Separate encoding of identity and similarity of complex familiar odors in piriform cortex
-
Edited by Linda M. Bartoshuk, Yale University School of Medicine, New Haven, CT, and approved August 24, 2006 (received for review May 25, 2006)
Abstract
Piriform cortical circuits are hypothesized to form perceptions from responses to specific odorant features, but the anterior piriform cortex (aPCX) and posterior piriform cortex (pPCX) differ markedly in their anatomical organization, differences that could lead to distinct roles in odor encoding. Here, we tested whether experience with a complex odorant mixture would modify encoding of the mixture and its components in aPCX and pPCX. Rats were exposed to an odorant mixture and its components in a go/no-go rewarded odor discrimination task. After reaching behavioral performance criterion, single-unit recordings were made from the aPCX and pPCX in these rats and in odor-naïve, control, urethane-anesthetized rats. After odor experience, aPCX neurons were more narrowly tuned to the test odorants, and there was a decorrelation in aPCX population responses to the mixture and its components, suggesting a more distinct encoding of the familiar mixture from its components. In contrast, pPCX neurons were more broadly tuned to the familiar odorants, and pPCX population responses to the mixture and its components became more highly correlated, suggesting a pPCX encoding of similarity between familiar stimuli. The results suggest aPCX and pPCX play different roles in the processing of familiar odors and are consistent with an experience-dependent encoding (perceptual learning) of synthetic odorant identity in aPCX and an experience-dependent encoding of odor similarity or odor quality in pPCX.
Perceptual learning induces an improvement in both perceptual acuity and in cortical encoding of familiar sensory information. Experience-dependent shifts in both perceptual acuity and central encoding of stimulus characteristics have been described in the somatosensory (1), visual, auditory, and olfactory systems. For example, in primary and higher order visual cortex, the orientation tuning of neurons becomes refined after learning or training (2, 3). Similarly, in the primary auditory cortex, the tuning to the conditioned frequency is narrowed (4, 5). In the olfactory system, exposure to, or training with, odors enhances behavioral discrimination (6, 7) and olfactory bulb encoding of those odorants (8, 9).
Perceptual learning and stimulus familiarity can modify stimulus encoding at many levels of a sensory pathway, with different ultimate consequences. For example, in the visual system, experience can refine single-unit tuning for stimulus orientation in early stages of visual cortex, whereas, in higher-order inferotemporal cortex, experience can refine single-unit encoding of more complex stimulus features and visual objects (10). Similarly, experience enhances encoding of familiar odorant features by olfactory bulb mitral cells (8), while enhancing synthesis of simple combinations of those features by neurons in anterior piriform cortex (aPCX) (11). Thus, the specific nature of the experience-induced changes within a sensory system component is influenced by the role of that component in the overall information processing of the system.
Several cortical structures are involved in olfactory perception, including the piriform cortex (PCX) and orbitofrontal cortex. The PCX receives direct afferent input from the olfactory bulb and can be subdivided into at least two major regions, the aPCX and posterior PCX (pPCX). These regions differ in their physiology (12) and in the patterning and density of afferent (13) and intracortical connectivity (14), with afferent inputs diminished and intracortical association connections predominating in pPCX. Recent human functional MRI data suggest different roles in odor perception for the aPCX and the pPCX, with aPCX encoding odorant structural identity (e.g., aldehyde or alcohol) and pPCX neurons encoding odor quality (e.g., lemon or vegetable) (15).
This distinction in the role of aPCX and pPCX in odor perception suggests that these two circuits may be differentially affected by perceptual learning. Associative learning modifies evoked field potentials in the PCX (16, 17) and in vitro paired-pulse facilitation and intracellularly recorded after-hyperpolarizations in PCX (18, 19). However, little is known about how odor information encoding is modified by odor experience in PCX. Here, we compared aPCX and pPCX encoding of a complex mixture and its components in odor-naïve and -experienced animals. We focused our attention on odorant mixtures, because most odors experienced in the natural world are complex mixtures of monomolecular odorants yet are generally perceived as unitary odor objects, and it has been demonstrated that experience is important in the synthesis of these complex mixtures into unitary percepts (for review see ref. 20). In fact, as odorant mixtures become more familiar, aPCX neurons develop enhanced ability to discriminate those mixtures from their components (11). The present results suggest differential effects of perceptual learning between aPCX and pPCX, and are consistent with a separation of odorant identity and odor quality coding in the PCX (15).
