Generalized vs. stimulus-specific learned fear differentially modifies stimulus encoding in primary sensory cortex of awake rats

Abstract

Experience shapes both central olfactory system function and odor perception. In piriform cortex, odor experience appears critical for synthetic processing of odor mixtures, which contributes to perceptual learning and perceptual acuity, as well as contributing to memory for events and/or rewards associated with odors. Here, we examined the effect of odor fear conditioning on piriform cortical single-unit responses to the learned aversive odor, as well as its effects on similar (overlapping mixtures) in freely moving rats. We found that odor-evoked fear responses were training paradigm dependent. Simple association of a condition stimulus positive (CS+) odor with foot shock (unconditioned stimulus) led to generalized fear (cue-evoked freezing) to similar odors. However, after differential conditioning, which included trials where a CS− odor (a mixture overlapping with the CS+) was not paired with shock, freezing responses were CS+ odor specific and less generalized. Pseudoconditioning led to no odor-evoked freezing. These differential levels of stimulus control over freezing were associated with different training-induced changes in single-unit odor responses in anterior piriform cortex (aPCX). Both simple and differential conditioning induced a significant decrease in aPCX single-unit spontaneous activity compared with pretraining levels while pseudoconditioning did not. Simple conditioning enhanced mean receptive field size (breadth of tuning) of the aPCX units, while differential conditioning reduced mean receptive field size. These results suggest that generalized fear is associated with an impairment of olfactory cortical discrimination. Furthermore, changes in sensory processing are dependent on the nature of training and can predict the stimulus-controlled behavioral outcome of the training.

the olfactory system involves a memory-based odor processing function that allows its acuity to be constantly shaped by experience (Wilson and Stevenson 2006). This function enables animals to not only identify myriad novel odors and odor combinations in the environment but also to associate odors with their ecological significance, which is critical for adaptive behavior. Experience-dependent perceptual changes (perceptual learning) and their underlying neural plasticity have been reported across phylogeny from the first to third order neurons in the olfactory system (Davis 2004).

In mammals, while robust experience-dependent plasticity is expressed in the olfactory bulb (Brennan et al. 1990; Doucette et al. 2011; Fletcher and Wilson 2003; Freeman and Schneider 1982; Mandairon et al. 2006; Ravel et al. 2003; Sullivan et al. 1989), the piriform cortex appears to play a special role in experience-dependent odorant perceptual synthesis and odor object formation (Barnes et al. 2008; Haberly 2001; Kadohisa and Wilson 2006; Li et al. 2008).

Work on olfactory perceptual learning has focused on enhanced acuity for learned or familiar odors (Cleland et al. 2002; Li et al. 2008; Moreno et al. 2009; Wilson 2003). Thus molecularly similar odorant molecules that initially cannot be perceptually distinguished by naïve animals become discriminable with appropriate experience (Cleland et al. 2002; Fletcher and Wilson 2002; Li et al. 2008). This improved acuity is associated with changes in the olfactory bulb (Doucette et al. 2011; Fletcher and Wilson 2003; Moreno et al. 2009) and piriform cortex (Chapuis and Wilson in press; Kadohisa and Wilson 2006; Li et al. 2008). Associative learning, for example, linking an odor with a specific biologically relevant meaning such as reward, can also modify olfactory bulb (Brennan et al. 1990; Doucette et al. 2011; Freeman and Schneider 1982; Moreno et al. 2009; Ravel et al. 2003; Sullivan et al. 1989) and piriform cortical odor coding and physiology (Calu et al. 2007; Kadohisa and Wilson 2006; Roesch et al. 2007; Saar and Barkai 2003; Saar et al. 2002).

However, in some situations acuity appears to be reduced by experience. For example, standard fear conditioning involving pairing of a condition stimulus positive (CS+) with a foot shock can induce a generalized fear response to many stimuli in some way similar to the CS+. This has been demonstrated in a variety of sensory modalities and paradigms (Armony et al. 1997; Honig and Urcuioli 1981; Pavlov 1927; Shepard 1987). Work in the auditory system suggests that this fear generalization is in part mediated by changes within the sensory system itself, with an apparent loss of sensory acuity mediating the generalized response to stimuli the same as, or similar to, the CS+ (Armony et al. 1997; Ito et al. 2009). Differential conditioning, in contrast, where both a CS+ (e.g., predicting foot shock) and a CS− (predicting no foot shock) are used during conditioning, induces much less generalization and instead induces learned responses that are highly specific to the stimulus (Ito et al. 2009; Pavlov 1927; Rescorla 1976).

