Presynaptic Inhibition of Olfactory Sensory Neurons: New Mechanisms and Potential Functions

Abstract

Presynaptic inhibition is the suppression of neurotransmitter release from a neuron by inhibitory input onto its presynaptic terminal. In the olfactory system, the primary sensory afferents from the olfactory neuroepithelium to the brain’s olfactory bulb are strongly modulated by a presynaptic inhibition that has been studied extensively in brain slices and in vivo. In rodents, this inhibition is mediated by γ-amino butyric acid (GABA) and dopamine released from bulbar interneurons. The specialized GABAergic circuit is now well understood to include a specific subset of GAD65-expressing periglomerular interneurons that stimulate presynaptic GABA_B receptors to reduce presynaptic calcium conductance. This inhibition is organized to permit the selective modulation of neurotransmitter release from specific populations of olfactory sensory neurons based on their odorant receptor expression, includes specialized microcircuits to create a tonically active inhibition and a separate feedback inhibition evoked by sensory input, and can be modulated by centrifugal projections from other brain regions. Olfactory nerve output can also be modulated by dopaminergic circuitry, but this literature is more difficult to interpret. Presynaptic inhibition of olfactory afferents may extend their dynamic range but could also create state-dependent or odorant-specific sensory filters on primary sensory representations. New directions exploring this circuit’s role in olfactory processing are discussed.

Introduction

Presynaptic inhibition is a circuit architecture in which the presynaptic terminal of a neuron receives inhibitory input that, when active, reduces neurotransmitter release from that terminal (Figure 1A). Unlike other inhibitory arrangements, presynaptic inhibition does not require changes in the global activity or firing rate of the target neuron, but instead it locally affects the target neuron’s synaptic output. This architecture can permit a brain region to modulate the activity of its afferent connections without the need for a reciprocal projection. Presynaptic inhibition can be implemented by a variety of ways, including the modulation of presynaptic calcium conductances by intracellular second messengers (the principal mechanism covered in this review) and the persistent depolarization of the axon terminal to inactivate sodium channels and reduce the magnitude of action potential-induced voltage transients. Some form of presynaptic inhibition has been observed in many sensory systems, especially early in the circuit. For instance, it has been reported on the synapses from retinal ganglion cells onto thalamic relay neurons (Chen and Regehr 2003), on primary auditory afferents (Baden and Hedwig 2010), on proprioceptive afferents to the spinal cord (Rudomin 1990; Lamotte d’Incamps et al. 1999), on thalamocortical projections in somatosensory system (Porter and Nieves 2004), and on olfactory sensory neurons (OSNs) (Nickell et al. 1994). However, the availability of in vivo data and recent evidence that the olfactory presynaptic inhibition system include modulation by centrifugal projections, specialization into tonic and feedback inhibition circuits, and organization into odorant feature–based functional units makes the olfactory circuit uniquely interesting for exploring the role of presynaptic inhibition in sensory information processing.

Figure 1.

Organization of presynaptic inhibition circuitry in the olfactory bulb. (A) Diagram of a subset of the glomerular circuit to illustrate the different configurations of presynaptic inhibition. The axons of OSNs (black) in the nasal epithelium enter the olfactory bulb and ramify within an individual glomerulus, where they make glutamatergic synapses onto external tufted cells (blue), GABAergic (GAD65-expressing) periglomerular cells (red), and GABAergic/dopaminergic (GAD67- and TH-expressing) SA cells (green), among other cell types. Most of the periglomerular cells (e.g., the lower red one) are under tonic excitatory drive from spontaneously bursting external tufted cells and thus mediate a tonic presynaptic inhibition onto the sensory neuron terminals (indicated by the straight line perpendicular to the OSN axon). A minority of periglomerular cells receive direct input from sensory neurons (e.g., the left-most red PG cell) and mediate a feedback presynaptic inhibition onto sensory neuron terminals. SA cells can be OSN-driven (as shown) or ET-driven and may or may not mediate direct dopaminergic presynaptic inhibition onto OSN terminals (indicated by question mark). For simplicity, only 1 sensory neuron axon is displayed and other glomerular cell types are omitted. The gap in the OSN axon denotes the very long distance traveled, which is truncated here for display purposes. (B) OSN axons converge to form glomeruli in the olfactory bulb (left, image of green fluorescent axons from a horizontal section of the olfactory bulb of an OMP-spH mouse), which are ringed by juxtaglomerular interneurons (middle, blue fluorescent cell bodies from a 4ʹ,6-diamidino-2-phenylindole nuclear stain) of various types. Right panel shows the overlay of the left and middle panels. Adapted from Czarnecki et al. 2012, used by permission. (C) Dopaminergic SA interneurons labeled with an antibody against TH after the manner of Moberly et al. (2012). Note the many fine projections into the middle of the glomerulus.

OSNs in the olfactory epithelium each selectively express just one out of hundreds of olfactory receptors, which determines the profile of odorants to which each responds (Zhao et al. 1998; Malnic et al. 1999; Bozza et al. 2002; Grosmaitre et al. 2006). OSNs project their axons via the olfactory nerve to the olfactory bulb, where they segregate by receptor type such that the axons of all the OSNs expressing a given receptor type converge into one or two specific glomeruli (Figure 1B; Mombaerts 2006). The positions of these glomeruli are relatively stereotyped (Soucy et al. 2009) and organized according to a broad chemotopy (Leon and Johnson 2006). The pattern of OSN input to olfactory bulb glomeruli thus represents the chemical identity of the odorant (Sharp et al. 1975) and maps onto the perception of odorant quality (Youngentob et al. 2006; Linster et al. 2001). A growing body of evidence shows that these primary sensory representations of odorants are shaped in potentially important ways by a presynaptic inhibition of OSN axonal terminals.

Within each glomerulus, OSN axons release glutamate (Berkowicz et al. 1994) onto postsynaptic cells including projection neurons (i.e., mitral and tufted cells) and a variety of local interneurons (Wachowiak and Shipley 2006). This synapse is specialized to have a very high release probability (Murphy et al. 2004). Each glomerulus is ringed by a population of interneurons (Figure 1B, middle and right) that project their axons and dendrites into the middle of the glomerulus (Figure 1C), where they are positioned to interact with the OSN synaptic terminals. Importantly, these interactions include not only receiving synaptic input from the OSNs but also chemical signaling to OSN presynaptic terminals to modulate their release of neurotransmitter. This presynaptic inhibition onto OSN terminals is thus the first opportunity for the central nervous system to directly shape the neural signals representing the olfactory sensorium. This review will first explore the neurobiological implementation of the presynaptic inhibition circuit in the mammalian olfactory system and then broadly review the potential functions of this inhibition in olfactory information processing.