Results
Behavioral Data.
Behavioral data were obtained from 31 animals trained to discriminate between the rewarded odor (mixture) and nonrewarded odors (three components and air). Fig. 1 shows the error ratio for each component of the mixture and air (false positive responses) across the last 14 sessions. Performance improved across the sessions and reached the criteria (error ratio ≤20%, at least in two sessions) by session 16 (average, 14.97 ± 1.10, n = 31). In the first two sessions, the ratios for some odors are >1.0 because animals accessed the water port more than once within the time allowed. One-way ANOVA showed no significant difference in discrimination performance for the last 14 sessions between air and the three components [F(3, 52) = 0.11, P = 0.96].
Behavioral discrimination (error ratio) of each component and air from the mixture for the last 14 sessions. Performance improved across the sessions. There was no significant difference in improvement between air and the three components.
Electrophysiological Data.
Data from a total of 74 aPCX neurons (40 control, 34 trained) from 32 animals (17 control, 15 trained) and 75 pPCX neurons (39 control, 36 trained) from 37 animals (21 control, 16 trained) are included in the analyses. Each of these cells responded to at least one of the components and to the mixture. It should be noted that, in contrast to recent immunohistochemical activity mapping (13), no cells were observed that responded only to the mixture and not to at least one of the components. The reconstructed positions of the neurons in aPCX and pPCX are shown in Fig. 2. Layer II is shown in gray. Most cells were obtained in layer II/III between 0.96 and −0.60 mm anterior to bregma in aPCX and between 2.8 and 4.24 mm posterior to bregma in pPCX. All cells in the aPCX were in the dorsal region. In pPCX, cells were scattered throughout the dorsal and ventral extent.
Reconstructed positions of aPCX (Left) and pPCX (Right) recording sites and representative single-unit responses in each region in this study. Layer II is shown in gray. Cells were located in layer II/III between 0.96 and −0.60 mm anterior to bregma in aPCX and between 2.8 and 4.24 mm posterior to bregma in pPCX. The neuron (100105#1-6668) was obtained at 0.0 mm anterior to bregma. This neuron responded to acetic acid (AA), eugenol (Eug), limonene (Lim), and Mix. The average spike waveform of this neuron is shown at the top of the column. Another neuron (012606#2-6355) was located at 3.3 mm posterior to bregma. This cell also responded to AA, Eug, Lim, and Mix. The average spike waveform of this neuron is shown at the top of the column.
The effects of experience on three measures of cortical odor processing were examined. First, we compared the effects of experience on simple odorant-evoked firing rates of individual cells in aPCX and pPCX. Second, we examined the effect of experience on breadth of tuning of single-units to determine whether individual cells in aPCX and/or pPCX increased or decreased the range of odorants to which they responded after familiarization. Finally, we examined population responses to the odorants to determine whether at the population level encoding of a mixture and its components became more or less distinct from each other with familiarity.
Effect of Experience on Response Magnitude.
To examine whether there was an effect of experience on the intensity of odorant-evoked activity, response magnitudes of aPCX and pPCX neurons to the mixture, its individual components, and a novel odorant (amyl acetate) were compared between control and trained animals. A two-way ANOVA between the stimuli and experience (control/trained) did not show any significant effect of stimuli and experience on single-unit response magnitude in either aPCX [F(3, 288) = 0.217, P = 0.885] or pPCX [F(3, 292) = 1.044, P = 0.373]. Although there was a trend for reduced response magnitudes in trained animals, there was no significant difference in response magnitude of either aPCX or pPCX neurons to any of the odorants between control and trained animals (t tests were not significant). A similar trend toward a nonselective attenuation of odorant-evoked activity has been shown in the olfactory bulb of odor-experienced rats (21).
Breadth of Tuning.
