Response kinetics of olfactory receptor neurons and the implications in olfactory coding

ORNs are embedded in the olfactory epithelium that covers the olfactory turbinates, a convoluted bone structure, located in the posterior nasal cavity in most mammals. These primary sensory neurons, like touch-sensitive neurons but unlike photoreceptors (see accompanying Perspectives by Bautista and Lumpkin, 2011, and Schwartz and Rieke, 2011) generate APs upon stimulation and send axons to the olfactory bulb, where they communicate with mitral and tufted (second-order) neurons. Odorants are carried to ORNs via the inhaled air in a rhythmic fashion, the frequency of which depends on the breathing frequency that can vary from 2 to 3 Hz at rest to up to 10 Hz during sniffing in rodents (Youngentob et al., 1987; Tankersley et al., 1994; Kepecs et al., 2007).

Olfactory transduction occurs in olfactory cilia, which extend from the dendritic knob at the end of the single dendrite into the mucus that covers the epithelium (Fig. 1 A). Transduction begins with the binding of odorants to odorant receptors (ORs) located in the ciliary membrane (Fig. 1 B). A given ORN selectively expresses one of ∼1,000 types of functional ORs found in the mouse genome. The OR triggers the activation of the olfactory G protein G_olf, which in turn activates adenylyl cyclase III (ACIII), leading to an increase in ciliary cAMP. Subsequent opening of the olfactory CNG channel by cAMP leads to an influx of Na⁺ and Ca²⁺, with Ca²⁺ gating a Ca²⁺-activated Cl⁻ channel (Kleene, 2008; Reisert and Restrepo, 2009; Su et al., 2009). As intracellular Cl⁻ is high in ORNs, a secondary excitatory anionic current ensues and carries up to 90% of the odorant-induced receptor current (Lowe and Gold, 1993; Reisert et al., 2005; Boccaccio and Menini, 2007). Fig. 2 shows the currents generated in an ORN that has been loaded with caged 8-Br-cAMP. Upon photolysis (Fig. 2, arrow) uncaged 8-Br-cAMP leads to activation of the CNG channel, giving rise to an initial small current that is followed by a large secondary current. This secondary current can be suppressed by niflumic acid, a blocker of the Ca²⁺-activated Cl⁻ channel.

The importance of the G_olf α subunit, ACIII, and the principle subunit of the CNG channel, CNGA2, in response to activation was demonstrated in mice engineered to lack these transduction components. Such mice showed hardly any odorant responses, and pups had poor survival rates because of their inability to feed. Although the knockout approach was very successful to demonstrate that these transduction components are required for olfactory transduction, the lack of response precluded more detailed analysis of their role in response kinetics. The physiological role of the last component in the activation chain, the olfactory Ca²⁺-activated Cl⁻ channel, has been known since the early 1990s but was only molecularly identified in 2009 to be Ano2 (Stephan et al., 2009; Hengl et al., 2010; Rasche et al., 2010). A recent loss-of-function study confirmed that Ano2 is indeed the olfactory Ca²⁺-activated Cl⁻ channel. Intriguingly, Ano2 knockout mice showed no behavioral deficit performing some Go/NoGo tasks (Billig et al., 2011), although humans with Ano2 mutations may suffer olfactory defects (Stephan et al., 2009). Although, again, knocking out the Cl⁻ channel will preclude a detailed investigation of its contribution to receptor current regulation, it will possibly reveal in more detail what the kinetics of the cAMP-induced current might be, as this will be the only remaining current.

