Tonic inhibition sets the state of excitability in olfactory bulb granule cells


C. Labarrera: The Silberman building 3th floor, room 343, Givat Ram, Neurobiology, The Hebrew University of Jerusalem, 91904 Jerusalem, Israel.  Email:

Key points

  • • Granule cells are the main source of inhibition in the olfactory bulb (i.e. the first station of odour processing in the mammalian brain), but very little is known about the inhibition that acts upon them.
  • • Using in vivo whole cell patch clamp recordings in anaesthetized mice we report the following new findings:
    • • We found odour-evoked responses to be rare (seen only in 18% of the odour presentations, and only in cells that showed also evoked excitatory responses to odours).
    • • We report for the first time the presence of tonic inhibition in the olfactory bulb.
    • • We show that tonic inhibition dominates over phasic synaptic inhibition evoked by odours, thereby being the key regulator shaping the granule cells spike output.
    • • Preliminary (in vivo) evidence suggests that sensory evoked phasic inhibition onto granule cells is provided by deep short axon cells in the olfactory bulb.

Abstract  GABAergic granule cells (GCs) regulate, via mitral cells, the final output from the olfactory bulb to piriform cortex and are central for the speed and accuracy of odour discrimination. However, little is known about the local circuits in which GCs are embedded and how GCs respond during functional network activity. We recorded inhibitory and excitatory currents evoked during a single sniff-like odour presentation in GCs in vivo. We found that synaptic excitation was extensively activated across cells, whereas phasic inhibition was rare. Furthermore, our analysis indicates that GCs are innervated by a persistent firing of deep short axon cells that mediated the inhibitory evoked responses. Blockade of GABAergic synaptic input onto GCs revealed a tonic inhibitory current mediated by furosemide-sensitive GABAA receptors. The average current associated with this tonic GABAergic conductance was 3-fold larger than that of phasic inhibitory postsynaptic currents. We show that the pharmacological blockage of tonic inhibition markedly increased the occurrence of supra-threshold responses during an odour-stimulated sniff. Our findings suggest that GCs mediate recurrent or lateral inhibition, depending on the ambient level of extracellular GABA.


action potential


deep short axon cells


evoked excitatory postsynaptic currents


granule cells


inhibitory postsynaptic currents


Inhibitory interneurons are integral elements of local microcircuits. In the olfactory bulb, the output of the principal neurons is gated by recurrent and lateral inhibition mediated by GABAergic granule cells (GCs) located in the innermost layer of the bulb. GCs have long been known to be excited by mitral cells and in turn inhibit mitral cells. These synaptic interactions are mediated by bidirectional dendro-dendritic synapses between GCs and mitral cell lateral dendrites. Calcium influx shapes transmitter release from GCs, and is distributed and regulated by complex regenerative response properties (Egger et al. 2003, Pinato & Midtgaard, 2005). The relationship between synaptic input, regenerative responses and GABA release remains unclear, but action potentials (APs) are not an absolute requirement of synaptic release (Jahr & Nicoll, 1980, Isaacson & Strowbridge, 1998).

Indirect evidence suggests that olfactory GCs are under some form of suppressive inhibitory drive. In vivo studies have shown that although GCs receive barrages of excitatory postsynaptic potentials, they rarely spike. Additionally, with prolonged odour stimulation, GC excitation adapts quickly between the first and second respiratory cycle (Cang & Isaacson, 2003, Margrie & Schaefer, 2003). Slice studies demonstrate that GCs are synaptically coupled to deep short axon cells (dSACs) in the GC layer (Pressler & Strowbridge, 2006, Eyre et al. 2008). How GCs are controlled by these inhibitory connections or whether they are controlled through alternative inhibitory mechanisms during odour processing remains unknown.

Recently, the synaptic activity of GCs has unequivocally been linked to neuronal processing time in odour discrimination in mice (Abraham et al. 2010). Odour recognition occurs within the time frame of a single respiratory cycle (<300 ms) in a manner that involves GC lateral circuits (Uchida & Mainen, 2003, Koulakov et al. 2005). However, the synaptic input to GCs and their responses during a respiratory cycle remain unexplored.