The relative distribution of responsiveness across the stimuli used for this study was measured by using a breadth of tuning metric developed by Smith and Travers (22). This metric ranges from 0.0, wherein a single-unit responds selectively to only one stimulus within a test set, to 1.0, wherein a single-unit responds equally well to all stimuli in the test set. In aPCX, the breadth of tuning in naïve control animals was 0.87 ± 0.02 (n = 32), whereas that in odor-experienced animals was significantly reduced [t(59) = 2.79, P = 0.007] to 0.77 ± 0.03 (n = 29) (Fig. 3 Upper). This result indicates that aPCX neurons were more narrowly tuned after odor experience. In contrast, pPCX neurons showed broader tuning after odor experience (Fig. 3 Lower). The mean breadth of tuning of pPCX neurons in control animals was 0.73 ± 0.04 (n = 33), whereas that in odor-experienced animals was 0.86 ± 0.02 (n = 29) [t(60) = −3.053, P = 0.003]. Thus, odor experience influenced the metric for the breadth of tuning of aPCX and pPCX neurons in opposite directions, with aPCX neurons becoming more narrowly tuned and pPCX neurons becoming more broadly tuned to familiar odorants.
Breadth of tuning (entropy) of aPCX (Upper) and pPCX (Lower) neurons in control and trained animals. In aPCX, the breadth of tuning in odor-experienced animals was significantly reduced (more selective) compared with that in control animals. In contrast, pPCX neurons showed significantly broader tuning (less selective) after odor experience compared with controls.
Similarities Between Stimuli Based on Evoked Firing Rates.
Finally, we investigated how the representation of mixtures and their components by cortical neuron populations changed after odor experience. Correlation matrices of unit responses (evoked response magnitude) to individual odorants and the mixture were determined for all aPCX and pPCX neurons, regardless of animal of origin. Fig. 4 A shows that, in control animals, PCX population responses to individual odorants were significantly less well correlated with other individual odorants than they were with responses to mixtures containing those odorants (intercomponent correlation, r = 0.46 ± 0.07, n = 6 odorant pairwise comparisons aPCX and pPCX; component-mixture correlation, r = 0.61 ± 0.06, n = 6; sign-test comparison of correlations, P = 0.031). This result indicates that in control, odorant-naïve animals, the responses to individual components are more highly correlated with the responses to the mixture than they are to each other. To confirm the generality of this effect, we recorded the response of 22 additional cells in different naïve animals (n = 9) to another triple odorant mixture, benzaldehyde + ethyl caprylate + alpha-pinene (Aldrich, St. Louis, MO; mixture ratio of 0.15:1:1, as in ref. 23). Again, population responses to the mixture were more highly correlated with the individual components in these naïve rats (r = 0.76) than intercomponent comparisons (r = 0.59). Similar results were obtained by us in a post hoc analysis of aPCX responses to components and binary mixtures of data (24). In that study, aPCX neuron responses to isoamyl acetate, pentane, eugenol, and binary mixtures of these odors were examined in odor-naïve rats. In four of six comparisons, aPCX population responses to individual odorants were more highly correlated (Pearson r) with binary mixtures that included that odorant than they were with unrelated odorants.
Correlation matrices of unit responses (evoked response magnitude) to individual odorants and the mixture. (A) PCX population responses to individual odorants were significantly less well correlated with other individual odorants than they were to responses to mixtures containing those odorants in naïve control animals. (B Left) The enhanced correlation between aPCX population responses to mixtures and their components was eliminated in experienced rats. Thus, the mixture was encoded as distinctly as the individual components in experienced rats. (B Right) In pPCX, both intercomponent and component-mixture correlations were significantly increased by experience.
Odorant familiarity, however, significantly modified these population measures (Fig. 4 B). That is, aPCX neurons encoded the mixture more differently from its components after the familiarization. In control aPCX, population responses to the components were significantly more correlated with the mixture than with other components (sign test, control intercomponent and component vs. mixture, P = 0.031). In contrast, in trained animals, the population response to the mixture was as poorly correlated with the components as individual components were with each other (sign test, trained intercomponent and component vs. mixture were not signficant). However, in pPCX, both intercomponent and component-mixture response correlations were significantly enhanced by experience (sign test, P = 0.031). Together, these results demonstrate an experience-related decorrelation between a mixture and its components in aPCX, and in pPCX enhanced correlations between odors experienced together, and between a mixture and its components.