Mechanisms of presynaptic inhibition of OSN synaptic terminals

Presynaptic inhibition of OSN terminals has been explored extensively, and its mechanisms are increasingly well understood. Juxtaglomerular (a general term that means “closely adjacent to the glomerulus” and replaces the older term periglomerular, which now refers only to a specific cell class) interneurons express the inhibitory neurotransmitters γ-amino butyric acid (GABA) and dopamine, and OSN terminals express GABA_B receptors and D₂ dopamine receptors. This organization suggests a straightforward circuit pharmacology that has in fact turned out to include some surprising complexities.

Neuronal origins of GABA and dopamine in juxtaglomerular circuitry

Juxtaglomerular interneurons expressing GABA and dopamine can be grouped into 2 main populations. Dopaminergic juxtaglomerular interneurons (usually identified by their expression of tyrosine hydroxylase (TH), the rate-limiting enzyme in dopamine synthesis; Figure 1C) almost always coexpress the GAD67 isoform of glutamic acid decarboxylase (GAD), the rate-limiting enzyme in GABA synthesis (Kosaka and Kosaka 2008; Kiyokage et al. 2010), at least in mice. These joint dopaminergic/GABAergic neurons have a short axon (SA) cell morphology, meaning that they distribute their processes across multiple glomeruli (Kiyokage et al. 2010). Most of these SA cells are oligoglomerular, meaning they innervate 5–12 glomeruli, whereas a minority is polyglomerular, meaning each cell innervates dozens of glomeruli (Kiyokage et al. 2010). By contrast, a second population of juxtaglomerular interneurons in the mouse olfactory bulb expresses the GAD65 isoform of GAD (sometimes also expressing GAD67) but not TH (Kiyokage et al. 2010). These neurons typically have a uniglomerular morphology (Shao et al. 2009; Kiyokage et al. 2010). Note that it has only recently been appreciated that the multiglomerular SA morphology corresponds to the expression of TH (Kosaka and Kosaka 2008; Kiyokage et al. 2010), so many papers in the literature erroneously refer to dopaminergic neurons as “periglomerular cells.” This review will refer to all TH-expressing neurons as SA cells.

Both the population of GABAergic periglomerular (PG) interneurons and the population of dopaminergic/GABAergic SA interneurons can be further divided into 2 distinctive subpopulations based on their synaptic input (Shao et al. 2009; Kiyokage et al. 2010): about one-third receive direct monosynaptic input from OSNs (e.g., the left-most red cell PG cell is shown in Figure 1A), whereas the remaining two-thirds (e.g., the right-most and lower red PG cell is shown in Figure 1A) receive input principally from burst-firing, OSN-entrained external tufted cells (Hayar et al. 2004) with limited direct input from the olfactory nerve. Ultrastructural data from rat OSN presynaptic terminals have shown that OSNs do not appear to receive conventional inhibitory synapses from PG cells (Pinching and Powell 1971), suggesting that they are sensitive to the local concentration of inhibitory neurotransmitter within the glomerular compartment.

Mechanisms of GABA_B-mediated presynaptic inhibition

OSNs express presynaptic GABA_B receptors (Bonino et al. 1999; Margeta-Mitrovic et al. 1999; Panzanelli et al. 2004), metabotropic receptors trigger a G protein–mediated second messenger cascade when activated (Bowery et al. 1980; Chalifoux and Carter 2011). When the GABA_B receptor agonist baclofen is applied to rat or mouse OSN terminals, neurotransmitter release from OSNs is reduced (Figure 2A, left). This effect has been shown in slice experiments using field potentials (Nickell et al. 1994; Aroniadou-Anderjaska et al. 2000), synaptopHluorin (spH) signals from gene-targeted olfactory marker protein (OMP)-spH mice expressing the fluorescent exocytosis indicator in OSN terminals (Bozza et al. 2004; Wachowiak et al. 2005), voltage-sensitive dyes (Keller et al. 1998), and electrophysiological recordings from neurons that receive monosynaptic input from the nerve (Nickell et al. 1994; Aroniadou-Anderjaska et al. 2000; Murphy and Isaacson 2003; Murphy et al. 2005). It has also been shown in vivo in anesthetized mice (McGann et al. 2005) using odorant stimuli to evoke neurotransmitter release (assayed by spH in OMP-spH mice), where a non-saturating concentration of baclofen (5 µM) suppressed odorant evoked release and the effect could be reversed by the addition of the GABA_B receptor antagonist CGP35348. The scale of this modulation is stunning—saturating concentrations of baclofen reduce stimulus-evoked neurotransmitter release by ~90% (Figure 2A, left; McGann et al. 2005; Wachowiak et al. 2005).

Figure 2.

GABA_B receptor-mediated presynaptic inhibition of presynaptic calcium influx and neurotransmitter release from OSNs. (A) Application of the GABA_B-receptor agonist baclofen greatly reduces nerve shock-evoked neurotransmitter release from OSN terminals, as indicated by the fluorescent exocytosis indicator spH in OMP-spH mice (left), as well as evoked presynaptic calcium influx, as indicated by the calcium-sensitive dye calcium green (right). Adapted from Wachowiak et al. (2005). (B) Calcium influx into OSN terminals evoked by single nerve shocks in vivo under control conditions (black line) and after blockade of GABA_B receptors with CGP35348 (grey line). The left and middle panels show examples from individual glomeruli, whereas the right panel shows the average across all glomeruli tested. These results demonstrate tonic presynaptic inhibition that is ongoing at the time of the initial sensory input. Adapted from Pírez and Wachowiak (2008), used by permission. (C) Paired olfactory nerve shocks 400ms apart evoke less neurotransmitter release (left, synaptopHluorin indicator) and less calcium influx (right, calcium green indicator) with the second stimulus under control conditions (black lines). This feedback inhibition can be partially relieved by blockade of GABA_B receptors with CGP55845 (grey lines). Left panel adapted from McGann et al. (2005); right panel adapted from Wachowiak et al. (2005). (D) EPSCs recorded from a GAD65-positive periglomerular cell that is monosynaptically driven by OSN output. Olfactory nerve stimulation normally evoked a large EPSC (first panel), but the magnitude of this nerve shock–evoked EPSC was suppressed when the nerve stimulation was preceded by a significant depolarization of the PG cell to evoke GABA release (second panel). Blockade of GABA_B receptors with CGP55845 increased the magnitude of the nerve shock–evoked EPSC relative to control (third panel), demonstrating a tonic presynaptic inhibition of sensory neuron output, and also prevented the suppression of sensory neuron output by the depolarization-induced release of GABA (fourth panel), demonstrating a GABA_B-mediated feedback inhibition as well. Adapted from Shao et al. (2009), used by permission.

