Lack of Pattern Separation in Sensory Inputs to the Olfactory Bulb during Perceptual Learning

Abstract

Recent studies revealed changes in odor representations in the olfactory bulb during active olfactory learning (Chu et al., 2016; Yamada et al., 2017). Specifically, mitral cell ensemble responses to very similar odorant mixtures sparsened and became more distinguishable as mice learned to discriminate the odorants over days (Chu et al., 2016). In this study, we explored whether changes in the sensory inputs to the bulb underlie the observed changes in mitral cell responses. Using two-photon calcium imaging to monitor the odor responses of the olfactory sensory neuron (OSN) axon terminals in the glomeruli of the olfactory bulb during a discrimination task, we found that OSN inputs to the bulb are stable during discrimination learning. During one week of training to discriminate between very similar odorant mixtures in a Go/No-go task, OSN responses did not show significant sparsening, and the responses to the trained similar odorants did not diverge throughout training. These results suggest that the adaptive changes of mitral cell responses during perceptual learning are ensured by mechanisms downstream of OSN input, possibly in local circuits within olfactory bulb.

Keywords: mitral cells; olfactory bulb; olfactory sensory neurons; perceptual learning; plasticity; two-photon imaging.

Figure 1.

Training mice to perform a difficult discrimination task. A, Experimental timeline. Water-restricted mice first undergo a pretraining period, where they become familiarized with the discrimination task using easy odorant pairs. After the pretraining period, glomerular responses are imaged while the mouse learns to perform a difficult discrimination task with two similar odor mixtures. B, Schematic of imaging setup (left) and trial structure (right). C, Schematic of behavioral paradigm. If a rewarded odorant is presented and the mouse responds with a lick, a water reward will be delivered through the lickport. If the unrewarded odor is delivered, no water reward will be given regardless of the mouse’s actions. D, Fraction of correctly answered trials (mean ± SEM of all mice, n = 13 mice) on each day of difficult discrimination training. Mice take 3 d on average to perform above an 80% success rate.

Figure 2.

Imaging glomerular odor responses during training. A, Schematic of the olfactory bulb. Two photon imaging of glomerular responses was performed in OMP-tTA::tetO-GCaMP6s mice, in which OSNs express GCaMP6s. B, An example of a typical glomerular field of view on the first day of imaging (day 1) and 6 d later (day 7). C, Examples of odorant responses (mean ± SEM) from individual glomeruli. Responses to the odorant 1 (S+, rewarded odorant) are shown in blue, and responses to odorant 2 (S-, unrewarded odorant) are shown in black. Odorant period is indicated by the thick horizontal black bar.

Figure 3.

Glomerular odor representations do not show an increased separation during training. A, top, Fractions of glomeruli classified as responsive (black) and divergent (magenta) are plotted for each day. The responsive fraction and divergent fraction remain stable across days (Pearson correlation; responsive, p = 0.97; divergent, p = 0.50). Bottom, Fraction of divergent glomeruli out of all responsive glomeruli does not significantly change during training (Pearson correlation; p = 0.73). B, Cumulative fraction distributions of the peak amplitude of excitatory glomerular odorant responses on day 1 and day 7. There is no difference in the distributions for either odorant between day 1 and day 7 (Kolmogorov-Smirnov test: odorant 1, p = 0.64; odorant 2, p = 0.99). C, Mean excitatory glomerular odorant responses on day 1 (black) and day 7 (red), normalized to peak amplitude. There are no significant differences between responses on day 1 and day 7 (Wilcoxon rank sum test for each imaging frame, false-discovery rate corrected: q = 0.05). D, The averaged d-prime of all divergent glomeruli does not change with difficult discrimination training (Pearson correlation; p = 0.87). E, The correlation coefficient between averaged ensemble odorant responses does not change with difficult discrimination training (Pearson correlation; p = 0.24). F, The decoder accuracy does not change during difficult discrimination training (Pearson correlation; p = 0.58). G–I, There were no significant changes in the (G) mean-squared distance between odor centroids nor (H) the total variance across trials for each odor (Pearson correlation; mean-squared distance, p = 0.68; total variance, p = 0.73). I, There was also no change in the variance along the axis of discrimination (the axis containing the line connecting Odor 1 and Odor 2 centroids) for the decoder (Pearson correlation; p = 0.50). Unless otherwise stated, mean ± SEM are shown in line plots with error bars.