Abstract
A key function of brain systems mediating emotion is to learn to anticipate unpleasant experiences. Although organisms readily associate sensory stimuli with aversive outcomes, higher-order forms of emotional learning and memory require inference to extrapolate the circumstances surrounding directly experienced aversive events to other indirectly related sensory patterns that were not part of the original experience. This type of learning requires internal models of emotion, which flexibly track directly experienced and inferred aversive associations. Although the brain mechanisms of simple forms of aversive learning have been well studied in areas such as the amygdala1,2,3,4, whether and how the brain forms and represents internal models of emotionally relevant associations are not known5. Here we report that neurons in the rodent dorsomedial prefrontal cortex (dmPFC) encode a flexible internal model of emotion by linking sensory stimuli in the environment with aversive events, whether they were directly or indirectly associated with that experience. These representations form through a multi-step encoding mechanism involving recruitment and stabilization of dmPFC cells that support inference. Although dmPFC population activity encodes all salient associations, dmPFC neurons projecting to the amygdala specifically represent and are required to express inferred associations. Together, these findings reveal how internal models of emotion are encoded in the dmPFC to regulate subcortical systems for recall of inferred emotional memories.
This is a preview of subscription content, access via your institution
Access options
Access Nature and 54 other Nature Portfolio journals
Get Nature+, our best-value online-access subscription
$32.99 / 30 days
cancel any time
Subscribe to this journal
Receive 51 print issues and online access
$199.00 per year
only $3.90 per issue
Buy this article
- Purchase on SpringerLink
- Instant access to full article PDF
Prices may be subject to local taxes which are calculated during checkout
Similar content being viewed by others
Data availability
All data used in this paper have been shared through the RIKEN CBS Data Sharing Platform (https://neurodata.riken.jp/id/20250401-001). Source data are provided with this paper.
Code availability
All codes used in this paper have been shared through the RIKEN CBS Data Sharing Platform (https://neurodata.riken.jp/id/20250401-001).
References
Herry, C. & Johansen, J. P. Encoding of fear learning and memory in distributed neuronal circuits. Nat. Neurosci. 17, 1644–1654 (2014).
Janak, P. H. & Tye, K. M. From circuits to behaviour in the amygdala. Nature 517, 284–292 (2015).
Tovote, P., Fadok, J. P. & Lüthi, A. Neuronal circuits for fear and anxiety. Nat. Rev. Neurosci. 16, 317–331 (2015).
Malezieux, M., Klein, A. S. & Gogolla, N. Neural circuits for emotion. Annu. Rev. Neurosci. 46, 211–231 (2023).
LeDoux, J. E. What emotions might be like in other animals. Curr. Biol. 31, R824–R829 (2021).
Barrett, L. F., Mesquita, B., Ochsner, K. N. & Gross, J. J. The experience of emotion. Annu. Rev. Psychol. 58, 373–403 (2007).
Phelps, E. A. & LeDoux, J. E. Contributions of the amygdala to emotion processing: from animal models to human behavior. Neuron 48, 175–187 (2005).
LeDoux, J. & Daw, N. D. Surviving threats: neural circuit and computational implications of a new taxonomy of defensive behaviour. Nat. Rev. Neurosci. 19, 269–282 (2018).
Poon, C.-S. & Merfeld, D. M. Internal models: the state of the art. J. Neural Eng. 2, E01 (2005).
Ito, M. Control of mental activities by internal models in the cerebellum. Nat. Rev. Neurosci. 9, 304–313 (2008).
Balleine, B. W., Dezfouli, A., Ito, M. & Doya, K. Hierarchical control of goal-directed action in the cortical–basal ganglia network. Curr. Opin. Behav. Sci. 5, 1–7 (2015).
Wikenheiser, A. M. & Schoenbaum, G. Over the river, through the woods: cognitive maps in the hippocampus and orbitofrontal cortex. Nat. Rev. Neurosci. 17, 513–523 (2016).
