Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2017 Jun;20(6):824-835.
doi: 10.1038/nn.4553. Epub 2017 Apr 24.

Amygdala Inputs to Prefrontal Cortex Guide Behavior Amid Conflicting Cues of Reward and Punishment

Affiliations
Free PMC article

Amygdala Inputs to Prefrontal Cortex Guide Behavior Amid Conflicting Cues of Reward and Punishment

Anthony Burgos-Robles et al. Nat Neurosci. .
Free PMC article

Abstract

Orchestrating appropriate behavioral responses in the face of competing signals that predict either rewards or threats in the environment is crucial for survival. The basolateral nucleus of the amygdala (BLA) and prelimbic (PL) medial prefrontal cortex have been implicated in reward-seeking and fear-related responses, but how information flows between these reciprocally connected structures to coordinate behavior is unknown. We recorded neuronal activity from the BLA and PL while rats performed a task wherein competing shock- and sucrose-predictive cues were simultaneously presented. The correlated firing primarily displayed a BLA→PL directionality during the shock-associated cue. Furthermore, BLA neurons optogenetically identified as projecting to PL more accurately predicted behavioral responses during competition than unidentified BLA neurons. Finally photostimulation of the BLA→PL projection increased freezing, whereas both chemogenetic and optogenetic inhibition reduced freezing. Therefore, the BLA→PL circuit is critical in governing the selection of behavioral responses in the face of competing signals.

