INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Fast oscillatory neuronal network activity is a prominent feature of sensory (Ribary et al. 1991), motor (Sanes and Donoghue 1993), and mnemonic (Fell et al. 2001) processing in many different brain regions. Recently, there have been marked advances in understanding the neuronal circuitry underlying high-frequency oscillations in cortical areas through both computational and experimental approaches (Traub et al. 1999). Studies in hippocampus (Whittington et al. 1995) and neocortex (Buhl et al. 1998) have shown that an autonomous network of inhibitory interneurons can synchronize firing in sub-populations of principal cells. In these cortical regions, oscillatory network activity is facilitated through electrical coupling between interneurons (Beierlein et al. 2000; Hormuzdi et al. 2001). Interestingly, mechanisms underlying cortical fast oscillations appear to have common features despite the great differences in the behavioral functions processed in these areas. However, it is unknown whether these same mechanisms also underlie fast oscillations in the olfactory bulb.

In the olfactory system, odorant presentation is associated with prominent, fast (20-70 Hz) electroencephalographic (EEG) activity that can be recorded in the olfactory bulb (OB) (Adrian 1950; Eeckman and Freeman 1990) or its invertebrate analog (Laurent and Naraghi 1994) in vivo. Odor-induced -frequency oscillations are associated with transient action potential (AP) synchrony between sub-populations of mitral cells (MCs), the olfactory bulb principal cells (Eeckman and Freeman 1990), or projection neurons, their insect counterparts (Wehr and Laurent 1996). Indirect evidence suggests that these oscillations and associated mitral cell synchrony are governed by inhibitory activity. Pharmacological (Stopfer et al. 1997) manipulation of inhibition impairs both local field oscillations and odor discrimination in honeybees. In rodents, genetic manipulation of granule cell excitability enhances field oscillations and alters discrimination (Nusser et al. 2001). In addition, in some species there is evidence that olfactory fast oscillations exhibit stimulus-evoked plasticity; repeated odor presentation leads to increased oscillatory activity which corresponds to decreased principal cell firing and increased spike synchrony between principal cell pairs (Stopfer and Laurent 1999).

While there is evidence for the behavioral importance of -oscillations in olfactory information processing, the cellular mechanisms underlying this network activity have not been defined. Theoretical and experimental studies have focused on how oscillatory activity can arise between mitral cells and inhibitory granule cells coupled through reciprocal dendrodendritic synapses (Freeman 1978; Isaacson and Strowbridge 1998; Rall et al. 1966). Other studies have examined how individual mitral cells can generate -frequency activity as a result of intrinsic electrical properties (Desmaisons et al. 1999). However, none of these studies examined how oscillatory activity occurs in populations of olfactory bulb cells and how elements in this network are synchronized. We developed an in vitro model of odor-evoked -oscillations to define the cellular mechanisms this activity. In acute olfactory bulb slices, we have found that repetitive glomerular stimulation induces persistent fast network oscillations and periodic inhibition of mitral cells. These network oscillations are mediated by both chemical and electrical synapses, providing the first physiological evidence for a role of gap junction coupling in olfactory sensory processing.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Horizontal slices (300-400 µm) were prepared from the olfactory bulbs of 11- to 26-day-old anesthetized (ketamine 150 mg/kg ip) Sprague-Dawley rats as previously reported (Isaacson and Strowbridge 1998). Four hundred micron thick slices were placed in an interface-type recording chamber where they were perfused with warmed (34-35.5°C), oxygenated solution (artificial cerebrospinal fluid, ACSF) containing the following (in mM): 124 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 1.3 MgSO₄, 26 NaHCO₃, 2.5 CaCl₂, and 10 dextrose in a humidified atmosphere saturated with 95% O₂-5% CO₂. Slices maintained in the interface chamber were used for experiments examining field and multi-unit activity. Mini-slices were made by dissecting away the glomerular and olfactory nerve layers under a dissecting microscope prior to incubation in the recording chamber. Three hundred micron thick slices were used in a submerged-type chamber and perfused with oxygenated ACSF (30-31°C). Slices maintained in this type of chamber were required for whole cell and cell-attached patch-clamp recording from identified cell types. Mitral and granule cells were visualized using an upright microscope (Axioskop FS, Carl Zeiss) equipped with infrared/differential interference contrast (IR/DIC) optics and a 63× water immersion objective.