Imamizu, H. & Kawato, M. Neural correlates of predictive and postdictive switching mechanisms for internal models. J. Neurosci. 28, 10751–10765 (2008).
Gershman, S. J. & Niv, Y. Learning latent structure: carving nature at its joints. Curr. Opin. Neurobiol. 20, 251–256 (2010).
Sadacca, B. F. et al. Orbitofrontal neurons signal sensory associations underlying model-based inference in a sensory preconditioning task. eLife 7, e30373 (2018).
Hart, E. E., Gardner, M. P. H. & Schoenbaum, G. Anterior cingulate neurons signal neutral cue pairings during sensory preconditioning. Curr. Biol. 32, 725–732.e3 (2022).
Miyashita, Y. Cognitive memory: cellular and network machineries and their top-down control. Science 306, 435–440 (2004).
Holmes, N. M., Parkes, S. L., Killcross, A. S. & Westbrook, R. F. The basolateral amygdala is critical for learning about neutral stimuli in the presence of danger, and the perirhinal cortex is critical in the absence of danger. J. Neurosci. 33, 13112–13125 (2013).
Miller, E. K. & Cohen, J. D. An integrative theory of prefrontal cortex function. Annu. Rev. Neurosci. 24, 167–202 (2001).
Baeg, E. H. et al. Dynamics of population code for working memory in the prefrontal cortex. Neuron 40, 177–188 (2003).
Matsumoto, K., Suzuki, W. & Tanaka, K. Neuronal correlates of goal-based motor selection in the prefrontal cortex. Science 301, 229–232 (2003).
Euston, D. R., Gruber, A. J. & McNaughton, B. L. The role of medial prefrontal cortex in memory and decision making. Neuron 76, 1057–1070 (2012).
Morrissey, M. D., Insel, N. & Takehara-Nishiuchi, K. Generalizable knowledge outweighs incidental details in prefrontal ensemble code over time. eLife 6, e22177 (2017).
Simon, J., Rudebeck, P. H. & Rich, E. L. From affective to cognitive processing: functional organization of the medial frontal cortex. Int. Rev. Neurobiol. 158, 1–28 (2021).
Tse, D. et al. Schema-dependent gene activation and memory encoding in neocortex. Science 333, 891–895 (2011).
Klein-Flügge, M. C., Wittmann, M. K., Shpektor, A., Jensen, D. E. A. & Rushworth, M. F. S. Multiple associative structures created by reinforcement and incidental statistical learning mechanisms. Nat. Commun. 10, 4835 (2019).
Bongioanni, A. et al. Activation and disruption of a neural mechanism for novel choice in monkeys. Nature 591, 270–274 (2021).
Gao, L. et al. Single-neuron projectome of mouse prefrontal cortex. Nat. Neurosci. 25, 515–529 (2022).
Takehara-Nishiuchi, K. Flexibility of memory for future-oriented cognition. Curr. Opin. Neurobiol. 76, 102622 (2022).
Courtin, J. et al. Prefrontal parvalbumin interneurons shape neuronal activity to drive fear expression. Nature 505, 92–96 (2014).
Laufer, O., Israeli, D. & Paz, R. Behavioral and neural mechanisms of overgeneralization in anxiety. Curr. Biol. 26, 713–722 (2016).
Salzman, C. D. & Fusi, S. Emotion, cognition, and mental state representation in amygdala and prefrontal cortex. Annu. Rev. Neurosci. 33, 173–202 (2010).
Likhtik, E., Stujenske, J. M., Topiwala, M. A., Harris, A. Z. & Gordon, J. A. Prefrontal entrainment of amygdala activity signals safety in learned fear and innate anxiety. Nat. Neurosci. 17, 106–113 (2014).
Burgos-Robles, A., Vidal-Gonzalez, I. & Quirk, G. J. Sustained conditioned responses in prelimbic prefrontal neurons are correlated with fear expression and extinction failure. J. Neurosci. 29, 8474–8482 (2009).
Jercog, D. et al. Dynamical prefrontal population coding during defensive behaviours. Nature 595, 690–694 (2021).