The present study had two primary goals. First, most work on changes in adult olfactory cortical single-unit responses to learned odors has relied on appetitive tasks (Calu et al. 2007; Kadohisa and Wilson 2006; Roesch et al. 2007; Saar et al. 2002). Here, using chronic recording of anterior piriform cortical (aPCX) single units, we examined whether aversive conditioning could similarly modify olfactory cortical odor responses. Second, by using both standard (CS+ only) and differential (CS+ and CS−) fear conditioning, we tested the hypothesis that piriform cortical odor coding could be shifted in two opposing directions, depending on the nature of the conditioning and in concert with the learned behavioral response to odor (generalized or selective odor-evoked fear). The results show that odor fear conditioning induced a decrease in single-unit spontaneous activity. In addition, each of the odor fear conditioning paradigms induced different changes in odor-evoked single-unit activity and receptive field size and the differences correspond to different learned behaviors in the rats.

MATERIALS AND METHODS

Subjects.

Twenty-seven male Long-Evans hooded rats (250–450 g) were used as subjects (Charles River Laboratories, Wilmington, MA). Animals were housed individually in polypropylene cages on a 12-h light-dark cycle, with food and water available ad libitum. Animal care protocols and all experiments were approved by the University of Oklahoma Institutional Animal Care and Use Committee and the Nathan S. Kline Institute Institutional Animal Care and Use Committee and in accordance with National Institutes of Health guidelines.

Odor stimulation.

Odors used were the monomolecular odorant limonene and a mixture (10C) that had 10 monomolecular odorants each at a concentration of 100 parts per million based on vapor pressure and dilution in mineral oil, as described in detail in Barnes et al. (2008) (see Fig. 1A for a list of the components and mixtures). These odorant mixtures were chosen due to previous work showing variability in ease of discrimination and unique characteristics of piriform cortex encoding (Barnes et al. 2008). Odorant components were either fruity or neutral odors based on descriptors provided at http://gara.bio.uci.edu/. Other odors included 10C-1 (10C with isoamyl acetate removed), 10C-2 (10C with isoamyl acetate and nonane removed), and 10CR1 (10C with isoamyl acetate replaced by 2-methyl-2-buten-1-ol), identical to those used previously (Barnes et al. 2008). Odors were presented (5 s) to the behavioral chamber via a flow-dilution olfactometer (1 LPM) that was controlled by a programmed script in Spike 2 (Cambridge Electronic Design, Cambridge, UK). Odors were presented with an interval ≥60 s and at random distances from the animals' noses given their free mobility. Thus precise timing of odor arrival at the rat snout was not possible. However, this allowed delivery of stimuli completely free of expectation on the subject's part, which is required for fear conditioning.

Fig. 1.

Olfactory fear conditioning with different paradigms. A: odor mixtures used in the experiments. 10C, an odor mixture that consists of 10 different odorants; 10C-1, 10C with 1 odorant missing; 10C-2, 10C with 2 odorants missing; 10CR1, 10C with 1 odorant replaced by a new odorant; NC, no change. B: odor-fear training paradigms. Standard training had 10 trials of 10C paired with an electric foot shock. Pseudotraining had 10 unpaired presentations of 10C and foot shock. Differential training had 10 trials of 10C [conditioned stimulus positive (CS+)] paired with a foot shock and 10C-2 (CS−) without a following foot shock. A retention test was carried out 24 h after the training session. All rats were randomly presented with 10C, 10C-1, 10C-2, 10CR1, and limonene 3 times. Odor-evoked freezing behavior of the rats was scored and recorded. C: average freezing of the nonimplanted rats in response to odor stimuli 24 h after training. Standard, differential, and pseudo represent different training paradigms. Bars that are marked with “a” are significantly different from bars that are marked with “b” and “c” [P < 0.05, Fisher's paired least significant difference (PLSD)]. D: results from a second behavioral test following 3 days of reminder training (3 group appropriate odor-shock pairings/day) as was also performed in the animals used for electrophysiological recording. No change in the overall pattern of results in the 3 groups was observed. Error bars represent SE.

Electrodes.

Extracellular recordings were obtained by using a drivable bundle of ten, 25-μm-diameter (38-μm-diameter with insulation) Formvar-insulated Nichrome wires (A-M Systems, Carlsborg, WA). A guide tube holding the wires was a 27-gauge thin wall cannula (Small Parts, Miami Lakes, FL). The electrode design was identical to those used previously (Roesch et al. 2007).