An increase in odorant concentration triggers, in a graded manner, APs with a shorter delay after stimulus onset and a higher firing rate. Response delays can be as short as ∼30 ms and maximal AP firing rates reach up to around 200 Hz in isolated mouse ORNs at saturating odorant concentrations. The dynamic range of ORNs is typically narrow, saturating within 2 log units of odorant concentration above threshold (Reisert and Matthews, 2001; Rospars et al., 2008). The number of APs fired does not show a monotonic behavior. Instead, the number of APs only increases up to intermediate odorant concentrations and decreases thereafter at higher concentrations down to only two to three APs. These APs are generated at the very onset of odorant stimulation during the rising phase of the receptor current (Shibuya and Shibuya, 1963; Gesteland and Sigwart, 1977; Reisert and Matthews, 2001; Reisert et al., 2007; Rospars et al., 2008). The odorant-induced receptor current can persist for much longer at higher odorant concentration for as long as the odorant is present with no further AP generation (see Fig. 3). Thus, under these conditions, no signaling of the presence of odorants would be sent to the brain by this particular ORN. Even a further increase in odorant concentration applied while a receptor current is still present will increase the receptor current but not cause additional AP firing (Reisert and Matthews, 1999). The shortening of the spike train is caused by a progressive decline of the AP amplitude (Fig. 3 A, inset), probably by inactivation during prolonged depolarization of voltage-gated Na⁺ and Ca²⁺ channels, which generate the olfactory AP (Trotier, 1994; Kawai et al., 1997). Thus, AP generation in ORNs and the duration of the spike train seem to be primarily controlled by the speed of activation (slope) of the receptor current rather than its overall time course. But termination of the receptor current is important for odorant perception during repeated stimulation, as shown in Fig. 3. ORNs were exposed twice in rapid succession for 1 s to the phosphodiesterase (PDE) inhibitor IBMX, which leads to an increase in ciliary cAMP and thus receptor current generation. In this kind of double-pulse stimulation paradigm, ORNs can only reliably generate APs to the second stimulation if the receptor current to the first odor stimulus has entirely terminated to allow the ORN to fully hyperpolarize to its resting membrane potential. For short interstimulus times (e.g., 0.25 s; Fig. 3, A and E, for time dependence of AP firing recovery), only around 50% of wild-type ORNs generate APs in response to the second stimulation but reliably do so when the interpulse time is lengthened to 1 s (Fig. 3 E). Or in other words, an increase in stimulation frequency (shorter interpulse time) progressively abolished AP generation.

Even a small response prolongation caused by altering the transduction cascade by removing a CaM-binding site on the CNG channel (CNGB1^ΔCaM mice; see below) can significantly alter AP firing during repetitive stimulation. Such CNGB1^ΔCaM ORNs were even less likely to generate APs compared with wild-type mice to the second stimulation at short (0.25-s) interpulse times (Fig. 3, B and E) because of the subtle response prolongation of the first response; but they can do so again once the interpulse interval is lengthened (Fig. 3 D). Furthermore, these changes are also conveyed to the olfactory bulb. Local field potential recordings from olfactory bulbs of freely breathing mice showed equivalent reductions in response magnitude to a second odorant exposure in CNGB1^ΔCaM mice compared with wild-type mice (Song et al., 2008).

During repetitive ORN stimulation designed to mimic sniffing, ORNs quite reliably generate APs to every stimulation when stimulated at a low (2 Hz; interpulse time, 400 ms) frequency at intermediate odorant concentrations. An increase to a stimulation rate more akin to sniffing (5 Hz; interpulse time, 100 ms) causes ORNs to fail to respond reliably with AP generation to every stimulation (Ghatpande and Reisert, 2011). At these higher frequencies, the interstimulus time is again too short and/or response termination mechanisms too slow to sufficiently terminate the receptor current to allow AP generation at the next stimulation. At high odorant concentrations, AP firing entirely stops during repetitive stimulation. This causes a reduction of information flow to the olfactory bulb from this particular ORN, an effect that is mirrored by reduced presynaptic Ca²⁺ transients in the olfactory bulb during high frequency sniffing (Verhagen et al., 2007). Inhibitory responses, seen as a suppression of the basal spike rate, have also been reported but do not form the focus of this Perspective.