The present study investigates the synaptic input and spiking behaviour of GCs during a 100 ms inhalation-aligned odour presentation in vivo. Our recordings of odour-evoked excitatory and inhibitory currents showed that GCs are predominantly excited by this sniff-like olfactory stimulus, whereas spiking is curbed by tonic rather than phasic inhibition. We demonstrated that the level of tonic conductance was 3-fold higher than the level of phasic synaptic inhibition and that blockage of tonic inhibition changed the GC input–output relationship.


In vivo electrophysiology

The animal procedures were conducted in accordance with the European Union's Council Directive 86/609/EEC. In vivo whole-cell recordings were performed as described by Margrie & Schaefer (2003). Female C57BL6/J mice (postnatal day 25–30; Taconic, Bomholtvej 10, 8680 Ry, Denmark) were anaesthetized with ketamine/xylazine (100/10 mg kg−1). The selection criteria for GCs were: recording depth >340 μm (434 ± 66 μm, n = 62), cellular capacitance <12 pF (8.6 ± 2.3 pF, n = 53) and input resistance >200 MΩ (Rin,normal internal = 384 ± 130 MΩ, n = 21; Rin,cesium internal = 1137 ± 638 MΩ, n = 50). All of the recordings were obtained using a Multiclamp 700B amplifier (Mollecular Devices, CA). The signal was filtered at 2 kHz It was digitized using ITC-18 (HEKA GmbH, Germany) and analysed using Neuromatics/NClamp software ( The pipette solution for current-clamp recordings contained 130 mm methanesulfonic acid, 10 mm Hepes, 7 mm KCl, 0.05 mm EGTA in KOH, 2 mm Na2ATP, 2 mm MgATP and 0.5 mm Na2GTP, pH 7.2, with KOH. A caesium-based solution was used for all voltage-clamp recordings: 140 mm methanesulfonic acid, 4 mm Tea-Cl, 2–5 mm QX-314, 15 mm Hepes, 4 mm NaCl, 1 mm EGTA in CsOH, 2 mm Na2ATP, 2 mm MgATP and 0.5 mm Na2GTP, pH 7.2, with CsOH. Furosemide and Gabazine (SR-95531) ( were each applied directly on the craniotomy in doses of 10–20 μl of 0.5 mm. Animals were terminated via cervical dislocation under ketamine/xylazine anaesthesia.

Odour stimulation

Respiration was monitored using a thermocouple placed in a 1 × 1 mm hole that accessed the nasal cavity. The slope of the respiratory signal just prior to inhalation was detected with a custom made window discriminator, and a Transitor-Transitor Logic (TTL) signal was utilized to open a final valve that controlled the flow of odorous air presented to the left nostril. To mimic natural sniffing, the odours were presented as individual short pulses that lasted 100 ms at the beginning of inhalation (Fig. 1A, inset). Odours (1:10 in mineral oil: isoamyl acetate (AA), ethyl butyrate (EB), nonanoic acid (NA), and a combination of cineole, lavender, carvone and benzaldehyde (Mix)) were presented with the inhalation using a custom-built olfactometer (Bodyak & Slotnick, 1999).

Figure 1.

Odour-evoked responses of olfactory bulb GCs in vivo
A, representative traces of the three observed output types in response to inhalation-aligned 100 ms odour presentation (red bar): subthreshold (grey), spikelets (black) and action potential (blue). Inset, odour delivery occurred during the square pulse, and the black sinusoidal is the respiration signal. Above, magnifications of each response, i.e. subthreshold, spikelets and an action potential. B, raster plots of three individual GCs presented with four different odour stimuli. The vertical grey bar represents the odour stimulus. Horizontal grey bars represent the time of peak amplitude of subthreshold responses. The colour code is the same as in A. C, distribution of output response types: subthreshold (sub) 58%, action potentials (AP) 22% and spikelets 14%. No response (non) was observed in 6% of the presentations (top). Below, distribution of response types for each specific odour; colour code as in A.