Surgical procedures.

Naïve animals were anesthetized and kept unconscious with an isoflurane anesthesia system (E-Z Systems, Palmer, PA) throughout the surgical process. The microwire bundle was chronically implanted in the left hemisphere and cemented on the rat's skull, with the tip slightly above or within aPCX (1.0 mm anterior to the bregma, 4∼4.5 mm laterally, and 5∼6 mm ventral to the surface of the brain). Immediately after the surgery, before recovery from anesthesia, an antibiotic and analgesic were subcutaneously injected in the rats (5 mg/kg enrofloxacin, 0.01 mg/kg buprenorphine for analgesia, and 12 h later). Animals were given 2 wk for recovery from surgery before initial training sessions.

Fear-conditioning paradigm.

Behavioral experiments were conducted in one of two custom chambers with a shock grid floor (Lafayette Instrument, Lafayette, IN; or Coulbourn Instruments, Whitehall, PA). Rats were randomly divided into three groups: standard, differential, and pseudotraining groups. The standard training group received 10 trials of a 5-s 10C (CS+) odor followed by a 1-s, 0.4–0.5 mA electric foot shock concurrent with the last second of odor. The intertrial interval for standard training was 120 s. The differential training group had a 5-s 10C (CS+) odor paired with a foot shock and a 5-s CS− odor (10C-2 or 10CR1 in different rats) with no foot shock. In differential conditioning, the rats received 10 CS+ and shock pairings as in standard conditioning but in addition received 30 CS− trials randomly interspersed with the CS+ trials. Intertrial interval in all cases was 120 s. The pseudotraining group received unpaired 5-s 10C and foot shock, and each was randomly presented for 10 times. For this group, 10C presentations were kept ≥120 s apart to prevent olfactory habituation. Behavioral and neural tests were performed ≥24 h after training.

Behavioral cue-evoked freezing tests.

For behavioral analyses, 24 h after training, a retention test was carried out in a transparent acrylic testing chamber (10“ W × 9.5” Lx 6“ H) different from the training chambers. During testing, all rats were randomly presented with 10C, 10C-1, 10C-2, 10CR1, and limonene three times each and odor-evoked freezing (motionless except for breathing, slightly arched posture) was scored in response to each odor stimulus. All behaviors were also monitored by video. For behavior-only rats, the video camera was shooting from the side of the chamber, and this angle provided good visualization of freezing behavior of the rats. For chronically recorded rats, the camera was positioned above the chamber to record the freezing behavior from the top. This angle caused problems of accurately visualizing freezing behavior in these rats. Thus behavioral data and neural data are from different animals, although training and testing protocols were identical.

Chronic single-unit recording.

For animals used for single-unit recordings, behavioral training was conducted following several days of baseline recordings. The same three conditioning groups were used for behavioral testing: standard training, differential training, and pseudotraining. Posttraining recordings sessions occurred over several days to increase recording yield; thus before each daily posttraining recording session, a reminder session (3 trials, paired, or unpaired odor shock) was given to the corresponding groups.

Data acquisition.

Neural data were collected by a Multichannel Acquisition Processor system (Plexon, Dallas, TX). Two weeks following the surgery, animals were placed in a custom testing chamber (10” W × 10“ L × 22” H) to record neural activity in piriform cortex. A DC fan was mounted on the wall and constantly drew air out of the chamber (see Fig. 3A). In a recording session, neural data were continuously recorded while the animal was repeatedly presented with test odors. Test odors were three mixtures, 10C, 10C-1, and 10CR1, and one single odorant, limonene. At least three of the four tested odors were given to the animals for at least three times. Before each session, recordings were screened for quality of unit activity. If no isolatable spikes were found across all electrodes, recording was terminated for the day. As described in Schoenbaum et al. (2003), the implanted wire bundles were lowered daily at the end of each session to a new recording site; thus individual recordings from individual cells were not maintained throughout the entire training and testing procedure. The daily advancing distance was ∼80 μm before reaching the depth of 6 mm, and it was then reduced to ∼40 μm thereafter. After the advance, rats had at least a 24-h rest before the next recording session (Schoenbaum et al. 2003). Recordings in aPCX (based on histological analyses) before odor training were used as the preconditioning baseline. Data collected during the odor-shock training session were not analyzed due to excessive electrical noise generated by the shock system and movement artifacts from the shocked rats. Posttraining recordings were conducted in the same way as baseline recordings, with electrode bundles advanced each day. The same recording procedures were used in all three conditioning groups.