Discussion
Cortical processing of complex odorant mixtures was markedly different in animals familiar with the mixture compared with animals for whom the mixture was unfamiliar. In aPCX, odor experience narrowed odor tuning and decorrelated population responses between the mixture and its components. In contrast, pPCX neurons developed broader tuning, and population responses to the mixture became more strongly correlated with the components. These results are consistent with recent interpretations of odor coding in human PCX. Human functional MRI data (15) suggest that aPCX encodes odorant structural identity. In this study, as a mixture becomes familiar and more distinctive (7), aPCX neuronal encoding of the mixture becomes more distinct from its components. That is, familiar mixtures come to be identified as odor objects distinct from their components (25), an effect that may contribute to the relatively poor ability to identify components within complex mixtures (23, 26). In contrast, pPCX neurons may encode odor quality (fruity, spicy, etc.; ref. 15). For example, both a fruity odorant component and a mixture containing that component may share a “fruity” quality and thus be encoded similarly. Furthermore, odors experienced together can come to share perceptual qualities (27). The enhanced correlation in pPCX population responses to familiar mixtures and their components is consistent with this higher-order perceptual quality or similarity encoding. The results further demonstrate that odorant experience is a factor in cortical encoding of odorant identity and odor quality.
No Reflection of Odor Familiarity in Response Magnitudes of aPCX and pPCX Neurons.
There was no significant difference in response magnitude of either aPCX or pPCX neurons to the odorants, although there was a trend for reduced response magnitudes of both regions in trained animals. An experience-dependent reduction of response magnitude also has been shown in rat olfactory bulb mitral cells (21). However, associative learning has been demonstrated to modify evoked field potentials (16, 17), in vitro paired-pulse facilitation and intracellularly recorded after-hyperpolarizations (18, 19), and c-fos immunoreactivity (28).
Effect of Odor Experience on Tuning of PCX Neurons.
The breadth of tuning of aPCX neurons in odor-experienced animals was significantly reduced compared with control animals, which suggests that odor experience led to a refinement in aPCX neuron coding. This experience-dependent fine tuning of neurons after learning has been observed in other parts of the olfactory system and in other sensory cortices. For example, odor exposure results in a shift of mitral cell receptive fields toward the familiar odorant (8). In the visual cortex, the orientation tuning of neurons becomes more precise with experience (2, 3, 10). Similarly, in the auditory cortex, responses to a conditioned frequency are increased with a concomitant reduction of responses to neighboring frequencies (4, 5). By sharpening of tuning, neurons decrease the overlap in their responses to a set of stimuli, potentially enhancing discrimination and identification of those stimuli (10).
In contrast, pPCX neurons became more broadly tuned in odor-experience animals compared with control animals. Broader tuning of pPCX neurons to odorants may reflect an experience-dependent enhancement in encoding of shared odor qualities (15). Again, in comparison with the visual system, as one moves from primary to high order cortices, receptive fields become more broad, for example, shifting from a line orientation at a particular visual field location to visual objects independent of field location (10).
Thus, odor familiarization may lead aPCX neurons to sharply tune to odorant identity (e.g., odorant A or odorant ABC), whereas pPCX neurons may become more broadly tuned as they begin to encode shared odorant qualities. These different effects of odor experience on aPCX and pPCX neurons would reflect different roles of both areas in the process of odor familiarization (or learning). The analysis of population response correlations supports this interpretation.
Similarities Between Odors Influenced by Experience.
Odors are encoded within the olfactory system by the activity of ensembles of neurons (29, 30). Our analysis of population responses to odorants suggest strong effects of odor familiarity on this coding and distinct roles of the aPCX and pPCX. In aPCX, odor experience eliminated the high correlation in population responses between odor mixtures and their components. This result suggests that familiarity enhances aPCX neurons' ability to discriminate mixtures from their components and is consistent with a view that familiar mixtures are encoded synthetically in aPCX (11). The decorrelation of odor responses combined with a narrowing of individual cell tuning should enhance discriminability and identification of familiar odors. Similarly, in the antennal lobe of the honey bee, odor learning decorrelates activity patterns between odorants, enhancing their discriminability (31).
