Hippocampal-Prefrontal θ Coupling Develops as Mice Become Proficient in Associative Odorant Discrimination Learning

Dual CA1-mPFC tetrode recording in mice undergoing the go-no go olfactory discrimination task

The goal of this study was to determine whether changes in coupled oscillations that occur in the OB as mice learn to discriminate odorants in a go-no go task are observed in downstream areas of the brain (mPFC and hippocampus). These downstream areas receive input originating from the OB and oscillations in the bulb are known to be coupled to the hippocampus and throughout the brain (Martin et al., 2007; Nguyen Chi et al., 2016; Tort et al., 2018). We also wanted to determine whether there was a difference between the CaMKIIα KO, CaMKIIα Hets, and the WT mice in tPAC that measures the amplitude of high bandwidth β and γ oscillations in different phases of slow θ oscillations. Furthermore, we determined whether the power of high bandwidth oscillations carry different information at the peak and trough high bandwidth amplitude phase of the θ oscillation for rewarded odorant trials. Evaluating tPRP allows for determining whether a downstream observer looking through these two θ phase windows receive different information on the stimulus. Lastly, we determined whether coordinated neural activity changed as the animal learned to discriminate odorants and we asked whether there was a difference in the relationship of oscillations between CA1 and mPFC between the different CaMKIIα genotypes.

In the go-no go odorant discrimination task, thirsty mice learned to lick on a spout in the presence of a rewarded odorant (S+) to obtain a water reward and refrained from licking in the presence of unrewarded odorant (S–) in a go-no go associative learning task (Losacco et al., 2020). The odorants were presented in pseudorandomized order in the go no-go task (Fig. 1A–D). The odorant pairs tested were different volatile compounds (Table 2) and we found no difference in any measurements between odorant pairs. Mice start the trial at will by licking on the lick port. The odorant is delivered at a random time 1–1.5 s after nose poke (the time course for the trial is shown in Fig. 1C). The mice learn to either lick a waterspout at least once during each 0.5-s segment in the 2-s response area to obtain a water reward for the rewarded odorant or refrain from licking for the unrewarded odorant. Mice refrain from licking during presentation of the unrewarded odorant because of the effort it takes to lick. Behavioral performance was termed naive or proficient when their performance estimated in a 20-trial window was below 65% for naive and above 80% for proficient. Mice were trained in the task for sessions of up to 200 trails. In the last session training ended after the animal achieved two 20-trial blocks of proficient performance; valence was reversed the next day or another odorant was tested. We recorded the LFP using two tetrodes (four electrodes per tetrode) implanted in the CA1 of the hippocampus and two tetrodes in ipsilateral mPFC, and we analyzed the data to determine whether the different genotypes differ in cross-frequency coupling for naive or proficient mice (Fig. 1E). The dataset is comprised of 747 recording sessions in 18 mice. The odorant pairs tested were different volatile compounds (or mixtures) whose nomenclature addresses the odorant names and the experimental set (e.g., APEB, see Table 2 for the nomenclature). Table 3 enumerates the total number of sessions per odorant pair, mouse, and experiment.

We performed tPAC analysis of the LFP recorded in the go no-go behavioral task. tPAC analysis is a cross-frequency coupling analysis to determine whether high-frequency oscillation bursts take place at specific phases of low-frequency θ oscillations (Tort et al., 2010). tPAC has been reported in brain areas such as the OB, hippocampus and PFC (Belluscio et al., 2012; Kaplan et al., 2014; Rojas-Líbano et al., 2014; Colgin, 2015; Scheffer-Teixeira and Tort, 2016; Losacco et al., 2020). Figure 1Ei–iv shows an example of high γ tPAC for the LFP recorded in CA1. Figure 1Ei shows the extracellular LFP sampled for hippocampus at 20 kHz and filtered between 1–750 Hz. The raw signal (Fig. 1Ei) was filtered with a 20^th order Butterworth filter into different LFP frequency bands [Fig. 1Eii–iv; θ, 6–14 Hz, adapted from Nguyen Chi et al. (2016) high γ, 65–95 Hz]. We used tPAC analysis (Tort et al., 2010) to evaluate the degree of coupling of the amplitude of the envelope of the β or high γ LFP on the phase of the θ LFP. Figure 1Eiii shows the θ phase and in Figure 1Eiv shows the envelope for the amplitude of the high γ LFP, both calculated with the Hilbert transform as detailed by Tort et al. (2010). Figure 1Eii,iv shows that the filtered high γ LFP changes amplitude in manner that appears coordinated with the θ phase.

The peak angle variance of tPAC increased for the unrewarded odorant as the animals became proficient differentiating between odorants

We proceeded to ask whether the strength of tPAC, quantified by the MI, changes as the animal learns to differentiate odorants in the go-no go task. MI is a measure of how localized high-frequency oscillation is within the phase of θ oscillations (Tort et al., 2010). Figure 2A–C illustrate an example of high γ tPAC recorded in CA1 for S+ and S– odorant trials in two sessions for naive and proficient mice. The phase amplitude plots for the trials are shown in pseudocolor in Figure 2A, for a WT mouse during the naive (Fig. 2Ai–iii) and proficient (Fig. 2Aii–iv) stages and the percent correct as a function of trial number is shown for the two sessions in Figure 2B. In this example, there appears to be an increase in the strength of tPAC when the animal becomes proficient differentiating between the odorants. Furthermore, there was an increase in peak angle variance for the unrewarded odorant as shown by the peak angle polar histograms in Figure 2Ci,ii. Figure 2Ci shows that for the unrewarded odorant the peak angle was near 90° during the naive stage, while Figure 2Cii shows the peak angle for the unrewarded odorant fluctuated widely during the proficient stage. This is in contrast with the rewarded odorant for which the peak angle remained near 90° during naive (Fig. 2Ciii) and proficient stages (Fig. 2Civ).

Figure 2D shows the differences in MI computed per mouse, per odorant pair between S+ and S– for naive and proficient mice for tPAC for hippocampus (Fig. 2Di) β, (Fig. 2Dii) high γ and mPFC (Fig. 2Diii) β, (Fig. 2Div) high γ. For the β tPAC MI for both the hippocampus and mPFC a GLM analysis found statistically significant differences between S+ versus S– and the interaction between S+ versus S– and naive versus proficient (p < 0.001, 188 observations, 184 df, F statistic = 51–67.7, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 2-1). GLM does not find significant differences for MI for high γ tPAC (p > 0.05). Figure 2E shows differences for peak angle variance between S+ and S– for naive and proficient mice for hippocampus tPAC (Fig. 2Ei) for β and (Fig. 2Eii) high γ and for mPFC tPAC for (Fig. 2Eiii) β and (Fig. 2Eiv) high γ. For both β and γ tPAC and for both the hippocampus and mPFC GLM found statistically significant differences between S+ versus S– and the interaction between S+ versus S– and naive versus proficient (p < 0.001, 188 observations, 184 df, F statistic = 19.5–26.4, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 2-1) and for naive versus proficient (p < 0.05). Asterisks in all figures show pairwise statistical significance (t test or ranksum test, p<pFDR, p value for significance corrected for multiple comparisons using the false discovery rate; Curran-Everett, 2000).

