Encoding and Representation of Intranasal CO₂ in the Mouse Olfactory Cortex

Abstract

Intranasal trigeminal sensory input, often perceived as a burning, tingling, or stinging sensation, is well known to affect odor perception. While both anatomical and functional imaging data suggest that the influence of trigeminal stimuli on odor information processing may occur within the olfactory cortex, direct electrophysiological evidence for the encoding of trigeminal information at this level of processing is unavailable. Here, in agreement with human functional imaging studies, we found that 26% of neurons in the mouse piriform cortex (PCX) display modulation in firing to carbon dioxide (CO₂), an odorless stimulant with known trigeminal capacity. Interestingly, CO₂ was represented within the PCX by distinct temporal dynamics, differing from those evoked by odor. Experiments with ascending concentrations of isopentyl acetate, an odorant known to elicit both olfactory and trigeminal sensations, resulted in morphing of the temporal dynamics of stimulus-evoked responses. Whereas low concentrations of odorant evoked responses upon stimulus onset, high concentrations of odorant and/or CO₂ often evoked responses structured to stimulus offset. These physiological experiments in mice suggest that PCX neurons possess the capacity to encode for stimulus modality (olfactory vs trigeminal) by differential patterns of firing. These data provide mechanistic insights into the influences of trigeminal information on odor processing and place constraints on models of olfactory-trigeminal sensory integration.

Introduction

Intranasal perception of airborne chemicals is mediated by both the olfactory and trigeminal systems. Indeed, while intranasal chemosensation is mostly attributed to olfaction, input from both of these sensory systems is critical for informing nutritional intake and avoidance of potentially harmful stimuli (Doty, 1995; Brand, 2006; Lundström et al., 2011). Major questions remain regarding the interaction of these systems in perceptually meaningful manners.

One reason why olfactory and trigeminal perceptual qualities are easily confused is that at high concentrations most odorants possess the capacity to elicit trigeminal sensations, which include intranasal burning, tingling, or prickling sensations (Doty et al., 1978, 1995). At low concentrations, however, most odorants fail to evoke trigeminal sensation. For example, nicotine is perceived to possess a musky odor at low concentrations, yet at high concentrations is a pungent nasal irritant (Edwards et al., 1987; Hummel and Kobal, 1992; Thuerauf et al., 1999). Further, the odorless trigeminal stimulant, carbon dioxide (CO₂), when presented with low-concentration odors, can impact pungency ratings of the odors (Cain and Murphy, 1980). Thus, there is considerable perceptual interplay between the olfactory and trigeminal systems that together result in our perception of odors (for review, see Brand, 2006).

The mechanisms whereby trigeminal sensation is encoded in the brain in manners capable of modulating olfaction remain unclear. Possibly subserving this, trigeminal sensory information converges into several olfactory structures (Yousem et al., 1997; Zald and Pardo, 2000; Billot et al., 2011; Lundström et al., 2011). There is direct anatomical innervation of the olfactory bulb by trigeminal fibers in the main olfactory epithelium (Schaefer et al., 2002). Trigeminal-evoked activity is observed within the olfactory epithelium (Tucker, 1971; Thürauf et al., 1991; Kratskin et al., 2000; Damann et al., 2006; Daiber et al., 2013) and the olfactory bulb (Hu et al., 2007; Gao et al., 2010). Human functional imaging studies have revealed that several structures display stimulus-dependent modulation in response to trigeminal stimulants, including the piriform cortex (PCX; Yousem et al., 1997; Zald and Pardo, 2000; Kollndorfer et al., 2013) and the olfactory tubercle (OT; Zelano et al., 2007). Related functional imaging studies have reported that even pure trigeminal stimulants, such as CO₂, evoke activity in the PCX (Hummel et al., 2009; Albrecht et al., 2010). This finding within the PCX, a structure considered critical for odor perception, including odor discrimination and odor learning (Haberly, 2001; Gottfried, 2010; Wilson and Sullivan, 2011), suggests nonolfactory sensory convergence in a manner that may underlie trigeminal influences on odor perception.

Some major questions remain, however, before being able to directly link trigeminal and olfactory physiological influences. At the most basic level, are CO₂ and other known trigeminal stimulants represented at the single-unit level in the olfactory cortex? If so, what are the spatial and temporal aspects of this encoding? Further, are concentration-dependent changes in trigeminal-elicited sensations also apparent among PCX neurons? To address these questions, here we performed in vivo electrophysiological multi- and single-unit recordings from anesthetized mice to explore the encoding and representation of trigeminal-evoked responses in the PCX. We focused our analyses mostly on activity elicited by CO₂, since this stimulus potently elicits trigeminal activity (Bensafi et al., 2008; Hummel et al., 2009; Wang et al., 2010; for review, see Luo et al., 2009), while sparing true olfactory sensation known to occur with numerous other trigeminal stimulants (e.g., ammonia, nicotine, menthol; Doty, 1995). Our results show that odor- and CO₂-evoked information converge onto a, respectively, large population of single units in the mouse olfactory cortex with distinct temporal dynamics that may underlie reports of olfactory-trigeminal integration.