The principal mechanism by which GABA_B receptor activation suppresses neurotransmitter release is the downregulation of presynaptic calcium influx (Figure 2A, right). There is excellent quantitative consensus across laboratories and studies about the non-linear relationship between presynaptic calcium signaling and exocytosis in the OSN terminal. In both mice and rats, it is well-described by a Hill function of extracellular calcium concentration with a co-operativity parameter of about 2 (this is unusually low; see Murphy et al. 2004) and a k_1/2 of about 1.2mM, as shown with whole-cell recordings from postsynaptic neurons during manipulation of extracellular calcium concentration in 2 different laboratories (Murphy et al. 2004; Wachowiak et al. 2005) and with spH imaging in a third laboratory (Wachowiak et al. 2005). Application of a saturating concentration of baclofen suppresses peak action potential–evoked presynaptic calcium influx into OSN terminals by about 50%, based on calcium imaging from slice experiments in mice (Wachowiak et al. 2005). Because presynaptic calcium is the proximal driver of exocytosis (Mulkey and Zucker 1993), this suppression directly reduces neurotransmitter release from the terminal. In experiments that visualized exocytosis via green-fluorescent spH and presynaptic calcium influx via red-fluorescent calcium-sensitive dyes in the same OSNs in OMP-spH mice, blockade of N-type calcium channels suppressed calcium influx by 62% and exocytosis by 85%, which is consistent with the non-linear relationship between calcium and exocytosis. Similar results were observed in separate experiments using baclofen, which reduced calcium influx by 46% and exocytosis by 83% (Wachowiak et al. 2005). The reduction in calcium influx is apparently selective to the N-type calcium conductance because after blockade of N-channels with ω-conotoxin GVIA, baclofen produces no further suppression of presynaptic calcium influx (Wachowiak et al. 2005).

What evokes GABA_B-receptor-mediated presynaptic inhibition?

This GABA_B-receptor-mediated presynaptic inhibition of OSN terminals appears to have both tonic and feedback components. In anesthetized mice whose OSN presynaptic terminals were loaded with calcium-sensitive dye, Pírez and Wachowiak (2008) used electrical stimulation to evoke single action potential volleys in OSN axons in vivo before and after the application of the GABA_B receptor antagonist CGP35348. They found that CGP35348 application increased the size of evoked presynaptic calcium signals by 38% (Figure 2B), which is evidence that a tonic GABA_B-mediated inhibition is ongoing even in the absence of explicit odorant presentation. Nearly identical results were observed under both pentobarbital and isoflurane anesthesia, demonstrating that this effect is unlikely to be a result of the anesthetic. Analogous experiments in mouse and rat olfactory bulb slices have also observed that GABA_B receptor blockade relieves a tonic presynaptic inhibition of release from OSN terminals onto periglomerular and external tufted cells (e.g., Aroniadou-Anderjaska et al. 2000; Shao et al. 2009). Several studies in slices have also failed to observe this tonic inhibition (e.g., Wachowiak et al. 2005). However, the reduced preparation of a brain slice may reduce the spontaneous activity of GABAergic interneurons in the bulb and differences in slice preparation methodology between laboratories may cause modest differences in slice health. Consequently, the observation of pronounced tonic presynaptic inhibition in vivo provides critical evidence that tonic activity is the normal state of the circuit, at least until data from awake, non-anesthetized mice become available.

GABA_B receptor-mediated inhibition of OSN terminals is also evoked as feedback in response to OSN stimulation. In slice experiments, electrical stimulation of OSN axon bundles evokes a robust, CGP35348-sensitive, presynaptic calcium-suppressing inhibition of subsequent neurotransmitter release (Figure 2C) that begins about 50ms after the initial stimulation, peaks about 200ms after the initial stimulation, and gradually recedes over 1–2 s (McGann et al. 2005; Wachowiak et al. 2005). When glutamatergic transmission is pharmacologically blocked with AMPA- and NMDA-type receptor antagonists to prevent the electrically evoked glutamate release from OSNs from exciting postsynaptic neurons, this GABA_B-receptor-mediated inhibition does not occur (McGann et al. 2005; Wachowiak et al. 2005). This demonstrates that it is a feedback inhibition evoked by the initial neurotransmitter release. This feedback presynaptic inhibition has been reported repeatedly in rat and mouse brain slice experiments (Aroniadou-Anderjaska et al. 2000; Murphy et al. 2005; Wachowiak et al. 2005). Using pairs of olfactory nerve shocks in vivo in OMP-spH mice as above, Pírez and Wachowiak (2008) showed that blockade of GABA_B receptors with CGP35348 increased the size of the presynaptic calcium signals evoked by the second shocks by ~14% at a 300- ms interval under pentobarbital anesthesia (though this effect was not replicated in 3 mice anesthetized with isoflurane). Importantly, they also showed that a single sniff of an odorant evoked feedback presynaptic inhibition that suppressed subsequent nerve shock–evoked presynaptic calcium signals by 23%, demonstrating for the first time that this feedback inhibition is evoked by olfactory stimuli in vivo.

In slice experiments from rat olfactory bulb, Murphy et al. (2005) patched single periglomerular cells and strongly depolarized them to evoke GABA release prior to the electrical stimulation of OSN afferents. They found that this stimulation of a single interneuron was sufficient to suppress the magnitude of nerve shock–evoked excitatory postsynaptic currents (EPSCs) in that neuron by about 30% and that this suppression could be entirely prevented by the blockade of GABA_B receptors with CGP55845. These data demonstrate that GABA released from PG cells does indeed inhibit neurotransmitter release from OSNs. This experiment was replicated in mice by Shao et al. (2009) in identified GAD65-positive periglomerular interneurons from the subpopulation that receives primarily direct OSN synaptic input, yielding a ~38% reduction in the magnitude of nerve shock–evoked EPSCs that could be blocked by CGP55845 (Figure 2D). Nerve shock stimulation of these neurons evoked GABA_B-receptor-mediated feedback inhibition that suppressed subsequent neurotransmitter release from the OSNs as above. Importantly, single stimulations of GABA release from the other subpopulation of GAD65-positive interneurons, which are driven by OSNs primarily indirectly via external tufted cells, did not further suppress nerve shock–evoked synaptic input, with no effect of GABA_B receptor blockade (Shao et al. 2009). The authors concluded that the PG cells driven directly by OSN input are likely responsible for the feedback component of the GABA_B-receptor-mediated presynaptic inhibition or OSNs, whereas the PG cells driven by external tufted cells are primarily responsible for the tonic inhibition of OSN synaptic terminals (Shao et al. 2009). This remarkable specialization suggests that the tonic and feedback components of the presynaptic inhibition of OSN synaptic output may have different functions in olfactory information processing (discussed below in Functions of Presynaptic Inhibition of OSN Presynaptic Terminals).