Cummings, K. A. & Clem, R. L. Prefrontal somatostatin interneurons encode fear memory. Nat. Neurosci. 23, 61–74 (2020).
Agetsuma, M. et al. Activity-dependent organization of prefrontal hub-networks for associative learning and signal transformation. Nat. Commun. 14, 5996 (2023).
Do-Monte, F. H., Quiñones-Laracuente, K. & Quirk, G. J. A temporal shift in the circuits mediating retrieval of fear memory. Nature 519, 460–463 (2015).
Dixsaut, L. & Gräff, J. The medial prefrontal cortex and fear memory: dynamics, connectivity, and engrams. Int. J. Mol. Sci. 22, 12113 (2021).
Rawson, T., O’Kane, M. & Talk, A. The medial prefrontal cortex and memory of cue location in the rat. Neurobiol. Learn. Mem. 93, 132–136 (2010).
Likhtik, E. & Paz, R. Amygdala–prefrontal interactions in (mal)adaptive learning. Trends Neurosci. 38, 158–166 (2015).
Anastasiades, P. G. & Carter, A. G. Circuit organization of the rodent medial prefrontal cortex. Trends Neurosci. 44, 550–563 (2021).
Maren, S. & Quirk, G. J. Neuronal signalling of fear memory. Nat. Rev. Neurosci. 5, 844–852 (2004).
Madarasz, T. J. et al. Evaluation of ambiguous associations in the amygdala by learning the structure of the environment. Nat. Neurosci. 19, 965–972 (2016).
Saez, A., Rigotti, M., Ostojic, S., Fusi, S. & Salzman, C. D. Abstract context representations in primate amygdala and prefrontal cortex. Neuron 87, 869–881 (2015).
Kyriazi, P., Headley, D. B. & Paré, D. Different multidimensional representations across the amygdalo-prefrontal network during an approach-avoidance task. Neuron 107, 717–730.e5 (2020).
Holmes, N. M., Wong, F. S., Bouchekioua, Y. & Westbrook, R. F. Not “either-or” but “which-when”: a review of the evidence for integration in sensory preconditioning. Neurosci. Biobehav. Rev. 132, 1197–1204 (2022).
Iordanova, M. D., Burnett, D. J., Aggleton, J. P., Good, M. & Honey, R. C. The role of the hippocampus in mnemonic integration and retrieval: complementary evidence from lesion and inactivation studies. Eur. J. Neurosci. 30, 2177–2189 (2009).
Bornstein, A. M. & Daw, N. D. Cortical and hippocampal correlates of deliberation during model-based decisions for rewards in humans. PLoS Comput. Biol. 9, e1003387 (2013).
Barron, H. C. et al. Neuronal computation underlying inferential reasoning in humans and mice. Cell 183, 228–243.e21 (2020).
Wimmer, G. E. & Shohamy, D. Preference by association: how memory mechanisms in the hippocampus bias decisions. Science 338, 270–273 (2012).
Jun, H. et al. Prefrontal and lateral entorhinal neurons co-dependently learn item–outcome rules. Nature 633, 864–871 (2024).
Rizley, R. C. & Rescorla, R. A. Associations in second-order conditioning and sensory preconditioning. J. Comp. Physiol. Psychol. 81, 1–11 (1972).
Gostolupce, D., Lay, B. P. P., Maes, E. J. P. & Iordanova, M. D. Understanding associative learning through higher-order conditioning. Front. Behav. Neurosci. 16, 845616 (2022).
Ghosh, K. K. et al. Miniaturized integration of a fluorescence microscope. Nat. Methods 8, 871–878 (2011).
Courtin, J. et al. A neuronal mechanism for motivational control of behavior. Science 375, eabg7277 (2022).
Wheeler, D. S., Sherwood, A. & Holland, P. C. Excitatory and inhibitory learning with absent stimuli. J. Exp. Psychol. Anim. Behav. Process. 34, 247–255 (2008).