Overall, we found an increase in peak angle variance for the unrewarded odorant as the animal became proficient. We also found small, but significant changes in the strength of β tPAC that differed between the S+ and S– odorants as the animal learned to discriminate odorants.

The odorant-elicited change in the θ phase-referenced β and γ power became negative for the unrewarded odorant as the mice became proficient discriminating between odorants

Wavelet power referenced to θ phase (peak or trough) was determined to evaluate whether it changes as the animal learns. This analysis is referred to as θ phase-referenced power (tPRP; Losacco et al., 2020). Figure 3A shows examples for CA1 of the time course during the trial for the average wavelet broadband LFP spectrograms for 30 trials during naive S+ (Fig. 3Ai), 27 trials during naive S– (Fig. 3Aii) 84 trials during proficient S+ (Fig. 3Aiii) and 84 trials during proficient S– (Fig. 3Aiv). After odorant onset, there was an increase in broadband power for the S+ odorant and a decrease in power for the S– odorant as the animal became proficient.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Changes in phase-referenced power as the animal became proficient. A, Examples of the time course per trial for the average wavelet broadband LFP spectrogram for (i) S+ naive (30 trials), (ii) S– naive (27 trials), (iii) S+ proficient (84 trials), (iv) S– proficient (84 trials). Pseudocolor scale for LFP power is shown in dB. The LFP was recorded from CA1. B, Average γ wavelet power referenced to θ peak and through for the same trials as shown in A (i) S– proficient, (ii) S+ proficient, red: peak, blue: through, shadow: confidence interval. C, Lick rate for the same trials (i) S– proficient, (ii) S+ proficient. D, E, Summary bar graphs showing that peak/trough tPRP for the odorant period (0.5–2.5 s after the odorant is diverted to the odor port) decreases for S– as the mice become proficient. Black points are per mouse per odorant averages. The gray symbols and lines are per mouse averages. D, Peak. E, Through. (i) θ/β and (ii) θ/γ hippocampus, (iii) θ/β and (iv) θ/γ PFC. GLM found statistically significant differences for S+ versus S– and for the interaction between S+ versus S– and naive versus proficient (p < 0.001, 376 observations, 368 df, F statistic = 72.7–103, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 3-1). Asterisks show significant p values (p<pFDR) for post hoc pairwise tests. Extended Data Figure 3-1 provides GLM and ANOVAN statistical analysis for the data in D, E.

Figure 3B shows for this example for LFP recorded in CA1 the time course during the trial for the average high γ tPRP referenced to the peak (red) or trough (blue) of θ for the proficient trials in Figure 3A. For the rewarded (S+) odorant the power increased for the peak and decreased for trough after addition of the odorant while for the unrewarded odorant both peak and trough tPRPs decreased. Figure 3C shows that, as expected, there was an increase in the lick rate for the rewarded odorant and a decrease for the unrewarded odorant. Finally, Figure 3D,E shows a per mouse per odorant pair analysis that indicated that tPRP became negative for S– as the mice became proficient. Figure 3E, peak; Figure 3F, through, (i) β and (ii) γ hippocampus; (iii) β and (iv) γ mPFC. GLM found statistically significant differences for tPRP between S+ versus S– and the interaction between S+ versus S– and naive versus proficient (p < 0.001, 376 observations, 368 df, F statistic = 72.7–103, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 3-1).

In addition, we asked whether odorant-induced changes in peak tPRP changed when odorant valence was reversed. Figure 4Ai,ii shows an example for CA1 LFP of the time course for the peak (red) and trough (blue) high γ tPRP when the mouse was proficient for a forward session where the rewarded odorant (S+) was acetophenone (AP) and the unrewarded odorant (S–) was ethyl benzoate (EB). Figure 4Aiii,iv shows that when the valence of the odorant was reversed (AP was S– and EB was S+) the response to EB resembled the response to AP in the forward sessions (compare Fig. 4Aii and iv) indicating that the response is a response to the contextual identity of the odorant (is it rewarded?) as opposed to the chemical identity of the odorant. Figure 4B,C shows a summary bar graph analysis per mouse per odorant pair of all reversal experiments indicating that for the proficient mouse the average peak and trough tPRP decreases for the unrewarded S– odorant regardless of the identity of the odorant. For high γ tPRP GLM found statistically significant differences for S+ versus S– and peak versus trough (p < 0.001, 188 observations, 180 df, F statistic = 61.8, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 4-1) and does not find a difference between forward and reverse sessions indicating that indeed the high γ tPRP responds to the contextual odorant identity. For β tPRP GLM found a statistically significant difference for S+ versus S– (p < 0.001, 188 observations, 180 df, F statistic = 87, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 4-1) and interestingly, GLM does find a small statistically significant differences between forward versus reverse (p < 0.05) for β tPRP. Although this is a relatively small difference, this indicates that β tPRP is not exclusively responsive to contextual odorant identity and that these brain regions may also encode information on the chemical identity of the odorant.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Effect of reversal of odorant valence on the phase-referenced power. A, Average γ wavelet power referenced to θ peak and through for the same trials when the mouse was proficient for the CA1 LFP example shown in Figure 3A. The γ tPRP is normalized by subtracting the power before odorant application. i, ii, Forward: (i) S– (ethyl benzoate, EB), (ii) S+ (acetophenone, AP). iii, iv, Reversed: (iii) S– (acetophenone, AP), (iv) S+ (ethyl benzoate, EB), red: peak, blue: through, shadow; confidence interval. B, C, Summary bar graphs showing that the average peak tPRP for the odorant period (0.5–2.5 s after the odorant is diverted to the odor port) decreases for S– on odorant application regardless of the identity of the odorant. Black points are per mouse per odorant averages. The gray symbols and lines are per mouse averages. B, Peak. C, Through. (i) θ/β and (ii) θ/γ hippocampus, (iii) θ/β and (iv) θ/γ PFC. For β tPRP GLM found statistically significant differences for forward versus reverse (p < 0.05) and S+ versus S– (p < 0.001, 188 observations, 180 df, F statistic = 87, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 4-1). For γ tPRP, GLM found statistically significant differences for S+ versus S– and peak versus trough (p < 0.001, 188 observations, 180 df, F statistic = 61.8, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 4-1). Asterisks show significant p values (p<pFDR) for post hoc pairwise tests. Extended Data Figure 4-1 provides GLM and ANOVAN statistical analysis for the data in B and C.