The “relative” contributions of the tonic and feedback modes of this inhibition thus appear to be dependent on whether the experiment is in vivo or in slices, on the anesthesia used in vivo, and whether it is evoked by electrical (and thus synchronous) or odorant (and thus less synchronous) stimulation of OSNs. However, neurotransmitter release evoked by odorant presentation in vivo is distributed in time over several seconds and so is presumably modulated by both components of this inhibition. Ignoring this distinction and simply blocking all GABA_B-receptor-mediated presynaptic inhibition in vivo with CGP35348 during odorant presentation causes about a 78% increase in total odorant-evoked spH signals (McGann et al. 2005; Vucinic et al. 2006) and a ~35% increase in odorant-evoked OSN presynaptic calcium signals (Pírez and Wachowiak 2008) in mice. The overall effects of this inhibition are thus large enough to substantially shape stimulus-evoked sensory input to the brain’s olfactory system.

Functional organization of GABA_B-mediated presynaptic inhibition

When GABAergic presynaptic inhibition is evoked as feedback to OSN stimulation, which OSNs are inhibited? The functional organization of this inhibition has been examined both in olfactory bulb slices and in vivo to determine if the feedback is selective to the OSNs that evoked it or whether the inhibition also affects other OSN afferents to the same glomerulus or to neighboring glomeruli (McGann et al. 2005; Murphy et al. 2005; Pírez and Wachowiak 2008). In slice experiments from OMP-spH mice (McGann et al. 2005), individual glomeruli were identified in which 2 separate OSN axon bundles entered from different directions, allowing the independent stimulation of each bundle with 2 different electrodes (Figure 3A, left). Stimulating 1 OSN axon bundle suppressed subsequent neurotransmitter release from the other bundle by about 58%, and this inhibition could be completely blocked by CGP35348 (Figure 3A), thus demonstrating that the GABA_B-receptor-mediated feedback inhibition affected other OSN axons innervating the same glomerulus. A weak effect (~12% suppression) was also observed on OSN axons innervating neighboring glomeruli in slices, but some or all of this effect could have been an artifact of direct electrical stimulation of SA cells projecting to neighboring glomeruli (Aungst et al. 2003; Liu et al. 2013). To test for effects on neighboring glomeruli more powerfully, an in vivo model was used that compared the magnitude of OSN inputs to a set of glomeruli when they were stimulated alone (by presentation of 1 test odorant) and when they were stimulated at the same time as their neighbors (by presentation of the test odorant with another odorant that activated nearby glomeruli). No inhibition of neighboring glomeruli was found across the 6 different odorant pairings that were tested (Figure 3B and C; McGann et al. 2005). Subsequent experiments in this model used electrical stimulation of OSN axon bundles in vivo to evoke presynaptic calcium signals in OSNs following odorant stimulation of those same OSNs or of OSNs innervating neighboring glomeruli (Pírez and Wachowiak 2008). These data demonstrated that odorant stimulation sufficient to evoke significant feedback inhibition of OSNs within a glomerulus did not evoke inhibition of OSNs innervating neighboring glomeruli. The GABA_B-mediated presynaptic inhibition is thus strictly intraglomerular, and consequently, it is specific to OSNs expressing a given receptor type (McGann et al. 2005). It could, therefore, potentially serve as the substrate for modulating primary sensory inputs for specific odorants or odorant features.

Figure 3.

Functional organization of GABA_B-mediated presynaptic inhibition of OSN. (A) In slice experiments in OMP-spH mice, glomeruli with bundles of olfactory neuron axons entering from different angles were identified so that the bundles could be stimulated independently (left). Single nerve shock stimulations of bundle 1 were found to evoke less neurotransmitter release from OSN terminals into the glomerulus when they were preceded by stimulation of bundle 2 than when bundle 1 was stimulated alone (middle). This effect could be entirely prevented by the blockade of GABA_B receptors with CGP55485 (right). Adapted from McGann et al. (2005), used by permission. (B) In in vivo experiments in OMP-spH mice, the odorants butyl acetate (left) and methyl valerate (middle) drive neurotransmitter release into neighboring sets of glomeruli, as shown in pseudocolor difference maps showing odorant-evoked increases in fluorescence across olfactory bulb glomeruli. Simultaneous presentation of both odorants evokes a pattern of neurotransmitter release that does not reveal any lateral presynaptic inhibition between neighboring glomeruli (right). The white line to the left is the outline of the olfactory bulbs, and the arrow denotes the glomerulus depicted in C. (C) Representative spH signals from the glomerulus marked with an arrow in part B under each of the 3 conditions depicted there. Note that the size of the response is not different when evoked by butyl acetate alone, when few of its neighboring glomeruli are also receiving sensory input, when it is evoked by butyl acetate and methyl valerate together, and when many of its neighboring glomeruli are receiving input, demonstrating a lack of lateral presynaptic inhibition. Grey bars indicate the time points subtracted to create the difference maps presented in part B. Parts B and C adapted from McGann et al. (2005), used by permission. (D) Difference maps showing the pattern of synaptic input to olfactory bulb glomeruli (spH signals) evoked by the odorant ethyl butyrate before (upper left) and after (lower left) the electrical stimulation of the dorsal raphe nucleus (DRN), which sends serotonergic projections into olfactory bulb glomeruli. A plot of individual glomerular responses after raphe stimulation compared with their own responses prior to stimulation showed about a 40% reduction in response amplitudes on average, with the glomeruli that made smaller initial responses being less affected than those that made larger initial responses (right). Adapted from Petzold et al. (2009), used by permission.

Evidence for presynaptic inhibition by dopamine

The olfactory bulb contains a very large number of dopaminergic neurons, substantially more than the entire substantia nigra and ventral tegmental area midbrain dopamine system (as noted by Ennis et al. 2001 in rats; compare Björklund and Lindvall 1984 and McLean and Shipley 1988). The cell bodies of these SA neurons are located within the juxtaglomerular interneuron shells and innervate groups of neighboring glomeruli (Figure 1C; Kosaka and Kosaka 2008; Kiyokage et al. 2010; Kosaka and Kosaka 2011). The presynaptic inhibition of OSN output by dopamine is considered well established, but a modern review suggests that this literature should be interpreted with caution.