Field potentials were recorded using glass micropipettes (10- to 15-µm tip diameter) and an AC-coupled amplifier with a differential headstage (ER-98, Cygnus). Interface chamber field recordings were filtered between 1 and 1000 Hz; submerged chamber field recordings were filtered between 1 and 300 Hz. Unit activity was recorded using tungsten microelectrodes (9-11 M) and band-pass-filtered at 0.1-10 kHz. Whole cell recording micropipettes (2-5 and 4-7 M for mitral and granule cells, respectively) contained the following (in mM): 115 Cs-methanesulfonate, 25 TEA-methanesulfonate, 4 NaCl, 10 HEPES, 1 EGTA, 4 MgATP, 0.3 Na₃GTP, 10 phosphocreatine, and 5 QX-314 to block fast Na⁺ currents (pH 7.3). In current-clamp and cell-attached experiments, 140 mM KMeSO₄ was substituted for Cs- and TEA-methanesulfonate and QX-314 was omitted. Series resistance, typically <8 M, was routinely compensated by >80%. Neurons with resting V_m < 55 mV or non-overshooting action potentials were excluded. After completion of cell-attached recordings, we routinely converted to whole cell mode to assess the intrinsic properties of that cell. Current-clamp and cell-attached recordings were low-pass-filtered at 5 kHz, while voltage-clamp recordings were low-pass-filtered at 2 kHz. All recordings were digitized using a 16-bit A/D board (Instrutech ITC-18) at 5-10 kHz.

Constant-current stimulation pulses (200-µs duration; typically 36-60 µA) were delivered using 25-µm-diameter concentric bipolar (interface chamber) or insulated tungsten monopolar electrodes (9-11 M, submerged chamber) at intensities 2-2.5 times threshold for evoking a field excitatory postsynaptic potential (EPSP). Simulated excitatory postsynaptic currents (EPSCs) for current injection were generated using an -function [I(x) = I_max  (e^x/1  e^x/2)/(_1  2); ₁ = 45 ms, ₂ = 35 ms]. Drugs were delivered focally using pressure pulses to a micropipette with a broken tip in the interface recording chamber. In submerged chamber experiments, drugs were applied by switching extracellular solutions. All drugs were obtained from Sigma except TTX (Calbiochem).

Analysis was performed off-line using custom Igor Pro (Wavemetrics) macros. Power spectral densities (PSDs) before and after the tetanus in each trial were computed from the average of eight overlapped epochs (0.5 s, overlapping every 0.25 s) windowed using a Welsh function. Prior to calculation of the PSD for each trial, these records were decimated to 256 Hz and digitally high-pass-filtered at 3 Hz. Relative changes in spectral density ( power) were calculated by subtracting the pretetanus PSD from the posttetanus PSD. Spectrograms shown in the figures reflect the pooled data from 10 to 20 responses from each slice. Unless noted, spectral power is reported as the integral of the PSD between 15 and 50 Hz. Example field potentials shown in the figures are the raw digitized records except they were smoothed using a sliding average function (box size = 9 samples or 1.8 ms) for clarity. In addition, trend lines created using a three-point sliding average are added to time plots shown in the figures for clarity. Statistical comparisons were made using paired, two-tailed Student's t-test except where noted. Means are reported as ±SE.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Neurons in the olfactory bulb are normally activated by excitatory afferents from the olfactory receptor neurons (Shepherd and Greer 1998). We mimicked this sensory synaptic input by applying a brief, focal tetanic stimulation (10 stimuli at 100 Hz; 36-60 µA) in the glomerular layer in acute rat olfactory bulb slices while recording extracellular field potentials in the border between the external plexiform and mitral cell layers (EPL/GCL). We found that glomerular tetanization led to a transient enhancement in the field power activity in raw field potential recordings from slices maintained in interface-style recording chambers (Fig. 1A). We initially analyzed this stimulus-evoked increase in field power using PSD plots (Fig. 1B). This analysis showed that tetanic stimulation caused a broad increase in spectral power between 5 and 50 Hz with a peak near 20 Hz. The increase in post-tetanus field power required five or more stimuli (mean  power between 15 and 50 Hz: 304 ± 110 pV²/s with tetanization compared with 17.0 ± 19 pV²/s in control trials without stimulation; means significantly different: P < 0.001; Students' t-test). A similar potentiation of ongoing field potential activity could also be recorded in slices maintained in submerged recording chambers though field potential activity was generally smaller in submerged slices (mean  power = 58.6 ± 27 pV²/s with tetanization; n = 19 slices). Pseudo-color images made from time plots of PSD data (Fig. 1C) showed that tetanus-evoked activity lasted for 1-2 s and tended to decrease in frequency near the end of the response.

