Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning

Report

Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning Highlights

Authors

d

Tone-odor conditioning induces theta phase reset in primate amygdala and dACC

Aryeh Hai Taub, Rita Perets, Eilat Kahana, Rony Paz

d

A directional phase locking develops between amygdala spikes and dACC theta

Correspondence

d

Information transfer from amygdala to dACC decreases once memory stabilizes

[email protected] (A.H.T.), [email protected] (R.P.)

In Brief Taub et al. report increased synchrony between amygdala’s output (spikes) and prefrontal’s input (local field potentials) that is locked to theta band and declines once memory is formed. Therefore, the primate amygdala/prefrontal pathway uses theta to transfer error-like information during fear learning.

Taub et al., 2018, Neuron 97, 1–8 January 17, 2018 ª 2017 Elsevier Inc. https://doi.org/10.1016/j.neuron.2017.11.042

Please cite this article in press as: Taub et al., Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning, Neuron (2017), https://doi.org/10.1016/j.neuron.2017.11.042

Neuron

Report Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning Aryeh Hai Taub,1,* Rita Perets,1 Eilat Kahana,1 and Rony Paz1,2,* 1Department

of Neurobiology, Weizmann Institute of Science, Rehovot 7610001, Israel Contact *Correspondence: [email protected] (A.H.T.), [email protected] (R.P.) https://doi.org/10.1016/j.neuron.2017.11.042 2Lead

SUMMARY

RESULTS

The contribution of oscillatory synchrony in the primate amygdala-prefrontal pathway to aversive learning remains largely unknown. We found increased power and phase synchrony in the theta range during aversive conditioning. The synchrony was linked to single-unit spiking and exhibited specific directionality between input and output measures in each region. Although it was correlated with the magnitude of conditioned responses, it declined once the association stabilized. The results suggest that amygdala spikes help to synchronize ACC activity and transfer error signal information to support memory formation.

Increased Theta Power in the Primate Amygdala and dACC We examined the development of theta oscillations in the basolateral amygdala and dACC of monkeys while they acquired novel tone-aversive odor associations on a daily basis (ndays = 19 + 13; Figures 1A–1C). We found increased theta power following the conditioned stimulus (CS) and before delivery of the aversive unconditioned stimulus (US) in both regions (Figures 1D and 1E; t test for mean power in theta during 1 s following CS onset versus mean of 5 consecutive seconds during the pre-CS, namygdala = 73, pamygdala = 0.007, D(CS/Baselineamygdala) = 1.18, ndACC = 76, pdACC = 1.6
13, D(CS/BaselinedACC) = 1.66). Keeping with this, autocorrelation of single units revealed an increased component between 4 and 8 Hz (t test, namygdala = 214, pamygdala = 2
9, D(CS/Baselineamygdala) = 2.85, ndACC = 217, pdACC = 5
6, D(CS/ BaselinedACC) = 2.54; Figure 1F). We observed an increase in the theta power throughout conditioning (Figure 2A and inset, one-way ANOVA, Famygdala = 4.54, dfamygdala = 4, pamygdala = 0.005, FdACC = 4.8, dfdACC = 4, pdACC = 0.007), with a difference between habituation and conditioning trials 11–20 (post hoc t test, p < 0.001). Taken together, increased power of theta oscillations and the pattern of spiking discharge suggest a role for theta activity in the primate amygdala and dACC.

INTRODUCTION Neuronal oscillations are considered a mechanism that synchronizes spiking activity across brain regions and can enhance information transfer (Buzsa´ki et al., 2012; Deffains et al., 2016). In rodents, much evidence suggests that oscillations at the theta range (4–8 Hz) support amygdala-prefrontal interactions and contribute to expression (Courtin et al., 2014), discrimination (Likhtik et al., 2014), and consolidation (Popa et al., 2010) of fear memories. In primates, spiking synchrony underlies persistent aversive memories (Livneh and Paz, 2012) and successful updating between fear and safety (Klavir et al., 2013). Yet, it remains unclear whether neural oscillations, specifically at the theta range, contribute to spiking synchrony in the primate amygdala-prefrontal pathway and, in turn, to successful formation of aversive memories. Moreover, the dynamics of such mechanism during learning and what kind of information the synchrony carries are unclear. We therefore examined the emergence of theta synchrony and resetting of theta phase in the amygdala and anterior cingulate cortex (dACC) of monkeys while they acquired novel aversive tone-odor associations on a daily basis. We further tested the relation between theta phase synchrony and single-unit activity in both regions and the directionality of information transfer during learning between the two regions. We provide here first evidence for the role of theta synchrony in the primate amygdala- prefrontal cortex circuitry and its dynamical role during aversive learning.

Tone-Odor Conditioning Induces Phase Resetting at the Theta Range Previous studies in rodents suggest that theta phase resetting can be used to organize local spiking activity and drive fear expression (Courtin et al., 2014; Karalis et al., 2016; Likhtik et al., 2014). In the primate, we observed that phase resetting develops in response to the tone during conditioning (Figures 2B– 2F). We found increased synchronization and decreased halfwidth theta phase distribution in the amygdala (n = 42, one-way ANOVA, Fsynchronization = 16.5, dfsynchronization = 4, psynchronization = 3
11, Fhalf-width = 19.18, dfhalf-width = 4, phalf-width = 2
13), as well as in the dACC (n = 57, Fsynchronization = 22.7, dfsynchronization = 4, psynchronization = 2
15, Fhalf-width = 13.92, dfhalf-width = 4, phalf-width = 2
10), with specific changes between habituation and conditioning trials 11–20 (post hoc t test, psynchronization < 0.03 and phalf-width < 0.001, Bonferroni corrected). In comparison, the same measure was not significant during the inter-trial interval (pITI amygdala > 0.5, pITI dACC > 0.5). Neuron 97, 1–8, January 17, 2018 ª 2017 Elsevier Inc. 1

Please cite this article in press as: Taub et al., Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning, Neuron (2017), https://doi.org/10.1016/j.neuron.2017.11.042