OSNs express D₂ receptors in their presynaptic terminals (Nickell et al. 1991; Koster et al. 1999; Gutièrrez-Mecinas et al. 2005), and the bath application of dopamine or D₂ receptor agonists such as quinpirole to olfactory bulb slices suppresses neurotransmitter release from OSNs in mice, rats, and turtles, typically by about 40% (Hsia et al.1999; Berkowicz and Trombley 2000; Ennis et al. 2001; McGann JP, Wachowiak, M, unpublished data in mouse). However, all these studies applied exogenous dopaminergic agents. The only experiment to show effects of endogenous dopamine on OSN neurotransmitter release to date treated mouse olfactory bulb slices with cocaine (Maher and Westbrook 2008), which disrupts dopamine reuptake and thus artificially increases its efficacy. In this experiment, cocaine greatly suppressed OSN-evoked postsynaptic currents. Although cocaine affects other monoamine systems and most dopaminergic neurons corelease GABA, the effect of cocaine could be completely reversed (on average) by the D₂ antagonist sulpiride, showing that it was indeed dopamine-specific. With the application of D₂ agonists or reuptake blockers, it is thus possible to observe dopaminergic modulation of neurotransmitter release from OSNs.

However, despite the fact that dopaminergic SA interneurons are spontaneously active, at least in mice (Pignatelli et al. 2005; Puopolo et al. 2005), there has never been a report of a D₂ antagonist substantively increasing nerve-shock- or odorant-evoked neurotransmitter release from OSNs. In whole-cell recordings from mitral cells in slices from young rats, Hsia et al. (1999) found that the D₂ antagonist sulpiride blocked the quinpirole-induced suppression of nerve shock–evoked EPSCs in mitral cells postsynaptic to OSNs, but by itself had very little effect (~5% increase in OSN output). In turtle slices, Berkowicz and Trombley (2000) found that sulpiride blocked the dopamine-induced suppression of nerve shock–evoked EPSCs recorded in mitral/tufted cells postsynaptic to OSNs, but that sulpiride alone had no effect. Similarly, Ennis et al. (2001) reported that the D₂ antagonist sulpiride could prevent the exogenously applied D₂ agonist quinpirole from suppressing the OSN synaptic component of field EPSCs in olfactory bulb slices, but found no effect of sulpiride alone. Maher and Westbrook (2008) found that sulpiride had no effect on OSN stimulation–evoked EPSC amplitudes in SA interneurons. In slices from OMP-spH mice, the addition of sulpiride did not relieve any further presynaptic inhibition of OSNs beyond what was already eliminated by application of GABA_B receptor and glutamate receptor antagonists (McGann et al. 2005). Either there was no tonic dopamine release in these slice experiments or if there was, it did not affect the synaptic output of OSNs.

Several experiments have attempted to evoke the release of endogenous dopamine using electrical stimulation while monitoring OSN synaptic output. In perhaps the most direct test of the D₂-mediated presynaptic inhibition hypothesis, Maher and Westbrook (2008) directly stimulated dopamine release from single patch-clamped TH-positive SA cells in mouse olfactory bulb slices by evoking trains of action potentials or inducing prolonged depolarizations (voltage steps to 0 mV for up to 100ms) prior to OSN stimulation, but no suppression of OSN stimulation–evoked EPSCs was observed. The lack of effect is of course not dispositive, as the activity of a single neuron might be insufficient to produce a measurable suppression, but an almost exactly analogous experiment performed in single PG interneurons does indeed produce a substantial suppression of OSN output (Murphy et al. 2005; Shao et al. 2009). In the experiment depicted in Figure 3A, McGann et al. (2005) observed robust presynaptic inhibition of OSNs when an OSN axon bundle innervating the same glomerulus was stimulated (as discussed above in Functional organization of GABA_B-mediated presynaptic inhibition), but GABA_B receptor blockade alone was sufficient to completely eliminate this stimulation-evoked presynaptic inhibition, suggesting that GABA alone was responsible for the observed suppression of release. These experiments were thus designed to evoke dopamine release, either directly from patched dopaminergic neurons or indirectly via nerve shock, but no effect of dopamine release was observed on neurotransmitter release from OSNs.

How can it be that artificial stimulation of D₂ receptors reduces OSN synaptic output but that this output is unaffected by D₂ antagonists or attempts to physiologically induce dopamine release under control conditions? Given the expression of D₂ receptors on OSN terminals, a likely explanation is that these experimental manipulations have simply failed to evoke appropriate dopamine release. For instance, most OSN stimulation paradigms have focused on OSN axon bundles innervating one or a few glomeruli (e.g., McGann et al. 2005), which might be insufficient to elicit action potentials in SA cells that typically innervate 5–12 glomeruli and many of which receive only disynaptic input from the olfactory nerve (Kiyokage et al. 2010). Alternatively, activity of SA cells may be partially driven by inputs extrinsic to the olfactory bulb (e.g., projections from the anterior olfactory nucleus [AON]; see below in Plasticity and extrinsic modulation of preynaptic inhibition) and are consequently deafferented and quiescent in a typical olfactory bulb slice preparation.

Another potential explanation would be that perhaps exogenous dopaminergic agonists are not acting directly on the OSN presynaptic terminal but are instead acting indirectly by stimulating GABA release from PG cells that in turn tonically suppress release from OSNs via presynaptic GABA_B receptors. Surprisingly, none of the experiments showing a presynaptic effect of dopaminergic agonists on OSN terminals were conducted in the presence of GABA_B receptor antagonists, so this indirect effect of dopamine agonists on presynaptic inhibition cannot be ruled out. Importantly, a recent study by Liu et al. (2013) used optogenetic stimulation of SA cells (using a TH-driven expression of channelrhodopsin-2) to show that their activity can increase firing in ET cells. This would be expected to increase the excitatory drive onto PG cells and thus enhance the tonic GABA_B-receptor-mediated inhibition of the OSN presynaptic terminal.

If the presynaptic inhibition induced by these agonists were in fact an indirect, GABA_B receptor-mediated effect, this might indicate that the presynaptic D₂ receptors do not mediate presynaptic inhibition of neurotransmitter release from OSNs. Instead, they may play a role in long-term presynaptic plasticity (Lee and Dong 2011; Kass et al. 2013a) or signaling through pathways like Akt-GSK-3 that do not directly modulate synaptic function (Beaulieu et al. 2007). Finally, dopamine has been shown to reduce odorant-evoked calcium transients in OSN cell bodies via D₂ receptors in slices of olfactory epithelium (Hegg and Lucero 2004), where dopamine is secreted in response to trigeminal stimulation by noxious stimuli (Lucero and Squires 1998), so it is at least imaginable that the expression of D₂ receptors on OSN presynaptic terminals is ectopic localization within the cell or somehow related to this peripheral function.