Fig. 3.

Anterior piriform cortex (aPCX) single-unit recordings from awake, freely moving rats. A: a rat implanted with a movable wire bundle was in the testing chamber for chronic unit recording and odor training. Gray bars in the chamber represent the metal grid floor that was part of the electric shocking system. B: procedures of chronic unit recordings. Note that each rat only was only trained once on a single day, while recording sessions were conducted for multiple days before and after the training day. C: a digital-filtered (bandpass, 300–3,000 Hz) trace showing spikes from a single microwire placed in aPCX. One larger and one smaller spike can be seen. Trace has a time base of 5 ms per division. D: extracted waveforms of units 1 and 2 from the trace in C. Signal-to-noise ratio of unit 1 (left) was 2.5:1; that of unit 2 (right) was 6:1. E: examples of peristimulus time histograms with raster plots of 2 single units (left and right) in response to 10CR1. Histograms showed cumulative spike count of three trials, with a 100-ms bin width. Both units showed a significant excitatory response to 10CR1. Horizontal black bars indicate odor delivery (5 s).

Data analysis.

Neural data recorded by the Plexon system were transformed into a format compatible with Spike2 and were analyzed offline using Spike2's template sorting features. Before template sorting, most channels were bandpass-filtered at 300 to 3,000 Hz using digital filters in Spike2. A threshold of spike amplitude was set to collect waveforms larger than 3:1 signal-to-noise ratio (Katz et al. 2002) (see Fig. 3D). Isolations of single units were initially done using the template-matching function in Spike2. Once the template matching was accomplished for all qualified waveforms in neural data of a recording session, nonaction potential waveforms, such as electrical noise and movement artifacts were manually removed from the templates using visual examination and the waveform-cutting function in Spike2. The identification of nonaction potential waveforms was based on shapes, amplitudes, and whether they were simultaneously observed in multiple channels. Waveforms in different templates were finally examined using principle component analysis, in which clusters of action potentials were shown in principle component space. The cluster-cutting algorithm in Spike2 was used at this stage to further identified single units. Each single unit was required to have an interspike interval of ≥1–2 ms. Typically, one to three units could be isolated on an active channel. Neural data collected 6 mm below the brain surface or deeper were considered to be from neurons in layer II/III of the piriform cortex and were confirmed by histological data (Fig. 2).

Fig. 2.

Representation of electrode tracks of 15 rats implanted with a movable microwire bundle. Black bars represent electrode tracks reconstructed from the brain sections. Recording sites were along the tracks. Black open rectangle represents possible recording sites across layer III of piriform cortex where track reconstruction was not available. Data suggested recordings were localized to layer II/III of anterior piriform cortex. Outlines are reproduced from Paxinos and Watson (2009) Copyright Elsevier, and represent sections ranging from 2.70 to 0.48 mm anterior to Bregma.

Odor-evoked activity was determined by spike counts from cumulative peristimulus time histograms based on three stimulus repetitions (Fig. 3E). The means and SD of spontaneous activity was calculated over ≥20 s for each cell with 5-s bin widths for direct comparison to the 5-s odor stimulus. Odor responses were defined as a change in activity during the 5-s odor of at least ±1 SD from spontaneous activity. This definition was chosen to provide a reliable estimate of odor-evoked activity that could be used to detect relative differences between pre- and posttraining conditions and across training groups. Odor-evoked activity in the olfactory system is often very sparse in awake animals (Rinberg et al. 2006); thus we chose 1 SD to be more inclusive of weaker responses. The large bin widths were chosen because in this paradigm odor stimuli were presented randomly to an animal that could be located anywhere within the conditioning chamber at stimulus onset. Thus precise timing of stimulus-response characteristics, as for example in odor-nose-poke paradigms (Calu et al. 2007; Roesch et al. 2007), was not possible. However, we were able to find reliable odor-evoked responses even in this more naturalistic paradigm.

Spike trains of units recorded from animals before and after training were analyzed for spontaneous activity rate, odor-evoked response probability, excitatory and suppressive response probability, and receptive field size (tuning breadth). Receptive field size of a single unit was calculated as the percentage of test odors to which a single unit showed a significant excitatory or suppressive response.

Histology.

Following the final day of recording, implanted rats were given an overdose of urethane and then perfused transcardially with 0.9% saline followed by 10% formaldehyde. Brains removed from the skulls were stored in a 30% sucrose/10% formaldehyde solution for later sectioning. The brains were sectioned coronally at 40 μm, mounted on the slides, and stained with cresyl violet or nuclear fast red. Electrode tracks and recording locations were verified under a light microscope, and images were acquired using a digital camera (Fig. 2).