Materials and Methods

Experimental subjects.

Adult male C57BL/6 mice (n = 27, 2–4 months of age) were obtained from Harlan Laboratories and maintained within the Case Western Reserve University School of Medicine animal facility. Food and water were available ad libitum. The mice were on a 12 h light/dark cycle with all experiments performed during the light cycle. All experiments were conducted in accordance with the guidelines of the National Institutes of Health and were approved by the Case Western Reserve University's Institutional Animal Care Committee.

In vivo electrophysiology.

Mice were anesthetized via urethane injection (1.0 mg/kg, i.p.) and mounted on a stereotaxic frame upon a water-filled heating pad (38°C). Anesthesia depth was verified by absence of toe-pinch reflex. A 0.05 ml injection of local anesthesia (1% lidocaine in 5% EtOH, s.c.) was then administered into the wound margin site before exposing the dorsal skull by removing the scalp. Two craniotomies were performed for electrode placement. First, a single hole was drilled from 0 mm anterior to bregma to 1.5 mm anterior to bregma for PCX and/or OT recording electrode placement. The second hole was drilled over the contralateral cortex to serve as entry for the reference electrode. Physiological (0.9%, 38°C) NaCl was applied to all craniotomy sites, which was replenished intermittently throughout the recording.

A tungsten electrode (0.01 in. o.d.; A-M Systems) was lowered into the reference craniotomy site ∼1.5 mm deep. A second tungsten electrode (same type as above) was lowered into the PCX or OT and used as the recording electrode. The recording electrode activity was digitized at 28 kHz along with respiration and stimulus presentation events using a Tucker-Davis Technologies amplifier and software.

After recording, mice were transcardially perfused with 4°C 0.9% NaCl and then 10% formalin (Fisher Scientific). Following perfusion, the brains were removed and stored in 30% sucrose formalin at 4°C.

Electrode placement verification.

All recording sites were verified by postmortem histological examinations of slide-mounted, 40 μm coronal brain sections stained with a 1% cresyl violet solution. PCX recording sites (layers i, ii, and/or iii; n = 26 from a total of 16 mice) were found across all coronal sections of the PCX (only the anterior regions of the PCX, containing the lateral olfactory tract, were used; Fig. 1). OT recording sites (n = 17 from a total of 11 mice) were found across all sections, but concentrated in the posterior area of the OT (layers i, ii, and/or iii; Fig. 1). Electrode tip locations were verified by multiple observers (K.C. and C.X.) with reference to a mouse brain atlas (Paxinos and Franklin, 2000).

Figure 1.

Electrode tip locations of PCX and OT recordings. Coronal stereotaxic panels showing the approximate location of the electrode tips following histological verification (n = 16 in PCX and 11 in OT). Total number of recording sites wherein units were collected and used for analysis = 26 in PCX and 17 in OT. Sections span from 2.0 to 0.5 mm anterior of bregma, in 0.5 mm intervals. Adapted from Paxinos and Franklin (2000).

Stimulus presentation.

Filtered, ultrahigh purity medical grade CO₂ (AirGas) and/or odor (isopentyl acetate, 1 Torr dilution unless otherwise stated; Sigma Aldrich) were presented through a Teflon odor-port using a custom air-dilution olfactometer at the total flow rate of 1 L/min in all experiments. Stimuli were presented in a counterbalanced order, ≥4 trials/stimulus, at a ≥20 s interstimulus interval. The exit of the odor port was held at a distance of 1 cm from the tip of the nose. All stimulus lines were independent up to the point of immediate entry into the Teflon odor port, eliminating the potential of cross-stimulus contamination. In all experiments, air dilution of CO₂ and odor were achieved using medical grade N₂. CO₂ stimuli (concentrations) arrived to the odor port by a single Teflon line wherein concentration of CO₂ (in N₂) was controlled between trials with an upstream flow controller. Stimulus onset (time to exit the odor port) was controlled across all stimuli by means of a vacuum line connected to a three-way solenoid valve. In our design, stimuli flowed toward and through the odor port for 20 s, in which time they were scavenged by the vacuum line. Computer-timed switching of the three-way solenoid valve resulted in a stimulus onset latency of <50 ms for all stimuli (measured with a piezo electric pressure-sensitive foil). Concentration of odor and CO₂ stimuli leaving the port were measured separately with a MiniRAE 3000 (used for odor; RAE Systems) and a TSI Q-Trak 7565 CO₂ meter. Measured values (averaged over 1 min, following application of instrument-specific correction factors) were 10 (odor), 3050 (50% CO₂), and 5820 ppm 100% CO₂. The 0% CO₂ (room air) was measured using the TSI Q-Trak at 450 ppm. Thus, this stimulus presentation paradigm resulted in dose-dependent concentrations of CO₂, all of which occurred within similar latencies. An analysis of respiration (via chest piezo sensor) from the present experiments failed to find an impact of CO₂ at either the 50% CO₂ (F_(1,12) = 0.00004, p = 0.99) or 100% CO₂ dilutions (F_(1,12) = 0.038, p = 0.849) on respiration rate (Fig. 2; 2 s prestimulus vs 2 s during stimulus, n = 7 mice, 8–12 trials/mouse).