In contrast, the correlation between a mixture and its components was increased in pPCX after odor familiarization. That is, odor experience led pPCX neurons to encode the mixture more similarly to its components, as well as encode the individual components as more similar. Odorants experienced together can come to acquire perceptual qualities of each other over time (27). For example, if a cherry odor is mixed with a smoky odor, on subsequent exposure to the cherry odor alone, human subjects rate that odor has having a smoky quality (27). Thus, after experience, both the mixture (cherry + smoky) and the components now share qualities that they initially did not. The present results suggest that pPCX may contribute to this emergent sharing of perceptual quality by (i) broadening individual cell tuning and (ii) producing highly correlated population responses between the familiar mixture and each of its components. As noted above, human functional MRI data suggest a similar role for the human pPCX in coding odor quality, whereas the aPCX encodes odorant structural identity (15).
Gilbert et al. (10) hypothesize that sensory coding has two features. First, local circuits act to decorrelate activity between responding neurons, enhancing the response to features that are unique to a stimulus and, by doing so, enhance the distinctiveness of activity evoked by a given stimulus across a population of neurons. At the same time, and seemingly in contrast, local circuits identify and encode predictable correlations present in the sensory environment. The results in this study suggest that, in olfaction, these two functions may be differentially represented across the aPCX and pPCX (14).
Materials and Methods
Subjects.
Male Long-Evans hooded rats (250–425 g at recording), obtained from Harlan Lab Animals (Indianapolis, IN), were used as subjects. Thirty-two animals were trained, and 38 animals were used as controls. Both groups were allowed to take food ad lib. Trained animals were given water for 30 min after training during the training period, and control group was deprived for 23 h before recording. Animal care and use conformed to National Institutes of Health guidelines and were in accordance with the University of Oklahoma Institutional Animal Care and Use Committee.
Animal Training.
To examine the effects of odorant exposure on cortical synthetic coding, animals were exposed to a triple odor mixture and its individual components. Behavioral, computational, and physiological data suggest that familiar mixtures are processed synthetically by the olfactory system (20), and the present experiment took advantage of this experience-dependent synthesis to examine differences in aPCX and pPCX complex odor processing. To ensure that animals attended to the odor presentations, odor exposure occurred in a discriminative operant task. The chamber used for the training was equipped with two ports, one for odor delivery and the other for water delivery. The odor port was connected to a flow-dilution olfactometer. Saturated odorant vapor was added at 0.1 liters per minute (LPM) to a nitrogen stream (1 LPM) via computer-controlled solenoids to produce an approximate dilution of 1:10 of saturated vapor. Water delivery was controlled via a solenoid valve to deliver per correct response.
The odorants used for the training were acetic acid (AA) (EM Science, Darmstadt, Germany), eugenol (Eug) (Aldrich), limonene (Lim) (Sigma, St. Louis, MO), and a mixture of these three components (3 μl of AA, 40 μl of Eug, and 40 μl of Lim), and air (nitrogen) was used as control. The concentrations of components in the mixture were chosen based on work by Laing and Francis (23) showing that humans were unable to identify individual components within such a mixture. Each odorant was applied to a piece of paper set in each bottle connected with the olfactometer. All odorants were prepared fresh each session. The training was a go/no-go water rewarded odor discrimination task with the mixture serving as the go (rewarded [S+]) odor. Rats initiated a trial by poking their nose in the odor sampling port to initiate odor onset. If the S+ odor was presented, the rat could approach and enter the water reward port to initiate water delivery. If no odor or a nonrewarded (S−) odor was delivered, the rat had to leave the odor sample port for at least 1 s before initiating a new trial. If rats entered the water port within 2–3 s after leaving the odor port, the response was scored as a go response; otherwise, it was scored as a no-go response. A go response on an S+ trial was scored as a correct response. A go response on an S− trial was scored as an error, and no water was delivered. Errors were also scored if the rat failed to enter the water port on S+ trials. The chance of S+ trail was 50%.
The training was conducted for 30 min in each session until criterion performance (error ratio <20%) was attained twice.
Recording and Odorant Stimulation.