A

D

B

E

C

F

Figure 1. Increased Theta Power in Response to Auditory Stimulus (A) Paired conditioning trials (green line): real-time detection of inhale elicited the presentation of a conditioned stimulus (CS, a pure tone) that predicted the release of unconditioned stimulus at the onset of the next inhale (US, an aversive odor, left and right gray rectangles mark the durations of CS and US, respectively). Aversive odors resulted in decreased inhale (UR) and increased inhale to the tone (CR). During habituation and extinction trials (blue line), the CS was not followed by a US. Shown are real data taken from one trial. y axis bar is 0.1 pressure at nose (AU). (B) Following conditioning trials (green line), mean size of the CR increased compared to both habituation (blue) and extinction (maroon), with the latter two not significantly different from each other. (C) Increased CRs, measured as area under the curve during 350 ms post-CS compared with mean inhale volume during pre-CS time (i.e., taken during the ITI), were observed during acquisition trials, with return to baseline levels during extinction trials. The red line is a fit for presentation purposes only. Inset shows individual session learning curves from three different days that are >7 sessions apart and normalized to previous-day extinction. All data are presented using a running window of 6 trials with overlap of 4 trials, separately for habituation, conditioning, and extinction. Data are represented as mean ± SEM. (D) An example of power spectral density analysis indicating an increase in theta power in response to the CS. Data are presented as log/log power with white band (non-shaded) highlighting the theta band. (E) Theta power increases in response to the CS, mean over all sessions and electrodes (horizontal dotted red lines mark theta band range). Right plots show statistics of the increase in power during the CS in comparison to pre-CS activity. Horizontal lines indicate theta band, and vertical line indicates significance level of <0.05. (F) Example of autocorrelation of single units recorded from the amygdala (top) and dACC (bottom) showing an increase in power between 4 and 8 Hz. Insets in log/log; mean power in theta band is presented for all units with white non-shaded background for during CS (solid red) and during inter-trial interval baseline activity (solid blue). Blue dashed line indicates 20 Hz.

Conditioning Increases Amygdala-dACC Theta Synchrony Next, to examine functional connectivity between the amygdala and the dACC during conditioning trials, we computed and found an increase in the correlation of theta power across the two structures (Figure 2F; Spearman’s correlation, n = 129, rhabituation = 0.14, phabituation = 0.1, rconditioning 1–10 =

2 Neuron 97, 1–8, January 17, 2018

0.36, pconditioning 1–10 = 0.0001), with an increase from habituation to conditioning trials 1–10 and 11–20 (Figure 2F, insets, p < 0.05 for both). The increased functional connectivity hypothesis was further supported by increased phase synchronization between the amygdala and dACC. Specifically, synchronization was calculated as a function of the lag (Figure 2G; two-way ANOVA, for Fconditioning stage = 14.22,

Please cite this article in press as: Taub et al., Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning, Neuron (2017), https://doi.org/10.1016/j.neuron.2017.11.042

A

F

C

B

G

D

H

E

I

Figure 2. Conditioning Induces Increased Theta Power and Phase Resetting (A) Mean theta power (±SEM) during 1,000 ms following CS onset increased significantly during conditioning. Inset illustrates mean theta power during 1,000 ms before and following CS onset along a daily session of habituation, conditioning, and extinction. (B and D) Phase reset developed during conditioning in both the amygdala (B) and the dACC (D). On the right, the distribution and mean (red line) of theta phases across electrodes and days of recordings are shown. (C and E) Phase reset was characterized by increased MRL score and decreased half-width theta phase distribution in the amygdala (C) and dACC (E), with some change in MRL score, but not half-width phase distribution, during pre-CS period (insets). Data are represented as mean ± SEM. (F) Within-trial correlation in theta power between the amygdala and the dACC increases during initial stages of conditioning and then decreases during the last ten trials of conditioning (not different than habituation and extinction, top inset). Bottom inset shows the change in correlation values along conditioning. (G) Phase synchrony as a function of lag between the two structures. Red line indicates the mean MRL score ± SEM and time of maximal MRL. (H) Imaginary coherence reveals specificity of conditioning-dependent increase in LFP power to the theta band and to trials 1–20. Data are represented as mean ± SEM. (I) Mean cross-correlation (±SEM) from 1,000 ms post-CS (red line indicates the mean, signed rank test of difference from zero, p < 0.0001). Histograms on right show cross-correlation peaks. The amygdala leads (
7.7 ms, p < 1
10, signed rank test). *p < 0.05, **p < 0.01, ***p < 0.001.

dfconditioning stage = 4, pconditioning stage = 3
9, Flag = 5.02, dflag = 40, plag = 8
14), with interaction of stage by lag (F = 1.37, df = 160, p = 0.001), but not during pre-CS activity (two-way ANOVA, for Fconditioning stage = 0.96, dfconditioning stage = 4, pconditioning stage = 0.43). To examine the band specificity of the phase synchronization, we used imaginary coherence analysis and found a specific increase in theta coherence during conditioning (Figure 2H; two-way ANOVA, Fband = 7.48, dfband = 21, pband = 0.001, Fstage = 7.48, dfstage = 3, pstage = 0.0001), with band by stage interaction effect (F = 1.41, df = 63, p = 0.02), and specific increase during conditioning trials 1
10 and 11–20 (post hoc, p < 0.05). No differences were observed in imaginary coherence during pre-CS period (twoway ANOVA, Fband = 1.06, dfband = 21, pband = 0.39). Taken together, the results suggest that theta oscillations can be used as a mean of long-range communication between the amygdala and the dACC during conditioning in the primate. Finally, the lag and the negative peak of cross-correlation (Figure 2I) suggest a directional flow of information from the

amygdala to the dACC. A Granger causality analysis (Cui et al., 2008; Ghosh et al., 2013) further supported this directionality assumption, showing a directional flow of information from the amygdala to the dACC (paired t test, p < 0.05; see STAR Methods). Given the above directionality findings, it is unlikely that amygdala-dACC synchrony was due to volume conduction alone. This possibility was further refuted by volume conduction analyses, where volume conductions of the inter-area local field potential (LFP) signal were computed using the phase lag index (PLI, values from 0 to 1, indicating deviation of the phase distribution from 0 or ±p) (Stam et al., 2007). This revealed high PLI values for all conditioning stages (>0.26; p < 0.001; compared to zero using two-tailed t test). Phase Locking Develops between Amygdala Spikes and dACC-Theta To further examine the directionality of amygdala-dACC interactions, we quantified the synchrony between the output of one

Neuron 97, 1–8, January 17, 2018 3

Please cite this article in press as: Taub et al., Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning, Neuron (2017), https://doi.org/10.1016/j.neuron.2017.11.042