Data analysis

The data were analysed using Neuromatics.v2/IgorPro.v5 and Excel software. A subthreshold odour-evoked response was analysed on the average trace from five trials and measured as the voltage difference (ΔV) between baseline and peak amplitude. Odour presentations with a ΔVstimulated cycleVspontaneous cycle ratio <1 were categorized as a non-response. Spikelets were detected at a threshold of 7 mV (Pinato & Midtgaard, 2005) using the threshold-above-baseline event detection function in Neuromatic.v2, and further characterized as dV/dt > 0.6 V s−1 (Zelles et al. 2006). The same approach was used to detect spontaneous and evoked excitatory postsynaptic currents (EPSCs) and inhibitory postsynaptic currents (IPSCs) using a threshold of 3–7 pA, depending on the peak-to-peak noise level (Figs 2B and 3A and B). The criteria of an evoked average odour response from our voltage-clamp recordings were set at an average peak of 5 pA above baseline for both excitatory and inhibitory inputs. Prior to averaging, the individually stimulated trials were aligned to the time of the peak of the stimulated respiration cycle. The time lag between inhibition and excitation was calculated using two methods. First, we calculated the time where the absolute value of the trace (IPSC or EPSC) reaches 10% of peak amplitude. In the second method we calculated the derivative of the current trace with respect to time for both inhibition and excitation. The delay between the two traces was calculated as the time difference between the time of the minimum derivatives of the excitation and the maximum derivative of the inhibition within a time window of 50 ms from onset. Action potential threshold was detected at 10 mV ms−1 on a phase plot (dV/dt vs. V; Naundorf et al. 2006). Statistical significance (P < 0.05) was assessed using a two-tailed Student's t test, and all data are reported as mean ± SD unless otherwise specified. Classification of recordings as originating from dSACs was based on a criterion of persistent spiking upon current injection or odour stimulation.

Figure 2.

Rare odour-evoked phasic inhibition in GCs 
A, schematic drawing of the local GC network. The mitral cell (MC) connects to the GC via excitatory (Ex) dendro-dendritic synapses on the MC lateral dendrites and via collaterals to local interneurons (dSAC, green; In). The arrow represents olfactory drive from the olfactory nerve (ON). B, odour-evoked excitatory (grey) and inhibitory (green) synaptic currents from six different cells recorded during voltage-clamping at –75 mV (a) and +20 mV (b). Below, corresponding peri-stimulus histograms (bin = 20 ms, five trials). Red bars denote the time of odour presentation. C, distribution of evoked and non-evoked excitatory and inhibitory odour responses, respectively. D, bar chart that shows the percentage of cells with non-evoked, excitatory-only and evoked inhibition–excitation responses below. Example of EPSC (grey) and IPSC (green) in response to odour presentation (red bar). To illustrate the time lag of inhibition, the EPSC has been horizontally mirrored (light grey). E, in vivo recording from dSACs in which odour stimulation induced long-lasting spike trains at frequencies of 2–4 Hz after the odour presentation (top left). No relationship was found between the spiking frequency during current injections and the persistent spiking of these suspected dSACs (r= 0.139; top right). Bottom left, recording and raster of IPSCs after an inhibitory evoked response in a GC. Bottom right, normalized frequency of spontaneous IPSCs before and after odour presentation in GCs that responded with evoked inhibition. The rate of IPSCs after the evoked response was analysed during a 1- to 5-s interval after the odour presentation.

Figure 3.

Tonic inhibition of GCs is mediated by furosemide-sensitive GABAA receptors and independent of phasic synaptic GABAergic transmission 
A, representative traces (left), time course of the experiment (middle) and all-points histograms from before and after blocking the GABAergic currents with gabazine. Note the shift in the holding current (ΔI= 24.3 ± 10.3 pA, P < 0.05; n= 4) and reduction in the current variance. B, as in A with furosemide (purple; ΔI= 42.0 ± 19.3 pA, P < 0.05; n= 4). C, left, overlay of control and furosemide sIPSCs (average of 10 events). Right, tonic conductance vs. rate of sIPSCs from each individual cell (r= 0.03). Data points are from blockade by both gabazine and furosemide. D, comparison of charge transfer from evoked synaptic and tonic currents measured over 1000 ms (evoked synaptic = 0.01 ± 0.009 pC s−1 (n= 26) vs. tonic = 0.032 ± 0.016 pC s−1 (n= 10), P < 0.01).