Figure 2.

Respiratory frequency is independent of acute CO₂ presentation. A, Example respiratory traces from an anesthetized mouse during presentation with “blank” (N₂), 50% CO₂, and 100% CO₂ stimuli (all at 1 L/min flow rate). B, Mean respiratory frequency 2 s prior (baseline, −2 to 0 s) and during (0–2 s) presentation with either 50% or 100% CO₂. n = 7 mice, 8–12 trials/mouse.

The primary set of stimuli used to screen units for basic odor- and CO₂-evoked responses consisted of 50% CO₂, 100% CO₂, and odor (10% air [N₂] dilution of a 1 Torr liquid dilution) each at 2 s duration. In separate experiments, CO₂ and odor were each presented at four discrete concentrations (CO₂: 25, 50, 75, and 100%; odor: 1, 2, 3, and 4 Torr) to explore representation of these stimuli by units at varying concentrations. Following the recordings of the concentration data, in the same mice, we explored the temporal structure of unit responses to CO₂ by presenting 50% CO₂ for a duration of 2 and 4 s separately.

Naris occlusion.

Acute unilateral nasal occlusion was performed by applying odorless TorrSeal epoxy (Loctite; Westlake) on the naris ipsilateral to the recording hemisphere.

Data analysis.

Waveform analysis and k-means cluster cutting were performed using Tucker-Davis Technologies software. Sorted data were verified to be independent units by an interspike interval analysis. For a unit to be considered a single unit, no more than 2% of all spikes could occur with an interspike interval < 2 ms. Putative single units that did not pass this criteria were omitted from further analysis.

The number of spikes (within 250 ms time bins) relative to the time of stimulus onset were extracted and organized by stimulus type. Custom macros written in Microsoft Excel were used to sort trials by stimulus type and calculate responsivity of neurons. A conservative p value (p < 0.01, two-tailed paired Student's t test) was used when comparing prestimulus (−2 to 0 s) to either during (0–2 s) or poststimulus epochs (2–4 s or 4–6 s) to identify stimulus-modulated units. These time epochs were selected after exploration of example traces to encompass both time periods in which we observed either odor-evoked (0–2 s, during stimulus) or CO₂-evoked activity (0–2, or 2–4 s after stimulus). For some analyses we calculated signal:noise ratios (s:n) as the average stimulus-evoked spike magnitude divided by the SD of the spontaneous firing (Payton et al., 2012; Varga and Wesson, 2013). This calculation differs from the standard ratio of stimulus-evoked spikes to all spikes, and was used to reduce variability in data due to trial-by-trial variability. Across region (OT vs PCX), stimuli (odor, 50% CO₂, 100% CO₂), and epoch (pre, during, post) comparisons were made using ANOVA followed by Fisher's PLSD. Statistical analyses were performed in Microsoft Excel or MATLAB (MathWorks). All data are reported as mean ± SEM unless otherwise noted.

Results

We sought to test whether single units in the PCX respond to a potent trigeminal stimulant, CO₂, and if so, to explore the nature of their stimulus-evoked dynamics. To initially address this, we recorded spontaneous and stimulus-evoked firing from cortical units in response to 50 and 100% CO₂ and a low concentration odor (isopentyl acetate). CO₂ was presented in a manner allowing its use as mostly a trigeminal (nonolfactory) stimulant (Hummel et al., 2009). Isopentyl acetate was selected due to its robust representation in both the PCX and OT (Payton et al., 2012). While this odorant is capable of eliciting trigeminal sensation at high concentrations (Doty et al., 1978, 1995; Porter et al., 2005), for our initial characterization of CO₂-evoked responses, we specifically used low concentrations of odor (∼10 ppm; see Materials and Methods), allowing this to serve as a simple olfactory-evoked comparison. Additionally, all physiological recordings were performed from urethane-anesthetized mice to control for effects of behavioral state on unit firing (Kay and Laurent, 1999; Rinberg et al., 2006; Doucette and Restrepo, 2008) and respiration (Youngentob et al., 1987; Wesson et al., 2008; Fig. 2).