Figures

Figure 1
Figure 1
Behavioral tasks to examine the discrimination and competition of reward and fear memories. (a) During discrimination, discrete conditioned stimuli (CS) predicted either sucrose or shocks. These CSs were termed the “CS-Suc” and “CS-Shock”, and their sensory modalities (light vs tone) were counterbalanced across animals. Sucrose was removed from the port if animals did not collect it by the end of the CS (vacuum; “Vac”). During competition, in addition to the individual CS-Suc and CS-Shock, animals were challenged by the co-presentation of these associations to induce conflicting motivational drives and “competition”. (b and c) For simplicity, we operationalize the term “Rew” to refer to port-entry behavior and “Fear” to refer to freezing. (b) Port entry responses per CS during the last discrimination session. Inset show the average time that animals spent in the port per CS (paired T-test: t15 = 20.3, ***P < 0.001, n = 16 animals). (c) Freezing responses per CS during the last discrimination session. Inset shows the average time that animals spent freezing per CS (paired T-test: t15 = 20.6, ***P < 0.001, n = 16 animals). (d) Port entry responses during the last competition session. Inset shows the average time in the port per CS (repeated measures one-way ANOVA: F2,30 = 107.6, P < 0.001, n = 16 animals; Bonferroni post-hoc tests: t15 > 6.85 and ***P < 0.001 for all comparisons). (e) Freezing responses during the last competition session. Inset shows the average time that animals spent freezing per CS (repeated measures one-way ANOVA: F2,30 = 89.0, P < 0.001, n = 16 animals; Bonferroni post-hoc tests: t15 > 6.01 and ***P < 0.001 for all comparisons). Error bands in line plots and error bars in insets represent s.e.m.
Figure 2
Figure 2
Correlated activity between the BLA and PL varied between the retrieval of reward- and fear-related memories. (a-b) Neural activity was simultaneously recorded in the BLA and PL (n = 12 animals). Representative CCs between BLA/PL neural pairs (25-ms bins) exhibit either excitatory correlations (i.e., the target cell showed increased firing when the reference cell fired) or inhibitory correlations (i.e., the target cell showed reduced firing when the reference cell fired). Significant peaks or troughs were detected within ±100 ms from the reference spikes. Correlations due to common input (i.e., zero lag) were excluded by examining CCs generated with smaller bin widths (5-ms bins; see Supplementary Fig. S3). Only a small fraction of cell pairs showed zero lag (see Supplementary Fig. S6c-d). (c-d) Excitatory correlations predominated over inhibitory correlations in all task epochs (Bonferroni-corrected chi-square tests: χ2 > 136.5, P < 0.001 in all task epochs). In the heatmaps, cell pairs were ordered based on the latency of peaks and troughs. Mean latency±s.e.m. of peaks for excitatory CCs: ITI, 3.9±2.4 ms; CS-Suc, 5.9±2.5 ms; and CS-Shock, 13.0±2.4 ms. Mean latency of troughs for inhibitory CCs: ITI, 12.5±5.3 ms; CS-Suc, 17.1±7.6 ms; CS-Shock, -1.4±5.2 ms. Numbers at the bottom of the heatmaps indicate the overall proportion of correlated cell pairs per epoch. Vertical dash lines represent the time of the reference spikes, and the horizontal dash lines represent the separation between putative BLA-led versus PL-led correlations. (e-f) Distinct populations of BLA/PL neural pairs exhibited correlated activity during different epochs. (g) Putative leading structure in the excitatory CCs, based on the latency of peaks. Cell pairs were deemed as BLA-led if the peak occurred within +2.5 to +100 ms from the reference, whereas they were deemed as PL-led if the peak occurred within -100 to -2.5 ms from the reference (n's reported within each bar). BLA leading was significantly greater than PL leading only during the CS-Shock (ITI, χ2 = 3.83, P = 0.14; CS-Suc, χ2 = 0.91, P = 0.71; CS-Shock, χ2 = 29.2, ***P < 0.001). The leading ratio during the CS-Shock was also significantly different than during other epochs (ITI vs CS-Suc, χ2 = 0.71, P = 0.78; ITI vs CS-Shock, χ2 = 10.4, **P = 0.004; CS-Suc vs CS-Shock, χ2 = 16.0, ***P < 0.001). (h) Putative leading structure in the inhibitory CCs. There was a trend towards higher BLA than PL leading during the CS-Suc (ITI, χ2 = 0.89, P = 0.72; CS-Suc, χ2 = 4.28, ~P = 0.11; CS-Shock, χ2 = 0.15, P = 0.97). The leading ratio during the CS-Suc was significantly different than during the CS-Shock (ITI vs CS-Suc, χ2 = 2.33, P = 0.35; ITI vs CS-Shock, χ2 = 1.37, P = 0.56; CS-Suc vs CS-Shock, χ2 = 5.70, *P = 0.05).
Figure 3
Figure 3