Rod photoreceptors are able to detect single photons because of the high amplification of their transduction cascade; a light-activated rhodopsin molecule stays active long enough to activate multiple G protein molecules. In contrast, ORNs lack this important amplification step, as an odorant molecule only stays bound to the OR for a very short time (Bhandawat et al., 2005), often too brief to activate even a single G protein. Thus, many OR activation events remain inconsequential. An activated OR does not, unlike rhodopsin in phototransduction, activate multiple G proteins, and an OR-binding event that does lead to G protein activation (successful odorant-binding event) only generates a small unitary current response of the order of 0.03 pA in frog ORNs (Bhandawat et al., 2010), too small to trigger APs. This raises the question of how many ORs have to be activated to reliably generate APs and signal to the olfactory bulb. Performing a series of experiments at low odorant concentrations and short exposure times, and equipped with the knowledge of the unitary response size, the chance of 50% AP generation per given odorant exposure was shown to require ∼40 successful odorant-binding events in frog ORNs (Bhandawat et al., 2010). The current required was estimated to be 1.2 pA. Broadly similar values have recently been shown for mouse ORNs (Ben-Chaim et al., 2011).

During sustained odorant exposure ORNs generate a receptor current that increases to its peak and thereafter declines, often back to near zero current levels even before stimulus termination. Alternatively, adaptation is often studied in olfaction in a double-pulse paradigm, with the second response being progressively suppressed with shorter interpulse periods.

What are the mechanisms involved in such manifestations of adaptation? In particular, in light of the short lifetime of an activated OR, what mechanism shuts off an activated G protein? Although little is known about the lifetime of GTP-bound G_olf or the G_olf–ACIII complex, a putative guanine nucleotide exchange factor for G_olf, Ric-8b, has been identified in ORNs (Von Dannecker et al., 2006). A possible candidate to influence response kinetics early in the transduction cascade is olfactory marker protein (OMP). Mice lacking OMP have a greatly slowed response time course (Buiakova et al., 1996), and OMP modulation of the ciliary transduction cascade has been limited to ACIII and upstream transduction events, although the precise mechanism remains to be clarified. As a consequence of the slowed and delayed receptor current time course, OMP-knockout ORNs generate APs only after a longer delay of ∼170 ms after stimulation onset compared with 50 ms in wild-type ORNs. Also, the number of fired spikes increases from three in wild-type to five in OMP-knockout ORNs when stimulated with 100 µM cineole, probably because of the slowed rising phase of the receptor current in the mutant ORNs (Reisert et al., 2007).

Mechanisms involved in adaptation have been shown to be dependent on Ca²⁺ influx through the CNG channel, as chelating extracellular Ca²⁺ or positive holding potentials under whole cell voltage-clamped recording conditions greatly reduced the adaptive current decline (Kurahashi and Shibuya, 1990; Zufall et al., 1991; Boccaccio et al., 2006). A promising candidate for adaptation is the CNG channel itself (Zufall et al., 1991; Kurahashi and Menini, 1997). Indeed, Ca²⁺, in the presence of calmodulin (CaM), reduces the channel’s affinity to cAMP (Chen and Yau, 1994; Bradley et al., 2001). In a heterologously expressed CNG channel comprised only of the principle subunit CNGA2 (of the native heterotrimeric channel), a basic amphiphilic α-helix motif near the N terminus has been shown to mediate the Ca²⁺-CaM–mediated desensitization (Liu et al., 1994). However, when all three olfactory CNG subunits, CNGA2 and the auxiliary CNGB1b and CNGA4, are expressed heterologously, two “IQ-type” CaM-binding sites on the N and C termini of the CNGB1b and CNGA4 subunits, respectively, mediate exogenous Ca²⁺-CaM fast channel desensitization. CaM actually pre-associates with the CNG channel at basal Ca²⁺ levels (Bradley et al., 2001). Channels that lack either auxillary subunit inactivate much more slowly. Perplexingly, in a channel that still has its endogenous CaM bound, the IQ site on the CNGB1b subunit seems to be sufficient to mediate Ca²⁺-CaM–mediated desensitization (Waldeck et al., 2009).