Robust representation of CO₂ among PCX neurons

In a first set of experiments, we recorded spontaneous and stimulus-evoked activity from 76 PCX neurons (n = 9 mice, 1–3 recording locations/mouse). We found that a substantial proportion of PCX units displayed robust modulations in firing in response to both odor and/or CO₂ (Fig. 3). As displayed in the example traces within Figure 3A, odor evoked a stereotypical increase in action potential firing upon stimulus onset, which decreased following stimulus offset. In contrast, this example unit displayed increased firing to 50% CO₂ upon stimulus offset, with no detectable firing during the stimulus. The 100% CO₂ failed to evoke firing in this unit, reflecting a level of concentration variance at least within this unit.

Figure 3.

Characteristics of CO₂-evoked activity in the PCX. A, Representative traces of multi-unit activity (MUA) from a single PCX recording during presentation with odor (isopentyl acetate), 50% CO₂, and 100% CO₂. In this example, MUA (spiking) increases during presentation with odor (isopentyl acetate), which decays following odor offset. In contrast, presentation of 50% CO₂ elicits increased spiking only after stimulus offset, which in this example is not also displayed in response to 100% CO₂. This MUA reflects a (or several) bimodal unit that responds to both odor and CO₂. B, PCX single-unit response distribution pie chart organized by significant stimulus type (p < 0.01, two-tailed paired t tests, total # of spikes −2 to 0 s vs 0–2 s) (n = 11 mice, 1–3 recording sites/mouse). Single units were considered CO₂ modulated if they were selective in responding to only with 50% or 100% CO₂, or if they responded to either 50% or 100% CO₂ and odor (isopentyl acetate); 68.4% of units failed to display stimulus-modulated activity (p > 0.01, two-tailed paired t tests).

With the above example data demonstrating CO₂-evoked action potential firing among PCX units, we next sought to determine the population-level representation of CO₂ within this structure. Units were considered significantly modulated by comparing the number of spikes prestimulus (−2 to 0 s) to the number of spikes during or after stimulus presentation across all trials (0–2 s or 2–4 s) (p < 0.01, two-tailed, paired Student's t test). A significant p value to either time bin comparison (0–2 s or 2–4 s) was used as an indicator of a modulated unit to not bias designation as modulated based upon temporal dynamics. These time epochs were selected after careful exploration of example traces to encompass both time periods during which we regularly observed either odor (0–2 s, during stimulus) or CO₂-evoked activity (2–4 s, after stimulus, see Fig. 3A).

We found that 26.3% of PCX units were modulated by at least one of the CO₂ stimuli (20/76; Fig. 3B). The majority (17.1%, 13/76) of CO₂-modulated neurons in the PCX were CO₂ selective, failing to also show modulation in response to odor. Of these CO₂-selective neurons, 46.2% (6/13) were modulated by 50% CO₂ and 53.8% (7/13) were modulated by 100% CO₂, reflecting similar representation of these concentrations at the population level (χ² (1) = 0.154, p = 0.695; Fig. 3B). While a modest number, 9.2% (7/76) of CO₂-modulated neurons were bimodal, displaying significant modulation to both CO₂ and the low-concentration odor, reflecting convergence at the single neuron level of odor- and CO₂-evoked information within the PCX.

Temporal dynamics of CO₂-evoked responses in the PCX

The example traces shown in Figure 3A suggest that CO₂ is represented among PCX neurons at a different time course than odor. We sought to quantify this by creating peristimulus histograms of the mean number of spikes (within 250 ms time bins) from all PCX neurons in response to each three stimuli (n = 9 mice, 76 units). As suggested by the traces in Figure 3A, a peristimulus plot containing only data from neurons modulated by each stimulus category supports these unique differences in temporal dynamics between odor and CO₂, with both 50 and 100% CO₂-evoked responses reaching peak magnitude ∼1 s post stimulus offset, ∼2 s following the peak in odor-evoked activity (Fig. 4A). These dynamics were displayed differentially by each neuron in the population, with detectable changes in firing during stimulus mostly observed only in response to odor (Fig. 4B).

Figure 4.

Distinct timescale of CO₂-evoked PCX single-unit activity. A, Average spike number across all PCX units significantly modulated (p < 0.01) by odor (isopentyl acetate), 50% CO₂, and 100% CO₂. Gray shaded box = stimulus duration. n = 76 units total in screen. B, 2D histogram displaying spiking of individual single units (same data as in A). C, Mean spike # across all modulated units (same as in A and B) prestimulus (“Pre,” −2 to 0 s), during stimulus (0–2 s), or poststimulus (“Post,” 2–4 s). *p < 0.05, pre versus during (odor). #p < 0.01, pre versus post (50 and 100% CO₂). ^&p < 0.0001 (50% CO₂), ^$p < 0.05 (100% CO₂), during versus post. All statistics are ANOVA followed by Fisher's PLSD.