Hall, G. Learning about associatively activated stimulus representations: implications for acquired equivalence and perceptual learning. Anim. Learn. Behav. 24, 233–255 (1996).
Yiu, A. P. et al. Neurons are recruited to a memory trace based on relative neuronal excitability immediately before training. Neuron 83, 722–735 (2014).
Zhou, Y. et al. CREB regulates excitability and the allocation of memory to subsets of neurons in the amygdala. Nat. Neurosci. 12, 1438–1443 (2009).
Cai, D. J. et al. A shared neural ensemble links distinct contextual memories encoded close in time. Nature 534, 115–118 (2016).
Rashid, A. J. et al. Competition between engrams influences fear memory formation and recall. Science 353, 383–387 (2016).
Zaki, Y. et al. Offline ensemble co-reactivation links memories across days. Nature 637, 145–155 (2025).
Mocle, A. J. et al. Excitability mediates allocation of pre-configured ensembles to a hippocampal engram supporting contextual conditioned threat in mice. Neuron 112, 1487–1497.e6 (2024).
Johansen, J. P. et al. Hebbian and neuromodulatory mechanisms interact to trigger associative memory formation. Proc. Natl Acad. Sci. USA 111, E5584–E5592 (2014).
Ozawa, T. et al. A feedback neural circuit for calibrating aversive memory strength. Nat. Neurosci. 20, 90–97 (2017).
Holmes, N. M., Fam, J. P., Clemens, K. J., Laurent, V. & Westbrook, R. F. The neural substrates of higher-order conditioning: a review. Neurosci. Biobehav. Rev. 138, 104687 (2022).
Carr, M. F., Jadhav, S. P. & Frank, L. M. Hippocampal replay in the awake state: a potential substrate for memory consolidation and retrieval. Nat. Neurosci. 14, 147–153 (2011).
Aly, M. H., Abdou, K., Okubo-Suzuki, R., Nomoto, M. & Inokuchi, K. Selective engram coreactivation in idling brain inspires implicit learning. Proc. Natl Acad. Sci. USA 119, e2201578119 (2022).
Ormond, J., Serka, S. A. & Johansen, J. P. Enhanced reactivation of remapping place cells during aversive learning. J. Neurosci. 43, 2153–2167 (2023).
Quirk, G. J., Armony, J. L. & LeDoux, J. E. Fear conditioning enhances different temporal components of tone-evoked spike trains in auditory cortex and lateral amygdala. Neuron 19, 613–624 (1997).
Maren, S. & Holmes, A. Stress and fear extinction. Neuropsychopharmacology 41, 58–79 (2016).
Milad, M. R. & Quirk, G. J. Fear extinction as a model for translational neuroscience: ten years of progress. Annu. Rev. Psychol. 63, 129–151 (2012).
Johansen, J. P., Cain, C. K., Ostroff, L. E. & LeDoux, J. E. Molecular mechanisms of fear learning and memory. Cell 147, 509–524 (2011).
Branco, T. & Redgrave, P. The neural basis of escape behavior in vertebrates. Annu. Rev. Neurosci. 43, 417–439 (2020).
Anderson, D. J. & Adolphs, R. A framework for studying emotions across species. Cell 157, 187–200 (2014).
LeDoux, J. E. & Pine, D. S. Using neuroscience to help understand fear and anxiety: a two-system framework. Am. J. Psychiatry 173, 1083–1093 (2016).
Paxinos, G. & Watson, C. The Rat Brain in Stereotaxic Coordinates (Academic Press, 2007).
Gründemann, J. et al. Amygdala ensembles encode behavioral states. Science 364, eaav8736 (2019).
Acknowledgements
We thank K. Hagihara, A. Lüthi and A. Uematsu for the analysis code used in our calcium imaging analyses; and Y. Ishizu for viral preparation; J. LeDoux, N. Holmes, T. Toyoizumi and L. I. Schmitt for comments on earlier versions of the manuscript; and Y. Kasuga, Z. Gungor and other Johansen laboratory members for comments on the manuscript and/or advice on experimental design. This work was partially supported by the Brain Initiative (1U01 NS122123 to J.P.J.) and KAKENHI (21K15212 to X.G.).