The accuracy for decoding the contextual identity of the odorants from θ phase-referenced power increased when the mice became proficient

We proceeded to determine whether we could decode contextual odorant identity from tPRP. Decoding was performed using a LDA to set a decision boundary hyperplane between binary stimulus classes (S+ vs S–; Vizcay et al., 2015). LDA was trained with tPRP from each electrode for each mouse (8 electrodes in the hippocampus and 8 electrodes in the mPFC per mouse) for all trials except one (the training dataset) and then the tPRP from the missing trial (test data) was classified as S+ or S–. This training was performed separately at both naive and proficient learning stages. As a control we shuffled the identity of trials in the training set.

Figure 5A shows an example of the time course during the trial for the decoding accuracy for one mouse for the LDA trained using CA1 tPRP for the EAPA odor pair for (Fig. 5Ai) naive stage β, (Fig. 5Aii) proficient stage β, (Fig. 5Aiii) naive stage γ, (Fig. 5Aiv) proficient stage γ. For the naive animal decoding accuracy increased slowly after the addition of the odorant, and increased further after the animal received the reward (Fig. 5Ai,iii). When the animal became proficient decoding accuracy increased rapidly beyond 80% after addition of the odorant (Fig. 5Aii,iv). For high γ tPRP the accuracy was higher for the proficient animal for peak-referenced tPRP compared with trough-referenced tPRP (Fig. 5Aiv). Figure 5B,C shows mean bar graphs for the mean accuracy for decoding contextual odorant identity calculated for the last second of the odorant epoch (1.5–2.5 s after diverting the odorant toward the mouse) for shuffled, naive, and proficient. Figure 5B shows the accuracy for the peak tPRP and Figure 5C shows the accuracy for trough tPRP. For all conditions the accuracy is significantly higher for the proficient stage compared with naive or shuffled. GLM analysis found for both CA1 and mPFC statistically significant differences between naive versus proficient and shuffled versus proficient for both β and γ tPRP (p < 0.001, 380 observations, 372 df, F statistic = 355–494, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 5-1) and for γ tPRP GLM found statistically significant differences between peak and trough (p < 0.05, 380 observations, 372 df, F statistic = 355–494, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 5-1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

LDA for decoding the contextual odorant identity from tPRP. A, Example for one mouse for the time course for the accuracy of odorant identity decoding by a LDA algorithm trained using tPRP calculated from CA1 LFP for the EAPA odor pair (i) naive θ/β, θ (ii) proficient θ/β, (iii) naive θ/γ, (iv) proficient θ/γ red: peak, blue: through, black: shuffled, shadow: confidence interval, black bar: odorant application. B, C, Bar graphs showing the differences in decoding accuracy between shuffled, naive, and proficient. B, Accuracy for peak tPRP for (i) θ/β in the hippocampus, (ii) θ/γ in the hippocampus, (iii) θ/β in mPFC, (iv) θ/γ in mPFC. C, Accuracy for through for (i) θ/β in the hippocampus, (ii) θ/γ in the hippocampus, (iii) θ/β in mPFC, (iv) θ/γ mPFC. The bars show the average accuracy, and the points are the accuracy per mouse per odor pair. The vertical bars show the confidence interval. The gray symbols and lines are per mouse averages. For β and γ tPRP for both prefrontal and hippocampus LDA, GLM found statistically significant differences for naive versus proficient and shuffled versus proficient (p < 0.001, 380 observations, 372 df, F statistic = 355–494, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 5-1). For γ tPRP for both prefrontal and hippocampus LDA, GLM found statistically significant differences between peak and trough (p < 0.05, 380 observations, 372 df, F statistic = 355–494, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 5-1). Asterisks show significant p values (p<pFDR) for post hoc pairwise tests. Extended Data Figure 5-1 provides GLM and ANOVAN statistical analysis for the data in B and C.

θ Phase-referenced power diverges between rewarded and unrewarded trials before divergence of lick behavior

θ Oscillations in mPFC are phase locked with licks in rats consuming liquid sucrose rewards and this θ range activity has been postulated to encode for the value of consumed fluids (Amarante et al., 2017; Amarante and Laubach, 2021). We proceeded to analyze the relationship between the tPRP time course in CA1 and mPFC and the time course for licks in the go-no go task. Figure 6A shows that licks are phase locked to the θ LFP for an example of licks aligned with θ LFP recorded from CA1 for a hit trial. Figure 6B shows the lick traces for the rewarded and unrewarded odorants for a proficient animal engaged in the go-no go task with the APEB odorant pair. For the rewarded odorant the animal licks continuously for several seconds after odorant application while for the unrewarded odorant the animal stops licking shortly after the odorant is delivered. The top panel shows that the p value calculated using a ranksum test of the difference in binary lick recordings between S+ and S– trials decreases sharply shortly after addition of the odorant reflecting divergence in lick behavior between the rewarded and unrewarded odorants. Figure 6C shows the mean lick rate for five mice undergoing the go-no go task for the APEB odorant pair.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Comparison of divergence between rewarded and unrewarded trial time courses for lick rate and tPRP. A, Example illustrating the relationship between licks and θ CA1 LFP for a Hit trial. The lick signal is binary with licks represented by an increase in voltage. The bar shows the 2.5-s odorant period divided in 0.5-s response areas. The mouse must lick at least once in each of the last four response areas to obtain a reward. B, Lick time course for a proficient mouse for 18 S+ and 18 S– trials (bottom) and for the p value for the ranksum test comparing the S+ and S– binary lick traces (calculated in 33-ms time bins). C, Average lick rate time course for the APEB odorant pair (±CI) for naive and proficient mice for the rewarded (S+) and unrewarded (S–) odorants (average of 5 mice). D, E, Time course for the average tPRP. The tPRP is z normalized by subtracting the mean tPRPfrom −2–0 s and dividing by the standard deviation in this period. The bounded line shows the average ztPRP bounded by the CI calculated from ztPRP traces per odorant per mouse (6 mice, 8 odor pairs). D is the peak ztPRP and E is the trough ztPRP. i, hippocampus θ/β, ii, hippocampus θ/γ, iii, mPFC θ/β and iv, mPFC θ/γ. Cyan is S+ and magenta is S–. GLM for the average β ztPRP between 0.5 and 2.5 s found statistically significant differences for both CA1 and mPFC for both naive versus proficient, rewarded versus unrewarded odorant and for the interaction between naive versus proficient and rewarded versus unrewarded odorant (p < 0.001, 376 observations, 368 df, F statistic = 110–104, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 6-1). GLM for the average γ ztPRP between 0.5 and 2.5 s found statistically significant differences for both CA1 and mPFC for both naive versus proficient (p < 0.001), rewarded versus unrewarded odorant (p < 0.001) and for the interactions between naive versus proficient and rewarded versus unrewarded odorant (p < 0.001), naive versus proficient and peak versus trough (p < 0.05) and rewarded versus unrewarded odorant and peak versus trough (p < 0.001, 376 observations, 368 df, F statistic = 88.9-94.5, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 6-1). Asterisks denote significant difference between rewarded and unrewarded odorants determined post hoc with t or ranksum tests (p<pFDR). F, G, i, ii, Time courses for the p value calculated with a ranksum test of the difference between rewarded and unrewarded trials for the ztPRP or the licks. The ranksum test was calculated in time bins of 33 ms. Shown are bounded lines of the average p value for ztPRP or licks (±CI) calculated per odorant per mouse. F (i) β CA1, F (ii) γ CA1, G (i) β mPFC, G (ii) γ mPFC. iii, iv, Divergence times between rewarded and unrewarded ztPRP and licks calculated from the ranksum p value time courses (see Materials and Methods). A GLM for the divergence times found statistically significant differences for both CA1 and mPFC between peak and licks and trough and licks (p < 0.001, 133 observations, 130 df, F statistic = 11.6–21, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 6-1). Asterisks denote significant differences determined post hoc with t or ranksum tests (p < pFDR). Extended Data Figure 6-1 provides GLM and ANOVAN statistical analysis for the data in D–G.