View larger version (34K): [in this window] [in a new window]   Fig. 1. Evoked fast network activity in the olfactory bulb in vitro. A: field potential recording of response to glomerular tetanus (10 @ 100 Hz; interface recording chamber). B: tetanization reliably enhances the power spectral density (PSD) between 10 and 30 Hz (black plot) over control levels (red plot), while TTX abolishes most of the spectral power (blue plot). The PSD plots were computed from the time periods indicated by the colored bars in A. C: plot of tetanus-evoked increase in spectral power from baseline ( power) vs. poststimulus time (s) for the same experiment shown in A.

We next asked whether glomerular stimulation evoked periodic activity that matched the approximately 20-Hz spectral peak we found in the PSD analysis. Autocorrelation analysis of field potential activity before and after stimulation (Fig. 2A) showed clear periodic sidebands in records after tetanization (Fig. 2B), suggesting that glomerular stimulation evokes periodic oscillatory network activity in the olfactory bulb. Similar periodic autocorrelograms were observed in field potential recordings from 38 olfactory bulb slices immediately after tetanization (mean frequency: 23.9 ± 1.0 Hz; range 14.8-38.1 Hz; Fig. 2E). We found that oscillatory network activity in the olfactory bulb in vitro is activated most strongly by stimulation of the glomerular layer; tetanic stimulation in the EPL or GCL evoked a much smaller enhancement of network activity (48.4 ± 24 and 59.2 ± 39 pV²/s, respectively; n = 8 slices).

View larger version (26K): [in this window] [in a new window]   Fig. 2. Fast inhibitory currents underlie oscillatory network activity in the olfactory bulb. A: brief tetanic stimulation (10 @ 100 Hz) of the on-beam glomerulus transiently increases field potential activity in 3 consecutive records. B: autocorrelograms before and after glomerular tetanization computed from the recordings in A over the indicated time periods. C: intracellular responses in mitral cells under voltage clamp to glomerular tetanization at different holding potentials (indicated above each trace). Tetanization evoked spontaneous IPSC activity that reversed polarity near 70 mV and was abolished by 50 µM picrotoxin (bottom). D: autocorrelograms of membrane currents before and after tetanization in the example recordings shown in C. E: summary of autocorrelation experiments showing that tetanization evoked periodic activity at approximately 22 Hz in field potential recordings in both interface (FP-Int) and submerged (FP-Sub) recording chambers and in intracellular recordings from mitral cells in submerged recording chambers (inhibitory postsynaptic current, IPSC).

We next sought to examine the cellular mechanism underlying the transient enhancement of network activity evoked by glomerular stimulation. Single tetani generated inhibitory current fluctuations in MCs recorded under voltage-clamp (Fig. 2C). These current fluctuations appeared to be inhibitory postsynaptic currents (IPSCs) since they were outward currents at depolarized holding potentials (50 mV; n = 6 cells), reversed polarity near 70 mV, and were blocked by picrotoxin (50 µM). Autocorrelograms demonstrated a periodicity in these IPSC fluctuations that was similar to that observed in the field potentials (current mean frequency: 23.9 ± 2.4 Hz; 12.0-47.6 Hz range; Fig. 2, D-E). These results suggest that the field potential oscillations and synaptic currents in mitral cells evoked by glomerular stimulation likely reflect activity in networks of inhibitory neurons, presumably granule cells.

We next used extracellular unit recordings to determine if spiking activity in GABAergic granule cells was correlated with the tetanus-evoked trains of IPSCs recorded in mitral cells. Following brief tetanization, we observed a large, transient increase in multi-unit granule cell activity (6.5 ± 1.6 times pretetanus spike frequency; n = 6 slices; P < 0.05; Fig. 3). The duration of increased granule cell activity was similar to the 2- to 3-s time course of tetanus-evoked enhancement in network oscillations shown in Figs. 1 and 2. These findings further support the idea that tetanus-evoked MC IPSCs and field oscillations are associated with increased granule cell network activity. In some slices, we also observed that repeated tetanization appeared to produce a persistent increase in granule cell unit activity (see Basal Firing Rate plot in Fig. 3B). This effect was difficult to analyze in complex multi-unit recordings where an unknown number of cells contribute to the spike raster plots. We turned instead to field potential and single-neuron measurements to determine if glomerular stimulation produced long-lasting changes in olfactory bulb network activity.

