Distinct roles of bulbar muscarinic and nicotinic receptors in olfactory discrimination learning

Abstract

The olfactory bulb (OB) and piriform cortex receive dense cholinergic projections from the basal forebrain. Cholinergic modulation within the piriform cortex has long been proposed to serve important functions in olfactory learning and memory. We here investigate how olfactory discrimination learning is regulated by cholinergic modulation of the OB inputs to the piriform cortex. We examined rats' performance on a two-alternative choice odor discrimination task following local, bilateral blockade of cholinergic nicotinic and/or muscarinic receptors in the OB. Results demonstrate that acquisition, but not recall, of novel discrimination problems is impaired following blockade of OB cholinergic receptors, although the relative contribution of muscarinic and nicotinic receptors depends on task difficulty. Blocking muscarinic receptors impairs learning for nearly all odor sets, whereas blocking nicotinic receptors only affects performance for perceptually similar odors. This pattern of behavioral effects is consistent with predictions from a model of cholinergic modulation in the OB and piriform cortex (de Almeida et al., 2013). Model simulations suggest that muscarinic and nicotinic receptors may serve complementary roles in regulating coherence and sparseness of the OB network output, which in turn differentially regulate the strength and overlap in cortical odor representations. Overall, our results suggest that muscarinic receptor blockade results in a bona fide learning impairment that may arise because cortical neurons are activated less often. Behavioral impairment following nicotinic receptor blockade may not be due to the inability of the cortex to learn, but rather arises because the cortex is unable to resolve highly overlapping input patterns.

Keywords: acetylcholine; cholinergic neuromodulation; olfactory bulb; olfactory cortex; rodent behavior.

Figure 1.

Experimental and computational experiments. A, Behavioral experiments. Rats are trained using an operant conditioning paradigm to discriminate between two novel odorants. Both odorants are rewarded and rats learn to associate the odorants with spatially separated reward ports. Learning is assessed by measuring the percentage correct performance for five consecutive (daily) sessions consisting of 200 trials each. Cholinergic modulation in the OB is manipulated by directly infusing antagonists bilaterally into the OBs. B, Simulations. To simulate odor discrimination learning, the OB and cortical network model is presented with a pair of odorants over a sequence of five training sessions. Olfactory information is presented to the bulbar network [olfactory sensory neuron (OSN) activation pattern], transformed and processed by the OB network and projected to the cortical network. The cortical representations are rendered more distinct during learning via plasticity at associational synapses, simulating behavioral discrimination learning. Manipulations of cholinergic modulation in the OB are simulated by changing OB network parameters. Learning is measured as the change in distance between cortical odor representations across training sessions. Heat maps show simulated average firing rates in the model represented in a 10 × 10 matrix, with warmer colors representing higher rates.

Figure 2.

Cannulae placement. Coronal section through the OB illustrating placement of guide cannula and infusion needle.

Figure 3.

Computational model of OB and piriform cortex. A, Simplified structure of the piriform cortex. The system receives excitatory input from Mi cell axons in the OB that connect to Pyr and Ff cells. Pyr cells work as an autoassociative network, synapsing onto other Pyr neurons and to Fb cells. Both Ff and Fb interneurons make inhibitory synapses on Pyr cells (Stokes and Isaacson, 2010). The autoassociative connections between Pyr cells are subject to activity-dependent plasticity (+). B, We defined four groups of OB outputs based on the impact of ACh on different receptors. B_i, When both nicotinic and muscarinic receptors are on, the Mi activation pattern exhibits both high sparseness and high coherence (synchronization). B_ii, When muscarinic receptors are off and nicotinic receptors are on, Mi output exhibits high sparseness but low coherence. B_iii, When muscarinic receptors are on and nicotinic receptors are off, the Mi output exhibits low sparseness but high coherence. B_iv, When both receptors are off, the Mi output has both low sparseness and low coherence. For ease of visualization, we here show a distribution centered on neuron #50. C, Average changes in network properties (±SEM) in each of the four configurations. The data in each panel are computed from 50 simulations with random odors. Despite the changes in sparseness (C_i) and coherence (C_ii), the average firing frequency of the Mi output (C_iii) was kept constant at ∼6 Hz. This was done to isolate the effects of manipulating sparseness and coherence from the firing rate of Mi cells.

Figure 4.