Author information
Authors and Affiliations
Contributions
X.G. and J.P.J. designed the experiments and wrote the manuscript. X.G. carried out the experiments and analysed the results.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature thanks Rony Paz and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data figures and tables
Extended Data Fig. 1 Behavioral data for aversive inference paradigms and baseline activity levels.
(a) Percentage of time spent freezing (‘freezing duration’, y-axis) during auditory and visual CSs in each trial of aversive conditioning (AC, visual-shock pairings) and recall for the paired (grey solid line) and unpaired (black dotted line) groups. BL: baseline. Two-way ANOVA (two-sided); ***: F(1,130) = 43.45, p = 9.9 × 10−10; n.s.: F(1,130) = 0.2, p = 0.65. A, auditory, V, visual. Data are presented as mean value ± SEM. (b) Median percentage (calculated for rats in each group) of time spent freezing during CSs for paired and unpaired groups for auditory (left) and visual (right) CS evoked memory recall. Dots represent individual animal values (filled = male, open = female). Mann-Whitney U test (two-sided); ***: z = 3.84, r = 287, p = 1.2 × 10−4; n.s.: z = 0.34, r = 211, p = 0.73. 2-way ANOVA after log-transformation, interaction (CSs vs groups), F(1,52) = 16.62, p = 1.6 × 10−4. (c) Median percent freezing during the auditory and visual stimulus periods for the paired (left) and unpaired (right) groups during sensory preconditioning. Dots represent individual animal values (filled = male, open = female). Kruskal-Wallis test (two-sided), n.s. (left): Χ2 (2,39) = 0.4, p = 0.82; n.s. (right): Χ2 (2,39) = 2.26, p = 0.32. Two-way ANOVA after log-transformation, interaction (CSs vs groups), F(2,78) = 1.07, p = 0.35. (d) Median percent freezing (calculated for rats in each group) during the auditory and visual stimulus periods for the paired (left) and unpaired (right) groups during memory recall after aversive conditioning. Dots represent individual animal values (filled = male, open = female). Mann-Whitney U test (two-sided), n.s.: z = 1.68, r = 240, p = 0.094; ***: z = 4.30, r = 297, p = 1.7 × 10−5; Two-way ANOVA after log-transformation, interaction (CSs vs groups), F(1,52) = 16.62, p = 1.6 × 10−4. (e) Schematic shows the aversive inference behavioral paradigm with 50% contingency between auditory and visual stimuli during Preconditioning. (f) Percentage of time spent freezing (‘freezing duration’, y-axis) during the auditory and visual CS period of each trial during aversive conditioning (AC) and recall for the paired (grey), unpaired (dashed) and partially paired (green) groups. Data are presented as mean value ± SEM. (g) Median freezing duration (calculated for rats in each group) for the paired, unpaired and 50% paired groups during auditory (left) and visual CS periods (right) at recall. Dots represent individual animal values (filled = male, open = female). Kruskal-Wallis test (two-sided), **: X2 (2, 17) = 9.51, p = 0.0086; n.s.: X2 (2, 17) = 1.31, p = 0.52; Post-hoc test, *: p = 0.049; ##: p = 0.009; n.s.: p = 0.90. (h, i) Schematic showing approximate positions of GRIN lens implantations for each rat in the paired and unpaired groups. (j) Percentage of time spent freezing (‘freezing duration’, y-axis) during auditory and visual CSs in each trial of aversive conditioning (visual-shock pairings) and recall for the paired (grey solid lines) and unpaired (black dotted lines) groups. BL: baseline. Two-way ANOVA (two-sided), **: F(1,50) = 8.24, p = 0.0060; n.s.: F(1,50) = 1.93, p = 0.17. Data are presented as mean value ± SEM. (k) Median percentage (calculated for rats in each group) of time spent freezing during CSs for paired and unpaired groups in the recall of auditory stimulus (left) and visual stimulus (right). Dots represent individual animal values (filled = male, open = female). Mann-Whitney U test (two-sided); *: r = 53, p = 0.026; n.s.: r = 44, p = 0.48. Two-way ANOVA after log-transformation, interaction (CSs vs groups), F(1, 20) = 2.85, p = 0.11. (l) Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for baseline activity during different behavioral stages in the paired (left) and unpaired (right) groups.