A

B

C

D

Figure 3. Phase Locking Depends on the Stage of Conditioning (A) A raster plot (top) and PSTH (bottom) of amygdala spikes overlaid with mean theta (±SEM) activity from dACC (purple). With conditioning, dACC theta became more phase locked to amygdala spikes. (B) The proportion of phase-locked amygdala units increased by 35% while that of the shuffled data and inter-trial interval baseline activity did not change significantly. The solid line depicts the percent of significant amygdala using a sliding window of five trials. In contrast, the proportion of dACC phase-locked units decreased during conditioning by 32% (inset). (C) Synchrony score (MRL) increased during conditioning. MRL score computed in pairs of electrodes that exhibited non-significant synchrony during both habituation and extinction stages (n = 33, p = 0.002) or habituation only (n = 28, p < 0.001). Purple solid line shows the change in MRL score in habituation-only units using a sliding window of five trials. Data are represented as mean ± SEM. (D) An example of spike-triggered averaging (STA; ±SEM) of theta activity averaged on spikes (arrow indicates spike time) taken from the 500 ms post-CS and the averaged Z score separately for the different learning phases (one-way ANOVA for learning stage, F = 7.84, df = 4, p = 3
6, post hoc analysis results are displayed as asterisks). *p < 0.05, **p < 0.01, ***p < 0.001.

region—measured as spiking activity and the input in the other region—approximated by the theta oscillations (Buzsa´ki et al., 2012) (Figures 3A–3D). This analysis revealed a phase locking between amygdala spikes and dACC oscillation, but not vice versa (Figure 3B). Specifically, the proportion of amygdala units that were locked to theta oscillations in the dACC increased during conditioning by about 35% (n = 234, from 11.25% to 17.4%), while there was no increase in synchronization during the intertrial interval or for the shuffled data (both < 0.6% significant cells throughout conditioning stages; Figure 3B). In these pairs of electrodes, the synchronization score increased during conditioning compared to habituation (Figure 3C; n = 28, one-way ANOVA, F = 6.81, df = 4, p < 0.001), with increased synchronization during trials 11–20 and 21–30 (post hoc p = 0.001 and p = 0.002, respectively). We further observed decreased half-width spike- theta phase distribution (n = 1,412, one-way ANOVA, F = 3.16, df = 4, p = 0.009), with specific changes between habituation and conditioning trials 11–20 (post hoc t test, p = 0.04). A spike-triggered LFP analysis revealed similar results (Figure 3D).

4 Neuron 97, 1–8, January 17, 2018

In contrast, the proportion of dACC units that were locked to theta oscillations in the amygdala decreased by about 32% (Figure 3B, inset, n = 289, from 12.32% to 8.33%), and synchronization score did not increase significantly compared to habituation levels. Phase Locking Is Inversely Correlated with Performance The amygdala-dACC phase synchrony increased during initial stages of conditioning and started to decline during later trials when behavior reached plateau and CS-US associations have already been formed (Figure 4A). We correlated the mean synchrony score in pairs of electrodes during the first 20 trials of acquisition with the conditioned response (CR) during these trials and found a positive relationship (Figure 4B; Spearman’s correlation, n = 65, r = 0.5, p = 6
5), suggesting that inter-area synchrony contributes to the formation of the associations. To examine whether the synchrony contributes more to the acquisition of CRs or to their maintenance, we correlated the day-byday number of trials to peak in synchrony with the number of

Please cite this article in press as: Taub et al., Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning, Neuron (2017), https://doi.org/10.1016/j.neuron.2017.11.042

A

B

C

Figure 4. Amygdala-dACC Synchronization Learning (A) Behavioral learning curve (blue dashed line), the percent of significant units (red dashed line), and half-width of spike-phase distributions (green dashed line). Solid lines represent exponent fit (Learning: r2 = 0.86, RMSE = 0.04; Synchrony: r2 = 0.83, RMSE = 0.11; half-width: r2 = 0.84, RMSE = 0.001). (B) Co-variation of synchrony magnitude and behavioral performance. (C) Left: number of trials to peak performance and to peak synchrony per session. Right: number of trials to peak performance and to minimal phase variance per session.

trials to behavioral plateau and the day-by-day number of trials to minimal variance in phase distribution with the number of trials to behavioral plateau. The positive correlation with the former and the negative correlation with the latter (Figure 4C) provide evidence that the cross-regional synchrony is related to the error signal and CR acquisition (McMains and Kastner, 2011; Rutishauser et al., 2010) (Spearman’s correlation, nbehavioral plateau = 65, r = 0.39, p = 0.003, median trial-to-peak behavior = 11.1 ± 1.07 SE, median-to-peak mean resultant length [MRL] = 14 ± 0.96, nvariance = 65, r =
0.33, p = 0.006, median trial-to-peak behavior = 11.1 ± 1.07 SE, median-to-min phase distribution = 15 ± 0.97). Finally, we calculated the slope of learning and the slope of synchrony during the first 20 trials of acquisition and found a ratio of 0.904 ± 0.097 (ndays = 32), suggesting that the synchrony slightly lags behavior and further supporting the error signal conclusion.

shown to facilitate neuronal responses to sensory stimuli in the primate (Lakatos et al., 2007) and suggested to optimize the processing of incoming information (Jutras and Buffalo, 2010), but the functional role during learning was demonstrated only in cortico-cortical circuits and not in amygdala-prefrontal circuits. Moreover, whereas in cortico-cortical circuits the synchrony was associated only with the memory strength, here we provide evidence that it is associated also with the acquisition of aversive associations and the representation of error signal and its dynamics (Li et al., 2011). The conditioning-dependent phase resetting in the amygdala and dACC and phase locking between amygdala spikes and theta oscillations in the dACC extend the notion that memory formation relies on amygdala inputs to prefrontal regions (Burgos-Robles et al., 2017; Klavir et al., 2017; Senn et al., 2014; Sotres-Bayon et al., 2012) and on theta synchronization (Rutishauser et al., 2010, 2015) and thus provide evidence for such a circuit-level mechanism in the primate.