Presynaptic inhibition by other neurotransmitters

Murphy and Isaacson (2003) showed that neurotransmitter release from OSN terminals is modulated by cyclic nucleotides in rats and mice. Modest concentrations (100–150 µM) of 8-Br-cyclic guanosine monophosphate (cGMP; a membrane permeable cGMP analog) greatly increased the frequency of spontaneous EPSCs recorded from neurons postsynaptic to OSN terminals, whereas higher concentrations (400–500 µM) also greatly suppressed the magnitude of evoked EPSCs under normal conditions via presynaptic depolarization block. This effect was confirmed to be pharmacologically independent of the GABAergic presynaptic signaling pathway, but increased cGMP could actually enhance OSN neurotransmitter release after the probability of release had been reduced by baclofen. It remains unclear how this cyclic nucleotide signaling could be endogenously activated in the bulb. There is growing evidence of activity-evoked production of nitric oxide in the bulb (Lowe et al. 2008; McQuade et al. 2010), which can alter cGMP levels.

A number of other transmitters are present in the olfactory bulb, but evidence of direct effects on the OSN terminal is sparse. Zinc has also been found to localize within vesicles in mouse olfactory bulb glomeruli (Jo et al. 2000) and could potentially presynaptically inhibit neurotransmitter release by modulating presynaptic calcium conductances (Horning and Trombley 2001), but this has not been demonstrated outside of a culture model. Serotoninergic fibers from the dorsal raphe richly innervate the olfactory bulb glomeruli, and direct application of serotonin can suppress neurotransmitter release from OSNs (Wachowiak et al. 2009), but this effect is indirect (see below; Petzold et al. 2009). The glomerular layer of the olfactory bulb is also innervated by cholinergic projections from the horizontal limb of the diagonal band of Broca (HDB), but a recent preliminary report indicated that this projection does not influence OSN synaptic output in mice (Rothermel et al. 2012).

Plasticity and extrinsic modulation of presynaptic inhibition

Presynaptic inhibition is a convenient substrate by which neural circuits can implement gain control (see Dynamic range extension below in), which could include not only short-term feedback but also long-term plasticity to adapt to the sensory environment. TH expression in SA interneurons is famously downregulated by olfactory sensory deprivation in mice and rats even in adult animals (Cho et al. 1996; Kass et al. 2013b), as is GAD67 expression (Parrish-Aungst et al. 2011) in those same neurons (Kiyokage et al. 2010), whereas GAD65 expression remains unchanged. This differential plasticity between uniglomerular, PG interneurons and multiglomerular SA interneurons suggests a differential role for the 2 populations in modulating neurotransmitter release from OSNs, but the functional consequences of this difference have not yet been demonstrated.

Another function of gain control can be to adapt the sensory system to different internal (McCormick and Bal 1994; Murakami et al. 2005) or behavioral (Gaudry and Kristan 2009) states of the animal. The glomerular layer of the olfactory bulb receives substantial centrifugal projections from other brain regions, including the dorsal raphe, locus coeruleus, the HDB, and the AON (Price and Powell 1970; de Olmos et al 1978; Macrides et al. 1981; Luskin and Price 1983; Záborszky et al. 1986; Salcedo et al. 2011). One of the few functions of these projections that has been definitively identified is the suppression of OSN synaptic output by stimulating the GABA_B-receptor-mediated presynaptic inhibition circuit. In a series of experiments in anesthetized OMP-spH mice, Petzold et al. (2009) demonstrated that activation of serotonin 2C receptors in the olfactory bulb by either direct application of agonist or by electrical stimulation of the dorsal raphe was able to suppress odorant-evoked neurotransmitter release from OSNs (Figure 3D), whereas blockade of these receptors increased odorant-evoked OSN output. Importantly, these serotonergic effects were not directly on the presynaptic terminal but instead required GABA_B receptors and could be blocked by CGP55845 application. Interestingly, the effects were disproportionately large in glomeruli that received the strongest initial OSN outputs (Figure 3D). The release of serotonin into olfactory bulb glomeruli appears to excite external tufted cells, thus increasing the excitatory drive onto the GABAergic PG cells that mediate tonic inhibition onto OSNs in mice (Liu et al. 2012).

The study from Petzold et al. (2009) is the most direct demonstration of extrinsic modulation of presynaptic inhibition of OSN synaptic output, but a body of indirect evidence is accumulating. The AON is a large, poorly understood structure located between the olfactory bulb and piriform cortex that sends substantial projections to the olfactory bulb, including to the glomerular layer of the olfactory bulb (Price and Powell 1970; Pinching and Powell 1972; Matsutani 2010; Markopoulos et al. 2012). Optogenetic stimulation of AON projections to the olfactory bulb has recently been shown to excite GABAergic PG interneurons in mice (Markopoulos et al. 2012). This innervation was shown to mediate inhibition of mitral cells (Markopoulos et al. 2012), but it could also plausibly drive either tonic or phasic presynaptic inhibition onto OSNs, though this was not tested.

Two recent reports have explored the effects of stimulating cholinergic projections from the HDB to the olfactory bulb in mice (Ma and Luo 2012; Rothermel et al. 2012). Ma and Luo (2012) reported that optogenetic HDB stimulation suppressed the spontaneous firing rate of PG interneurons but increased their responsiveness to odorants. This result suggests that corresponding presynaptic effects could be observed in OSN presynaptic terminals, but a preliminary report from Rothermel et al. (2012) showed no effect of HDB stimulation (on average across glomeruli) on odorant-evoked neurotransmitter release in OSNs in OMP-spH mice. It may be that HDB stimulation indeed has no effect on OSN presynaptic function, or that the stimulation paradigm used by Rothermel et al. (2012) did not excite PG cells in the same way as the paradigm used by Ma and Luo (2012), or simply that the bulb’s cholinergic system has heterogeneous effects under different circumstances. Moreover, acetylcholine has been shown to suppress the firing of dopaminergic SA cells in mice (Pignatelli and Belluzzi 2008), so the effects of HDB stimulation may be difficult to interpret given the complexities of the dopaminergic circuit in the bulb (Liu et al. 2013). Despite the present uncertainties about the mechanisms, it is clear that the shaping of OSN synaptic output by extrinsic projections could have tremendous import for models of olfactory coding (see Presynaptic inhibition as a substrate for tuning perceptual filters below).

Conclusions

• Glutamate release from OSNs is strongly modulated by a GABA_B-receptor-mediated presynaptic inhibition that is organized on an intraglomerular basis.

• GABAergic presynaptic inhibition of OSNs has both a tonic, “always on” component and a phasic, “feedback” component that may be mediated by different populations of GABAergic PG interneurons.

• OSNs also receive a D₂ dopamine–receptor-mediated presynaptic inhibition that is organized on a multiglomerular basis, but the mechanisms of this inhibition and circumstances under which it is engaged remain poorly understood.

• The behavior of these presynaptic inhibition circuits can vary as a result of extrinsic modulation from other brain regions and experience-dependent plasticity in olfactory bulb circuitry.