View larger version (28K): [in this window] [in a new window]   Fig. 3. Tetanization increases granule cell unit activity. A: schematic diagram showing the location of the field potential electrode (FP), extracellular electrode (GC-MU), and stimulating electrode (Stim). B: simultaneous field potential and granule cell layer extracellular recording before and after tetanic glomerular stimulation. The frequency of multi-unit activity in the granule cell layer is transiently increased following tetanic stimulation at the same time as the increased activity in the FP recording. The time course of increased granule cell spiking activity is resolved in the raster records (responses to 31 consecutive trials are shown) and in the spike probability graph. Repetitive tetanization appears as increased basal granule cell unit activity (gray bars left of raster plot).

Repeated odor presentation leads to a persistent enhancement in oscillatory field potentials in olfactory brain areas in vivo (Stopfer and Laurent 1999). In olfactory bulb slices, in addition to the immediate effect of stimulation, we found that repeated (interval = 40-45 s) tetanic stimulation led to a pronounced build-up of inter-stimulus field potential activity (Fig. 4, A and C). We quantified the build-up in network activity with conditioning using slices maintained in a submerged recording chamber (Fig. 4B). Slices maintained in this type of recording chamber exhibited spontaneous oscillations, albeit with a lower basal field power (5.3 ± 1.4 pV²/s) than similar slices maintained in interface chambers (67.0 ± 11 pV²/s; n = 19 slices). Although the majority of this difference is due to the lower extracellular resistance of the submerged slices, the reduced number of neurons in the thinner submerged slices may contribute to decreased network activity. We observed a statistically significant increase in spontaneous and tetanus-evoked field power after 5 min of conditioning (229 ± 49 and 240 ± 46% of baseline, respectively; P < 0.05; n = 19 slices). Spontaneous field potential activity continued to increase with additional stimulation until it reached a maximal level (388 ± 52% of preconditioning power; n = 19 slices; Fig. 4B) after 10 min of stimulation. Spontaneous field potential activity was stable in unstimulated slices (101 ± 3.1% of control power over 5 min; not statistically different; n = 8 slices; data not shown). Repeated stimulation also increased inter-stimulus power in slices maintained in interface recording chambers (160 ± 19% of the preconditioning power; P < 0.05; n = 8 slices; Fig. 4, C and E). The smaller percentage increase observed in interface chamber recordings is likely due to the higher basal (unstimulated) field power in these slices.

View larger version (34K): [in this window] [in a new window]   Fig. 4. Repeated tetanization leads to a persistent increase in field potential activity. A: build-up of basal field power with conditioning stimulation in the submerged recording chamber. Glomerular tetani were repeated every 45 s from 0 to 20 min. Moving average-filtered trace (solid line) is superimposed on basal power measurements (recorded immediately before stimulation in each record). Average prestimulus power is indicated by the dashed line. Example traces shown above before repeated tetanization and following 20 min of tetanization. B: summary of persistent increase in basal (prestimulus) and poststimulus power with repeated tetanization. C: persistent increase in field potential power by repeated tetanization in an interface recording chamber. Repeated tetanization (every 45 s for 30 min) increases basal field potential power approximately twofold. Field potential power remained elevated for 60 min until 2-amino-5-phosphonovaleric acid (APV; droplet concentration 0.5 mM) was applied focally to the slice. Note breaks in the time axis. Moving average (solid line) superimposed over individual basal power measurements. D: field potential activity before and after bath application of TTX (1 µM). E: summary of average power during baseline periods (before stimulation), during conditioning (with tetani repeated every 45 s), 30 min after the final tetanus, and after focal application of APV or 1,2,3,4-tetrahydro-6-nitro-2,3-dioxo-benzo[f]quinoxa- line-7-sulfonamide (NBQX). *P < 0.05; **P < 0.01.