In order to compare the divergence of lick behavior with tPRP we compared the time course for the decrease of p value estimating divergence between S+ and S– for licks versus tPRP. Figure 6D,E shows the time course for the z normalized peak and trough tPRP (ztPRP) for naive and proficient animals for the different bandwidths for CA1 (Fig. 6Di,ii, Ei,ii) and mPFC (Fig. 6Diii,iv, Eiii,iv). For proficient mice ztPRP diverged between rewarded and unrewarded odorant shortly after odor onset. Consistent with the results in Figure 3, a GLM analysis for the average β ztPRP between 0.5 and 2.5 s found statistically significant differences for both CA1 and mPFC for naive versus proficient, rewarded versus unrewarded odorant and for the interaction between naive versus proficient and rewarded versus unrewarded odorant (p < 0.001, 376 observations, 368 df, F statistic = 110–104, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 6-1). GLM for the average γ ztPRP between 0.5 and 2.5 s found statistically significant differences for both CA1 and mPFC for naive versus proficient (p < 0.001), rewarded versus unrewarded odorant (p < 0.001) and for the interactions between naive versus proficient and rewarded versus unrewarded odorant (p < 0.001), naive versus proficient and peak versus trough (p < 0.05), and rewarded versus unrewarded odorant and peak versus trough (p < 0.001, 376 observations, 368 df, F statistic = 88.9–94.5, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 6-1).

To estimate the time for divergence of ztPRP between rewarded and unrewarded trials in proficient mice, we calculated the p value for a two tailed t test for ztPRP for each mouse for each odorant pair and compared it to the time course for the p value for divergence of lick behavior. As shown for CA1 and mPFC in Figure 6Fi,ii, Gi,ii, there was a sharp decline in the p values shortly after addition of the odorant and the decrease in p value took place at earlier times of ztPRP compared with licks. Divergence time was computed as the time point after odorant onset where the p value dropped below 0.005 for ≥1.2 s after addition of the odorant. Figure 6Fiii,iv, Giii,iv shows the time for divergence for peak and trough ztPRP compared with lick behavior. A GLM found statistically significant differences for both CA1 and mPFC for divergence time between peak ztPRP and licks and trough ztPRP and licks (p < 0.001, 133 observations, 130 df, F statistic = 11.6–21, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 6-1). Asterisks denote significant differences determined post hoc with t or ranksum tests (p<pFDR).

The time for divergence between rewarded and unrewarded trials differs between pre-lick-referenced and post-lick-referenced tPRP