Spiking activity in GCs upon odour presentation

In vivo whole-cell patch-clamp recordings were performed from inhibitory GCs in the olfactory bulb in anaesthetized mice presented with a 100 ms inhalation-aligned odour pulse (Fig. 1A). Despite the central role of GCs in controlling the final output from the bulb, their spontaneous spike rate in vivo is low (<0.4 Hz, n= 20) (Cang & Isaacson, 2003, Margrie & Schaefer, 2003). In our experiments, full somatic spikes were recorded in 22% of the odour presentations, while subthreshold depolarization was the most widespread response type with 58% (207 of 355; spontaneous vs. stimulated; peak depolarization: –55.8 ± 6.3 vs.–64.1 ± 6.8 mV, respectively, P < 0.001; ΔVm: 10.6 ± 4.9 vs. 4.4 ± 3.0 mV, respectively, P < 0.001; area: 2376 ± 1273 vs. 23 ± 118 mV ms, respectively, P < 0.001; n= 15 cells; Fig. 1A). The suprathreshold responses observed were in the form of full-amplitude APs (77 of 355 presentations; Fig. 1A, blue) or spikelets (50 of 355 presentations; Fig. 1A, black), 21 out of 355 presentations did not evoke a response; Fig. 1C. The spikelets were characterized by a fast rise and small amplitude and often occurred in bursts at a frequency of ∼15 Hz (Rise time10−90%: 1.2 ± 0.19 ms; Amp: 10.5 ± 0.2 mV; dV/dt > 0.6).

Evoked responses in GCs were not selective to the specific type of odour (Fig. 1C, bottom). Rather, the response form, evoked by one odour, was also the most likely type of response to the additional odours presented, suggesting that GCs have a preferred response profile (Fig. 1B).

GC excitatory and inhibitory synaptic currents in response to odour stimulation

Rare and delayed spiking in GCs is partially determined by their intrinsic properties (Schoppa & Westbrook, 1999, Kapoor & Urban, 2006) but may also reflect synaptic inhibition. Previous studies showed that it is possible to successfully voltage-clamp olfactory GCs (Schoppa & Westbrook, 1999, Balu et al. 2007). Thus, we applied the reversal potential clamp technique to isolate and separate odour-evoked inhibitory and excitatory synaptic responses in GCs. To reduce distortion of synaptic currents due to space-clamp (Williams & Mitchell, 2008, Poleg-Polsky & Diamond, 2011), we used a caesium-based internal solution that contained ion channel blockers to pharmacologically minimize the active intrinsic properties of the dendrites (see Methods). The excitatory synaptic input evoked by sniff stimulation was recorded at –75 mV, whereas the inhibitory currents were monitored at approximately +20 mV. A stimulus-related EPSC burst was observed in response to most of the presentations (135 of 189 presentations; Fig. 2 Btop and C). In contrast, a low occurrence of inhibitory evoked responses was observed (35 of 189 presentations; Fig. 2 BBottom and C). In rare cases of evoked IPSCs, the currents were broadly activated and only observed in GCs with evoked excitation. Comparisons of the timing of the excitatory and inhibitory currents revealed a delay of inhibition relative to excitation (see Methods; analysed at 10% of the peak amplitude and using maximum derivatives: IPSC time lag (mean ± SEM), 19.9 ± 5.17 and 11.3 ± 6.4 ms, respectively, N= 10 cells, n= 24 stimulations; Fig. 2D, lower panel).