DISCUSSION In the present study, we show for the first time the importance of theta oscillations in the primate amygdala-dACC circuit during aversive learning. We found within-regional phase reset and further inter-regional phase synchronization, where the phase resetting in the amygdala was followed by a later reset in the dACC. The oscillatory phase synchronization was accompanied by phase locking of dACC theta oscillations to amygdala spikes, but not vice versa. These findings expand on previous results in the rodent amygdala-prefrontal circuit (Adhikari et al., 2010; Berry and Seager, 2001; Courtin et al., 2014; Karalis et al., 2016; Likhtik et al., 2014; Popa et al., 2010) and on corticocortical communication findings in primates (Jutras and Buffalo, 2010; Liebe et al., 2012; Rutishauser et al., 2010). Together, our results point to a specific circuit and oscillatory mechanism in the primate brain that contributes to learning and formation of aversive memories. The findings support a local and circuit-level role for theta synchronization in the primate brain (Lee et al., 2005; Paz et al., 2008; Seidenbecher et al., 2003). Theta phase resetting was

Development of Theta Phase Resetting in the Amygdala and dACC during Aversive Conditioning Theta oscillations are associated with active processing of information and stimulus-related information (Buzsa´ki, 2002; Jones and Wilson, 2005; Jutras et al., 2013), and several recent rodent studies suggested a functional role for oscillatory phase synchronization in spatial behavior, motor, and emotional conditioning and memory consolidation (Berry and Seager, 2001; Buzsa´ki, 2002; Karalis et al., 2016; Likhtik et al., 2014; Popa et al., 2010). In the amygdala and medial prefrontal cortex of rodents, theta power and synchrony were shown to increase in response to tone CS and just ahead of the expected unconditioned aversive stimulus (Adhikari et al., 2010; Courtin et al., 2014; Pare´ and Collins, 2000; Paz et al., 2008; Seidenbecher et al., 2003). In line with these findings, we found increased theta power and a development of theta oscillation phase reset in the amygdala and dACC of primates during aversive conditioning. Taken together, the cross-species phase resetting suggests a functional mechanism that contributes to learning and a general phenomenon in fear learning that is robust across circuits.

Neuron 97, 1–8, January 17, 2018 5

Please cite this article in press as: Taub et al., Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning, Neuron (2017), https://doi.org/10.1016/j.neuron.2017.11.042

Inter-area Synchrony Increases during Conditioning Accumulating evidence suggests that adequate aversive conditioning requires coordinated activity among groups of distant neurons. In line with this suggestion, hippocampal-medial prefrontal cortex synchrony was positively correlated with anxiety behavior in mice (Adhikari et al., 2010), and increased amygdala-prefrontal theta synchrony was reported in response to the CS+, but not to the CS
, in another rodent study (Likhtik et al., 2014). Here, we observed increased theta synchrony between the amygdala and dACC of primates that emerges during fear conditioning. Similar to previous findings in the oscillation domain in rodents (Likhtik et al., 2014) and spike domain in primates (Klavir et al., 2013), we found a clear amygdala to dACC directionality in theta oscillatory activity in response to the CS. This finding is also in line with the hippocampus-to-prefrontal direction of information transfer reported previously in rodents (Jones and Wilson, 2005; Siapas et al., 2005) and suggests a similar mechanism in the amygdala-to-prefrontal circuit in primates. The temporal lag between theta oscillations in the amygdala and that of the dACC was at the range of 10–20 ms in both measures used in this study (the cross-correlation and MRL). Such a delay is likely to imply inter-area communication mechanism rather than a shared common input (Fries, 2005; Liebe et al., 2012). Taken together, the increased inter-area oscillatory phase synchronization can serve as a mean for reducing trial-by-trial variability and facilitating inter-area communication. Spike-LFP Synchronization It was recently suggested that theta oscillations are used to organize and synchronize inter-area spiking activity (Benchenane et al., 2010; Courtin et al., 2014). Our results indicate this to be the case in the primate brain as well, due to the finding that amygdala spikes are more phase locked to dACC theta oscillations than vice versa (i.e., dACC spikes to amygdala oscillations). Our result is also consistent with visual (v4)-prefrontal cortex spike-LFP phase locking reported in primates (Liebe et al., 2012) and with the amygdala to dACC directionality of spiking activity we reported in a previous study (Klavir et al., 2013), but unlike a recent work in rodents (Likhtik et al., 2014). Such an amygdala-to-dACC mechanism may imply the amygdala output to be more effective in driving oscillatory activity and using it to synchronize activity in the dACC. We previously showed that in primates, spiking synchrony in the amygdala-prefrontal network underlies persistent aversive memories (Livneh and Paz, 2012), and in line with this, we suggest that presynaptic amygdala output can serve to drive postsynaptic oscillations in the dACC and increase the probability for spike in the dACC. Such a well-timed synchronized activity can facilitate learning and memory through decreased trial-by-trial variability (Darling et al., 2011; Taub et al., 2014) and increased postsynaptic excitability (Sehgal et al., 2014; Yiu et al., 2014). Synchrony as a Mechanism to Transfer Error Signal during Learning We found that synchrony between the amygdala and the dACC increased during the initial stages of conditioning, and together

6 Neuron 97, 1–8, January 17, 2018

with the positive relationship found between synchrony and performance on a daily basis, it suggests that inter-area synchronization contributes to development of associations and memory formation. This finding is in line with previous findings in non-human and human primates (Liebe et al., 2012; Rutishauser et al., 2010). Yet, whereas such mechanisms were documented only in cortico-cortical circuits (Liebe et al., 2012), our work demonstrates that this mechanism underlies similar function in subcortical-cortical circuits for aversive learning. Interestingly, we found that synchrony started to decline after behavior reached a plateau, yet when paired, CS-US trials are still presented. Indeed, there was a tight correlation between the time behavior reached plateau in an individual day and the time of maximal synchrony during that day (or the time of minimal phase variance). Additionally, we found that synchrony lagged behavior during learning, but only slightly. Taken together, these findings point to an error signal that is large during the early stages of learning and diminishes as the association is formed, as suggested by many models of learning theory (Rescorla and Wagner, 1972; Roesch et al., 2012). This signal is mediated, at least in part, by oscillations in the primate amygdala-prefrontal circuit (Costa et al., 2016) and implies that synchrony in the amygdala-dACC circuit contributes more to acquisition of new behaviors and less for maintenance (Li et al., 2011; Roesch et al., 2010; Rudebeck et al., 2013). Finally, our findings lend more support to the notion that learning and memory formation rely on spike-theta synchronization (Rutishauser et al., 2010, 2015). In conclusion, we show that theta oscillations serve as a neuronal mechanism for long-range communication and directional information transfer between subcortical and cortical structures during emotional learning in the primate. This mechanism can further be used to suggest a specific target for intervention in human psychopathologies (Averbeck and Chafee, 2016; Etkin et al., 2015; Salzman and Fusi, 2010). STAR+METHODS Detailed methods are provided in the online version of this paper and include the following: d d d d

KEY RESOURCES TABLE CONTACT FOR REAGENT AND RESOURCE SHARING EXPERIMENTAL MODEL AND SUBJECT DETAILS B Animals METHOD DETAILS B Recordings B Behavior B Data Analysis

ACKNOWLEDGMENTS We are grateful to Dr. Uri Livneh for experimental design and data collection. We thank Yossi Shohat for major contribution to the work and welfare of the animals and Dr. Edna Furman-Haran and Nachum Stern for MRI procedures. This work was supported by ISF #26613 and ERC-FP7-StG #281171 grants to R. Paz.