In order to understand the relationship between licks and the power of β and γ referenced to the peak and trough of θ (tPRP) we sorted the time of occurrence of θ oscillation peaks and troughs with respect to the time of onset of adjacent licks for proficient animals. Figure 7A,B show the probability density (PD) for peak (Fig. 7A) and trough (Fig. 7B) occurrence timed with respect to adjacent licks. Consistent with studies in mPFC (Amarante et al., 2017; Amarante and Laubach, 2021) the average peaks tend to occur near the lick while the trough probability density show a bimodal distribution with peaks before and after the lick. We then calculated β and γ ztPRP time courses for peaks and troughs that occur before and after the lick (pre-lick-referenced and post-lick-referenced ztPRP). We found that the time course for pre-lick-referenced tPRP (Fig. 7C,E) tended to diverge less between rewarded and unrewarded odorants than the post-lick-referenced tPRP (Fig. 7D,F) and that the divergence was sustained for post-lick-referenced ztPRPs and transient for pre-lick-referenced ztPRPs (Fig. 7C–F). GLM for the average β or γ ztPRP between 0.5 and 2.5 s found statistically significant differences for both CA1 and mPFC for naive versus proficient (naive are not shown; Fig. 7), rewarded versus unrewarded odorant and pre-lick versus post-lick (p < 0.001, 752 observations, 736 df, F statistic = 91–123, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 7-1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Lick-referenced tPRP for proficient mice. A–D, Plots of the probability density (PD) for the time of peaks (A) or troughs (B) of θ oscillations referenced to the time for adjacent lick onset shown for CA1 (i, ii) and mPFC (iii, iv) for S+ (Ai, Bi, Aiii, Biii) and S– (Aii, Bii, Aiv, Biv). The plots show a bounded line (average ± CI) calculated from all the per mouse per odorant data. C–F, Time course for the average lick-referenced tPRP. C and D are for peak lick-referenced tPRP and E and F are for trough lick-referenced tPRP. C and E are pre-lick lick-referenced tPRP calculated for peaks and troughs that took place before the reference lick, and D and F are post-lick lick-referenced tPRP calculated for peaks and troughs that took place after the reference lick. The tPRP is z normalized by subtracting the mean tPRPfrom −2–0 s and dividing by the standard deviation in this period. The bounded line shows the average ztPRP bounded by the CI calculated from ztPRP traces per odorant per mouse (6 mice, 8 odor pairs). i, hippocampus θ/β, ii, hippocampus θ/γ, iii, mPFC θ/β, and iv, mPFC θ/γ. Cyan is S+ and magenta is S–. GLM for the average β or γ ztPRP between 0.5 and 2.5 s found statistically significant differences for both CA1 and mPFC for naive versus proficient (naive are not shown in the figure), rewarded versus unrewarded odorant and pre-lick versus post-lick (p < 0.001, 752 observations, 736 df, F statistic = 91–123, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 7-1). Asterisks denote significant difference between rewarded and unrewarded odorants determined post hoc with t or ranksum tests (p < pFDR). G, H, Divergence times between rewarded and unrewarded lick-referenced ztPRP and licks calculated from the p value time courses (see Materials and Methods). G is pre-lick-referenced ztPRP, and H is post-lick-referenced ztPRP. i and iii are for β lick-referenced ztPRP, and ii and iv are for γ lick-referenced ztPRP. A GLM for the divergence times for pre-lick tPRP found no statistically significant differences between licks and either peak or trough for all bandwidths for mPFC (p > 0.05, 100 observations, 97 df, F statistic = 0.006–0.65, p > 0.05, 6 mice, 8 odor pairs; Extended Data Fig. 7-1). A GLM for the divergence times for pre-lick tPRP found no statistically significant differences between licks and either peak or trough for γ for CA1 (p > 0.05, 100 observations, 97 df, F statistic = 0.086, p > 0.05, 6 mice, 8 odor pairs; Extended Data Fig. 7-1) and found a statistically significant difference between both peak and trough and licks for β CA1 (p < 0.05, 100 observations, 97 df, F statistic = 4, p < 0.05, 6 mice, 8 odor pairs; Extended Data Fig. 7-1). In contrast, for post-lick ztPRP, divergence for all bandwidths and for both CA1 and mPFC GLM found a statistically significant difference between both peak and trough and lick divergence (p < 0.001, 124 observations, 121 df, F statistic = 17.8–31.8, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 7-1). Asterisks denote significant differences determined post hoc with t or ranksum tests (p < pFDR). Extended Data Figure 7-1 provides GLM and ANOVAN statistical analysis for the data in C–H.

Additionally, we found an interesting difference between pre-lick-referenced ztPRP and post-lick-referenced ztPRP when we assessed the time for divergence between rewarded and unrewarded trials for the time courses for lick-referenced ztPRP. For post-lick-referenced ztPRP the time for divergence was smaller than the time for divergence for lick behavior for all bandwidths for both peak and trough for both CA1 and mPFC (Fig. 7H). In contrast, for pre-lick-referenced ztPRP in mPFC the time for divergence for both bandwidths and peak and trough did not differ from the time for divergence for lick behavior (Fig. 7Giii,iv). For CA1 the time for divergence for pre-lick-referenced ztPRP for both peak and trough differed from the time for divergence for lick behavior for β, but not for γ (Fig. 7Gi,ii). A GLM for the divergence times for pre-lick tPRP found no statistically significant differences between licks and either peak or trough for all bandwidths for mPFC (p > 0.05, 100 observations, 97 df, F statistic = 0.006–0.65, p > 0.05, 6 mice, 8 odor pairs; Extended Data Fig. 7-1). A GLM for the divergence times for pre-lick tPRP found no statistically significant differences between licks and either peak or trough for γ for CA1 (p > 0.05, 100 observations, 97 df, F statistic = 0.086, p > 0.05, 6 mice, 8 odor pairs; Extended Data Fig. 7-1) and found a statistically significant difference between both peak and trough and licks for β CA1 (p < 0.05, 100 observations, 97 df, F statistic = 4, p < 0.05, 6 mice, 8 odor pairs; Extended Data Fig. 7-1). In contrast, for post-lick ztPRP divergence for all bandwidths and for both CA1 and mPFC GLM found a statistically significant difference between both peak and trough and lick divergence (p < 0.001, 124 observations, 121 df, F statistic = 17.8–31.8, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 7-1).

Coordinated hippocampal-prefrontal neural activity decreased for the unrewarded odorant as the animal became proficient

Coordinated hippocampal-prefrontal neural activity supports the organization of brain rhythms and is present during a range of cognitive functions presumably underlying transfer of information between these two brain regions (Colgin, 2011; Gordon, 2011; Headley and Paré, 2017; Lisman et al., 2017). We proceeded to determine whether there were changes in coordinated neural activity between dorsal CA1 and mPFC as the animal learned to discriminate the odorants. We used two complementary methods to quantify coordinated neural activity by calculating imaginary coherence (iCoherence) and PLV, measures of coherence that are independent of volume conduction (Nolte et al., 2004; Bastos and Schoffelen, 2016; Namburi, 2021).