The potentiation in field potential activity induced by conditioning was persistent and did not change significantly 30 min after conditioning (Fig. 4E; interface recording chamber; n = 4 slices). However, both ongoing and potentiated field potential activity could be eliminated by bath application of TTX (1 µM; n = 3 slices; Fig. 4D), suggesting that repeated tetanization induced persistent Na⁺-based spiking activity in a subpopulation of olfactory bulb neurons. Mitral cells are known to interact with granule cells via glutamatergic synapses that contain both AMPA and NMDA receptors (Aroniadou-Anderjaska et al. 1999; Isaacson and Strowbridge 1998). Focal application of either NMDA receptor or AMPA receptor antagonists transiently abolished most of the persistent spontaneous network activity (18.8 ± 4.7 and 9.0 ± 4.9% of control preconditioning field potential power, respectively; n = 7 and 4 slices; P < 0.01; Fig. 4, C-D). These results suggest that persistent network oscillations in the olfactory bulb require glutamate receptor activation.

The previous results suggest that repeated olfactory bulb stimulation causes a persistent increase in the spiking activity of inhibitory neurons. To test whether this change in inhibitory tone had a functional effect on principal cells, we examined mitral cell responses to simulated EPSCs (100- to 200-pA peak amplitude, 2 Hz). The mitral cell membrane potential was adjusted until a simulated EPSC evoked an AP for the majority of the time (spike probability, P = 0.59 ± 0.1; n = 3 cells; Fig. 5A, left). Following 15 min of conditioning, there was a significant reduction in the probability of the same simulated EPSC in evoking an AP (P = 0.14 ± 0.05; P < 0.05; Fig. 5A, middle, 5B) without a significant change in mitral cell membrane potential (from 58 ± 5 to 57 ± 4 mV; not statistically different). We verified that following conditioning stimulation, mitral cells could still generate APs by applying DC depolarization. With this additional depolarization, EPSC injection evoked full-height APs following conditioning (Fig. 5A, right). Conditioning slightly reduced the mean mitral cell input resistance (from 107 ± 4 to 96 ± 7 M; not statistically different; Fig. 5C). To confirm that repeated tetanic stimulation induced a persistent increase in mitral cell inhibition, we performed cell-attached mitral cell recordings. In mitral cells that fired spontaneously under resting conditions, one to four tetanic stimuli significantly reduced this baseline activity (20.1 ± 1.4% of baseline AP rate; n = 3 cells; P < 0.05; Fig. 5D). We also confirmed that potentiation of persistent network activity was associated with an increase in spontaneous granule cell firing rate using cell-attached recordings. In five slices where conditioning evoked a persistent increase in field activity, basal firing rate was also enhanced in all granule cells tested (from 0.7 ± 0.3 to 6.9 ± 1.6 APs/s after 10 min; n = 5 cells; P < 0.05; Fig. 5E). These findings suggest that the persistent inhibitory network activity induced by repeated stimulation is sufficient to alter the responsiveness of mitral cells to synaptic inputs.

View larger version (32K): [in this window] [in a new window]   Fig. 5. Increased functional inhibition associated with persistent network activity. A: simulated excitatory postsynaptic currents (EPSC) injection (200 pA peak amplitude, 2 Hz) adjusted to evoke mitral cell action potentials in the majority of trials at baseline conditions (left). Middle: after 15 min of conditioning, the same stimulus failed to evoke action potentials in most trials at the same membrane potential. Right: firing occurs if the mitral cell is directly depolarized following conditioning. B: plot of the probability of evoking an AP with a simulated EPSC vs. time. C: summary of spike probability, input resistance (Rin), and membrane potential (Vm) in three slices before (empty bars) and after conditioning (filled bars). *P < 0.05. D: decrease in mitral cell spontaneous firing (cell-attached recordings) during conditioning. Decrease in mitral cell firing rate parallels the potentiation of oscillations in the field potential recording. E: conditioning-evoked increase in basal firing rate in cell-attached recordings from granule cells. Example traces are taken just prior to conditioning and after 4, 7, and 14 min of conditioning.