Quantitative comparisons of mean unit firing among only significantly modulated neurons revealed that significant changes in firing during the stimulus epoch were only observed for odor (F_(1,16) = 8.275, p = 0.011; pre vs during; Fig. 4C). Following stimulus offset, both 50 and 100% CO₂ evoked significant increases in firing compared with both spontaneous (50% F_(1,24) = 28.103, p < 0.0001; 100% F_(1,16) = 9.076, p = 0.0083) and during stimulus epochs (50% F_(1,24) = 22.92, p < 0.0001; 100% F_(1,16) = 4.925, p = 0.041; Fig. 4C). Interestingly, within their respective modulatory time points (during stimulus for odor and post for CO₂), both 50% (F_(1,20) = 0.008, p = 0.93) and 100% CO₂ (F_(1,16) = 0.166, p = 0.69) evoked similar amounts of spiking compared with odor. These results demonstrate that CO₂ is represented among PCX single units by unique temporal dynamics compared with odor.

Inverse relationship between odor and CO₂ intensity on PCX unit responses

Several questions are raised based upon these initial observations. First, as suggested by the surprisingly similar responses of modulated units to 50 and 100% CO₂ (Fig. 4), do PCX units as a population display concentration variance of CO₂ sensory input? Related, and second, are the unique temporal dynamics observed in response to CO₂ a specific characteristic of CO₂-evoked activity at this level, or a more general activity pattern that could be similarly evoked by other trigeminal stimulants?

To address these questions, we recorded PCX unit activity in a separate cohort of animals (n = 7 mice) throughout exposure to a concentration series of both CO₂ (25, 50, 75, and 100%) and odor (1, 2, 3, and 4 Torr). Histograms of s:n responses (see Materials and Methods) to odor and CO₂ from three example units are displayed in Figure 5A. These example units reflect an inverse relationship between odor and CO₂ encoding at the level of PCX units, in terms of their responses during, compared with after, stimulus. In one dramatic example (Fig. 5A, far left) CO₂ is represented by increased s:n following but not during stimulus duration, up to the point that the concentration of the CO₂ reaches 100%, at which time the large offset response characteristic of 25–75% CO₂ disappears. In contrast, this unit represents increasing concentrations of odor with increasing s:n during the stimulus, which at highest concentrations (4 Torr) is also accompanied with a respectfully large offset response. This increased offset response for 4 Torr odor resembles offset responses observed to CO₂, and is not displayed by every unit (Fig. 5A, far right), suggesting this is not an artifact of failure for the nose to clear odor at this concentration. Interestingly, some units (Fig. 5A, middle) failed to display major increases in s:n for the 4 Torr odor, instead displaying a substantially greater s:n during offset in these presentations (Fig. 5B). Results across all spontaneously active units (n = 18), displayed in Figure 5C, reflect that as a population, PCX units represent odor and CO₂ in temporally divergent manners, yet at qualitatively similar magnitudes depending upon the time epoch (during vs poststimulus). The overall profile of responses across all concentrations results in an inverted U-shape distribution during odor onset, but contrastingly, during offset for CO₂. This effect is summarized in Figure 5D, and demonstrates that (1) PCX units as a population represent CO₂ in a concentration variant fashion and (2) offset responses among PCX units appear as a generalized response property of these cells to putative trigeminal stimuli as a stimulus class, not just CO₂.

Figure 5.

Inverse relationship between odor and CO₂ intensity on PCX unit responses. A, S:n histograms of three example PCX units throughout concentration series of both odor (isopentyl acetate) and CO₂. Gray shaded bar = 0–2 s during stimulus. Unshaded epoch after stimulus epos = 2–4 s post stimulus. The first two example units displayed increased offset firing to a high concentration of odor (vertical downward arrow), which was not present at low concentrations of odor nor was it displayed by the third example unit. The third example unit reflects a case wherein the most potent odor stimulus failed to evoke a response comparable to the previous three (weaker) odor conditions. v.p., Vapor pressure. B, Example trace from middle unit shown in B in response to a single trial of 4 Torr odor. C, Mean s:n across all units (n = 18 spontaneously active units, 7 mice) to ascending concentrations of odor and CO₂. D, Same data as in C, but arranged within stimulus types, by “during” and “offset” epochs. These data reflect that PCX units represent increasing concentrations of odor in similar manners as increasing intensity of CO₂, but with highly different temporal dynamics (CO₂-evoked changes occurring poststimulation). MUA, multi-unit activity.