Please cite this article in press as: Taub et al., Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning, Neuron (2017), https://doi.org/10.1016/j.neuron.2017.11.042

AUTHOR CONTRIBUTIONS A.H.T. and R. Paz designed and performed the study. E.K. contributed to experiments. A.H.T. analyzed the data. R. Perets contributed to data analyses. A.H.T. and R. Paz wrote the manuscript.

Deffains, M., Iskhakova, L., Katabi, S., Haber, S.N., Israel, Z., and Bergman, H. (2016). Subthalamic, not striatal, activity correlates with basal ganglia downstream activity in normal and parkinsonian monkeys. eLife 5, 5. €chel, C., and Gross, J.J. (2015). The neural bases of emotion reguEtkin, A., Bu lation. Nat. Rev. Neurosci. 16, 693–700.

DECLARATION OF INTERESTS

Fries, P. (2005). A mechanism for cognitive dynamics: neuronal communication through neuronal coherence. Trends Cogn. Sci. 9, 474–480.

The authors declare no competing interests.

Geweke, J. (1982). Measurement of linear dependence and feedback between multiple time series. J. Am. Stat. Assoc. 77, 304–313.

Received: August 29, 2017 Revised: October 31, 2017 Accepted: November 30, 2017 Published: December 28, 2017 REFERENCES Adhikari, A., Topiwala, M.A., and Gordon, J.A. (2010). Synchronized activity between the ventral hippocampus and the medial prefrontal cortex during anxiety. Neuron 65, 257–269. Averbeck, B.B., and Chafee, M.V. (2016). Using model systems to understand errant plasticity mechanisms in psychiatric disorders. Nat. Neurosci. 19, 1418–1425. Barnett, L., and Seth, A.K. (2014). The MVGC multivariate Granger causality toolbox: a new approach to Granger-causal inference. J. Neurosci. Methods 223, 50–68. Benchenane, K., Peyrache, A., Khamassi, M., Tierney, P.L., Gioanni, Y., Battaglia, F.P., and Wiener, S.I. (2010). Coherent theta oscillations and reorganization of spike timing in the hippocampal- prefrontal network upon learning. Neuron 66, 921–936. Berens, P. (2009). CircStat: a MATLAB toolbox for circular statistics. J. Stat. Softw. 31, 1–21. Berry, S.D., and Seager, M.A. (2001). Hippocampal theta oscillations and classical conditioning. Neurobiol. Learn. Mem. 76, 298–313. Brovelli, A., Ding, M., Ledberg, A., Chen, Y., Nakamura, R., and Bressler, S.L. (2004). Beta oscillations in a large-scale sensorimotor cortical network: directional influences revealed by Granger causality. Proc. Natl. Acad. Sci. USA 101, 9849–9854. Burgos-Robles, A., Kimchi, E.Y., Izadmehr, E.M., Porzenheim, M.J., RamosGuasp, W.A., Nieh, E.H., Felix-Ortiz, A.C., Namburi, P., Leppla, C.A., Presbrey, K.N., et al. (2017). Amygdala inputs to prefrontal cortex guide behavior amid conflicting cues of reward and punishment. Nat. Neurosci. 20, 824–835. Buzsa´ki, G. (2002). Theta oscillations in the hippocampus. Neuron 33, 325–340. Buzsa´ki, G., Anastassiou, C.A., and Koch, C. (2012). The origin of extracellular fields and currents–EEG, ECoG, LFP and spikes. Nat. Rev. Neurosci. 13, 407–420. Costa, V.D., Dal Monte, O., Lucas, D.R., Murray, E.A., and Averbeck, B.B. (2016). Amygdala and ventral striatum make distinct contributions to reinforcement learning. Neuron 92, 505–517. Courtin, J., Chaudun, F., Rozeske, R.R., Karalis, N., Gonzalez-Campo, C., Wurtz, H., Abdi, A., Baufreton, J., Bienvenu, T.C.M., and Herry, C. (2014). Prefrontal parvalbumin interneurons shape neuronal activity to drive fear expression. Nature 505, 92–96. Cui, J., Xu, L., Bressler, S.L., Ding, M., and Liang, H. (2008). BSMART: a Matlab/C toolbox for analysis of multichannel neural time series. Neural Netw. 21, 1094–1104.