BLA neurons biased towards encoding the CS-Sucrose (Rew) or CS-Shock (Fear) exhibited similar proportions of correlated activity with PL, whereas a greater proportion of “fear”-biased PL cells exhibited correlated activity with BLA. (a-b) BLA populations based on the response to the reward-associated cue (i.e., CS-Suc; “R”), fear-associated cue (i.e., CS-Shock; “F”), or to both cues (“R” and “F” combinations). There were no significant differences in proportions across these populations (Bonferroni-corrected chi-square tests: χ2 < 9.50 and P > 0.056 for all comparisons). (c) Separation of reward and fear biased BLA cells, based on the peak response to each cue. Cells in the light-gray zones exhibited larger responses to the reward-related cue, and were deemed as “reward biased”. Cells in the dark-gray zones exhibited larger responses to the fear-related cue, and were deemed as “fear biased”. Inset shows the average proportion of reward and fear biased cells in the BLA per subject (paired T-test: t11 = 0.61, P = 0.56, n = 12 animals). (d) Reward and fear biased BLA cells showed similar proportions of cross-correlated activity with simultaneously recorded PL cells. These values were normalized to the total number of correlated neural pairs per subject per cue (repeated measures two-way ANOVA: cells, F1,22 = 0.01, P = 0.91; cue, F1,22 = 1.48, P = 0.24; interaction, F1,22 = 0.04, P = 0.85, n = 12 animals). (e-f) PL populations based on the response to the cues. Asterisk in the F+ bar indicates that this population was significantly larger than most other populations (χ2 = 8.86 and P = 0.079 compared to R+F+; χ2 > 11.06 and P < 0.024 compared to all other populations). (g) Separation of reward and fear biased PL cells, based on the peak response to each cue. A greater proportion of PL cells exhibited fear bias than reward bias (t11 = 3.03, *P = 0.011, n = 12 animals). (h) A greater proportion of fear-biased PL cells showed cross-correlated activity with simultaneously recorded BLA cells (cells, F1,22 = 13.5, P = 0.0013; cue, F1,22 = 3.66, P = 0.069; interaction, F1,22 = 1.23, P = 0.28; CS-Suc, t11 = 3.32, **P = 0.0026; CS-Shock, t11 = 3.73, ***P < 0.001, n = 12 animals). Error bars represent s.e.m.
Figure 4
Figure 4
The majority of BLA-PL cells recorded showed selective excitations to the shock-predictive cue. An optogenetic approach was used to photoidentify these cells (“BLA-PL population”). (a-e) Assessment of photoresponse latencies in slices. (a) Ex vivo whole-cell patch-clamp recordings were performed after selectively expressing ChR2 in BLA-PL cells using a Cre-dependent viral system (n = 7 animals). Expressing cells (n = 6 cells) and non-expressing neighbors (n = 24 cells) were recorded while stimulating with blue light (5-ms pulses at 1 Hz). (b) Representative traces from a ChR2+∷BLA-PL cell and 4 non-expressing neighbors. (c) Distribution of all cells sampled with whole-cell patch-clamp recording. (d) Representative traces show the latency of photoresponses at various light power densities (power range, 0.5-84 mW/mm2). Latencies were calculated from light onset to action potential peaks. (e) Distribution of photoresponse latencies for the BLA-PL cells. Dots represent individual cells, and error bar represents s.e.m. (f-h) Photo Identification of BLA-PL cells in behaving animals. (f) Optrodes were chronically implanted in the BLA for neural recordings after selectively expressing ChR2 in BLA-PL cells. Optimal ChR2 expression and detection of photoresponses was achieved in a subset of animals (n = 2/6 animals, 33%). (g) BLA-PL cell displaying photoresponses in vivo (bin-width, 20 ms). (h) Assessment of photoresponse latencies in vivo. Latencies were calculated from laser onset to the time at which cells exhibited a significant increase in firing frequency. Eleven out of sixty cells (18%) were deemed as BLA-PL cells as they displayed photoresponse latencies shorter than 12 ms, which was the longest latency observed in slices. One cell displayed latencies greater than 12 ms (white-filled bin) and it was excluded from further analyses. (i-j) Response profile of photoidentified BLA populations. Error bands in line plots represent s.e.m. (i) BLA-PL population. A greater proportion of these cells displayed selective excitatory responses to the fear-associated cue (“F+”, n = 6/11 cells, 55%). (j) An additional BLA population that exhibited significant inhibition during ChR2 stimulation. These cells thus did not terminate in PL, and perhaps received inhibitory influence from the BLA-PL network. These cells were deemed as network-inhibited cells (“Netw(-)”, n = 8/60 cells, 13%). During the discrimination task, the majority of these cells exhibited either inhibitory responses to the fear cue (“F-”, 3/8, 38%) or excitatory responses to the reward cue (“R+”, 3/8, 38%).
Figure 5
Figure 5