CO₂-evoked PCX unit activity requires inhalation

While the example traces in Figure 3 suggest that PCX unit firing in response to CO₂ displays respiratory coupling, it is also possible that CO₂-evoked activity is entering the PCX following transduction by a variety of putative CO₂-sensitive channels or receptors on the eyes, face, and/or mouth (Guimaraes and Jordt, 2007; Gerhold and Bautista, 2009). To directly test whether inhalation of CO₂ is required for CO₂-evoked PCX unit firing, we took advantage of the fact that mice are obligate nasal-air breathers and performed ipsilateral naris occlusion during the course of recording multi-unit activity in response to 50% CO₂ (n = 7 mice, same as used in Fig. 5). As shown in the example traces in Figure 6A, stimulus-offset multi-unit (and also local field potential) responses to 50% CO₂ were abolished with ipsilateral naris occlusion. Quantification of CO₂-evoked spiking across all mice (n = 15 units) revealed that naris occlusion significantly decreases spiking during stimulus-offset compared with the activity previously elicited by CO₂ while the naris was open (Fig. 6B; F_(1,24) = 18.236, p = 0.0003). These results demonstrate that in this context, CO₂-evoked activity enters the PCX following intranasal inhalation/transduction.

Figure 6.

Ipsilateral naris occlusion abolishes CO₂-evoked excitation. A, Multi-unit activity (MUA), local field potential (LFP; 0—100 Hz second-order bandpass filter), and respiration traces displaying activity of a CO₂-modulated unit during presentation with 50% CO₂ before (left) and ∼1 min after (right) ipsilateral naris occlusion (see Materials and Methods). B, Mean spike # across units prestimulus (Pre, −2 to 0 s), during stimulus (0–2 s), or poststimulus (Post, 2–4 s), both before naris occlusion (“open”) and after (“occluded”). The occlusion was not reversible, thus each animal contributed only 1 trial/condition. n = 7 mice. **p < 0.001, open versus occluded, ANOVA followed by Fisher's PLSD.

Distribution of CO₂-evoked responses in the olfactory cortex

Is the PCX uniquely involved in representing trigeminal information at this level of processing? To explore this question, we performed additional recordings of stimulus-evoked responses from neurons in the neighboring OT (n = 11 mice, 57 units, 1–3 recording locations/mouse). Similar to the PCX, the OT receives dense olfactory sensory input from mitral and tufted cells in the olfactory bulb, but also other sensory information from association structures (Wesson and Wilson, 2011). We found that only 3.5% (2/57) of OT units were modulated by CO₂ (Fig. 7; 1–50% CO₂, 1–100% CO₂). Among the two CO₂-modulated units, none were bimodal in responding significantly to both odor and CO₂. The PCX possessed a significantly greater number of CO₂-modulated neurons compared with the OT (χ² (1) = 10.676, p = 0.0011; Fig. 7). Thus, at the population level, the PCX plays a unique role in the representation and processing of intranasal trigeminal information compared with the OT.

Figure 7.

Representation of CO₂-evoked activity throughout the olfactory cortex. OT (left) single-unit response distribution pie chart organized by significant stimulus type (p < 0.01, two-tailed paired t tests, total # of spikes −2 to 0 s vs 0–2 s). Single units were considered CO₂ modulated if they were selective in responding to only 50% or 100% CO₂, or if they responded to either 50% or 100% CO₂ and odor (isopentyl acetate). Also displayed for comparison purposes is the PCX single-unit response distribution (right, same as in Fig. 3B). Whereas 28.9% of PCX units were CO₂ modulated, only 3.5% of those in the OT were. No bimodal neurons were found within the OT.

CO₂-evoked responses are structured to stimulus offset

CO₂-sensitive PCX neurons appear to structure their response excitation based upon CO₂ stimulus offset. To test if indeed CO₂-responsive neuronal responses are structured to stimulus offset, we performed additional experiments (n = 7 mice, same as used for Fig. 5) wherein we presented 50% CO₂ for either 2 s or 4 s durations. We later compared the latency to peak firing among PCX single units relative to stimulus onset between the 2 s CO₂ and 4 s CO₂ conditions (only from neurons displaying significant modulation during stimulus offset, n = 19 [2 s], n = 17 [4 s]). As shown in Figure 8, across all significantly modulated (excitatory) units in both stimulus conditions, spiking did not increase until following CO₂ offset. At the individual unit levels, the probability of a single unit reaching maximum firing (spike # within 250 ms time bin) increased dramatically post, yet not before, stimulus offset (Fig. 8B). These data demonstrate that CO₂-evoked dynamics are shaped by the duration of CO₂ stimulation in a manner that might impact the temporal interaction between odor-evoked and trigeminal-evoked information at this level.