Extended Data Fig. 2 Neuronal activity changes induced by learning and Inhibitory responses following preconditioning and aversive learning.
(a) Heatmaps of calcium responses for all imaged neurons in the paired (top) and unpaired (bottom) groups during preconditioning (PC), aversive conditioning (AC) and recall. Each row represents the averaged calcium responses for each trial type of one imaged neuron. The heatmaps of preconditioning and recall are aligned separately for each trial type according to the calcium responses during the auditory stimulus for PC and recall and visual aligned during AC. Dashed lines indicate onset and offset of sensory stimuli. (b) Heat-maps of calcium responses for all imaged neurons in the paired (top) and unpaired (bottom) groups during preconditioning and recall. Each row represents the averaged calcium responses of an individual neuron across each trial type. The heat-maps of preconditioning and recall are aligned separately for each trial type according to the calcium responses during the visual stimulus. Dashed lines indicate onset and offset of sensory stimuli. (c) Top: peri-event time histograms showing population averaged auditory evoked calcium responses (lines) and SEM (shaded) of activated neurons without baseline subtraction during preconditioning for paired (red) and unpaired (grey) groups. Bottom: similar to top panel, but for evoked calcium responses during recall. (d, e) Top: peri-event time histograms showing population averaged auditory-evoked calcium responses (lines) and SEM (shaded) of activated neurons during preconditioning (grey) and recall (red) in the paired (d) and unpaired (e) groups. Dashed lines indicate the onset and offset of sensory stimuli. Bottom: box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium responses during auditory CSs from peri-event time histograms in the top panel. Mann-Whitney U test (one-sided); *: z = 2.08, r = 94077, p = 0.019; n.s.: z = 0.27, r = 55985, p = 0.39. (f, g) Similar to d, e, but for visual responses in the paired (f) and unpaired (g) groups. Mann-Whitney U test (one-sided); * (f): z = 1.84, r = 94640, p = 0.033; * (g): z = 1.94, r = 86503, p = 0.026. (h, i) Top: peri-event time histograms showing population averaged auditory-evoked calcium responses (lines) and SEM (shaded) of inhibited neurons in the unpaired (grey) and paired (red) during preconditioning (h) and recall (i). Bottom: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium responses during auditory CSs from the peri-event time histograms in the top panels. Mann-Whitney U test (two-sided); n.s.: z = 0.29, r = 13345, p = 0.77; n.s.: z = 0.32, r = 12015, p = 0.75. (j, k) Similar to h, i, but for visual CSs during preconditioning (j) and recall (k). Mann-Whitney U test (two-sided); n.s.: z = 1.86, r = 13156, p = 0.063; n.s.: z = 1.29, r = 11203, p = 0.20.
Extended Data Fig. 3 Changes in baseline decoding accuracy following preconditioning and aversive learning.