In hippocampus, recent studies (e.g., Hormuzdi et al. 2001) have shown a prominent role for gap junction coupling between interneurons in generating -frequency network activity. Anatomical studies suggest that gap junctions may be present between granule cells (Reyher et al. 1991) as well as between mitral cells and granule cells (Landis et al. 1974; Miragall et al. 1996). We used a series of pharmacological agents that alter gap junction coupling to test whether olfactory bulb network oscillations are mediated by electrical synapses. We found that halothane (5 mM) significantly decreased persistent field activity in six of seven slices tested (49.4 ± 9.7% of control; P < 0.05; Fig. 6, A-C). In addition, halothane also significantly reduced the tetanus-evoked enhancement in network activity in all seven slices (55.7 ± 17% of control; P < 0.05). Persistent and transient network oscillations were also reduced by octanol (1 mM; spontaneous: 26.2 ± 9.4%; evoked: 24.1 ± 9.5%; n = 4) and carbenoxolone (100 µM; spontaneous: 48.6 ± 14.3%; evoked: 47.7 ± 9.8%; n = 3; Fig. 6D), two other agents that decrease gap junction coupling (Davidson and Baumgarten 1988; Spray et al. 1985). By contrast, alkalinization with NH₄Cl (10 mM added to the ACSF), which increases gap junction coupling (Spray et al. 1985), enhanced both spontaneous and evoked network activity (156 ± 35 and 179 ± 21%, respectively; n = 4 slices).

View larger version (34K): [in this window] [in a new window]   Fig. 6. Gap junction blockers reduce network activity in the olfactory bulb. A: field potential records showing that halothane (5 mM) reversibly reduces both pre- and post-tetanus field potential activity. B: plot of the effect of halothane (application indicated by bar) on pretetanus and  field potential power. C: plot of reduction of prestimulus power for 7 slices exposed to halothane. Halothane reduced spontaneous field activity to 49.4 ± 9.7% (P < 0.05) of control power. D: summary of the effects of gap junction modulators on pre- and post-tetanus field potential power.

We next attempted to define the role of gap junction-mediated olfactory bulb network activity at the cellular level. In addition to reducing tetanus-evoked field oscillations, halothane also significantly reduced the frequency of post-stimulus IPSCs in five of six mitral cells tested (P < 0.05, Fig. 7A). The initial synaptic current evoked directly by the tetanus was not greatly altered by halothane (Fig. 7A, bottom). Gap junction blockers also reversed the persistent increase in inhibitory tone induced by conditioning. In cell-attached recordings from mitral cells (n = 3; Fig. 7, B-C), halothane reversibly increased spontaneous firing. These effects are unlikely to be due to nonspecific effects of halothane since halothane had no significant effect on mitral cell input resistance in naïve slices (n = 5 cells).

View larger version (31K): [in this window] [in a new window]   Fig. 7. Gap junction blockers decrease inhibitory activity in the olfactory bulb. A: halothane reduces the frequency and duration of posttetanus inhibitory postsynaptic currents (IPSCs) in voltage-clamped mitral cells (VH = 10 mV). Inset: stimulus-evoked EPSC is minimally affected by halothane and is abolished by APV (50 µM) and NBQX (10 µM). B: halothane reversibly increases basal mitral cell firing rate (cell-attached recordings). C: plot of the increase in mitral cell spontaneous firing rate with halothane (cell-attached recordings). D: focal application of halothane (50 mM, open arrow left of raster plot) reversibly blocks the tetanus-evoked (filled arrow) increase in granule cell extracellular unit activity as well as reducing basal firing rate (interface chamber recording). E: example records of multi-unit granule cell activity before, immediately after, and following washout of the focal halothane application. F: correlation between tetanus-evoked  field potential power and the tetanus-evoked increase in granule cell firing rate for the records shown in the raster plot above (r = 0.79, P < 0.0001).

Finally, we used extracellular recordings to assess the role of gap junctions in inhibitory networks. Focal application of halothane (50 mM) in the granule cell layer in four slices reversibly reduced both the persistent and the immediate tetanus-evoked increase in granule cell unit activity (Fig. 7, D-F). These effects of halothane on the field power and granule cell firing rate occurred in parallel and were consistent with the hypothesized role of granule cell activity in mediating network oscillations. Our results suggest that persistent, synchronous network activity in the olfactory bulb occurs through a combination of electrical and chemical synaptic interactions.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In this study, we demonstrate persistent approximately 20-Hz network activity in the olfactory bulb in vitro that is transiently enhanced following tetanization and potentiated with repeated stimulation. Both the transient and the long-term increases in field oscillations are associated with increased firing in inhibitory granule cell networks, leading to enhanced inhibition onto mitral cells. Potentiation of spontaneous, persistent activity with repeated stimulation enhances the functional inhibition of mitral cells. This fast inhibitory network activity depends on both chemical and electrical synapses, presumably onto granule cells.