The role of Ca²⁺-CaM–mediated CNG channel inhibition in adaptation was investigated in mice lacking either the CNGA4 or CNGB1b subunit. Indeed, ORNs from such mice showed odorant-induced responses that terminated much more slowly (Munger et al., 2001; Michalakis et al., 2006), and, when ORNs were stimulated twice in succession, a less reduced (less adapted) response to the second stimulation was recorded. This is consistent with the notion that, because of a lack of an auxillary CNG channel subunit, channel desensitization is slowed, causing a reduced adaptation of the odorant response. A confounder in this interpretation is that the knocking out of an entire auxiliary CNG subunit not only slows Ca²⁺-CaM–mediated desensitization but also greatly reduces the overall sensitivity of the CNG channel to cAMP, and greatly reduces the ciliary channel density of the CNG channels due to reduced ciliary channel trafficking (Michalakis et al., 2006).

We chose a more limited and targeted genetic manipulation to address the involvement of CNG channel desensitization and odorant adaptation by generating a mouse line that only lacked the IQ-binding site in the N terminus of the CNGB1b subunit (Song et al., 2008). In this CNGB1^ΔCaM mouse, neither the cAMP sensitivity of the CNG channel nor the ciliary channel density was changed. But as designed, Ca²⁺-CaM–mediated CNG channel desensitization was slowed ∼100-fold. Surprisingly, we did not see adaptational defects in receptor current generation as expected using a double-pulse stimulation protocol (see Fig. 3), suggesting that the role of Ca²⁺-CaM–mediated adaptation on the CNG channel might not be as important as previously thought for ORN adaptation. Instead, as mentioned above, we observed a response prolongation of around 60 ms, which could arise from prolonged Ca²⁺ influx through the CNG channel as a result of reduced Ca²⁺-CaM–mediated desensitization.

In addition to the CNG channel, ACIII has been suggested to be a feedback target of Ca²⁺ for adaptation, particularly adaptation that is induced by long odor exposures (Leinders-Zufall et al., 1999). Ca²⁺, entering cilia through CNG channels, is thought to activate CaM-dependent kinase II (CaMKII), which then phosphorylates ACIII to reduce its activity and thus down-regulates cAMP production. This hypothesis was mainly derived from experiments conducted in heterologous expression systems (Wayman et al., 1995; Wei et al., 1996) and experiments using CaMKII inhibitors in olfactory cilia preparations and in isolated ORNs (Wei et al., 1998; Leinders-Zufall et al., 1999). Phosphorylation on a single serine residue, serine¹⁰⁷⁶, of ACIII was identified to account for Ca²⁺/CaMKII-mediated inhibition (Wei et al., 1996, 1998). In a recent experiment (unpublished data), we again took a targeted genetic approach to address the potential involvement of this feedback mechanism in adaptation by generating a mouse line, which should lack CaMKII-mediated ACIII phosphorylation because of a single mutation of serine¹⁰⁷⁶ to alanine. Surprisingly, electrophysiological recordings showed no differences in the responses between wild-type and mutant mice to single short (subsecond) or long (a few to tens of seconds) stimulations, or in several adaptation-stimulation paradigms. ORNs in the mutant mice show a similar slowing of the activation kinetics in responses after a prolonged stimulation as in wild-type mice, a phenomenon that is blocked by CaMKII inhibitors and has been suggested to be a result of inhibition of ACIII (Leinders-Zufall et al., 1999). The results in the mutant mice thus suggest that Ca²⁺/CaMKII-mediated feedback inhibition of ACIII by phosphorylation is unlikely to play a significant role in adaptation.

Thus, the two historically main suspects for adaptational mechanisms, Ca²⁺-CaM–negative feedback of the CNG channel via the IQ-binding site of the CNGB1b subunit and ACIII phosphorylation of serine¹⁰⁷⁶, are not playing their expected role, at least investigated individually. Currently, no other clear candidates for adaptation are waiting in the wings. Interestingly, Ca²⁺-independent adaptation is also observed (Reisert et al., 2005), but its overall contribution to adaptation and its mechanism remains to be clarified.