Functions of presynaptic inhibition of OSN presynaptic terminals

The modulation of the synaptic output of OSNs can be fundamentally conceptualized as “primary sensory gain control” because it alters the relationship between a given physical stimulus and the corresponding neural input to the central nervous system. Although similar effects could be produced by suppressing OSN firing rates or reducing the gain of response amplification during olfactory transduction, the anatomy of the olfactory system does not include extensive projections to the olfactory epithelium where the OSN cell bodies and dendritic cilia are located. Modulation of OSN signaling at their synaptic output in the olfactory bulb may thus be the most anatomically convenient way to regulate this input gain. Moreover, because the axons of spatially distributed OSNs expressing a given receptor converge to form the olfactory bulb glomerulus (Mombaerts 2006), inhibition at the glomerular layer potentially enables gain control of the aggregate output of many functionally similar OSNs at once.

A number of functions have been proposed for presynaptic gain control in OSN terminals, including the preservation of OSN synaptic resources, extension of the dynamic range of responses, tuning of sensory inputs across glomeruli, and the “clearing” of OSN responses between sniffs. The feedback and tonic components of this inhibition could have considerably different effects and thus correspondingly different functions.

Preservation of OSN synaptic resources

The olfactory system is exquisitely sensitive, meaning that even very modest numbers of odorant molecules can produce meaningful OSN input to the olfactory bulb and corresponding olfactory perceptions (Devos et al. 1990). Presumably to facilitate this sensitivity, the OSN synapse is specialized to have an unusually high release probability (Murphy et al. 2004). As a consequence, the OSN synapse is also quite susceptible to paired-pulse depression of neurotransmitter release, even when all known sources of presynaptic inhibition have been pharmacologically blocked, apparently as a result of vesicle depletion (Murphy et al. 2004).

Paradoxically, the inhibition of neurotransmitter release by GABA_B receptors can sometimes actually facilitate that release over the course of a train of synaptic inputs. In the avian auditory system, the auditory nerve fibers make end bulbs that also contain strongly depressing glutamatergic synapses that have unusually high release probabilities, and they also receive presynaptic inhibition via GABA_B receptors (Brenowitz et al. 1998). The application of baclofen at this synapse reduces electrically evoked transmitter release from the presynaptic terminal in this system just as it does in the OSN terminal, but greatly improves the terminal’s ability to repeatedly release transmitter over the course of a train of high-frequency electrical stimuli (Brenowitz et al. 1998). Although these experiments have not been replicated in the olfactory system, data from several laboratories have shown that baclofen increases the paired-pulse ratio in paired-shock experiments (Murphy and Isaacson 2003; Wachowiak et al. 2005), consistent with this potential role. Given that OSNs can fire action potentials at up to 200 Hz (Reisert and Matthews 2001), the preservation of synaptic resources during prolonged stimulation may be an important function of the GABAergic presynaptic inhibition circuit.

Dynamic range extension

One of the principal challenges facing the olfactory system (like most sensory systems) is that the range of ecologically relevant physical stimulus magnitudes can greatly exceed the dynamic range of OSN responses. Individual OSNs typically require at least some minimum odorant concentration to evoke a response, increase their firing rate with increasing odorant concentration over some range (called the dynamic range), and then “saturate” at some odorant concentration above which no further increase in firing rate is possible. Bozza et al. (2002) found that the dynamic range of individual M71-expressing OSNs was often less than 1 log unit for a given odorant in dissociated mouse cells. Grosmaitre et al. (2006) found that the dynamic range of MOR23-expressing OSNs in intact mouse olfactory epithelium spanned up to 3 log units, which they described as an unexpectedly broad dynamic range. However, the range of odorant concentrations encountered in the environment and discriminable by human subjects is much greater than that.

By suppressing neurotransmitter release from OSNs, presynaptic inhibition has been proposed to stave off saturation as the odorant concentration increases (Wachowiak and Cohen 1999; Ennis et al. 2001; Wachowiak et al. 2005), thus extending the dynamic range. The feedback mode of presynaptic inhibition would be especially well suited to this function because it could provide inhibition proportional to the synaptic output of the neuron. Figure 4A illustrates this model, where the concentration-response function of the OSN is depicted as a Hill function whose dynamic range is truncated by a hard limit on actual synaptic output (e.g., transmitter depletion, depolarization block, and so on). The addition of a proportional feedback inhibition reduces the magnitude of synaptic output at all concentrations but greatly increases the range of concentrations that evoke non-saturating responses from the OSN.

Figure 4.

Potential functions of presynaptic inhibition of OSN. (A) Diagram illustrating the potential expansion of OSN dynamic range by proportional presynaptic inhibition. The dashed line indicates the maximum synaptic output that limits neurotransmitter release in the absence of presynaptic inhibition, whereas the light grey line above the limit indicates the pattern of output that would occur if release were unlimited. The dark and light grey bars indicate the dynamic range of OSNs without and with proportional feedback inhibition, respectively. (B) Diagram illustrating the potential shaping of odorant representations by tonic presynaptic inhibition. The left-most box represents the relative activity of OSN activity in afferents from the epithelium (darker indicates more activity), whereas the middle boxes represent 2 potential patterns of presynaptic inhibition in those populations (darker represents stronger inhibition). The right-most boxes depict the final synaptic input to the brain after the OSN input is convolved with the presynaptic inhibition. Note the quite different representations of the same odorant between the 2 different states (upper and lower).

This model has received some experimental support in work in the glomeruli of the Drosophila antennal lobe (Root et al. 2008). However, this dynamic range extension would only apply if neurotransmitter release is the limiting factor in the OSN response across odorant concentrations. Given that all the above studies observed narrow dynamic ranges in the odorant-evoked responses of OSN cell bodies in the olfactory epithelium (Bozza et al. 2002; Grosmaitre et al. 2006), well upstream of the synapse, it is not clear whether presynaptic inhibition of OSNs would be effective at preventing response saturation within an individual OSN. However, the overall circuit includes the convergence of thousands of OSN axons into each glomerulus, each of which is potentially coming from a neuron with a different sensitivity to odorants (Bozza et al. 2002). Because the inhibition evoked by output from 1 set of OSN afferents affects other OSN afferents within the same glomerulus (Figure 3A), the effect of this inhibition across the population of OSNs innervating a given glomerulus could be complex.

Presynaptic inhibition as a substrate for tuning perceptual filters

Dynamic range extension is most effective with a feedback inhibition, but there is increasing evidence that the tonic mode of inhibition may be more significant in vivo (Pírez and Wachowiak 2008). This suggests that a significant portion of the presynaptic inhibition of OSNs is determined prior to odorant onset. Consequently, this circuit could serve as a global gain control or gate, limiting the strength of olfactory sensory input as a function of the animal’s behavioral state, as it does in the leech mechanosensory system (Gaudry and Kristan 2009). The global modulation of this presynaptic inhibition circuit by serotonergic inputs from the dorsal raphe likely serves precisely this purpose (Petzold et al. 2009), providing a broad “volume control” that gates olfactory input as a whole, perhaps especially the strongest inputs (Figure 3D).