Figure 8A shows an example of a spectrogram for the time course for iCoherence for a session where a mouse was engaged in the go-no go task. The pseudocolor plots show the average iCoherence time course per trial for (Fig. 8Ai) S+ naïve, (Fig. 8Aii) S– naïve, (Fig. 8Aiii) S+ proficient, (Fig. 8Aiv) S– proficient. For S+ proficient there is an increase in iCoherence after odorant addition compared with naive. This positive iCoherence indicates that coherent oscillations that take place in CA1 before they ensue in mPFC. Another example for a different electrode pair from the same session for the proficient animal shows that the rewarded odorant elicits a decrease in θ iCoherence to a negative value (oscillations take place earlier in mPFC; Fig. 8B). Figure 8C shows the histogram for odorant-elicited changes in θ iCoherence for all electrode pairs for the proficient mouse in this session. The distribution of the changes in iCoherence was broad and on the average the rewarded odorant elicited an increase in θ iCoherence and the unrewarded odorant elicited a smaller increase.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Odorant-elicited change in iCoherence and PLV decreased for the unrewarded odorant as the animal became proficient. A, Example of pseudocolor plots of the average mPFC/CA1 iCoherence average trial time course for (i) S+ naive (30 trials), (ii) S– naive (27 trials), (iii) S+ proficient (33 trials), (iv) S– proficient (35 trials). Positive iCoherence means that CA1 oscillatory activity precedes the activity in mPFC because iCoherence is referenced to an electrode in the hippocampus. B, iCoherence time course for the same mPFC electrode referenced to a different CA1 electrode for the same proficient trials as in A. (i) S+ proficient (33 trials), (iv) S– proficient (35 trials). In this case mPFC precedes CA1. C, Odorant-elicited change in θ iCoherence for all 64 electrode pairs in the same session as in A iii, iv. (i) S+ proficient (33 trials), (ii) S– proficient (35 trials). D, Distribution of average iCoherence during odorant application per mouse per odorant pair. The distribution is normalized to the peak. Pink is the distribution for iCoherence during the odorant and black is the distribution for iCoherence calculated after shuffling trials. Shown is the mean bounded by the 95% CI of the distributions calculated per odorant per mouse. All conditions were included in this distribution: naive and proficient, S+ and S–. Vertical bars are the mean iCoherence values. E, Summary for changes in odorant-elicited changes in iCoherence (Δ iCoherence) as the animal learns. Average Δ iCoherence is shown for the different bandwidths (per mouse per odorant pair): (ii) θ, (iii) β, (iv) γ (6 mice, 8 odor pairs, the vertical line is the 95% CI). GLM found for all bandwidths a statistically significant difference for naive versus proficient (p < 0.001, 188 observations, 184 df, F statistic = 6–12, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 8-1) and for the interaction between naive versus proficient and rewarded versus unrewarded (p < 0.05, 188 observations, 184 df, F statistic = 6–12, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 8-1). F, Percent significant odorant-elicited changes in iCoherence per odorant per mouse. GLM found for all bandwidths a statistically significant difference for naive versus proficient (p < 0.001, 188 observations, 184 df, F statistic = 18.7–29.1, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 8-1). G, Bar graphs showing odor-elicited change in average odorant-elicited change in phase-locking value (Δ PLV) per mouse per odor pair for (i) θ, (ii) β, (iii) γ (6 mice, 8 odor pairs, the vertical line is the 95% CI). GLM found statistically significant differences between naive versus proficient (p < 0.05) and for S+ versus S– and the interaction between S+ versus S– and naive versus proficient (p < 0.001, 192 observations, 188 df, F statistic = 29.5–46.5, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 8-1). Extended Data Figure 8-1 provides GLM and ANOVAN statistical analysis for the data in E–G.

We proceeded to characterize iCoherence and the odorant-elicited changes in iCoherence for all sessions. Figure 8D, magenta bounded lines, show the distribution of average θ, β, and γ iCoherence per mouse per odor pair during the odorant administration period (0.5–2.5 s). The mean θ iCoherence illustrated by a vertical magenta line is positive, indicating that on the average θ oscillations take place earlier in CA1, consistent with findings by other investigators (Adhikari et al., 2010). Shuffling the trials results in a narrow symmetrical distribution for iCoherence centered at zero (Fig. 8D, black bounded lines). Figure 8E shows the change of iCoherence (Δ iCoherence) elicited by the rewarded (S+) and unrewarded (S–) odorants for naive and proficient mice calculated as the mean per odorant pair per mouse. Δ iCoherence decreases for the unrewarded odorant when the animal becomes proficient. A GLM analysis found for all bandwidths statistically significant differences for Δ coherence for S+ versus S– and the interaction between S+ versus S– and naive versus proficient (p < 0.001, 188 observations, 184 df, F statistic = 12.6–32.4, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 8-1).

Figure 8G shows the odor-elicited change in average PLV (Δ PLV) per mouse per odor pair for (Fig. 8Gi) θ, (Fig. 8Gii) β, (Fig. 8Giii) γ. Here, we also found a negative Δ PLV for the unrewarded odorant for the proficient mice. For Δ PLV GLM found statistically significant differences for Δ PLV between naive versus proficient (p < 0.05), S+ versus S– and the interaction between S+ versus S– and naive versus proficient (p < 0.001, 192 observations, 188 df, F statistic = 29.5–46.5, p < 0.001, 6 mice, 8 odor pairs). In conclusion, both the Δ iCoherence and Δ PLV measures indicate that as the mouse becomes proficient there was a decrease in coordinated hippocampal-prefrontal neural activity for the unrewarded odorant.

Decreased performance for homozygote and heterozygote CaMKIIα knock-out mice in the go-no go task

Since CaMKIIα is a protein involved in LTP, we asked whether behavioral performance differed between the different CaMKIIα genotypes (WT, CaMKIIα Het and CaMKIIα KO). We trained mice from the three genotypes in the go-no go task. There was no difference in the number of sessions to criterion (Fig. 9A, GLM p > 0.05, 24 observations, 21 df, F statistic = 0.26, p > 0.05, 8 odor pairs; Extended Data Fig. 9-1). However, for proficient mice the percent correct was higher for WT compared with both CaMKIIα Het and CaMKIIα KO (Fig. 9B, GLM p < 0.05, 137 observations, 134 df, F statistic = 3.7, p < 0.05, 6 mice, 8 odor pairs; Extended Data Fig. 9-1) and the intertrial interval was larger for CaMKIIα Het (Fig. 9B, p < 0.001, 137 observations, 134 df, F statistic = 13.5, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 9-1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Analysis of the behavioral performance for mice of the three genotypes. A, Number of sessions to criterion per odorant pair (2 blocks of 20 trials with performance >80%). There was no difference between genotypes in the number of sessions to criterion (GLM p > 0.05, 24 observations, 21 df, F statistic = 0.26, p > 0.05, 8 odor pairs). B, Percent correct for proficient mice per odorant per mouse. GLM found a statistical difference for WT compared with both CaMKIIα Het and CaMKIIα KO (GLM p < 0.05, 137 observations, 134 df, F statistic = 3.7, p < 0.05, 6 mice, 8 odor pairs). C, Intertrial interval (ITI) for proficient mice. GLM analysis found that the ITI for CaMKIIα Het differed from both CaMKIIα Het and CaMKIIα KO (p < 0.001, 137 observations, 134 df, F statistic = 13.5, p < 0.001, 6 mice, 8 odor pairs). Asterisks denote statistically significant differences evaluated with either t test or ranksum corrected for multiple comparisons (p < pFDR). Extended Data Figure 9-1 provides GLM statistical analysis for the data in A–C.