OB cholinergic modulation affects behavioral performance and cortical discrimination. A, Behavioral results. A_i, Average performance (±SEM) as a function of test session for saline-infused, scopolamine-infused, MLA-infused, and combination mixture-infused rats, pooled across odor sets (n = 13). A_ii, Average PI (±SEM) for each of the four treatment groups. Rats infused with scopolamine alone or in combination with MLA have significantly lower PIs than saline-infused rats, whereas the difference between MLA-infused and saline-infused rats is not significantly different. B, Computational results. B_i, Average change in cortical discrimination (measured relative to Euclidean distance between odor pairs before training) following training using different bulbar network configurations. In the control condition with all bulbar ACh receptors on, corresponding to saline condition in behavioral experiments, rats attain higher levels of discrimination than all other simulations. Bii, Cortical PI (degree of improvement in odor discrimination in the cortical network) for networks trained with the four bulbar configurations. Similar to behavioral results, networks trained with bulbar ACh on perform better than networks trained on inputs from the bulbar network with any combination of blocked ACh receptors.

Figure 5.

Nicotinic receptor blockade varies across odor sets. A, Behavioral results. Performance as a function of test session for the four treatment groups on two individual odor sets (A_i, odor set 7; A_ii, odor set 9). Each curve represents data obtained from one rat; the four curves in each panel, representing the four drug conditions, were obtained from different rats. The effects of MLA infusion are highly variable, ranging from no noticeable effect (A_i) to complete impairment (A_ii). B, Computational results. Performance on two individual odor sets depicted as change in discrimination between the odors in the pair across test sessions when cortical network is trained with bulbar input in each of the four network configurations. Note that in one odor pair (B_i) performance under nicotinic receptor blockade is similar to that of controls, whereas in a second odor set (B_ii), performance under the same condition is notably impaired.

Figure 6.

Determining perceptual similarity of odor sets. A, Behavioral approach. A_i, Average investigation time (±SEM) pooled across odor sets for each 50 s odor exposure trial in the olfactory habituation task. One odor from an odor set was randomly assigned as the habituation odor, and was presented during four 50 s trials (H1–H4) separated by 5 min intervals, followed by two 50 s probe trials in which animals were presented the same odor (Hab) or the second odor in the pair (Novel), in random order. A_ii, Average investigation time (±SEM) on the two probe trials (Hab and Novel) for each of the odor sets. A_iii, Behavioral dissimilarity index for each odor set, averaged across rats (±SEM). Data are displayed in ascending order; labels on the abscissa correspond to the odor sets in A_ii and Table 1. Odor sets were classified into similar (filled circles) and dissimilar (open circles) groups based on the behavioral dissimilarity index. B, Example simulated bulbar input patterns for similar and dissimilar odorants. The graphs show a heat map of a 10 × 10 representation of olfactory sensory neuron (OSN) responses to three odorants with varying degrees of similarity. These are representative of patterns chosen to simulate similar and dissimilar odorants in the model. Note that in these heat maps, the odor responses are artificially ordered for ease of visualization (see Materials and Methods); in the actual simulations, the spatial distribution of active OSNs is random. Heat maps are calculated from the average OSN firing rates during odor stimulation for each odorant; warmer colors correspond to higher firing rates. Pattern dissimilarity was calculated as the Euclidean distance between two 100-dimensional vectors of average response rates, as specified in Materials and Methods. C, Correspondence between behavioral dissimilarity index and bulbar odor response patterns (reproduced with permission from Cleland et al., 2002). The graph shows the behavioral dissimilarity index measured using a habituation/dishabituation task and an odor–reward associative task (Cleland et al., 2002) in response to aliphatic acids of varying carbon chain length, as a function of the dissimilarity measured from 2DG activation maps in response to these same odorants (Johnson et al., 1999). Each data point shows the average behavioral dissimilarity index as a function of the average 2DG dissimilarity index. Data from two types of behavioral experiments are depicted (1 nonassociative and 1 associative; for details, see Cleland et al., 2002); the lines represent regression lines for each experiment. Red circles indicate the range of behavioral dissimilarities used in the present experiments (A_iii), which defined the odor activation dissimilarities used in the corresponding simulations (B). Odorants used in the present experiments can be divided into two classes of dissimilarity (C, red circles; A_iii, dark and light circles) and simulation odorants were chosen to have the corresponding degree of overlap (B, 0.272 and 0.904).

Figure 7.

Role of bulbar cholinergic receptors in acquisition of novel odor discrimination problems depends on perceptual similarity of odors. Ai, Average performance (±SEM) as a function of test session for rats performing the acquisition task using dissimilar odor sets (n = 9). Note the high performance in MLA-infused rats. Aii, Average performance (±SEM) as a function of test session for rats performing the acquisition task using similar odor sets (n = 4). Note the low performance in MLA-infused rats. B, PI for each of the drug treatment groups plotted separately for dissimilar and similar odor sets. *p < 0.005, **p < 0.001. C, Simulated cortical PI for networks trained with bulbar input under the four experimental conditions for odor inputs with low (dissimilar) and high (similar) overlap. Similar to behavioral results, the effect of turning off nicotinic receptors depends on odor set similarity, such that the cortical network exhibits poorer discrimination of similar odors following removal of bulbar nicotinic receptors.

