Dynamic sensory representations in the olfactory bulb: modulation by wakefulness and experience

Abstract

How are sensory representations in the brain influenced by the state of an animal? Here we use chronic two-photon calcium imaging to explore how wakefulness and experience shape odor representations in the mouse olfactory bulb. Comparing the awake and anesthetized state, we show that wakefulness greatly enhances the activity of inhibitory granule cells and makes principal mitral cell odor responses more sparse and temporally dynamic. In awake mice, brief repeated odor experience leads to a gradual and long-lasting (months) weakening of mitral cell odor representations. This mitral cell plasticity is odor specific, recovers gradually over months, and can be repeated with different odors. Furthermore, the expression of this experience-dependent plasticity is prevented by anesthesia. Together, our results demonstrate the dynamic nature of mitral cell odor representations in awake animals, which is constantly shaped by recent odor experience.

Figure 1. Long-term imaging of mitral cell activity in awake mice

(A) Olfactory bulb schematic. OSN: olfactory sensory neurons, Glom: glomeruli. Each color represents OSNs that express a unique odorant receptor. (B) Top, fixed thin section of the dorsal olfactory bulb four weeks after injection of AAV-hSyn-FLEX-GCaMP3 into the olfactory bulb of a PCdh21-Cre mouse. Note the dense labeling in the mitral cell layer. All glomeruli in the field of view contain apical dendrites of mitral cells expressing GCaMP3. GCaMP3 fluorescence in the external plexiform layer (EPL) largely reflects dendrites of mitral cells with additional contributions from tufted cells which also express Cre. GCaMP3 fluorescence in deeper bulb layers represents expression in mitral cell axons. Bottom, Expression of GCaMP3 in the mitral cell layer co-localizes with immunolabeling of Tbx21, a mitral cell-specific marker. (C) Top, two-photon single plane image of the mitral cell layer in an olfactory bulb slice from a GCaMP3-expressing PCdh21-Cre mouse illustrating recording configuration and region of line scan (yellow). Bottom, average fluorescence transients (3 trials) evoked by two, four, and six action potentials (APs) at 50 Hz (1 nA, 3 ms current steps) in the same cell. (D) Top, average GCaMP3 responses (2–5 trials) from a representative cell in response to a range of APs elicited at 100 Hz. Bottom, summary of the increase in dF/F (n = 9 cells) for varying numbers of APs elicited at 100 Hz indicates that GCaMP3 is a relatively linear reporter of AP firing. Black line is average response and gray lines represent responses of individual cells. (E) In vivo imaging configuration. (F) Two-photon imaging of GCaMP3-expressing mitral cells in vivo. Left; imaging ~50 μm under pial surface in the glomerular layer shows that all glomeruli (~60) in the field of view contain GCaMP3-expressing dendrites. Middle; imaging ~200 μm deeper in the mitral cell layer (zoomed in from the yellow square on the left) reveals ~100 GCaMP3-expressing mitral cell somata. Right; the same sets of mitral cells imaged 37 days later. (G) Different odors activate overlapping but distinct ensembles of simultaneously imaged mitral cells. Gray: seven individual trials; black: average trace.

Figure 2. Mitral cell odor responses in the awake state are more sparse, temporally dynamic and efficient compared to the anesthesized brain state

(A) Mitral cell odor-evoked activity maps (pseudocoloring represents odor-evoked GCaMP3 dF/F response averaged across seven trials) from a mouse recorded before and after the induction of anesthesia. Mitral cell ensemble response is weaker in the awake state (middle) compared to the anesthestized state (right). The left panel shows all imaged mitral cells in white. The top traces show example EEG traces during both states. (B) Average (solid lines) and SEM (shading) of odor-evoked GCaMP3 responses from five mitral cells in (A) during awake and anesthesized states show an increase in odor-evoked activity after the induction of anesthesia. Images at the bottom show that each mitral cell was clearly identified under both conditions. (C) Summary data shown in cumulative frequency plot of odor tuning (the number of odors eliciting responses out of the seven tested odors) for 340 mitral cells in the awake state compared to after induction of anesthesia (n = 6 mice, 3 urethane-anesthetized, 3 ketamine-anesthetized). Anesthesia causes a rightward shift (i.e. broader tuning) in the anesthetized state. (D) Cumulative frequency plot of the peak response amplitude for all mitral cell-odor pairs that were responsive in both the awake and anesthetized state (567 pairs). Anesthesia causes a rightward shift (i.e. larger responses) in the anesthetized state. (E) The fraction of trials with correct odor classification plotted against the number of responsive cell-odor pairs used for the classification. In the awake state, fewer responsive cell-odor pairs are required for high levels of correct classification. Dotted line represents the chance level of classification. (F) Mitral cell responses are more temporally dynamic in the awake state. Top: heat maps of the activity of all responsive mitral cell-odor pairs in awake and anesthetized states (awake: 824 cell-odor pairs; anesthetized: 1218), sorted by their onset times. Bottom: histograms of the onset times of all responsive cell-odor pairs in the two states show that responses are temporally more diverse in the awake state, while responses under anesthesia are more time-locked to odor onset. Dashed lines indicate the odor period. (G) Average (solid) and SEM (shading) of correlation coefficients of ensemble activity patterns and the initial patterns (average of 0.75 – 1.25 sec after odor onset) as a function of time in all mouse-odor pairs (42 pairs: six mice, seven odors). Awake responses become progressively more decorrelated from the initial patterns over time than responses under anesthesia. (H) Principal component analysis (PCA) reveals that odor representations become more distinct over time in the awake state compared to anesthetized brain states. The ensemble activity patterns of mitral cells pooled across animals (n = 6 mice) from 49 trials (seven trials / seven odors, four second odor application) each in the awake and anesthetized states are plotted in the space of the first three principle components defined by concatenating responses in both states (Experimental Procedures). Plots show ensemble activity under baseline conditions (Pre-odor), early during odor application (0–1 s), and during the last second of odor application (3–4 s). Activity patterns elicited by different odors are more distinct in the awake state, especially during the later phase of odor stimulation. Points represent individual trials. Green: isoamyl acetate; red: ethyl butyrate; black: heptan-4-one; light blue: heptanal; yellow: butyric acid; purple: cyclohexanone; blue: ethyl tiglate. (I) The ability of cell ensembles to correctly classify odors improves over time during odor presentation. The temporal improvement is greater in the awake state. Odor onset = 0 s. See also Figures S1 and S2.