Termination of the response requires cAMP and Ca²⁺ levels to return to their pre-stimulus levels to allow the CNG and the Cl⁻ channel to close. Termination kinetics should be largely determined by whichever is slower, either the rate of cAMP removal or the rate of Ca²⁺ removal from cilia. In ORNs, cAMP is hydrolyzed by PDEs. So far, two PDEs, PDE1C and PDE4A, have been found to be expressed in ORNs. PDE1C is localized in the cilium, whereas PDE4A is present in the rest of the ORN including the dendritic knob, but not in the cilia (Juilfs et al., 1997). After being generated by ACIII in cilia upon odor stimulation, cAMP could theoretically either be degraded locally by PDE or diffuse into the dendrite, where it would be degraded by PDE4A. Although little is known about the fate of ciliary cAMP, recent knockout studies in mice revealed that removal of ciliary cAMP is not a rate-limiting factor for termination kinetics. Knockout of either PDE1C or PDE4A alone does not result in a prolonged termination phenotype (Cygnar and Zhao, 2009), suggesting that either PDE1C in the cilia or PDE4A outside the cilia is sufficient to allow rapid removal of ciliary cAMP and wild-type termination kinetics. PDE activity also does not contribute to fast adaptation during a double-pulse protocol (Boccaccio et al., 2006). A potential mechanism that could contribute to a ciliary cAMP reduction (and hence response termination) is diffusional escape of cAMP from the cilia into the cell body (Chen et al., 1999; Boccaccio et al., 2006; Flannery et al., 2006). This is supported by the observation that mice that lack both PDE1C and PDE4A do show slowed response termination (Cygnar and Zhao, 2009).

Accumulating evidence suggests that the rate of ciliary Ca²⁺ removal dictates the termination kinetics, and that Na⁺/Ca²⁺ exchange is the major mechanism for ciliary Ca²⁺ removal in mice. Because ORN cilia lack intraciliary vesicular organelles, ciliary Ca²⁺ removal is thought to be achieved mainly by plasma membrane Ca²⁺ transporters including Na⁺/Ca²⁺ exchangers and ATP-dependent Ca²⁺ pumps. The presence of Na⁺/Ca²⁺ exchanger(s) in ORNs has been suggested since the mid-1990s. Preventing Ca²⁺ extrusion by reducing extracellular Na⁺, and thus abolishing the driving gradient that fuels Na⁺/Ca²⁺ exchange, prolongs the odorant-induced receptor current by seconds by generating a prolonged Ca²⁺-activated Cl⁻ current (Reisert and Matthews, 2001; Antolin and Matthews, 2007). These results indicate that Na⁺/Ca²⁺ exchange is critically important in mice for response termination. The very Na⁺/Ca²⁺ exchanger(s) that is responsible for ciliary Ca²⁺ extrusion has not been determined (Pyrski et al., 2007). In a recent experiment, we revealed that the potassium-dependent Na⁺/Ca²⁺ exchanger 4 (NCKX4) is present in olfactory cilia preparations (Stephan et al., 2009) and that it likely functions for such purpose (Stephan et al., 2010). The NCKX4 knockout ORNs display prolonged response termination by up to several seconds, a phenotype mimicking what is seen in wild-type ORNs when replacing extracellular Na⁺ to prevent Na⁺-dependent Ca²⁺ extrusion. The lesser prolonged termination seen in NCKX4 heterozygous mice compared with the knockout supports that the rate of ciliary Ca²⁺ removal is the rate-limiting factor for the termination kinetics. It is interesting that NCKX4 knockout ORNs display normal response amplitude and activation kinetics to single-odorant pulses, suggesting that the basal level of ciliary Ca²⁺ is not greatly affected by the lack of NCKX4, maybe because during resting conditions, diffusional escape of ciliary Ca²⁺ in the dendrite is sufficient to keep ciliary Ca²⁺ levels low. But in a double-pulse stimulation paradigm, NCKX4 knockout ORNs show a severe reduction in the response to the second stimulus, much more pronounced than that seen in CNGB1^ΔCaM ORNs. Consequently, the ability of AP generation to repeated stimulation is also greatly reduced, emphasizing the importance of NCKX4 in rapid response termination in olfactory coding during repeated stimulation.