Ghosh, S., Laxmi, T.R., and Chattarji, S. (2013). Functional connectivity from the amygdala to the hippocampus grows stronger after stress. J. Neurosci. 33, 7234–7244. Granger, C.W.J. (1969). Investigating causal relations by econometric models and cross-spectral methods. Econometrica 37, 424–438. Gregoriou, G.G., Gotts, S.J., Zhou, H., and Desimone, R. (2009). High-frequency, long-range coupling between prefrontal and visual cortex during attention. Science 324, 1207–1210. Jones, M.W., and Wilson, M.A. (2005). Theta rhythms coordinate hippocampal-prefrontal interactions in a spatial memory task. PLoS Biol. 3, e402. Jutras, M.J., and Buffalo, E.A. (2010). Synchronous neural activity and memory formation. Curr. Opin. Neurobiol. 20, 150–155. Jutras, M.J., Fries, P., and Buffalo, E.A. (2013). Oscillatory activity in the monkey hippocampus during visual exploration and memory formation. Proc. Natl. Acad. Sci. USA 110, 13144–13149. Karalis, N., Dejean, C., Chaudun, F., Khoder, S., Rozeske, R.R., Wurtz, H., Bagur, S., Benchenane, K., Sirota, A., Courtin, J., and Herry, C. (2016). 4-Hz oscillations synchronize prefrontal-amygdala circuits during fear behavior. Nat. Neurosci. 19, 605–612. Klavir, O., Genud-Gabai, R., and Paz, R. (2013). Functional connectivity between amygdala and cingulate cortex for adaptive aversive learning. Neuron 80, 1290–1300. Klavir, O., Prigge, M., Sarel, A., Paz, R., and Yizhar, O. (2017). Manipulating fear associations via optogenetic modulation of amygdala inputs to prefrontal cortex. Nat. Neurosci. 20, 836–844. Lachaux, J.P., Rodriguez, E., Martinerie, J., and Varela, F.J. (1999). Measuring phase synchrony in brain signals. Hum. Brain Mapp. 8, 194–208. Lakatos, P., Chen, C.M., O’Connell, M.N., Mills, A., and Schroeder, C.E. (2007). Neuronal oscillations and multisensory interaction in primary auditory cortex. Neuron 53, 279–292. Lee, H., Simpson, G.V., Logothetis, N.K., and Rainer, G. (2005). Phase locking of single neuron activity to theta oscillations during working memory in monkey extrastriate visual cortex. Neuron 45, 147–156. Li, J., Schiller, D., Schoenbaum, G., Phelps, E.A., and Daw, N.D. (2011). Differential roles of human striatum and amygdala in associative learning. Nat. Neurosci. 14, 1250–1252. Liebe, S., Hoerzer, G.M., Logothetis, N.K., and Rainer, G. (2012). Theta coupling between V4 and prefrontal cortex predicts visual short-term memory performance. Nat. Neurosci. 15, 456–462, S1–S2. Likhtik, E., Stujenske, J.M., Topiwala, M.A., Harris, A.Z., and Gordon, J.A. (2014). Prefrontal entrainment of amygdala activity signals safety in learned fear and innate anxiety. Nat. Neurosci. 17, 106–113. Livneh, U., and Paz, R. (2012). Amygdala-prefrontal synchronization underlies resistance to extinction of aversive memories. Neuron 75, 133–142. McMains, S., and Kastner, S. (2011). Interactions of top-down and bottom-up mechanisms in human visual cortex. J. Neurosci. 31, 587–597.

Daffertshofer, A., and Stam, C.J. (2007). Influences of volume conduction on phase distributions. Int. Congr. Ser. 1300, 209–212.

Nolte, G., Bai, O., Wheaton, L., Mari, Z., Vorbach, S., and Hallett, M. (2004). Identifying true brain interaction from EEG data using the imaginary part of coherency. Clin. Neurophysiol. 115, 2292–2307.

Darling, R.D., Takatsuki, K., Griffin, A.L., and Berry, S.D. (2011). Eyeblink conditioning contingent on hippocampal theta enhances hippocampal and medial prefrontal responses. J. Neurophysiol. 105, 2213–2224.

Pare´, D., and Collins, D.R. (2000). Neuronal correlates of fear in the lateral amygdala: multiple extracellular recordings in conscious cats. J. Neurosci. 20, 2701–2710.

Neuron 97, 1–8, January 17, 2018 7

Please cite this article in press as: Taub et al., Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning, Neuron (2017), https://doi.org/10.1016/j.neuron.2017.11.042

Paz, R., Bauer, E.P., and Pare´, D. (2008). Theta synchronizes the activity of medial prefrontal neurons during learning. Learn. Mem. 15, 524–531. Popa, D., Duvarci, S., Popescu, A.T., Le´na, C., and Pare´, D. (2010). Coherent amygdalocortical theta promotes fear memory consolidation during paradoxical sleep. Proc. Natl. Acad. Sci. USA 107, 6516–6519. Rescorla, R.A., and Wagner, A.R. (1972). A theory of Pavlovian conditioning: variations in the effectiveness of reinforcement and nonreinforcement. In Classical Conditioning II: Current Research and Theory, A.H. Black and W.F. Prokasy, eds. (Appleton-Century-Crofts), pp. 64–99. Roesch, M.R., Calu, D.J., Esber, G.R., and Schoenbaum, G. (2010). Neural correlates of variations in event processing during learning in basolateral amygdala. J. Neurosci. 30, 2464–2471. Roesch, M.R., Esber, G.R., Li, J., Daw, N.D., and Schoenbaum, G. (2012). Surprise! Neural correlates of Pearce-Hall and Rescorla-Wagner coexist within the brain. Eur. J. Neurosci. 35, 1190–1200. Rudebeck, P.H., Mitz, A.R., Chacko, R.V., and Murray, E.A. (2013). Effects of amygdala lesions on reward-value coding in orbital and medial prefrontal cortex. Neuron 80, 1519–1531. Rutishauser, U., Ross, I.B., Mamelak, A.N., and Schuman, E.M. (2010). Human memory strength is predicted by theta-frequency phase-locking of single neurons. Nature 464, 903–907. Rutishauser, U., Mamelak, A.N., and Adolphs, R. (2015). The primate amygdala in social perception - insights from electrophysiological recordings and stimulation. Trends Neurosci. 38, 295–306. Salzman, C.D., and Fusi, S. (2010). Emotion, cognition, and mental state representation in amygdala and prefrontal cortex. Annu. Rev. Neurosci. 33, 173–202.