Figure 3. Granule cell activity is enhanced during wakefulness

(A) Left: schematic of granule cell imaging. Right: a field of granule cells imaged in vivo. (B) Spontaneous activity of a population of granule cells (n = 43) imaged in the awake and anesthetized states over 60 seconds in each condition. In the awake state, many granule cells show spontaneous increases in GCaMP fluorescence, which is virtually absent in the anesthetized state. (C) Granule cell odor-evoked activity maps (pseudocoloring represents odor-evoked GCaMP3 dF/F response averaged across seven trials) from a mouse recorded in both the awake and anesthetized state. Granule cell ensembles respond more strongly to the same odor in the awake state (middle) compared to under anesthesia (right). The left panel shows all imaged granule cells in white. Top traces are example EEG traces in both states. (D) Average (solid lines) and SEM (shading) of odor-evoked GCaMP3 responses from five granule cells in (C) during awake and anesthesized states show a decrease in odor-evoked activity after the induction of anesthesia. Images at the bottom show that each granule cell was clearly identified under both conditions. (E) Summary data shown in cumulative frequency plot of odor tuning (the number of odors eliciting responses out of the seven tested odors) for 336 granule cells in the awake state compared to after induction of anesthesia (n = 6 mice, 3 urethane-anesthetized, 3 ketamine-anesthetized). Anesthesia causes a leftward shift (i.e. narrower tuning). (F) Cumulative frequency plot of the peak response amplitude of all granule cell-odor pairs that were responsive in both the awake and anesthetized state (222 pairs) shows a leftward shift (i.e. smaller responses) in the anesthetized state. (G) Temporal dynamics of granule cell odor responses. Top: heat maps of the activity of all responsive granule cell-odor pairs in awake and anesthetized states (awake: 819 cell-odor pairs; anesthetized: 300), sorted by their onset times. Bottom: histograms of the onset times of all responsive cell-odor pairs in the two states. Dashed lines indicate the odor period. See also Figure S2.

Figure 4. Odor experience leads to a gradual and odor-specific reduction of mitral cell responses