The strength of tPAC and peak angle variance differed between the CaMKIIα genotypes

We then asked whether tPAC in the hippocampus and mPFC differs between the different genotypes. When we compared strength of tPAC, measured by the MI, and the peak angle variance between genotypes (WT, CaMKIIα Het and CaMKIIα KO), we found significant differences. Figure 10A,B shows the average MI per mouse per odorant pair for β (Fig. 10Ai), high γ (Fig. 10Bi) for hippocampus and for β (Fig. 10Aii) and γ (Fig. 10Bii) for PFC. For β tPAC GLM found a statistically significant difference for MI for the interaction between WT versus CaMKIIα KO and S+ versus S– in both the hippocampus and mPFC (p < 0.001, 544 observations, 532 df, F statistic = 10.3–34.3, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 10-1). For high γ tPAC there was an increase in MI for CaMKIIα Het and a decrease in MI for CaMKIIα KO compared with WT and GLM analysis found a statistically significant difference between WT versus CaMKIIα KO and WT versus CaMKIIα Het for both hippocampus and mPFC (p < 0.001, 544 observations, 532 df, F statistic = 10.3–34.3, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 10-1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 10.

tPAC and peak angle variance per genotype. A, B, Bar graphs showing average MI per mouse per odor pair for each genotype, S+ and S– in naive and proficient mice for (A) θ/β and (B) θ/γ for (i) hippocampus and (ii) mPFC. The bars show the average MI, and the points are the MI per mouse per odor pair. For θ/β tPAC MI GLM found a statistically significant difference for S+ versus S–, for the interaction of WT versus KO and S+ versus S– and the interaction of WT versus KO and naive versus proficient (p < 0.001, 544 observations, 532 df, F statistic = 10.3–34.3, p < 0.001, 6 mice, 8 odor pairs). For θ/γ tPAC MI GLM found a statistically significant difference between WT versus KO and WT versus Het (p < 0.01) and for the interaction of WT versus Het and S+ versus S– (p < 0.05, 544 observations, 532 df, F statistic = 10.3–34.3, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 10-1). C, E, Bar graphs showing average peak angle variance per mouse per odor pair for each genotype, S+ and S– in naive and proficient mice for (A) θ/β and (B) θ/γ for (i) hippocampus and (ii) mPFC. For β tPAC MI GLM found a statistically significant difference for WT versus Het (p < 0.001) and WT versus KO (p < 0.05), S+ versus S–, (p < 0.001) and naive versus proficient (p < 0.05, 544 observations, 532 df, F statistic = 20.6, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 10-1). For γ tPAC MI GLM found a statistically significant difference for WT versus Het (p < 0.05) and WT versus KO (p < 0.05), S+ versus S–, (p < 0.001) and naive versus proficient (p < 0.05, 544 observations, 532 df, F statistic = 18.8, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 10-1). Extended Data Figure 10-1 provides GLM statistical analysis for the data in A–D.

Figure 1C,D shows average peak angle variance per mouse per odorant pair for β (Fig. 1Ci), high γ (Fig. 1Ei) in the hippocampus and (Fig. 1Cii) β, (Fig. 1Eii) high γ in PFC. For proficient mice, the peak angle variance for the rewarded odorant decreased for CaMKα Het and increased for CaMKα KO. GLM analysis found a statistically significant difference for WT versus CaMKIIα Het, WT versus CaMKIIα KO, naive versus proficient and S+ versus S– for θ/β and θ/γ for both the hippocampus and mPFC (p < 0.05, 544 observations, 532 df, F statistic = 20.6, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 10-1).

The accuracy for decoding the contextual identity of the odorant decreased in the CaMKIIα knock-out mouse and was correlated with percent correct discrimination

The differences in tPAC between CaMKIIα genotypes raises the question whether decoding of contextual identity from tPRP is altered in the CaMKIIα KO and the CaMKIIα Het mice. Figure 11 shows the results of our comparison of decoding accuracy between genotypes. Figure 11A,B shows examples for proficient mice of the time course for the accuracy of decoding of odorant contextual identity by LDA trained using tPRP for the EAPA odor pair [Fig. 11A, θ/β tPRP; Fig. 11B, θ/γ tPRP, (i) WT, (ii) Het, (iii) KO]. For the WT and CaMKIIα Het mice the accuracy increases monotonously during the odorant epoch (Fig. 11Ai,ii, Bi,ii) while for the CaMKIIα KO mouse the accuracy reaches a maximum value after 1 s and then decreases slightly for the rest of the odorant epoch (Fig. 11Aiii, Biii). Figure 11C,D shows the differences in decoding accuracy between the different genotypes assessed in the window from 1.5 to 2.5 s after addition of the odorant. For both β and γ peak or trough tPAC decoding for both the hippocampus and the mPFC decoding accuracy was lowest for CaMKIIα KO. In addition, for γ trough tPRP decoding for mPFC the accuracy was highest for WT, and decreased for both CaMKIIα Het and CaMKIIα KO mice (Fig. 11Div). For β tPRP LDA GLM found statistically significant differences for decoding accuracy between WT and KO for all conditions (p < 0.001) and between WT and Het for peak γ tPRP in the hippocampus, trough γ tPRP in mPFC (p < 0.05, 137 observations, 134 df, F statistic = 9.9–34.3, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 11-1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 11.

The accuracy for decoding the contextual identity of the odorant from tPRP decreased in the CaMKIIα knock-out mouse and was correlated with percent correct discrimination. A, B, Examples for one mouse of the time course for the accuracy of odorant identification by LDA trained using CA1 tPRP for the EAPA odor pair. A, β tPRP. B, γ tPRP. (i) WT. (ii) Het. (iii) KO. Shadow: confidence interval, black bar: odorant application. C, D, Bar graphs showing the differences in discriminant accuracy between the different genotypes. C, Accuracy for peak tPRP for (i) θ/β in the hippocampus, (ii) θ/γ in the hippocampus, (iii) θ/β in mPFC, (iv) θ/γ in mPFC. D, Accuracy for through for (i) θ/β in the hippocampus, (ii) θ/γ in the hippocampus, (iii) θ/β in mPFC, (iv) θ/γ mPFC. The bars show the average accuracy, and the points are the accuracy per mouse per odor pair. The vertical bars show the confidence interval. For β tPRP LDA GLM found statistically significant differences between WT and KO for all conditions (p < 0.001) and between WT and Het for γ trough tPRP (p < 0.05, 756 observations, 744 df, F statistic = 9.9–34.3, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 11-1). Asterisks show significant p values (p<pFDR) for post hoc pairwise tests. E, Relationship for proficient mice between percent correct in the go-no go behavior and accuracy of odor identification by the LDA decoding algorithm shown per mouse per odor pair (6 mice, 8 odor pairs). The correlation coefficients were: E (i) 0.3, E (ii) 0.24, E (iii) 0.29, E (iv) 0.23, F (i) 0.3, F (ii) 0.36, F (iii) 0.29, F (iv) 0.40, and the p value for significance was p < 0.01. Lines are best fit lines. Extended Data Figure 11-1 provides GLM statistical analysis for the data in C–F.