BLA cells terminating in PL more accurately predicted the animal's behavioral response during competition. (a) Photo Identification of the distinct BLA populations: “BLA-PL” (8/57, 14%), population terminating in PL; “Netw(-)” (8/57, 14%), population that showed inhibition during photostimulation; and “Unidentified” (41/57, 72%), population that did not respond to light stimulation. (b) Schematic of the competition task in which in addition to CS-Suc and CS-Shock trials, animals were challenged by the co-presentation of these associations to induce behavioral competition. Below, trial-by-trial behavioral output for a representative animal during each trial type. (c-d) Support vector machine (SVM) model to predict behavioral responses during 20 competition trials. The SVM model was trained using neural activity during the CS-Suc and CS-Shock trials. Data for the entire 20 s of CS presentation was used to classify neural activity. The model was then tested during competition trials to predict behavioral responses based on neural activity. For this given example, this BLA-PL cell accurately predicted behavioral responses on 85% of the competition trials. (e) Mean decoding accuracy for the distinct BLA populations. Superimposed dots represent individual cells (n's per population are reported in the bars). All BLA populations showed averaged decoding accuracies that were significantly higher than chance (Bonferroni-corrected paired T-test comparisons against scrambled data are represented by the asterisks above the number of cells per population: BLA-PL, t7 = 3.31, *P = 0.013; Netw(-), t7 = 3.74, **P = 0.007; Unidentified, t40 = 2.29, *P = 0.028). Furthermore, the BLA-PL population but not the Netw(-) population showed significantly higher decoding accuracy than unidentified cells (one-way ANOVA: F2,54 = 3.36, P = 0.042; Bonferroni post-hoc tests: BLA-PL vs Unidentified, t47 = 2.74, *P = 0.017; Netw(-) vs Unidentified, t47 = 1.22, P = 0.23). (f) Mean decoding accuracy for the BLA populations, when their activity was paired with the activity of simultaneously recorded PL cells with which they showed either uncorrelated activity (“Unc”) or significantly correlated activity (“Corr”). Superimposed dots represent BLA/PL neural pairs (number of cell pairs per population are reported within the bars). All populations showed decoding accuracies that were significantly higher than chance (BLA-PL Unc, t38 = 2.91, **P = 0.006; BLA-PL Corr, t38 = 6.76, ***P < 0.001; Netw(-) Unc, t32 = 5.87, ***P < 0.001; Netw(-) Corr, t68 = 10.5, ***P < 0.001; Unidentified Unc, t284 = 8.36, ***P < 0.001; Unidentified Corr, t124 = 4.24, ***P < 0.001). Furthermore, the BLA-PL cells showed significantly higher decoding accuracy when its activity was paired with correlated PL activity (one-way ANOVA: F5,584 = 11.1, P < 0.001; Bonferroni post-hoc tests: “BLA-PL”, Unc vs Corr, t76 = 2.64, *P = 0.011; “Netw(-)”, Unc vs Corr, t100 = 0.68, P = 0.50; “Unidentified”, Unc vs Corr, t408 = 1.18, P = 0.24). Error bars represent s.e.m.
Figure 6
Figure 6
Stimulation of BLA inputs to PL facilitated fear-related behavior, and biased behavioral responses towards fear during competition. (a) Optogenetic strategy to stimulate BLA inputs to PL. The BLA was unilaterally transduced with either eYFP (n = 10 animals) or ChR2 (n = 8 animals), and an optical fiber was chronically implanted in the dorsal regions of PL to locally stimulate BLA inputs. (b) Schematic of the discrimination task in which half of the trials were paired with 20-Hz blue light stimulation. The trial and laser sequences were pseudorandom. (c) Freezing behavior during CS-Shock trial, illustrated as the difference score in the percentage of time spent freezing, relative to laser-OFF. Stimulation of BLA inputs to PL significantly enhanced freezing responses (repeated measures two-way ANOVA: group, F1,16 = 11.4, P = 0.004; laser, F1,16 = 2.88, P = 0.11; interaction, F1,16 = 11.4, P = 0.004; eYFP vs ChR2 during laser-ON: t16 = 4.78, ***P = 0.0002). (d) Port entry behavior during CS-Suc trials, illustrated as the difference score in the percentage of time spent in the sucrose port, relative to laser-OFF. No significant differences were detected for port entry responses (group, F1,16 = 0.95, P = 0.34; laser, F1,16 = 0.13, P = 0.72; interaction, F1,16 = 0.95, P = 0.34). (e) Pharmacology experiment to rule out the possible contribution of stimulation of fibers of passage. After unilateral transduction of the BLA with ChR2 (n = 8 animals), a cannula was chronically implanted above PL to allow for the infusion of either ACSF or the AMPA and NMDA receptor antagonists NBQX and AP5 ~10-15 min prior to inserting an optical fiber for optical stimulation and behavioral testing. (f) Experimental design for drug treatment and schematic of the competition task in which half of the trials were paired with light stimulation. The trial and laser sequences were pseudorandom. (g) Freezing behavior during CS-Shock trials. Ruling out the possibility of stimulation of fibers of passage, the NBQX+AP5 treatment abolished the stimulation effect on freezing observed after the ACSF treatment (drug, F1,14 = 4.88, P = 0.044; laser, F1,14 = 7.64, P = 0.015; interaction, F1,14 = 4.88, P = 0.044; ACSF vs NBQX+AP5 during laser-ON: t7 = 