RESULTS

Three odor-fear conditioning paradigms were used to train 12 rats (3 groups, 4 in each group) for behavioral analysis. Animals used for chronic recording received the same kind of training, but due to difficulties in accurately visualizing freezing behavior while recording in this system, behavioral and neural data are from different rats. Neural activity was recorded in aPCX from 15 freely moving rats (standard conditioning, n = 6; differential conditioning, n = 4; and pseudoconditioning, n = 5). A total of 528 aPCX single units were isolated and analyzed.

Olfactory fear generalization is dependent on nature of the training.

Rats were trained in standard, differential, or pseudo odor-shock conditioning and odor-evoked freezing was measured in a different context 24 h after training (Fig. 1B). A mixed ANOVA was performed, with group (standard, differential, and pseudo) and odor (10C, 10C-1, 10C-2, 10CR1, and limonene). The three conditioning groups showed distinct odor-evoked fear responses (Fig. 1C) with significant main effects of odor [F(4, 36) = 11.73, P < 0.01], training groups [F(2,36) = 5.90, P < 0.05], and a significant odor × group interaction [F(8,36) = 2.46, P < 0.05]. Pseudoconditioned rats did not show significant freezing in response to any test odor (Fig. 1C, pseudo). The standard conditioning group, which had 10C as the CS+, showed significant freezing responses to 10C and the overlapping mixtures 10C-1, 10C-2, and 10CR1, as well as to limonene [P < 0.05, Fisher's paired least significant difference (PLSD); Fig. 1C, standard]. In contrast, the differential conditioning group, which was trained to associate 10C (CS+) with foot shock and 10CR1 (CS−) with safety (no shocks), showed significant freezing to 10C, at the same level as in the standard conditioning group. However, freezing evoked by 10C-1, 10CR1, or limonene was significantly <10C (P < 0.05, Fisher's PLSD). The results suggest that differential conditioning induced highly odor-specific fear, while standard conditioning induced highly generalized odor fear.

Because the animals used for single-unit recordings described below received reminder conditioning trials over the course of several days after conditioning (see materials and methods), the animals used for behavior testing were also subjected to reminder trials for 3 days after the 24-h freezing test to determine the stability and selectivity of the fear memory. Thus day 1 was initial training, day 2 was the 24-h memory test, and on days 3–5 the animals received reminder trials (3 group-appropriate CS+ shock and CS− trials). At least 2 h after the last reminder trial, they were given a final memory test in a new context. As shown in Fig. 1D, the pattern of generalized or odor-specific freezing was maintained. Similar to the 24-h test, after several days of reminder trials the three training groups showed distinct patterns of odor-evoked fear responses, with significant main effects of odor [F(4,36) = 17.36, P < 0.01], training group [F(2,36) = 61.34, P < 0.01], and a significant odor × group interaction [F(8,36) = 2.98, P < 0.01]. In separate animals we examined whether these disparate behavioral outcomes were associated with changes in cortical odor processing.

Neural responses.

A total of 528 single units were isolated from 15 animals. In the standard conditioning group, we recorded from 95 units before training and 121 units after training. In the pseudoconditioning group, we recorded from 54 units before and 55 units after training. In the differential conditioning group, we recorded from 122 units before and 87 units after training. Both spontaneous and odor-evoked activity were compared within each group before and after training. Mean number of single units isolated from individual animals >30.

Nonspecific decreases in aPCX spontaneous activity associated with olfactory aversive learning.

We first examined spontaneous aPCX single-unit activity, and found a main effect of conditioning groups [F(5, 528) = 5.39, P < 0.0001, one-way ANOVA]. There was no significant difference in spontaneous activity across the three conditioning groups pretraining. However, both standard and differential conditioning induced a significant decrease (Fig. 4) in average spontaneous firing rate in aPCX neurons (standard group, P < 0.0001; differential group, P < 0.005, Fisher's PLSD). No significant change was observed in pseudoconditioned rats. Thus, regardless of the specific paradigm, odor fear conditioning induced a decrease in aPCX spontaneous activity that was not observed after pseudoconditioning.

Fig. 4.

Effects of odor-fear conditioning on spontaneous activity of aPCX neurons. Standard and differential training induced a significant decrease in spontaneous firing rate of neurons in the aPCX. This decrease was not significant in the pseudotrained group. There was no significant difference in spontaneous activity of the animals before training. Error bars represent SE. *P < 0.01.