8 Neuron 97, 1–8, January 17, 2018

Sehgal, M., Ehlers, V.L., and Moyer, J.R., Jr. (2014). Learning enhances intrinsic excitability in a subset of lateral amygdala neurons. Learn. Mem. 21, 161–170. Seidenbecher, T., Laxmi, T.R., Stork, O., and Pape, H.-C. (2003). Amygdalar and hippocampal theta rhythm synchronization during fear memory retrieval. Science 301, 846–850. €ndemann, J., Senn, V., Wolff, S.B.E., Herry, C., Grenier, F., Ehrlich, I., Gru €ller, C., Letzkus, J.J., and Lu €thi, A. (2014). Long-range connecFadok, J.P., Mu tivity defines behavioral specificity of amygdala neurons. Neuron 81, 428–437. Siapas, A.G., Lubenov, E.V., and Wilson, M.A. (2005). Prefrontal phase locking to hippocampal theta oscillations. Neuron 46, 141–151. Sotres-Bayon, F., Sierra-Mercado, D., Pardilla-Delgado, E., and Quirk, G.J. (2012). Gating of fear in prelimbic cortex by hippocampal and amygdala inputs. Neuron 76, 804–812. Stam, C.J., Nolte, G., and Daffertshofer, A. (2007). Phase lag index: assessment of functional connectivity from multi channel EEG and MEG with diminished bias from common sources. Hum. Brain Mapp. 28, 1178–1193. Stujenske, J.M., Likhtik, E., Topiwala, M.A., and Gordon, J.A. (2014). Fear and safety engage competing patterns of theta-gamma coupling in the basolateral amygdala. Neuron 83, 919–933. Taub, A.H., Segalis, E., Marcus-Kalish, M., and Mintz, M. (2014). Acceleration of cerebellar conditioning through improved detection of its sensory input. Brain Comput. Interfaces (Abingdon) 1, 5–16. Yiu, A.P., Mercaldo, V., Yan, C., Richards, B., Rashid, A.J., Hsiang, H.-L.L., Pressey, J., Mahadevan, V., Tran, M.M., Kushner, S.A., et al. (2014). Neurons are recruited to a memory trace based on relative neuronal excitability immediately before training. Neuron 83, 722–735.

Please cite this article in press as: Taub et al., Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning, Neuron (2017), https://doi.org/10.1016/j.neuron.2017.11.042

STAR+METHODS KEY RESOURCES TABLE

REAGENT or RESOURCE

SOURCE

IDENTIFIER

Chemicals, Peptides, and Recombinant Proteins Propionic acid

Sigma-Aldrich, Israel

Cat# P1386; CAS ID: 79-09-4

Mineral oil

Sigma-Aldrich, Israel

Cat# M5904-500ML; CAS ID: 8042-47-5

B.F.C. Monkey Breeding Farm, Israel

N/A

Experimental Models: Organisms/Strains Adult male macaca fascicularis Software and Algorithms MATLAB (v.R2016b)

MathWorks

RRID: SCR_001622

Circular Statistics Toolbox

Berens, (2009)

N/A

MVGC Multivariate Granger Causality MATLAB Toolbox

http://users.sussex.ac.uk/ lionelb/MVGC/

RRID: SCR_015755

Alpha Omega; We-Sense

N/A

Other Glass/Narylene-coated tungsten microelectrodes

CONTACT FOR REAGENT AND RESOURCE SHARING Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Aryeh Taub ([email protected]). EXPERIMENTAL MODEL AND SUBJECT DETAILS Animals Two adult male macaca fascicularis (4–7 kg) were implanted with a recording chamber above the right amygdala and dACC. All surgical, experimental procedures and housing in the non-human primate colony in the Weizmann Institute were approved and conducted in accordance with the regulations of the Weizmann Institute Animal Care and Use Committee (IACUC), following NIH regulations and with AAALAC accreditation. Anatomical magnetic resonance imaging (MRI) scans with positioning electrodes were acquired before, during and after the recording period using a 3-T MRI scanner (MAGNETOM Trio, Siemens). Scans were used to determine the depth and location of the regions. METHOD DETAILS Recordings Recording locations were confirmed by MRI scans with calibrating electrodes performed before, during, and after recordings, and overlapped the regions shown in Klavir et al. (2013) and Livneh and Paz (2012). Based on known extracellular properties and anatomy when lowering the electrodes, and based on the calibration electrodes, we can reliably estimate that > 90% of our neurons were recorded from the BLA (homogenously), but we cannot rule out that few neurons came from the Ce. Each day, three to six microelectrodes (0.6–1.2 MU glass/Narylene-coated tungsten, Alpha Omega or We-Sense) were lowered through a metal guide (Gauge 25xxtw, OD: 0.51 mm, ID: 0.41 mm, Cadence) into the amygdala and the dACC according to electrophysiological and anatomical markers using a head tower and electrode-positioning system (Alpha Omega). Electrode signals were pre-amplified, 0.3–6 kHz band-pass filtered, and sampled at 25 kHz, and online spike sorting was performed using a template-based algorithm (Alpha Lab Pro, Alpha Omega). At the end of the recording period, offline spike sorting was performed for all sessions to improve unit isolation (offline sorter, Plexon). For LFP signal, signals were pre-amplified, 0.5 – 125 Hz band-pass filtered, and sampled at 781.25 Hz. Behavior Each daily experimental session consisted of an habituation phase of ten presentations of a pure tone CS (randomly chosen from 1,000–2,400 Hz, with minimal gap of 500 Hz. between consecutive days) that was followed by an acquisition phase of 30 trials of paired CS-US, where the US is an aversive odor (propionic acid diluted in mineral oil; Sigma-Aldrich). CS where q was triggered by real-time detection of inhale onset, and odor (US) was released at the following inhale onset (but not before 1000 ms elapsed). Twenty unpaired CSs were used in order to extinguish the acquired CS - US association. Extinction data was analyzed for the

Neuron 97, 1–8.e1–e3, January 17, 2018 e1

Please cite this article in press as: Taub et al., Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning, Neuron (2017), https://doi.org/10.1016/j.neuron.2017.11.042

last ten trials in order to ensure full extinction of behavioral and neuronal CRs (Livneh and Paz, 2012). Behavioral responses were computed as the area under the curve 0–350 ms. from inhale onset and normalized by the habituation inhale volume. Data Analysis Data were analyzed using a combination of custom-written MATLAB scripts (MathWorks, MA) and Circular Statistics Toolbox (Berens, 2009). To evaluate LFP power across different frequencies we looked at spectrograms by obtaining the multitaper power spectral density estimation (Thomson multitaper method) of unfiltered LFP data during one second during baseline and CS periods (781 samples, time-bandwidth product of 4 and NFFT of 1024). Log-log change in LFP power was computed for CS period as above and mean of five consecutive seconds during baseline activity. Spike autocorrelation was calculated by computing the temporal autocorrelation of the spike train during 500 and 1000 ms. following CS onset or during baseline activity. Spikes were binned (5 ms.) and number of spikes was normalization by the number of bins, with the center peak removed. Increased in theta autocorrelation was defined by calculating the power spectral density estimate of the cross-correlation vector for CS and baseline. Estimation of change in theta power during conditioning was done for each trial using a multitaper power spectral density estimation as above, during 3 s. ahead and 2 s. following the CS onset, with sliding window of 50 ms and normalized to the max. Power spectral density estimation was done separately for pre and post CS to avoid smearing of changes in theta power between preand post-CS time bins. Following the above, power was summed and a mean was calculated for power between 4 - 8 Hz. In order to compute the absolute change in theta power in response to the CS along conditioning, we calculated the relative change in theta power in response to the CS versus pre-CS using the following: Theta power =