Fast oscillations in the olfactory bulb

In this study, we demonstrate for the first time that synchronous, fast network oscillations can be evoked in the mammalian olfactory bulb in vitro. These oscillations are enhanced by afferent stimulation, mimicking a prominent response of the olfactory bulb to sensory input. Odor-evoked local field potentials in vivo display a fast oscillatory component as well as a slower, respiration-linked rhythm (Onoda and Mori 1980). These slow responses were not observed in our reduced preparation as they most likely require mechanisms extrinsic to the OB. The similarity of the stimulus-evoked fast oscillations in this study to previously published in vivo recordings strongly suggests that the circuitry underlying these odor-evoked oscillations must be intrinsic to the olfactory bulb. Odor presentation is associated with 20- to 70-Hz field oscillations in mammalian (Eeckman and Freeman 1990) and nonmammalian species (Laurent and Naraghi 1994) in second-order olfactory areas. Oscillatory activity may be generated as a result of intrinsic cellular currents, network properties, or a complex interaction between both mechanisms. Our results show that oscillatory activity in the olfactory bulb is associated with increased activity in inhibitory granule cells. These network interactions that activate granule cells appear to involve both chemical and electrical synapses. These oscillations provide a potential mechanism to regulate the precise timing of mitral cell APs during olfactory responses.

Other types of synchronous, oscillatory activity also occur in the olfactory bulb. Slower field oscillations (2-8 Hz) can also be recorded in vivo and are tightly linked to respiration and sniffing activity. These oscillations are thought to be, in part, mediated by centrifugal inputs, although the substrate for synchronizing mitral cell populations at slow rates may be the olfactory bulb local circuitry. Recent in vitro studies have suggested that slow (1-2 Hz) synchrony between mitral cell pairs is mediated by excitatory interactions in the glomerulus (Carlson et al. 2000; Schoppa and Westbrook 2001). These slow oscillations are distinct from the oscillations reported here since they do not appear to require activity of interneurons in the granule cell layer.

Cellular mechanisms underlying olfactory bulb network oscillations

How are fast oscillations generated in the olfactory bulb? In the insect olfactory bulb analog, -oscillations have been proposed to depend, in part, on reciprocally coupled inhibitory local neurons (Laurent et al. 2001). There is at present no evidence for granule cell-granule cell synaptic interactions in the mammalian olfactory bulb. There are, however, anatomical studies, suggesting that these interneurons are interconnected through electrical synapses (see below). In the neocortex, gap junction connections appear to mediate oscillatory activity in interneuronal networks (Beierlein et al. 2000). The block of sustained oscillations in olfactory bulb slices by glutamate receptor antagonists suggest that oscillatory activity does not result from an autonomous interneuronal network interconnected through gap junctions but instead requires synaptic interactions between GABAergic and glutamatergic olfactory bulb neurons. While some authors (Freeman 1978; Rall et al. 1966) proposed that reciprocal synaptic interactions between mitral cells and granule cells underlie odor-induced oscillations, it is uncertain whether this synaptic feedback inhibition plays a major role in the oscillations observed here. The activation of reciprocal inhibition in this synapse appears to be dependent on NMDA receptor activation (Halabisky et al. 2000; Isaacson and Strowbridge 1998; Schoppa et al. 1998), while we find that olfactory bulb oscillations require both AMPA receptors and NMDA receptors. There are several possible sources of this fast excitation. It is possible that granule cells require a tonic depolarization that occurs through temporal summated glutamatergic EPSPs. These EPSPs need not be necessarily synchronized with the inhibitory oscillation evoked in mitral cells. Alternatively, each cycle in the oscillation may be driven by fast excitatory synaptic input, as likely occurs in the hippocampus (Traub et al. 2000). The sensitivity of persistent olfactory oscillations to glutamate-receptor antagonists is somewhat surprising given that the effect of these oscillations appears to be increased inhibitory tone in mitral cells, the primary source of excitatory drive to granule cells. It is not clear whether the glutamate-receptor sensitivity of persistent network activity results from blockade of tonic or from slowly modulating excitatory current, perhaps due to ambient glutamate, or instead reflects a role of pacing from a subpopulation of excitatory neurons that are not inhibited by olfactory oscillations.