Conditioning paradigm-dependent changes in aPCX odor-evoked activity.

For odor-evoked activity, we first assessed the proportion of single units that showed a significant odor-evoked response (±1 SD change from spontaneous activity) averaged across all test odors before and after fear conditioning (Fig. 5A). There was a significant group × time interaction in total proportion of units responsive to odors [F(2,18) = 10.50, P < 0.01]. While there was no significant difference across conditioning groups in the proportion of units responsive to odor pretraining, there was a significant increase in odor responsive units poststandard conditioning and a significant decrease in odor responsive units postdifferential conditioning (P < 0.05, Fisher PLSD). No change was observed after pseudoconditioning.

Fig. 5.

Proportion of single units responsive to odors before and after training in the three groups. A: proportion of single units showing response to odor (excitation + suppression) was enhanced after standard conditioning and decreased after differential conditioning, while pseudoconditioning induced no change. B: standard and differential training induced proportionally less excitatory single-unit odor responses to the test odors, while pseudoconditioning induced no change. C: standard training induced proportionally more suppressive odor responses, with no change after either pseudo or differential conditioning. The change in proportion of units responding to odors [total (D), excitatory (E), and suppressive (F)] after standard conditioning was not odor specific. Error bars represent SE. *P < 0.05. For calculation of single-unit odor responses, see materials and methods.

Total odor-evoked responses were then divided into excitatory (>1 SD increase over spontaneous activity) and suppressive (>1 SD decrease over spontaneous activity) responses. The mean proportion of single units showing excitatory responses across odors was significantly modified by standard and differential conditioning. There was a significant main effect of time in the proportion of units showing excitatory odor responses [F(1,18) = 12.77, P < 0.01], with a significant decrease in excited units in both standard and differential conditioning groups (P < 0.05, Fisher PLSD). There was no change in the proportion of excitatory responses after pseudoconditioning (Fig. 5B).

In contrast to excitatory responses, the proportion of units showing suppressive responses was significantly enhanced after standard conditioning [Fig. 5C; a significant main of time, F(1, 18) = 10.80, P < 0.01; and group × time interaction, F(2,18) = 5.01, P < 0.05, mixed ANOVA]. Post hoc Fisher PLSD tests revealed a significant increase in the proportion of units suppressed by odor after standard conditioning (P < 0.05), with no change after differential or pseudoconditioning. Thus the opposing effects of standard and differential conditioning on the proportion of total responsive units (Fig. 1A), largely reflects the large increase in suppressive responses in the standard group, which does not occur after differential conditioning.

The training-induced changes in the proportion of units showing odor-evoked responses did not appear due to odor-specific changes in either the standard or differential conditioning paradigms. For example, as shown in Fig. 5D, there were increases in the proportion of units responding to all test odors. The increase in units showing suppressive responses after standard conditioning was also not stimulus specific. Similar nonspecific effects were observed after differential conditioning (not shown).

Training paradigm-dependent changes in aPCX single-unit receptive field breadth.

In addition to examining the proportion of odor-responsive units, we also examined the odor selectivity of individual units. Receptive field size of each aPCX single unit before (pre) and after (post) training was calculated as the proportion of the test odors that evoked a significant response (±1 SD change from spontaneous activity).

Odor fear conditioning significantly modified single-unit receptive field breadth in a conditioning paradigm specific manner. There was a significant group × time interaction in total receptive field breadth [combined excitatory and suppressive responses, F(2,528) = 5.37, P < 0.01]. In post hoc comparisons, there was no difference across groups in total receptive field size pretraining (P < 0.05, Fisher PLSD; Fig. 6A). Furthermore, receptive field size was not modified postpseudoconditioning. However, standard conditioning (which induces generalized odor fear) resulted in a significant increase in mean single-unit total receptive field size in aPCX compared with pretraining levels (P < 0.05, Fisher PLSD). This increase was predominantly due to an increase in receptive fields for suppressive responses [Fig. 6C; main effect of time, F(1, 528) = 13.85, P < 0.01; and group × time interaction, F(2, 528) = 6.67, P < 0.01], even though standard conditioning did induce a significant decrease in excitatory receptive field size [Fig. 6B; main effect of time, F(1, 528) = 7.04, P = 0.01]. In contrast, differential conditioning, which led to odor-specific fear, induced a decrease in both mean total receptive field size and mean excitatory receptive field size (P < 0.05, Fisher PLSD). Thus standard conditioning, which induced generalized, nonodor-specific fear, also resulted in a decrease in odor selectivity of aPCX single units, while differential conditioning, which induced relatively odor-specific fear, resulted in an increase in odor selectivity of aPCX single units. Pseudoconditioning induced no detectable change in receptive fields.

