
LTSs could be observed in JG neurons with morphological characteristics
of GABAergic interneurons, suggesting that activation of some JG
neurons should produce inhibition, not excitation. Taken together our
data are most consistent with LTSs arising from the activation of LVA
calcium channels.
Morphology of neurons with LTSs
The morphological studies of Pinching and Powell
(1971a-c
) classified JG neurons into different categories: PG,
ET, and short axon neurons. The PG neurons, which are thought to be
inhibitory (Duchamp-Viret et al. 1993
; Pinching
and Powell 1971a
-c
; Puopolo and Belluzzi 1998
;
Shepherd 1971
; but see Freeman 1974a
;
Siklos et al. 1995
), have small cell bodies, spine-like
processes (gemmules), and dendrites that usually did not completely
fill a glomerulus. Short axon cells, also thought to be inhibitory,
have spineless dendrites with processes that avoided entering
glomeruli. The excitatory ET cells exhibit dendrites with few spines.
These three cell types are richly interconnected, with PG cells forming
local dendrodendritic and lateral axodendritic connections with ET, mitral-tufted, and other PG cells. ET cells form dendrodendritic connections with PG cells and lateral axonal connections onto both
short axon and PG cells. Importantly, axons of ET cells synapse on
olfactory bulb neurons in other layers, including granule cells (for
review, see Shipley and Ennis 1996
). Short axon neurons
appeared to be interneuron selective, only forming axonal connections
with PG cells.
Although our results show that examples of all three classes that could occasionally produce an LTS, almost all (86%) of JG cells producing LTSs were either external tufted or periglomerular cells. Thus both excitatory and inhibitory JG neurons appear to be capable of producing an LTS. Given that these JG neurons are richly interconnected to one another, and to neurons in other regions of the olfactory bulb, the LTSs in these cell types may have important physiological functions in olfactory processing within the olfactory bulb.
Physiological implications of the LTS in JG neurons
We found that the LTSs in JG cells could be activated by
physiological phenomenon such as IPSPs and EPSPs; the LTSs also endowed JG cells with the ability to spontaneously oscillate. In vivo studies
have previously demonstrated that superficial cells in the bulb produce
bursts of spikes (Freeman 1974a; Schneider and Scott 1983
; Shepherd 1963
). Subsequent studies
demonstrated that at least a subset of these neurons were PG neurons
and that these fired in bursts or plateaus (Wellis and Scott
1990
). Interestingly, these cells could be activated by
stimulating anywhere in the sensory nerve layer, outside the area that
would directly stimulate sensory input to the recorded cell, which
implies extensive interconnections among these bursting superficial
cells (Freeman 1974a
,b
; Wellis and Scott
1990
). Superficial cells could also be rhythmically activated
during an artificial sniff cycle, even in the absence of an odor
(Onoda and Mori 1980
). Interestingly, these superficial cells fired either in phase, or out of phase with inspiration (Onoda and Mori 1980
). The out of phase cells were
spontaneously active when the receptor neurons were not activated,
whereas the in phase cells were silent. On inspiration and stimulation
of the receptor neurons, the in-phase cells burst, and the out of phase
became silent, possibly due to the inhibition of the out of phase cells
by the in phase cells. This suggests that a network oscillation driven
by the inspiration cycle may occur between neurons in the glomerular
layer. Furthermore, the granule cell layer produced a synchronous
population oscillation during the inspiration that was delayed relative
to the activation of in phase JG cells. In contrast, the output mitral
and tufted cells showed only a small amount of firing, which was rarely
correlated with the inspiration cycle or was very complicated relative
to the glomerular cells and granule cell rhythm.
There is a significant amount of evidence suggesting that oscillations
occurring in the deeper layers arise from the synaptic interactions
between mitral/tufted cells and granule cells (Desmaison et al.
1999; Freeman 1974a
; Kashivadini et al.
1999
; Mori and Takagi 1978
). However, the
experiments of Onoda and Mori (1980)
show that
mitral/tufted cells do not show coherent activity when oscillations
occur in the granule cell layer. The question arises, is this minimal
activity in mitral/tufted cells alone enough to generate the
rhythmicity in granule cells? We suggest that because activation of JG
neurons during inspiration precedes oscillations in the granule cell
layer and that many bursting neurons in the glomerular layer have axons
that project toward the granule cells, that the projection of bursting
neurons to deeper layers of the olfactory bulb may generate
oscillations in the granule cell layer.
This rhythmic synchronization of action potentials appears to play an
important function in olfactory processing in the olfactory bulb.
Synchronization from distributed principal cells in a variety of
different organisms is associated with olfactory discrimination tasks
(for reviews, see Laurent 1996, 1999
; Mori et al.
1999
). Slow oscillations, such as those that the LTS in JG
cells can produce, are evident in the olfactory bulb of several species (Delaney and Hall 1996
; Lam et al. 2000
).
In rats, spatially distributed principal cells synchronize at 30-60
Hz, much faster than the bursting of JG cells. However, these fast
oscillations are modulated by the slower rhythms associated with the
breathing rhythm or superimposed on top of the slower rhythm
(Freeman 1978
; Freeman and Viana Di Prisco
1986
).
Interestingly, the properties of the LTS in JG cells, the phenomena
that activates this LTS, the cell types that produce an LTS, and the
rich synaptic interactions between cells that produce an LTS are
similar to the thalamus. It is these properties of the thalamic neurons
that allow them to oscillate and generate biologically important
rhythms such as spindle waves and slow wave sleep (for review see
McCormick and Bal 1997). Given the similarities of JG
neurons to rhythmic thalamic neurons, that JG neurons may rhythmically
burst in vivo, and that JG neurons with an LTS project extensively
throughout the bulb, we propose that the LTSs in JG cells may be
important for contributing to the generation of synchronous rhythms
throughout the olfactory bulb during olfactory processing.
![]() |
ACKNOWLEDGMENTS |
---|
The authors thank M. Gray for assistance with the histology and Drs. Leo Belluscio and Tom Tucker for helpful discussions.
![]() |
FOOTNOTES |
---|
Address for reprint requests: L. C. Katz, Duke University Medical Center, Dept. of Neurobiology, Box 3209, Bryan Research Building, Research Dr., Durham, NC 27710 (E-mail: larry{at}neuro.duke.edu).
Received 8 January 2001; accepted in final form 26 June 2001.
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|