The role of gap junction coupling in network oscillations

Experimental and theoretical studies in other brain regions have demonstrated a prominent role for gap junction coupling in many types of fast inhibitory network oscillations. Electrical coupling between populations of interneurons has been described in several cortical (Deans et al. 2001; Galarreta and Hestrin 1999; Gibson et al. 1999; Hormuzdi et al. 2001) and subcortical (Landisman et al. 2002) regions where they play a prominent role in oscillatory network behavior. In these brain areas, interneuron gap junction coupling is mediated predominantly by Cx36 (Deans et al. 2001; Hormuzdi et al. 2001; Landisman et al. 2002).

There is anatomical evidence for the presence of electrical synapses onto interneurons in the olfactory bulb. Gap junction coupling between granule cells has been shown with both electron microscopy and dye-coupling (Reyher et al. 1991). Also, there is prominent labeling for Cx36, the connexin which mediates interneuronal coupling in cortical networks, expression in the mouse olfactory bulb, especially in the granule cell layer (Condorelli et al. 1998). Freeze-fracture electron microscopy (Landis et al. 1974; Miragall et al. 1996) and dye-coupling (Paternostro et al. 1995) studies suggest that gap junctions also exist between mitral cell and granule cell membranes. Based on immunohistochemical and in situ hybridization labeling, the gap junction protein forming these mitral cell-granule cell electrical synapses is likely to be Cx43 (Miragall et al. 1996; Paternostro et al. 1995), although there is also some recent evidence for Cx45 expression in mitral cells as well (Zhang and Restrepo 2002). It is unknown, however, whether both granule cell-granule cell and mitral cell-granule cell gap junctions observed in anatomical studies are functional. Nicoll (1972) demonstrated that many types of anesthetics, including halothane, enhanced recurrent inhibition in vivo, as assayed by paired lateral olfactory tract stimuli. While possibly suggesting a role for gap junctions in dendrodendritic inhibition, it is difficult to determine in in vivo experiments such as these whether the effects of halothane (provided as a mixture 1.5-3.0% in oxygen) relate to a inhibition of electrical synapses.

Recent in vitro and modeling studies in cortical regions have elucidated the various mechanisms by which electrical coupling facilitates network oscillations. Gap junction coupling between interneurons plays an important role in stabilizing neocortical (Tamas et al. 2000) and hippocampal (Hormuzdi et al. 2001; Traub et al. 2001) -frequency oscillations that are generated by subpopulations of reciprocally connected GABAergic cells (Traub et al. 1999). Periodicity in this interneuron network activity is an emergent property of these electrical and chemical interactions and is independent of oscillations in fast excitatory input (Whittington et al. 1995). Synchronous inhibitory oscillations at approximately 10 Hz can occur in networks of electrically coupled cortical interneurons depolarized with mGluR agonists in the absence of chemical synaptic interactions (Beierlein et al. 2000). Some forms of gap junction-mediated interneuron oscillations, however, may rely on synchronization of excitatory inputs onto interneuron populations. For instance, carbachol-evoked persistent -band activity in CA3 depends on AMPAR-mediated activation of interneurons by reverberant activity in an electrically coupled pyramidal cell axonal plexus (Traub et al. 2000). In the olfactory bulb, electrical synapses may function to facilitate propagation of regenerative electrical activity (Na⁺-based action potentials or Ca²⁺ spikes) through networks of granule cells, thereby generating large-amplitude IPSPs in mitral cells.

Stimulus-evoked persistent changes in olfactory network dynamics

Stopfer and Laurent (1999) observed that repeated odor presentation altered the response of the antennal lobe network. In this study, successive odor stimulation increased field oscillations as well as decreased firing in projection neurons. The authors proposed this augmentation to be a neuronal correlate of short-term olfactory memory; this mechanism may increase an animal's ability to identify and discriminate the conditioned odorant. We observe similar conditioning-dependent changes in the mammalian olfactory bulb slice that lasted for over 30 min. The mechanism by which repeated afferent stimulation and therefore repeated activation of granule cell networks by mitral cells is unclear. The slow time course of this potentiation may reflect slow adaptations in olfactory bulb local circuit dynamics or may indicate a potential role for second-messenger systems. Activity-dependent second messengers such as cAMP have been shown to enhance gap junction coupling between neurons (Gladwell and Jefferys 2001) and could enhance network synchrony.

In conclusion, our findings demonstrate that population activity in local inhibitory networks coupled by electrical and chemical synapses play an important role in generating the fast oscillations that are likely to be involved in odor discrimination. This novel in vitro model may facilitate further studies addressing the relationship between inhibitory oscillations and mitral cell firing patterns during odor encoding (Laurent et al. 2001).