Fig. 6.

Single-unit receptive field (RF) width before and after training in the three groups. A: standard training induced a significant increase in mean RF size of anterior piriform cortex neurons. In contrast, differential training induced a significant decrease in mean RF size. No change was induced by the pseudotraining. B: standard and differential training induced a significant decrease in excitatory RF size. No change was observed after pseudotraining. C: standard training induced a significant increase in suppressive RF size, while no change in suppressive RFs was observed after pseudo or differential conditioning. Error bars represent SE. *P < 0.05. For receptive field calculation, see materials and methods.

DISCUSSION

The present results demonstrate that increases and decreases in apparent odor perceptual acuity are associated with specific, opposing changes in anterior piriform single-unit receptive fields recorded in freely moving rats. Thus standard odor-shock conditioning, which results in generalized freezing responses to odorant mixtures overlapping with the CS+, was associated with an increase in piriform cortical single-unit odor receptive field width. In contrast, differential odor-shock conditioning, using both a CS+ and CS−, induced freezing responses relatively specific to the CS+ odor and a decrease in single-unit receptive field width. Pseudoconditioning induced no freezing and no receptive field changes. In addition to these paradigm-specific changes, both standard and differential conditioning but not pseudoconditioning reduced spontaneous firing rates, as recorded in the conditioning chamber. This reduction could enhance signal-to-noise ratios of odor-evoked activity, although given the observed decrease in odor-evoked excitatory responses, the effect on signal-to-noise ratios is unclear. Although different cells were recorded before and after conditioning, electrodes were advanced daily in all groups, thus conditioning-dependent changes do not simply reflect changes in cell populations sampled but rather true experience-dependent plasticity. Together, the results suggest that cortical and behavioral olfactory acuity can be either increased or decreased, based on experience, and that generalized fear may reflect not only changes in emotion circuits such as in the amygdala but also changes within early sensory pathways.

The difference in neural and behavioral outcomes between the standard and differential conditioning paradigms has been suggested to reflect the fact that comparative judgments between stimulus pairs are not required in the single-CS (standard training) paradigms, whereas they are an essential ingredient for the discriminative (differential training) paradigms (Ganz 1962; Guttman and Kalish 1956; Honig and Urcuioli 1981). The comparative nature of differential conditioning presumably involves higher centers that utilize olfactory information for associative memory and decision making, such as the amygdala (Roesch et al. 2010; Rosenkranz and Grace 2002; Sullivan et al. 2000), orbitofrontal cortex (Schoenbaum et al. 2003), entorhinal cortex (Kay and Freeman 1998), and hippocampal formation (Knafo et al. 2005; Wiebe and Staubli 2001). For example, odor-shock conditioning has been demonstrated to modify CS+ odor-evoked responses in the basolateral amygdala (Hegoburu et al. 2009; Rosenkranz and Grace 2002; Sullivan et al. 2000). In fact, the learned changes at synapses within lateral amygdala can be stimulus specific (Debiec et al. 2010; Diaz-Mataix et al. 2011), suggesting local changes within the amygdala may directly contribute to the stimulus specificity of learned fear. Furthermore, these networks presumably interact to help shape odor processing in the piriform cortex via feedback connections (see below).

Chapuis and Wilson (in press) have recently observed similar task-dependent enhancement and reduction of behavioral and piriform cortical sensory acuity using appetitive tasks. Although not directly comparable, both this study and Chapuis and Wilson (in press) showed strong behavioral generalization and narrowing of receptive field width after specific training protocols. Whether the mechanisms of these changes induced by appetitive and aversive conditioning are the same is unclear. Work in humans suggests stimulus generalization is greater after aversive conditioning than after appetitive conditioning (Schechtman et al. 2010). Further work will be required to explore differences between these paradigms. It should also be noted that given the variability in stimulus time of arrival at the subject's snout, we were limited to very broad temporal windows of analysis. More precise timing in the future may allow analysis of changes in temporal structure of activity (Calu et al. 2007; Roesch et al. 2007), rather than just general activity rate.