Power during
Power before Power during + Power before

Estimation of LFP phase - phase synchrony was done by extracting phase information from the Hilbert transform of the thetafiltered (4 - 8 Hz, FIR filter) signal. We next generated histograms of the phase distributions and the half-width of these distributions was calculated from the circular standard deviation. Phase - phase locking value was defined as the mean resultant length (MRL) and was calculated using the following N 1 X jexpðjqðt; nÞÞ j N n=1

where q(t, n) denotes the phase difference 41(t, n) - 4 2(t, n) (Lachaux et al., 1999). Thus, MRL values ranged between zero (no phaselocking) and one (perfect phase-locking). For analyzing inter-area power correlation we computed the mean theta power for each trial as above and calculated the mean Spearman’s correlation averaged over group of ten successive trials. Differences in correlation scores were computed using Fisher r-to-z transformation. To analyze the directionality of amygdala – dACC activity we calculated phase – phase MRL as a function of lag between the amygdala and dACC for
200 to 200 ms. centered at CS onset in steps of 10 ms. Granger causality analysis in the frequency domain (Geweke, 1982; Granger, 1969) was done using the MVGC multivariate Granger causality MATLAB toolbox (Barnett and Seth, 2014), as it has been previously used in primates and rodents studies to analyze the directionality of interarea interactions in the brain (Brovelli et al., 2004; Ghosh et al., 2013; Gregoriou et al., 2009; Popa et al., 2010; Stujenske et al., 2014). To this end we estimated the relative magnitude of directional amygdala – dACC influences using a time series (bivariate autoregressive models). In general, we conducted the same analysis as in Ghosh et al. (2013), with data downsampled to 390 Hz. and preprocessed by subtracting the ensemble’s mean from each 3 s. single-trial (2 s. ahead and 1 s. following CS-onset) time series and divided by the SD to remove 1st order non-stationarities and to get an ensemble of single trial time series considered as zero mean stochastic process. Development of directional amygdala-dACC influence was estimated in running window of 200 ms. with overlap of 7 ms. The model order for the bivariate autoregressive model was chosen as 2 based on the minimal values of the Bayesian information criteria (Barnett and Seth, 2014). Model stability was confirmed (Ghosh et al., 2013) and the Granger spectral (4 – 8 Hz.) coefficients between the amygdala and dACC at a given window was compared in respect to the significance limit obtained from surrogate datasets (n = 100 datasets). Each dataset (220 trials) was composed of amygdala and dACC traces randomly shifted in time by 250 – 700 ms. (per each trial) and computing the directional coefficient. Later a significance limit (95% confidence interval, normal fit) was calculated using these coefficients (n = 100). Directional influence was estimated as the first of the five consecutive time points significantly higher for one direction than the reverse direction (paired t test, p < 0.05). Volume conduction of the interareal LFP signal was computed using the phase lag index (PLI) (Daffertshofer and Stam, 2007; Stam et al., 2007). PLI measures the symmetry of phase distributions, bounded between 0 and 1 of the occurrence of the relative amygdala-dACC phase difference from 0 or ± p jhsign½DBðtk Þi j

e2 Neuron 97, 1–8.e1–e3, January 17, 2018

Please cite this article in press as: Taub et al., Oscillations Synchronize Amygdala-to-Prefrontal Primate Circuits during Aversive Learning, Neuron (2017), https://doi.org/10.1016/j.neuron.2017.11.042

where, DBðtk Þ is the phase differences between the two time series. PLI was computed per conditioning stage for 6 s signal (5 s. ahead and 1 s following CS onset) for n = 220 trials and compared for each experimental stage, i.e., conditioning 1 – 10, 11 – 20, 21 – 30 and extinction. Inter-area cross-correlation was calculated by computing the mean of the correlation during one second starting at CS onset with maximal lag of 100 ms. for each consecutive 10 trials (i.e., habituation, three stages of conditioning and last ten trials of extinction) and using a running window of 5 trials with 4 trials overlap. For calculation of imaginary coherence, we computed the absolute imaginary cross power spectral density estimate using Welch’s method over one second starting at CS onset using the following P jImð binsSxyÞ j iCoh = pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P P binsðSxxÞ  binsðSyyÞ; where Sxy denotes the cross-spectrum, Sxx and Syy denotes the auto-spectra and summation is done over the spectrogram bins corresponding to the quantified time bin (Karalis et al., 2016; Nolte et al., 2004). For spike – LFP phase synchrony, MRL value was calculated using the above by assigning each spike event during 500 ms. after CS onset and during baseline its corresponding theta phase value. Spike – LFP phase MRL was calculated for each consecutive 10 trials and for running window of 5 trials with 4 trials overlap. MRL of shuffled data was obtained by shuffling the location of spikes in time during the CS 1000 times. Criteria for significant MRL value were obtained by calculating the MRL value for 10 randomly drawn normally distributed numbers between ± Pi for 10000 times. Criteria for significant MRL value were defined as > 95th percentile of the distribution. For percent units above criteria we computed the percent of amygdala spike – dACC LFP pair of electrodes with averaged MRL above criteria (criteria values were obtained as above). MRL was computed over 10 consecutive trials i.e., habituation, conditioning 1 – 10, 11 – 20, 21 – 30 and extinction. Conditioning dependent change in spike – LFP synchrony was computed for pairs of electrodes with MRL below criteria either during habituation or habituation and extinction that that exhibited MRL above criteria during at least one of the three epochs of conditioning trials, i.e., conditioning 1 – 10, 11 – 20 or 21 – 30. We next generated histograms of the spike - LFP - phase distributions and the half-width of these distributions was calculated from the circular standard deviation, similar to that obtained for LFP phase distribution above. For synchrony – percent CR correlation we calculated the mean daily MRL score using a running window of 5 trials with overlap of 4 trials over the first 20 trials. Behavioral performance was calculated as percent trials with significant response to the CS. For correlation between peak MRL and behavioral plateau we correlated the day by day number of trials for peak in amygdaladACC synchrony with the number of trials for obtaining a maximal percent conditioned response. All results are presented with correction for multiple comparisons.

Neuron 97, 1–8.e1–e3, January 17, 2018 e3