Finally, we did not find differences between CaMKIIα genotypes for divergence times between rewarded and unrewarded ztPRP and or lick time courses (Fig. 12, divergence times were calculated as in Fig. 6). GLM for the divergence times did not find statistically significant differences between genotypes for licks (p > 0.05, 127 observations, 124 df, F statistic = 1.5, p > 0.05, 6 mice, 8 odor pairs; Extended Data Fig. 12-1). A GLM for the divergence times did not find statistically significant differences between genotypes for β or γ ztPRP (p > 0.05, 252 observations, 246 df, F statistic = 0.5–1, p > 0.05, 6 mice, 8 odor pairs; Extended Data Fig. 12-1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 12.

Divergence times between the ztPRP time courses for rewarded and unrewarded trials for the different CaMKIIα genotypes. Divergence times between rewarded and unrewarded ztPRP and licks were calculated from the ranksum p value time courses as shown in Figure 6 (also see Materials and Methods). A, Divergence times for licks. A GLM for the divergence times did not find statistically significant differences between genotypes for licks (p > 0.05, 127 observations, 124 df, F statistic = 1.5, p > 0.05, 6 mice, 8 odor pairs; Extended Data Fig. 12-1). B, C, Divergence times for ztPRP. B, CA1. C, mPFC. (i) β tPRP. (ii) γ tPRP. A GLM for the divergence times did not find statistically significant differences between genotypes for β or γ ztPRP (p > 0.05, 252 observations, 246 df, F statistic = 0.5–1, p > 0.05, 6 mice, 8 odor pairs; Extended Data Fig. 12-1). Extended Data Figure 12-1 provides GLM and ANOVAN statistical analysis for the data in A–C.

Relationship between tPRP decoding and behavior and decision-making times across genotypes

We then asked whether there was a relationship between contextual odorant identity decoding accuracy and percent correct performance for proficient mice in the go-no go task. Figure 11E,F shows that there were statistically significant correlations between decoding accuracy and percent correct performance for all the different conditions. The correlation coefficients were as follows: 0.3 for hippocampal peak β tPRP, 0.24 for hippocampal peak γ tPRP, 0.29 for mPFC peak β tPRP, 0.23 for mPFC peak γ tPRP, 0.3 for hippocampal trough β tPRP, 0.36 for hippocampal trough γ tPRP, 0.29 for mPFC trough β tPRP and 0.40 for mPFC trough γ tPRP, and the p value for significance of the correlation coefficient was p < 0.01. This indicates that the decoding accuracy obtained with tPRP is related to behavioral performance.

Coherent hippocampal-prefrontal neural activity differed between CaMKIIα genotypes

We asked whether there were differences for coordinated hippocampal-prefrontal neural activity for the different CaMKIIα genotypes. Figure 13A shows the odorant-elicited changes in Δ iCoherence for the different genotypes. For θ Δ iCoherence GLM found a statistically significant difference between WT and KO (p < 0.05, 544 observations, 532 df, F statistic = 5.35, p < 0.001, 6 WT mice, 7 Het mice and 5 KO mice, 8 odor pairs; Extended Data Fig. 13-1). However, these were relatively small differences of θ Δ iCoherence between genotypes and post hoc tests did not yield significant differences between CaMKIIα KO and WT. Furthermore, the percent of electrode pairs that showed a significant Δ iCoherence was higher for Hets compared with WT for the rewarded odorant for β and γ Δ iCoherence (Fig. 13Bii,iii). For β Δ iCoherence GLM found a statistically significant difference between WT and Het (p < 0.05, 544 observations, 532 df, F statistic = 18.1, p < 0.001, 6 WT mice, 7 Het mice and 5 KO mice, 8 odor pairs; Extended Data Fig. 13-1). For γ Δ iCoherence, GLM found a statistically significant difference for the interaction between WT and Het and rewarded versus unrewarded odorant (p < 0.05, 544 observations, 532 df, F statistic = 14.7, p < 0.001, 6 WT mice, 7 Het mice and 5 KO mice, 8 odor pairs; Extended Data Fig. 13-1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 13.

There were small differences in odorant-elicited change in iCoherence between CaMKIIα genotypes. A, Summary bar graphs comparing the odorant-elicited change in iCoherence (Δ iCoherence) for the different genotypes. (i) θ, (ii) β, (iii) γ. The bars show the average Δ iCoherence, and the points are Δ iCoherence per mouse per odor pair. The vertical bars show the confidence interval. For θ Δ iCoherence GLM found a statistically significant difference between WT and KO (p < 0.05, 544 observations, 532 df, F statistic = 5.35, p < 0.001, 6 WT mice, 7 Het mice and 5 KO mice, 8 odor pairs; Extended Data Fig. 13-1). For γ Δ iCoherence, GLM found a statistically significant difference for the interaction between WT and KO and rewarded versus unrewarded odorant (p < 0.05, 544 observations, 532 df, F statistic = 2.76, p < 0.05, 6 WT mice, 7 Het mice and 5 KO mice, 8 odor pairs; Extended Data Fig. 13-1). Asterisks show significant p values (p<pFDR) for post hoc pairwise tests. B, Percent significant odorant-elicited changes in iCoherence per odorant per mouse. For β Δ iCoherence GLM found a statistically significant difference between WT and Het (p < 0.05, 544 observations, 532 df, F statistic = 18.1, p < 0.001, 6 WT mice, 7 Het mice and 5 KO mice, 8 odor pairs; Extended Data Fig. 13-1). For γ Δ iCoherence, GLM found a statistically significant difference for the interaction between WT and het and rewarded versus unrewarded odorant (p < 0.05, 544 observations, 532 df, F statistic = 14.7, p < 0.001, 6 WT mice, 7 Het mice and 5 KO mice, 8 odor pairs; Extended Data Fig. 13-1). C, Summary bar graphs comparing change in Δ PLV for the different genotypes. (i) β, (ii) θ, (iii) γ. For Δ PLV GLM found no statistically significant differences between WT and KO, and for the interaction of WT versus KO and S+ versus S– (p > 0.05, 512 observations, 500 df, F statistic = 25.5–39.9, p < 0.001, 6 mice, 8 odor pairs; Extended Data Fig. 13-1). Extended Data Figure 13-1 provides GLM and ANOVAN statistical analysis for the data in A–C.

Δ PLV did not show any differences between genotypes (Fig. 13C). The GLM found no statistically significant differences for Δ PLV between WT and KO or WT and Het (p > 0.05, 512 observations, 500 df, F statistic = 25.5–39.9, p < 0.001, 6 mice, 8 odor pairs).