Immense progress has been made over recent years in understanding odorant responses in the olfactory bulb, using either imaging or electrophysiological approaches in the behaving mouse or rat. But little is known about how mammalian ORNs respond in the live animal during natural breathing. A successful approach to start addressing this problem has been to record AP extracellularly in anesthetized rats (Duchamp-Viret et al., 1999) using sharp electrodes inserted through a small hole in the nasal bone, with odorant application so far being restricted to long (e.g., 2-s) odorant exposures directly to the olfactory epithelium. Alternatively, the electroolfactogram, which records the ensemble activity of ORNs, can be recorded from medial or lateral turbinates, with electrodes inserted through holes in the nasal bones (Scott-Johnson et al., 2000) in an anesthetized, artificially ventilated rat. An important issue to consider in these recording configurations is the unique (probably with the exception of taste receptors in the mouth) environment ORNs have to function in. Although the cell bodies reside within the olfactory epithelium in a normal ionic environment, cilia, where transduction takes place, have to function outside of the body in the environment of the olfactory mucus. The ion concentrations in mucus are significantly different compared with interstitial fluid and might also change, especially in amphibians. The few measurements to determine the mucosal ion concentrations suggest that Cl⁻ might be as low as 50 mM (Chiu et al., 1988; Reuter et al., 1998), a concentration that would considerably increase the excitatory Cl⁻ current by facilitating Cl⁻ efflux. Also, mucosal Na⁺ is low (Chiu et al., 1988; Reuter et al., 1998) and close to the K_d of the (amphibian) Na⁺/Ca²⁺/(K⁺?) exchanger (Antolin and Matthews, 2007). This could begin to slow response recovery due to slowed Ca²⁺ removal from the cilia. Indeed, at least in isolated ORNs, preventing or slowing Ca²⁺ extrusion can entirely suppress AP generation during repetitive stimulation (Reisert and Matthews, 1998), even at stimulation frequencies as low as 2 Hz (Ghatpande and Reisert, 2011). This slowed Ca²⁺ removal can further limit ORNs to respond faithfully to rapid stimulation.

So how many sniff cycles are required to determine the identity of an odorant? Well-trained mice and rats can distinguish odorant in <200 ms (Uchida and Mainen, 2003; Abraham et al., 2004; Rinberg et al., 2006), the equivalent of a single sniff at 5 Hz. But the odorant sampling duration progressively lengthens to up to 500 ms when mice have to perform more complex odorant discrimination tasks (Rinberg et al., 2006), suggesting that under such conditions, repeated odorant exposures might be required. Thus, unraveling how ORNs encode for odorants during repeated stimulations at breathing rates at rest or during high frequency sniff bouts in a freely breathing animal will be an important question to understand and interpret the olfactory code. As will understanding how each of the aspects of the olfactory response (e.g., timing, frequency, number of APs, synchronicity across ORNs targeting the same glomeruli, etc.) contributes specifically to the olfactory code.

This Perspectives series includes articles by Farley and Sampath, Schwartz and Rieke, Bautista and Lumpkin, and Zhang et al.

The authors would like to thank Drs. C.-Y. Su and G. Lowe for critical and constructive reading of the manuscript.

This work was supported by National Institutes of Health grants DC007395 (to H. Zhao), DC009613 (to J. Reisert), and DC009946 (to H. Zhao and J. Reisert), and a Morley Kare Fellowship (to J. Reisert).

Robert S. Farley served as guest editor.