Persistently firing interneurons mediate rare synaptic inhibition of GCs

While performing blind in vivo patch-clamping at a depth of >400 μm we occasionally encountered neurons with AP characteristics which clearly differed from GCs. We suspect these cells to be dSACs (GCs vs. dSACs; APthres: –39.2 ± 6.2 vs.–46.6 ± 3.7 mV, P < 0.01; APamp: 31.0 ± 6 vs. 38.1 ± 10.5 mV, P < 0.05; n= 30 random spikes, data not shown). In these suspected dSACs the firing probability in response to odour presentation was high and the AP number during a brief depolarization was high compared with GCs (dSACs vs. GCs; APPo= 0.90 vs. 0.22; AP number per cycle: 3.8 ± 1.8 vs. 2.87 ± 2.3, P < 0.05; n= 45 random stimulations from each group). Most importantly these interneurons exhibited long-lasting spiking activity upon current injection or odour stimulation (500 ms, 10–50 pA or sniff-like odour presentation). Notably, the frequency of persistent firing was independent of the amplitude of current injected to the cell or of the firing rate evoked by the current injection (r= 0.139; Fig. 2E, top right) (Pressler & Strowbridge, 2006). In slice studies it has been shown that persistently firing interneurons in the GC layer mediate monosynaptic feed-forward inhibition onto GCs (Pressler & Strowbridge, 2006). These findings led us to analyse the post-stimulus frequency of IPSCs in GCs responding with IPSC bursts. We found that the few GCs with inhibitory evoked potentials exhibited an increase in single IPSCs frequency after an odour-evoked response (IPSCspont= 2.2 ± 2.9 vs. IPSCafter-stim= 4.4 ± 8.4 Hz, P < 0.05; n= 32 presentations in 11 cells; Fig. 2E, bottom right). This increase in IPSCs was confined to GCs with evoked inhibitory responses and not observed in the large group of GCs that lacked inhibitory evoked bursts upon odour presentation (IPSCspont= 1.8 ± 1.5 vs. IPSCafter-stim= 1.28 ± 0.72 Hz, P > 0.05; n= 23 presentations in 15 cells; data not shown).

These results may suggest that dSACs with a persistent pattern synapse onto GCs.

Tonic inhibition in GCs

The GABAergic origin of IPSC events in GCs was verified by applying gabazine (Fig. 3A). Gabazine expectedly removed transient synaptic currents but surprisingly also significantly decreased the holding current (ΔI= 24.3 ± 10.3 pA, P < 0.05, n= 4; Fig. 3A). Additionally, the current shift was associated with a reduction of the baseline peak-to-peak noise level (Fig. 3A, right panel, Gaussian widths), consistent with the blockade of background inhibitory tonic conductance (Bright & Brickley, 2008).

Previous studies have shown that inhibitory tonic conductance is mediated by extrasynaptic GABAA receptors, often expressing α4 and δ subunits (Mody, 2001). Evidence at the mRNA level suggests that olfactory bulb GCs express extrasynaptic α4 and δ GABAA subunits. Furthermore, GABAAα4 and δ subunit mRNA was exclusively found in the GC layer and only on GCs (Laurie et al. 1992). To investigate the possible impact of these extrasynaptic GABAA receptors on GCs, we applied the GABAAα46 subtype-selective antagonist furosemide (Korpi et al. 1995, Korpi & Lüddens, 1997, Mody, 2001). Furosemide significantly reduced the baseline holding current (ΔI= 42.0 ± 19.3 pA; P < 0.05, n= 4; Fig. 3B), whereas no change in the waveform characteristics and frequency was observed for synaptic inhibitory events (IPSCrise ctrl= 1.84 ± 0.66 vs. IPSCrise furo= 2.67 ± 0.81 ms, P > 0.05; IPSCdecay ctrl= 22.27 ± 3.17 vs. IPSCdecay furo= 22.75 ± 4.09 ms, P > 0.5; IPSCamp ctrl= 49.25 ± 6.4 vs. IPSCamp furo= 42 ± 5.89 pA, P > 0.05, IPSCfreq . ctrl= 3.13 ± 3.18 vs. IPSCfreq . furo= 4.0 ± 4.4 Hz, P > 0.05; n= 4, Fig. 3C, overlay). Moreover, a clear increase in the input resistance of GCs was observed, supporting our hypothesis that the change in holding current resulted from blockade of tonic inhibition mediated by extrasynaptic GABAA receptors (Rctrl= 339 ± 49 vs. Rfuro 496 ± 189 mΩ, P < 0.05; n= 7, data not shown).

The presence of tonic inhibition in GCs in the olfactory bulb is novel and the cellular mechanism behind it is as yet unknown. In other systems the most widespread explanation for the origin of extrasynaptic GABAergic activity is spillover from neighbouring synapses (Farrant & Nusser, 2005; Duguid et al. 2012). However, tonic inhibition of GCs in the cerebellum has also been shown to involve glial cells (Lee et al. 2010). Another physiologically viable option is that reversed activity of sodium chloride symporters dynamically regulates the level of extrasynaptic GABA (Richerson & Wu, 2003). In accordance with previous studies, we analysed the magnitude of tonic conductance compared with the frequency of an individual IPSC to investigate the potential role of spillover as a source of extrasynaptic GABA (Bright & Brickley, 2008). As no correlation was found (r= 0.03; Fig. 3C), this suggests that tonic inhibition is not a consequence of synaptic spillover from the GCs own GABAergic input. However, it does not rule out the possibility of spillover from the surrounding population of GCs.