Figure 8.

Behavioral control experiments. A, Dose–response curve for dissimilar odor sets. Ai, Average performance (±SEM) as a function of test session for rats performing the novel odor acquisition task using dissimilar odor sets (n = 6 odor sets) and receiving either a high dosage of cholinergic antagonists (scopolamine, 22 mm; MLA, 19 mm), a low dosage at one-fifth the original concentration (scopolamine, 4.4 mm; MLA, 3.6 mm), or saline infusions. Aii, PI for each of the drug treatment groups. **p < 0.005, ***p < 0.001. B, Control for impairment of general odor processing and perception. The graph shows the average performance as a function of daily sessions for three experimental groups. All groups received saline infusions for the first 4 d of training, followed by a fifth day in which they received either saline, scopolamine (22 mm), or MLA (19 mm) infusions. These graphs indicate that once an odor discrimination task has been acquired, manipulation of bulbar cholinergic modulation does not affect discrimination performance.

Figure 9.

Odor processing and learning in olfactory network. A, Heat maps depicting 10 × 10 arrays of odor response patterns in the OB and piriform cortex. Each image shows the average responses of Mi cells or Pyr cells during a 2 s odor presentation with warmer colors representing higher responses. Each column shows responses to one simulated odor. The odors in the left and middle columns together form a dissimilar pair, while the odors in the middle and right columns constitute a similar odor pair, according to the definitions in Figure 5. Odor representations conveyed to the OB are processed in the OB network to create bulbar output. The three upper rows show examples of bulbar output patterns under control conditions (ACh ON) and partial receptor blockade conditions (Ni OFF or Mu OFF). In the bulb, average response patterns are modulated by nicotinic, but not muscarinic, receptor activation. Bulbar output patterns are conveyed to the cortical network (PC before learning), where these are modulated by synaptic plasticity in the cortical association fibers (PC after learning). In the piriform cortex, neural response patterns are transformed by learning. Learned patterns in response to OB patterns under the different drug conditions are shown in the last three rows of panels. Note that learning under control conditions (ACh ON) leads to well defined and separated pairs of odor patterns, whereas learning with Ni OFF (in the OB) creates high-overlap pattern and learning with Mu OFF (in the OB) leads to very reduced learning. B, Average change in pairwise distances between dissimilar and similar bulbar output patterns relative to the distance between these patterns at the input to the OB. Any changes within a category are due to bulbar processing. The graph shows distances between odor representations conveyed to the bulb (OB input), and at the output of the OB under full or partial modulation (OB output). Bulbar output representations are rendered more dissimilar when nicotinic receptors are active (OB output, ACh ON, Mu OFF). In contrast, at this level of representation—based on firing rate only—blockade of muscarinic receptors does not affect bulbar output (OB output, Mu OFF). C, Average change in pairwise Euclidean distance of simulated cortical representations for dissimilar and similar odorants relative to the prelearning distance between dissimilar odorants. The graph shows pairwise distances between cortical odor representations before learning (Pre learning) or after learning with bulbar modulation on (Post learning, OB Ach ON) or partially on (Ni OFF, Mu OFF). Cortical learning of dissimilar odorants is only impaired by blockade of muscarinic receptors in the OB (OB Mu OFF), whereas cortical learning of similar odorants is impaired by blockade of both muscarinic and nicotinic receptors (Ni OFF, Mu OFF).

Figure 10.

Learning in the cortical network. A_i–C_i, Raster plots of Pyr cell firing before learning in response to a sample odor under the three modulatory conditions: (A) ACh ON, (B) ACh ON/Ni OFF, and (C) ACh ON, Mu OFF. A_ii–C_ii, Raster plots of Pyr cell firing after learning of the same odor. Note that Pyr cell firing is enhanced after learning with bulbar nicotinic receptors off, but not with muscarinic receptors off. A_iii–C_iii, Example activity of two Pyr cells in the network, one undergoing changes in response to learning and the other not. A_iv–C_iv, Heat maps of synaptic weights in the model (100 × 100) before and after learning for each condition. The heat maps show synaptic weights between all pairs of Pyr cells before and after odor learning with warm colors indicating higher synaptic weights. D, Graph of average synaptic weights normalized relative to the highest weight reached during learning. E, Average change in Pyr cell firing rate during learning relative to the rate before learning under each condition.