Details of single-unit recording and odorant-response characterization techniques for layer II/III aPCX and pPCX neurons have been reported in detail (32). Briefly, animals were anesthetized with urethane (1.5 g/kg) and were freely breathing with the respiratory cycle monitored through a piezoelectric device. The single-unit nature of the recordings was verified by at least a 2-ms refractory period in interval histograms. Layer II/III aPCX and pPCX neurons were identified by lateral olfactory tract-evoked responses and/or histological confirmation. After isolation of a single-unit 2-s test, stimulus presentations were delivered for each odorant to test for responsiveness.
Odorants were delivered with a flow-dilution olfactometer, with a constant, 1-LPM flow of charcoal-filtered, humidified air presented 1–2 cm from the animal's nose. Saturated odorant vapor was added at 0.1 LPM to the clean air stream by means of computer-controlled solenoids to produce an approximate dilution of 1:10 of saturated vapor. Odorant stimulus onset was triggered off the respiratory cycle to coincide with the transition from inhalation to exhalation, and test stimulus duration was 2 s with at least a 60-s interstimulus interval. Stimuli used during recordings were the same odorants with the same concentration as those used for the training and isoamyl acetate (Aldrich) as a novel stimulus (control stimulus). Odorized airflows entered a 3-cm-long chamber before delivery to the animal.
Data Analysis.
Behavioral performance was analyzed with two-way ANOVA (training day × stimulus) over the last 14 sessions.
The response magnitudes of aPCX and pPCX neurons to odorants were quantified as the difference in number of spikes evoked during the 2-s stimulus compared with a 2-s prestimulus period. Odor-responsive neurons were defined by firing rates with a 30% odor-evoked increase above spontaneous activity (33).
To examine how reward-association learning was reflected in aPCX and pPCX neuron odor response magnitude, two-way ANOVA was conducted across the odorants used in this study and experience (control/trained). Further, Student's t test was performed between control and trained animals for comparisons of response magnitude to the stimuli.
Correlation coefficients (Pearson r) were calculated for responses across stimulus pairs for aPCX and pPCX neurons to examine the representation of similarity of stimuli by populations of neurons. Sign tests were conducted on the population response correlations of aPCX and pPCX neurons in control animals between components and between each component and the mixture to examine whether aPCX and pPCX neurons encode the mixture differently from its components and whether experience modified this encoding.
The breadth of tuning metric of Smith and Travers (22) was calculated as follows. The proportion of a neuron's total response that is devoted to each of five stimuli used in this
study was used to calculate its coefficient entropy (H). The entropy is calculated as
where H is the breadth of tuning, K is the scaling constant (K = 1.43 for five stimuli), and Pi is the proportional response to each of the stimuli. When the neurons respond equally to all stimuli, H = 1.0. The coefficient ranges from 0.0, representing total specificity to one of the stimuli, to 1.0, indicating an equal
response to all of the stimuli. Entropy was calculated separately for aPCX and pPCX neurons. To examine the differences in
breadth of tuning for aPCX and pPCX neurons between control and trained animals, Student's t test was performed. As the breadth-of-tuning metric cannot involve negative numbers because of its logarithmic form, neurons
with suppressive responses to the stimuli were excluded from this specific analysis.
Histology.
After recording, animals were overdosed with anesthetic and transcardially perfused with saline and 4% paraformaldehyde, and the brains were subsequently sectioned coronally at 40 μl and stained with cresyl violet for determination of electrode positions.
Acknowledgments
This work was funded by National Institute on Deafness and Other Communication Disorders Grant R01 DC3906 (to D.A.W.).
Footnotes
- *To whom correspondence should be addressed. E-mail: miki.kadohisa{at}ou.edu
-
Author contributions: D.A.W. designed research; M.K. performed research; M.K. and D.A.W. analyzed data; and M.K. and D.A.W. wrote the paper.
-
The authors declare no conflict of interest.
-
This paper was submitted directly (Track II) to the PNAS office.
-
See Commentary on page 14985.
- Abbreviations:
- PCX,
- piriform complex;
- aPCX,
- anterior PCX;
- pPCX,
- posterior PCX;
- LPM,
- liters per minute.
- © 2006 by The National Academy of Sciences of the USA