The most dramatic changes observed here were decreases in single-unit spiking in the conditioned groups, including decreases in spontaneous activity, decreases in odor-evoked excitatory responses, and in the standard conditioning paradigm, increases in the probability of suppressive responses. Furthermore, the odor response changes were not odor specific. How does this combination of outcomes contribute to the observed behavioral change and what mechanisms may underlie these changes? For the differential conditioning paradigm, which results in odor-specific fear, the relationship between physiology and behavior appears relatively clear. A reduction in excitatory receptive field breadth compared with preconditioning leads to more precise encoding of odors and enhanced discriminability of those odors. This could contribute to stimulus specific fear. The results in the standard conditioning paradigm are more complex. An increase in receptive field breadth can lead to impaired perceptual acuity, which could contribute to the generalized fear response.

However, the greatest contributor to receptive field broadening after standard conditioning was the enhancement in suppressive responses, which seems counterintuitive for decreasing acuity. Nonetheless, the increase in suppressive responses may suggest an important role for inhibitory interneurons in this plasticity. There are a variety of different classes of GABAergic interneurons in the aPCX each of which contribute differently to circuit function (Suzuki and Bekkers 2010). In naïve animals, inhibitory interneurons in piriform cortex are much more broadly odor responsive than pyramidal cells (Poo and Isaacson 2009). Whether the enhancement in suppressive response tuning after standard conditioning was due to further broadening of inhibitory interneuron receptive fields or increased effectiveness of this inhibition on an already suppressed firing rate is not clear and is under investigation. Furthermore, inhibitory interneurons are a target of some descending inputs to the piriform cortex (Luna 2011; Mouly and Di Scala 2006), which help contribute to experience-dependent shaping of cortical odor responses. Thus a contribution from either changes in local inhibitory circuitry (Poo and Isaacson 2009; Suzuki and Bekkers 2007; Zhang et al. 2006) or higher order centers such as the amygdala (Luna 2011), orbitofrontal cortex (Cohen et al. 2008), or entorhinal cortex (Ferry et al. 1996) may contribute to the constellation of learning induced changes observed here. Manipulation of the described learned changes in aPCX sensory physiology will be required to determine their necessary and/or sufficient contribution to the behavioral changes.

Although previous recordings in awake rats using a differential association appetitive task (go/no-go test) have shown odor-specific changes in aPCX-evoked activity (Roesch et al. 2007), the differences in aPCX responses we recorded before and after training appear to be nonodor specific. This difference may result from different aPCX plasticity requirements to reach behavioral criteria for the paradigms. Rapidly acquired aversive conditioning can quickly modify odor-evoked postsynaptic potentials in neurons of the lateral nucleus of the amygdala (Rosenkranz and Grace 2002), which may contribute to odor-evoked freezing behaviors. When fast responses are needed, precise CS-unconditioned stimulus association may not be required in the piriform cortex. In contrast, the slowly acquired olfactory discrimination tasks often involve complex rule learning and spatial learning (Cleland et al. 2002; Saar et al. 1998, 1999) and require more specific odor and perhaps multimodal associative encoding in the piriform cortex to reach the behavioral criteria (Calu et al. 2007; Roesch et al. 2007; Schoenbaum and Eichenbaum 1995).

Summary.

The present results showed that standard and differential fear conditioning induced different levels of odor-fear generalization in rats. There were corresponding training paradigm-dependent differences in aPCX single-unit activity, suggesting that plasticity of piriform cortical networks may contribute to odor generalization and discrimination induced by aversive conditioning. Both aversive learning paradigms induced a significant reduction in spontaneous activity that was not observed in pseudoconditioned rats. This reduction may increase signal-to-noise ratio in the aPCX and presumably enhance cue odor detection in a dangerous environment. Together, the results suggest that changes within the primary sensory system may contribute to differing outcomes of fear conditioning. Generalized fear responses may in part be due to reduced selectivity of piriform cortical odor responses.

GRANTS

Funding for this study was provided by National Institute on Deafness and Other Communication Disorders Grants DC-03906 and DC-008982 (to D. A. Wilson) and DC-009910 (to R. M. Sullivan).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

C.-F.F.C. and D.A.W. conception and design of research; C.-F.F.C. and D.C.B. performed experiments; C.-F.F.C., D.C.B., and D.A.W. analyzed data; C.-F.F.C. and D.A.W. interpreted results of experiments; C.-F.F.C. and D.A.W. prepared figures; C.-F.F.C. drafted manuscript; C.-F.F.C. and D.A.W. edited and revised manuscript; C.-F.F.C., D.C.B., and D.A.W. approved final version of manuscript.