Tonic inhibition, not synaptic inhibition, affects GC activity

To examine the impact of tonic conductance compared with phasic inhibitory input on GC activity an analysis of charge contribution was performed. We found that the average current associated with tonic conductance was 3-fold greater than the average current of evoked phasic inhibitory events (32 ± 16 vs. 10 ± 9 pC s−1; n= 8 and 26, respectively; Fig. 3D). Blockage of tonic inhibition altered the evoked output properties of GCs such that cells with a subthreshold response to odour presentation prior to furosemide application often responded with full-amplitude APs after applying furosemide (Fig. 4A). The proportion of responses that included full-amplitude APs increased from 10.2 ± 31.6 to 40.1 ± 51.6% in the presence of furosemide (Fig. 4B, left; P < 0.05 paired t test; n= 10 cells, one cell per mouse, average and standard deviations across mice; four odours presented five times each). Moreover, the number of APs evoked during a single stimulated respiratory cycle increased (0.8 ± 1.2 vs. 2.4 ± 3.4, P < 0.05; n= 10, Fig. 4B, right). This indicates that tonic conductance plays a central role in determining the spiking probability of GCs.

Figure 4.

Blockade of tonic GABAergic current increases GC excitability 
A, odour-evoked responses in control (grey) and after application of furosemide (blue). Below, rasters from five odour trials. B, left, distribution of output response types in control and with furosemide: subthreshold (grey), spikelets (black) and action potential (blue). Right, average number of spikes in five trials during an odour-evoked sniff in control and with furosemide (Ctrl vs. furosemide: 1.1 ± 2.2 vs. 3.5 ± 2.8 APs, P < 0.01; n= 35). C, schematic drawing illustrating how the level of tonic inhibition could regulate the extent of inhibition mediated by the GC (black) on to the mitral cells. Principal neurons are illustrated via their long lateral dendrites (grey and blue: lateral inhibited). We propose that the ambient level of tonic inhibition can act as a regulator between recurrent and lateral inhibition mediated by the GC. A back-propagating AP in the mitral cells lateral dendrite will give rise to local recurrent inhibition only during high levels of tonic inhibition (C, upper). Under low tonic inhibitory conditions a similar input from the mitral cell might cause global lateral inhibition (C, lower); see Discussion.


It is striking that although GCs rarely spike, they play a key role for odour processing in the bulb (Abraham et al. 2010). In the present study, we investigated synaptic input and firing responses of GCs. We found that excitatory synaptic currents were broadly evoked in 71% of odour presentations. In contrast, inhibitory evoked responses were rare and only observed in 18% of the presentations. The low occurrence of evoked inhibitory responses shows that phasic synaptic inhibition plays only a minor role in shaping the output of GCs during odour processing. We find that subthreshold responses are the most common response type, accounting for 58% of responses, followed by spikes (22%) and spikelets (14%).

Furthermore, we identified a tonic inhibitory conductance in GCs and provided evidence of the involvement of furosemide-sensitive GABAA receptors, probably a high-affinity combination of α4 and δ subunits (Laurie et al. 1992, Olsen & Sieghart, 2009). This tonic inhibition showed a strong suppressive effect on the excitability of GCs and ultimately determined whether a GC elicited a sub- or suprathreshold response during odour-evoked sniffing. We propose that tonic inhibition provides the GCs with the ability to shift between local (dendro-dendritic) and global release forms (Fig. 4C).