In contrast to this global function, a prominent feature of this GABA_B-mediated presynaptic inhibition is that it is glomerulus- and thus olfactory receptor-specific (McGann et al. 2005). Consequently, by separately modulating the PG circuitry of each glomerulus, the brain could differentially modulate the brain’s inputs from different OSN populations, potentially “tuning” the system in ecologically appropriate ways. This model is illustrated in Figure 4B, which depicts the relative activation of the OSN populations innervating olfactory bulb glomeruli based on odorant binding in the epithelium (left), but which is then modulated by presynaptic inhibition that is stronger in some glomeruli than others (middle column, darker circles indicate glomeruli with stronger tonic presynaptic inhibition). This differential presynaptic inhibition serves to shape the overall patterns of OSN synaptic input the olfactory bulb such that it reflects a combination of the primary sensory activity driven by the odorant and the pre-existing pattern of tonic inhibition across glomeruli. Different neural or behavioral states could induce different patterns of inhibition (upper vs. lower rows), resulting in perceptual filters that induce quite different primary representations of the same odorant under different circumstances.

One neurobiological limitation on this model is that brain regions that serve as the origin of these extrinsic fibers do not obviously recapitulate the mapping of olfactory receptors onto glomeruli. In the AON pars externa, different odorants do evoke spatially distinct patterns (Kay et al. 2011), suggesting that a spatial mapping for extrinsic modulation of bulbar interneurons is possible (Markopoulos et al. 2012), perhaps based on the “patchy” projection patterns to the olfactory bulb (Matsutani 2010). However, it is unclear how diffuse projections from neuromodulatory structures like the dorsal raphe or the HDB could implement a glomerulus-specific presynaptic inhibition. Of course, it may be true that these projections have more glomerulus-specific structure than is currently appreciated, but alternatively, this representational mismatch could conceivably be resolved by a system in which all glomeruli receive the same neuromodulatory input, but the intraglomerular circuitry responds to it differently in different glomeruli (e.g., based on differences in neuromodulator receptor expression in ET cells across glomeruli).

This model predicts that the brain could potentially release the tonic inhibition in an odorant-specific set of glomeruli to become more sensitive to the corresponding odorant, or increase the inhibition to suppress responding to it or differentiate it from related odorants. The olfactory system exhibits considerable effects of expectation and attention (Dalton 1996; Kay and Laurent 1999; Distel and Hudson 2001; de Araujo et al. 2005; Doucette et al. 2011) and receives massive centrifugal inputs from higher brain regions (Price and Powell 1970; de Olmos et al. 1978; Luskin and Price 1983; Laaris et al. 2007), suggesting that presynaptic inhibition onto OSN terminals could indeed serve as a substrate for these more cognitive effects. Because it may require extrinsic input from other brain regions, the role of presynaptic inhibition on OSNs in establishing such perceptual filters is presumably only testable in vivo, possibly only in awake mice.

Temporal features of presynaptic inhibition

The time course of the feedback component of the GABA_B receptor-mediated presynaptic inhibition roughly corresponds to the frequency of sniffing observed in rats and mice (Youngentob et al. 1987; Uchida and Mainen 2003; Wesson et al. 2008). It has been suggested that this feedback inhibition serves to temporally filter OSN inputs accordingly (Aroniadou-Anderjaska et al. 2000; Ohliger-Frerking et al. 2003), presumably by suppressing residual inputs from one sniff in favor of new inputs from a second sniff. This hypothesis was tested in a study that experimentally varied the sniffing frequency of OMP-spH mice by performing a double tracheotomy (to decouple the mouse’s respiration from its sniffing) and using intratracheal suction to artificially induce inspirations through the nose at frequencies from 1–5 Hz (Pírez and Wachowiak 2008). A 5-Hz frequency corresponds to an inhalation every 200ms, with each inhalation occurring at roughly the peak of the GABA_B receptor-mediated feedback presynaptic inhibition evoked by the previous inhalation, whereas the 1-Hz frequency is slow enough to permit the inhibition to mostly dissipate between inhalations (McGann et al. 2005). However, it was found that the effect of GABA_B-receptor blockade on odorant-evoked neurotransmitter release from OSNs was equivalent across sniffing frequencies. An alternative model is that the rapid feedback component of this presynaptic inhibition plays a role in temporal odor coding within an individual sniff, which also occurs on the scale of tens of milliseconds (Smear et al. 2011).

Clinical applications

GABA_B receptor antagonists have been a popular target for drug discovery recently because of growing evidence that they can produce a general cognitive improvement in animal models (Bowery et al. 2002). One such antagonist, SGS742 (CGP36742), was used in a Phase II clinical trial in human patients with mild cognitive impairment (Froestl et al. 2004), resulting in significantly improved attentional function and reaction time on sensory-cued testing. The results reviewed above suggest that GABA_B receptors could also be a worthwhile target to try to improve sensory function in certain dysosmias. For instance, the blockade of GABA_B receptors might improve hyposmia associated with normal aging (Lasarge et al. 2009) or exposure to environmental toxicants (Sułkowski et al. 2000; Baker and Genter 2003; Bushnell et al. 2007; Czarnecki et al. 2011; Czarnecki et al. 2012).

Conclusions

• Presynaptic inhibition of OSN axon terminals serves as a primary sensory gain control that scales the magnitude of neural input to the brain for a given olfactory stimulus.

• This inhibition could serve to maintain OSN responsivity when faced with strong olfactory stimuli, either by preserving synaptic resources or by extending the system’s dynamic range.

• The shaping of the presynaptic inhibition of OSNs by extrinsic inputs from other brain regions and sensory activity suggests that these circuits could be used to “tune” the brain’s primary olfactory input based on prior experience.

• This presynaptic inhibition could be a useful target for pharmacological manipulation of olfactory function in the clinic.

Summary

The presynaptic inhibition of neurotransmitter release from OSNs is robust both in vivo and in slices and includes both feedback and tonic components. Much of this inhibition is mediated by presynaptic GABA_B receptors driven by GABA arising from juxtaglomerular interneurons. Although dopamine can also presynaptically modulate transmitter release from OSNs, the mechanisms and circumstances of this modulation remains unclear. Presynaptic inhibition may serve to establish perceptual filters, preserve synaptic resources, or extend the system’s dynamic range.

Funding

This work was funded by National Institute on Deafness and Other Communication Disorders [R00 DC009442].