3.12, **P = 0.0075). (h) Port entry behavior during CS-Suc trials. No significant differences were detected (drug, F1,14 = 1.27, P = 0.28; laser, F1,14 = 4.58, P = 0.0504; interaction, F1,14 = 1.27, P = 0.28). (i) Freezing during competition trials. Stimulation of BLA inputs to PL also enhanced freezing during competition trials under the ACSF treatment, and this effect was abolished by the NBQX+AP5 treatment (drug, F1,14 = 6.79, P = 0.02; laser, F1,14 = 1.89, P = 0.19; interaction, F1,14 = 6.79, P = 0.02; ACSF vs NBQX+AP5 during laser-ON: t7 = 3.69, **P = 0.0024). (j) Port entry behavior during competition. There was a trend towards reduced port entry responses during competition (drug, F1,14 = 2.18, P = 0.16; laser, F1,14 = 1.83, P = 0.20; interaction, F1,14 = 2.18, P = 0.16; ACSF vs NBQX+AP5 during laser-ON: t7 = 2.09, ~P = 0.056). Error bars represent s.e.m.
Figure 7
Figure 7
The BLA-PL pathway is necessary for expression of the fear-associated memory, but not for reward-seeking behavior. (a) Optogenetic strategy to inhibit BLA inputs to PL. The BLA was bilaterally transfected with either GFP (n = 6 animals) or the opsin ArchT (n = 6 animals), and optical fibers were chronically implanted just above PL to silence BLA inputs locally. (b) Competition paradigm in which half of the trials were paired with constant yellow light to silence BLA inputs to PL. The trial and laser sequences were pseudorandomized. (c) Freezing during CS-Shock trials. Silencing of BLA inputs to PL significantly impaired freezing responses (repeated measures two-way ANOVA: group, F1,10 = 5.64, P = 0.039; laser, F1,10 = 2.75, P = 0.14; interaction, F1,10 = 5.64, P = 0.039; GFP vs ArchT during laser-ON: t10 = 3.36, **P = 0.0072). (d) Port entries during CS-Suc trials. No significant differences were detected on port entry responses (group, F1,10 = 2.53, P = 0.14; laser, F1,10 = 1.70, P = 0.22; interaction, F1,10 = 2.53, P = 0.14). (e) Freezing during competition trials. Significant group differences were detected for freezing during competition (group, F1,10 = 5.20, P = 0.046; laser, F1,10 = 4.37, P = 0.063; interaction, F1,10 = 5.20, P = 0.046; GFP vs ArchT during laser-ON: t10 = 3.23, **P = 0.0091). (f) Port entries during competition trials. Significant group differences were also detected for port entries during competition (group, F1,10 = 10.5, P = 0.009; laser, F1,10 = 9.73, P = 0.011; interaction, F1,10 = 10.5, P = 0.009; GFP vs ArchT during laser-ON: t10 = 4.58, ***P = 0.001). (g) Chemogenetic strategy to selectively silence BLA cell that terminate in PL (i.e., selective inhibition of the BLA-PL population). Using a Cre-dependent dual-virus method, BLA-PL cells were bilaterally transduced with either mCherry (n = 7 animals) or M4D-Gi (n = 7 animals), which is a Gi-coupled designer receptor that induces neuronal silencing upon activation with the designer drug clozapine-N-oxide (CNO). (h) Experimental design to treat animals with either vehicle (5% DMSO in 0.9% saline, i.p.) or CNO (10 mg/kg, i.p.) ~15-20 min prior to behavioral testing. (i) Freezing behavior during CS-Shock trials. Silencing the BLA-PL population significantly impaired freezing responses (group, F1,12 = 3.41, P = 0.09; drug, F2,24 = 7.96, P = 0.0022; interaction, F2,24 = 6.31, P = 0.006; mCherry vs M4D(Gi) during CNO: t12 = 3.66, **P = 0.0033). (j) Port entry behavior during CS-Suc trials. No significant differences were detected (group, F1,12 = 0.13, P = 0.72; drug, F2,24 = 0.69, P = 0.51; interaction, F2,24 = 0.57, P = 0.58). (k) Freezing behavior during competition trials. Silencing of the BLA-PL population impaired freezing responses (group, F1,12 = 1.45, P = 0.25; drug, F2,24 = 0.09, P = 0.91; interaction, F2,24 = 2.67, P = 0.09; mCherry vs M4D(Gi) during CNO: t12 = 2.56, *P = 0.025). (l) Port entry behavior during competition trials. No significant differences were detected (group, F1,12 = 0.13, P = 0.72; drug, F2,24 = 0.60, P = 0.55; interaction, F2,24 = 0.17, P = 0.84). Error bars represent s.e.m.

Similar articles

See all similar articles

Cited by 52 articles

See all "Cited by" articles

References

    1. Maren S, Quirk GJ. Neuronal signalling of fear memory. Nat Rev Neurosci. 2004;5:844–852. - PubMed
    1. Janak PH, Tye KM. From circuits to behaviour in the amygdala. Nature. 2015;517:284–292. - PMC - PubMed
    1. Rogan MT, Stäubli UV, LeDoux JE. Fear conditioning induces associative long-term potentiation in the amygdala. Nature. 1997;390:604–607. - PubMed
    1. Paton JJ, Belova MA, Morrison SE, Salzman CD. The primate amygdala represents the positive and negative value of visual stimuli during learning. Nature. 2006;439:865–870. - PMC - PubMed
    1. Tye KM, Stuber GD, de Ridder B, Bonci A, Janak PH. Rapid strengthening of thalamo-amygdala synapses mediates cue-reward learning. Nature. 2008;453:1253–1257. - PMC - PubMed

MeSH terms

LinkOut - more resources

Feedback