The local inhibitory network of the GC

In the present study we revealed a marked discrepancy in the occurrence of synaptically evoked excitation and inhibition in GCs. Concurrent excitation and inhibition are observed in many neurons during network activity, but GCs display prominent excitation and a striking paucity of synaptic inhibition during odour presentation. Interestingly, we only observed evoked inhibition in combination with evoked excitation. Moreover, our recordings of GCs revealed that evoked synaptic inhibition was followed by an increase in the rate of spontaneous IPSCs >1000 ms after odour presentation. Assuming that dSACs are the source of phasic inhibition, the fact that a GC with phasic inhibition alone was never encountered suggests that the dSACs receive input from the same population of principal neurons as the GCs. As a consequence, the same AP that back-propagates into the lateral mitral cell dendrite and excites the GC may also evoke phasic inhibition via activation of an inhibitory dSAC that synapses onto the GC. Thus, via inhibition of the GC the principal neuron activates its own disinhibition. Among the interneurons in the GC layer, the Blanes cells, a subtype of dSACs, mediate feed-forward synaptic inhibition onto GCs and respond in vitro with long trains of AP in response to brief depolarization. A more hyperpolarized threshold level of AP initiation compared with GCs also characterizes Blanes cells, characteristics observed in the dSACs we recorded as well. Interestingly, we also found that the probability of obtaining an evoked inhibitory response (18%) was coherent with the likelihood of obtaining Blanes cell/GC monosynaptically coupled pairs in slices (Pressler & Strowbridge, 2006).

Lastly, inhibition of GCs may be of local origin, although input from centrifugal inhibitory fibres is also possible (Kunze et al. 1992, Gracia-Llanes et al. 2010).

Tonic inhibition regulates the output properties of GCs

The discovery of tonic inhibition in GCs may explain the hyperpolarized resting membrane potential, the low spike rate and why GCs need strong, synchronous input to reach threshold (Cang & Isaacson, 2003, Margrie & Schaefer, 2003). Our findings showed that tonic inhibition has a strong impact on the output of GCs and dominates over phasic inhibition. Although loop diuretic drugs such as furosemide have been shown to block the potassium chloride cotransporter 2 (KCC2) in rat hippocampal in vitro studies at concentrations of 0.5 and 1 mm (Viitanen et al. 2010), our results do not show indications of such interaction, as no change of IPSC waveforms were observed (Fig. 3C). The fact that odour-evoked subthreshold responses were changed to full-amplitude APs, after application of furosemide, suggests that the dynamic level of tonic inhibition regulates the extent of inhibition, i.e. possibly from recurrent to lateral (Fig. 4C).

It is possible that high levels of tonic inhibitory conductance maintain olfactory GCs in a state that favours local graded regenerative responses in dendritic spines and branches to enrich and differentiate GABA release, similar to the isolated signal processing that occurs in amacrine cells (Grimes et al. 2010). The morphology of GCs, distribution of release sites and bidirectional dendro-dentritic synapses appear well suited to allow release sites in the same neuron to be regulated by the level of tonic inhibition, thereby playing a key role in the dynamics of dense synchronous synaptic release versus a temporal spatial GABA release from the GC dendritic tree. Thus, strong shunting via high tonic conductance would limit activation to local dendritic branches only. In contrast, low levels of tonic inhibition would permit activation in local branches/spines to spread throughout the dendritic tree causing lateral (global) inhibition of the connected population of mitral cells.

We propose that the spatiotemporal evolution of circuit activity during odour discrimination depends on these properties provided by tonic extrasynaptic inhibition in GCs.

The source of GABA that activates extrasynaptic receptors is unknown. There was no clear relationship between the magnitude of tonic conductance and spontaneous IPSC frequency in individual GCs, supporting the hypothesis that synaptic and extrasynaptic GABAA receptors are recruited by GABA from separate sources (Bright & Brickley, 2008). Yet, this does not entirely rule out spillover as a source of extrasynaptic GABA. One obvious possibility is the numerous and densely packed release sites from neighbouring GCs, which could provide a source of volume-conducted GABA in the extracellular space.


Author contributions

C.L. performed all experiments. K.A. and C.L performed the data analysis with help from M.L. M.L. provided laboratory facilities and inputs to the second round of experiments, data analysis and figures. C.L. and K.A. wrote the manuscript, with inputs from M.L. All authors approve the final version for publication.


We thank Elad Ganmor, Jørn Hounsgaard and Andreas T. Schaefer for commenting on the manuscript. This work was supported by The Danish Council for Independent Research (K.A.), The Carlsberg Foundation, The Oticon Foundation and The Lundbeck Foundation.