(A) Mitral cell odor-evoked activity maps (pseudocoloring represents odor-evoked GCaMP3 dF/F response averaged across seven trials) from a mouse imaged on two consecutive days. Mitral cell ensembles respond more weakly to the same series of novel odors on the second day of testing. The top right panel shows all imaged mitral cells in white. (B) Odor experience and chronic imaging protocol. (C) Activity maps of mitral cell population responses from one mouse show that responses to experienced (odor set A) odors become weaker after seven days (Day 7), while responses to less-experienced (odor set B) odors remain relatively stable. (D) Average (solid) and SEM (shading) of Change Index (CI, see text) for each trial across days. On Day 7, responses to experienced odors are significantly weaker than those to less-experienced odors. ***p < 0.001. (E) Correlations of odor representations across days for each odor set, measured by Pearson’s correlation between population vectors which were constructed by concatenating responses of each odor-cell pair. For both odor sets, the measurements (black dots) were higher than shuffled data (10,000 repetitions in which ROI labels were shuffled within each odor-animal pair) which are shown in box plots where whiskers represent most extreme values within 1.5 × I.Q.R. These data indicate that ensemble response patterns are conserved after odor experience. (F) Individual mitral cells show odor-specific plasticity, such that responses to experienced odors are generally reduced, while responses to less-experienced odors are largely unaffected. Average (solid lines) and SEM (shading) of odor-evoked GCaMP3 responses from two mitral cells in the same mouse on the first day of testing (Day 1, black) and after seven days of experience with a subset of odors (Day 7, red). (G) Tuning curves of the two cells in (F) (odors ranked for each cell according to the response strength on Day 1) indicate a shift in tuning towards less-experienced odors on Day 7. (H) Summary of odor-specific response plasticity of individual mitral cells that responded to both odor sets A and B (n = 134). Each black point represents an individual mitral cell, whose y and x positions show the average CI values for experienced and less-experienced odors, respectively. Red cross represents average of all cells. If the cells show the same amount of plasticity for experienced and less-experienced odors (i.e. a simple gain change), they would fall along the dashed unity line. The distribution shows a clear separation away from the unity line (p < 0.001, bootstrap, 10,000 repetitions) with CI values < 0 for experienced odors and CI values ≈ 0 for less-experienced odors, indicating that a specific tuning shift occurs in a large number of mitral cells. Histograms indicate the distribution of CI values for the experienced (right) and less-experienced (top) odor groups for each cell. (I) The fraction of trials classified correctly for four experienced odors (odor set A) on Day 1 and Day 7 plotted against the number of responsive cell-odor pairs used for the classification. Odor coding efficiency by mitral cell ensembles is unchanged after seven days of odor experience. n = 5 mice. Dotted line represents the chance level of classification. See also Figures S3–S7.

Figure 5. Sensory neuron input to the olfactory bulb is unaltered by repeated odor experience

(A) Schematic illustrating expression of optical reporter of transmitter release in OSNs. (B) Odor experience and chronic imaging protocol. (C) Odor-evoked glomerular activity maps from an OMP-spH mouse reveal that OSN transmission is unaltered by odor experience. (D) Average (solid lines) and SEM (shading) of odor-evoked synapto-pHluorin responses from two glomeruli in the same mouse on the first day of testing (Day 1, black) and after seven days of experience with a subset of odors (Day 7, red). (E) Summary data (n = 79 glomeruli from 4 mice) indicating that synapto-pHluorin responses remain stable (Change Index near 0) throughout the course of the experiment. See also Figures S1 and S8.

Figure 6. Dynamic modulation of sensory representations and mitral cell tuning by recent experience

(A) Schematic of the months-long experimental protocol, where responses of the same mitral cell populations were imaged during odor set A experience, recovery, and subsequent odor set B experience. (B) Mitral cell activity maps from a representative mouse over the course of the experiment. Imaging time points correspond to days illustrated in (A). Recovery, 2 months. Recently-experienced odors (set A on Day A7 and set B on Day B7) are more weakly represented than novel odors (sets A and B on Day 1) or odors with distant experience (set B on Day A7 and set A on Day B7). (C) Summary of Change Index (CI) values, defined for responses on each day, over the time course of the experiment (n = 132 mitral cells from 3 mice). (D) After recovery, odor representations are similar to those measured on Day A1. See legend for Figure 4E. (E and F) Tuning of individual mitral cells is dynamically modulated by recent experience. (E) Odor preference for individual cells (n = 43, black points) for set A or set B odors shifts with experience. Only cells that responded to at least one odor from each odor set on Day A1 and at least one odor from either set at the other times are included. Red marks indicate average and SEM. (F) Tuning curves for odors in each set ranked based on the response strength on Day A1 (black). Odor set A experience initially shifts the tuning curve towards odor set B. After recovery, odor set B experience shifts the tuning curve towards odor set A. (G) Average fraction of responsive mitral cells to odor set B on Day B7 (recently experienced) is significantly less (p < 0.001, bootstrap, 10,000 repetitions) than the fraction of cells responsive to odor set A (remotely experienced). n = 12 and 9 mouse-odor pairs for remote and recent, respectively. **p < 0.01, ***p < 0.001.

Figure 7. The expression of the experience-dependent plasticity of mitral cell activity depends on wakefulness

(A) Change Index (CI) for each trial within a day in awake (black) and anesthetized (red) animals. Responses under anesthesia are significantly more stable compared to those in awake animals, which show significant weakening. Error bars represent SEM. ***p < 0.001. (B) Activity of mitral cell populations imaged over a seven-day experience protocol. Additionally, activity of the same mitral cell populations to the same odors was imaged under ketamine/xylazine anesthesia on days 1 and 7. Summary of Change Index (CI) values, defined for responses on each day, over the time course of the experiment shows that the odor-specific reduction in mitral cell activity observed during wakefulness on Day 7 was abolished when subsequently tested under anesthesia. Error bars (SEM) are contained in symbols.

Figure 8. Schematic of dynamic odor representations

In the awake state, odors are represented by sparse and temporally dynamic patterns of mitral cell ensemble activity. Odor experience leads to a further sparsening of representations of experienced odors. Anesthesia broadens mitral cell responses and reduces temporal dynamics, resulting in a greater overlap of representations of different odors and a reduced efficiency of odor coding.