Figure 8.

CO₂-evoked responses are structured to stimulus offset. A, Average spike number across PCX units significantly modulated (p < 0.05, pre vs offset) in response to 2 s 50% CO₂ (red; n = 19 units) or 4 s 50% CO₂ (blue; n = 17 units). Horizontal bars = duration of CO₂. Vertical dashed lines = stimulus offset. n = 7 mice. B, Cumulative distribution comparing the latency of CO₂-modulated (excitatory) unit responses to 2 and 4 s stimuli (same as in A). Latency = time from stimulus onset to 250 ms bin with peak number of spikes. The majority (>85%) of single units do not reach peak firing until stimulus offset in both 2 and 4 s conditions.

Discussion

Trigeminal sensation, including the burning and tingling sensations elicited by intranasal CO₂ and high-concentration odorants, are well known to modulate olfactory perception (Cain and Murphy, 1980; Silver and Finger, 1991; Doty, 1995; Brand, 2006). In the present studies we used two separate stimulants with known trigeminal capacity: CO₂ and high-concentration isopentyl acetate. While both stimuli are capable of stimulating the trigeminal system, it is important to clarify that neither are entirely selective (Doty, 1995; Luo et al., 2009). Nevertheless, our findings provide direct electrophysiological evidence for intranasal CO₂-evoked responses at the single-unit level within the olfactory cortex. These responses were observed among 26.3% of PCX single units and at statistically greater amounts than observed in the neighboring OT. A considerably small subset of PCX single units displayed multimodal convergence of CO₂ and a low concentration odor. The present results build upon and extend previous functional imaging studies demonstrating the presence of CO₂-evoked responses in the PCX (Hummel et al., 2009; Albrecht et al., 2010; Bensafi et al., 2012) and provide physiological evidence that the known perceptual interplay between odor and trigeminal stimuli (Doty et al., 1978, 1995; Cain and Murphy, 1980; Livermore and Laing, 1996; Brand, 2006; Boesveldt et al., 2007; Bensafi et al., 2008; Kleemann et al., 2009; Wise et al., 2012) may occur at the level of the single neuron within the olfactory cortex.

Unique spatial and temporal representation of CO₂ in the olfactory cortex

Here we found that CO₂ elicited increases in neural activity largely within the PCX. This robust level of activity evoked by CO₂ in the PCX, but not the OT, perhaps underlies reports of CO₂-evoked activity in human PCX (Albrecht et al., 2010). In contrast, similar inter-regional differences were not found in response to odor (Fig. 7; Payton et al., 2012). This striking, and significant difference in the recruitment of olfactory cortex neurons by CO₂ stimulation suggests major inter-regional differences in the processing of CO₂. This finding further suggests fundamental differences in anatomical innervation of these two structures by CO₂-sensitive neurons, or by neurons that receive synaptic input from CO₂-sensitive neurons.

CO₂-evoked responses in the PCX consisted of delayed transient dynamics–with little to no increases in excitatory firing displayed during stimulus presentation. Instead, the bulk of stimulus-evoked firing occurred following stimulus offset. The temporal dynamics of these putatively trigeminal-mediated patterns of PCX firing are the direct opposite of most odorant-evoked firing observed within the olfactory cortex. Classically, odorant-evoked responses in the olfactory cortex occur upon the first inhalation of odorant (Fig. 3) (Schoenbaum and Eichenbaum, 1995; Wilson, 2000; Rennaker et al., 2007; Chapuis and Wilson, 2011; Payton et al., 2012; Rampin et al., 2012). Some examples, however, of delayed excitatory responses to odorants among PCX neurons are present (Rennaker et al., 2007) perhaps reflecting that at their respectfully strong concentrations used, these odorants possessed a trigeminal component (Doty et al., 1978). In contrast, here we used a considerably low concentration of a single odorant, isopentyl acetate, known at higher concentrations to recruit a large portion of OT and PCX neurons (Payton et al., 2012), though at the concentration used herein recruited considerably less units (albeit while sparing likely trigeminal responsivity, as suggested by a significantly reduced representation of odor vs CO₂ in both structures (Fig. 7). Interestingly, upon increasing the concentration of this odorant, we showed that the temporal dynamics of stimulus-evoked firing in some PCX neurons can be altered from showing onset (low-concentration odorant) to offset responses (high-concentration odorant) (Fig. 5). Thus, based upon these data it is highly likely that single neurons possess the capacity to encode for stimulus modality (olfactory vs trigeminal) by differential patterns of firing (onset vs offset).

Based upon the present findings, an important goal for future research will be to identify the mechanisms for the delayed response onset for CO₂ observed among PCX neurons. The substantial differences in latency to spiking between odor and CO₂ suggests the stimulus-evoked information from these two stimuli enters the PCX by means of distinct routes or quite possibly by distinct biochemical transduction elements. Regarding the latter, the transduction of CO₂ upon nasal inhalation may occur by TRPA1 channels on trigeminal ganglia (Guimaraes and Jordt, 2007; Wang et al., 2010). TRPA1 is activated by CO₂ via the gating of intracellular protons (Wang et al., 2011). CO₂ is known to specifically activate a subset of trigeminal neurons in a manner dependent upon TRPA1 gene expression (Wang et al., 2010). In mice, CO₂ transduction may also be mediated by specialized olfactory sensory neurons within the nasal cavity that express guanylyl cyclase D (Hu et al., 2007), which is also being largely regulated by extracellular or intracellular pH that is tightly controlled by CO₂ delivery (Luo et al., 2009). Indeed, nasal pH decreases with CO₂ stimulation (Shusterman and Avila, 2003), following which, offset of the CO₂ stimulus would allow pH levels to rebound and thereby possibly increase the spiking of specific neurons through altered transduction rates. The neural mechanisms of CO₂ sensing have received considerable recent attention in both vertebrates and invertebrates (Hu et al., 2007; Luo et al., 2009; Turner and Ray, 2009; Wang et al., 2011) and future work addressing this question at the level of the PCX will be highly informative in translating this work into a better understanding of possible olfactory-trigeminal integration.

Trigeminal-olfactory multisensory convergence in the olfactory cortex

While only using a single odorant, and thereby limiting the absolute understanding of bimodal convergence in this structure, our finding that CO₂- and odor-evoked information converges at the single-unit level in the mouse olfactory cortex adds to a growing body of literature regarding the role of the olfactory cortex in multisensory integration. PCX single units in rodents display both gustatory-evoked (Maier et al., 2012) and auditory-evoked activity (Varga and Wesson, 2013). Further, PCX activity is modulated by visual sensory-input in humans (Gottfried and Dolan, 2003), in a manner that even modulates olfactory perception (Gottfried and Dolan, 2003; Jadauji et al., 2012). In addition to the PCX, the OT also displays olfactory-auditory sensory convergence (Wesson and Wilson, 2010). Thus, together, these results reflect that the olfactory cortex, particularly the PCX and OT, do not serve as only unimodal olfactory processing centers, but in fact are polysensory structures. Indeed, other primary sensory cortices, including those of audition and vision, are also increasingly becoming understood to integrate multisensory information in manners that may underlie the enhancement of perception and behavior in polysensory contexts (Foxe et al., 2002; Schroeder and Foxe, 2005; Ghazanfar and Schroeder, 2006; Kayser and Logothetis, 2007; Lakatos et al., 2007; Cohen et al., 2011; Iurilli et al., 2012). Future studies exploring the principles of trigeminal-olfactory convergence as reported herein will be informative in elucidating neural mechanisms underlying the modulation of odor perception by trigeminal stimuli. This will be largely made possible by using additional stimuli with known trigeminal properties (nicotine, ammonia, etc.) in behaving animals in combination with physiological recordings.

Implications of convergence within the cortex on perception

The considerably large representation of CO₂-evoked responses in the PCX, and particularly the convergence of trigeminal (CO₂ and high concentration odorant) sensory information onto single “bimodal” units within the PCX, suggests a likely major influence of trigeminal information on the processing of odors in the PCX. At the most basic level, these results support the impact of trigeminal sensory input on odor processing within the cortex. The PCX (Gottfried, 2010; Wilson and Sullivan, 2011) and OT (Wesson and Wilson, 2011) are both hypothesized as critical for olfactory perception. Thus, disruption of normal odor coding by single neurons within these structures by incoming trigeminal-evoked neural activity would impact odor learning and odor perception. Indeed, CO₂ alters odor perception in humans (Cain and Murphy, 1980; Kleemann et al., 2009; Daiber et al., 2013). Relatedly, based upon the present findings that the bulk of CO₂-evoked activity initiated upon stimulus offset, we predict that the critical window for trigeminal sensory input to impact the processing of odors in the PCX would occur following the immediate inspiration of the odor. In this case, trigeminal stimulation might affect the lingering, persistent sensation of odors and consequently reporting of odor perceptual qualities–throughout which the entire time somatosensory components of the trigeminal stimuli (burning, tingling) are being perceived likely through impact on somatosensory cortex (Hummel et al., 2009). This differs from the common assumption that for these stimuli to have cross-modal associations, they would likely modulate activity among single neurons simultaneously. Tests of this prediction with experiments to modulate the precise timing of odor inhalations as they relate to CO₂ delivery could provide insights into this question and further our understanding of mechanisms whereby trigeminal information may modulate odor perception.