(a) Peri-event time histograms showing population averaged auditory evoked calcium responses (lines) and SEM (shaded) of activated (solid) and suppressed (dashed) neurons without baseline subtraction during preconditioning (left) and recall (right) for paired group. (b, c) Top: Peri-event time histograms showing decoding accuracy for distinguishing auditory stimuli from baseline activity based on dmPFC population activity (including all neurons) during preconditioning (PC, b) and recall (c) in the paired (‘P’, red, solid) and unpaired (‘unP’, grey, dashed) groups as well as corresponding shuffled paired (‘PS’) and unpaired (‘unPS’) control datasets. Bottom: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) of decoding accuracy for auditory stimuli from the peri-event time histograms in the top panels. Kruskal-Wallis test (two-sided), *: X2 = 11, p = 0.012; ***: X2 = 25.72, p = 1.1 × 10−5; Post-hoc test, p(a,unP-PS) = 0.012; p(b,P-unP) = 0.0013; p(b,P-PS) = 0.0013; p(b,P-unPS) = 1.5 × 10−5; (d, e) Similar to b, c, but for visual CSs during preconditioning (d) and recall (e). Kruskal-Wallis test (two-sided), n.s.: X2 = 4.76, p = 0.19; ***: X2 = 26.85, p = 6.3 × 10−6; Post-hoc test, p(d,P-PS) = 1.6 × 10−4; p(d,unP-PS) = 0.0028; p(d,P-unPS)=0.001; p(d,unP-unPS)=0.013; (f) Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for decoding accuracy (auditory response vs baseline) during preconditioning and recall for the paired group. Mann-Whitney U test; *: z = 2.30, r = 1285, p = 0.022. (g) Similar to f, but for decoding accuracy in the unpaired group. Mann-Whitney U test; *: z = 2.14, r = 1273, p = 0.032. (h) Decoding accuracy (visual response vs baseline) during preconditioning and recall in the paired group. Mann-Whitney U test (two-sided); ***: z = 4.11, r = 1426.5, p = 3.9 × 10−5. (i) Similar to h, but for decoding accuracy in the unpaired group. Mann-Whitney U test (two-sided); **: z = 2.64, r = 1312, p = 0.0082. (j) Auditory (red) and visual (blue) stimulus evoked calcium responses of three example neurons during memory recall. Each trace shows responses on a single trial. (k, l) Spearman correlations between averaged calcium responses in single neurons in response to auditory and visual stimuli for the paired (left) and unpaired (right) groups during preconditioning (k) and recall (l). Auditory-only cells (red), visual-only cells (blue) and co-responsive cells (purple) are denoted by different colors. Each dot represents one neuron. (m) The distribution of correlation coefficients in the unpaired group between auditory and visual CS-evoked calcium responses during preconditioning by bootstrap resampling. The correlation coefficient in the paired group (red line) is within the 95% confidence interval (dashed lines). (n) Similar to m, but for the correlation coefficients of neural response during recall which falls outside of the 95% confidence interval.
Extended Data Fig. 4 Freezing behavior does not alter learning induced changes in dmPFC sensory representations.
(a) Cumulative distribution of auditory (A) and visual (V) selectivity during preconditioning (PC) and recall for paired and unpaired groups. Kolmogorov–Smirnov test (two-sided); **: k = 0.11, p = 0.0057; ##: k = 0.11, p = 0.0024; ***: k = 0.13, p = 4.9 × 10−4. (b) Illustrative schematic diagram showing peri-event time histograms of the activity of 3 exemplar neurons (top) and of the representation of their combined population activity in high dimensional space (bottom). (c) The distribution of angular distances in the unpaired group between auditory and visual CS-evoked population vectors during preconditioning by bootstrap resampling. The angular distance of the paired group (Red line) falls outside of the 95% confidence interval (dashed lines). (d) Similar to c, but for angular distance of population vectors during recall. (e, f) Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for auditory CS-evoked calcium responses in paired and unpaired groups taken from time periods during auditory CSs when rats were not freezing. Auditory: Mann-Whitney U test (two-sided); **: r = 96134, z = 2.76, p = 0.0058; Visual: Mann-Whitney U test; n.s.: r = 94292, z = 0.58, p = 0.56. (g, h) Box plots of median (red lines) and interquartile range (blue) showing decoding accuracy for auditory (g) and visual (h) CSs or corresponding shuffled data relative to corresponding pre-CS baseline in paired and unpaired groups. Auditory: Kruskal-Wallis test (two-sided), ***: X2 = 73.89, p = 6.3 × 10−16; Post-hoc test, p(P-unP) = 3.9 × 10−9; p(P-PS) = 3.8 × 10−9; p(P-unPS) = 3.8 × 10−9; Visual: Kruskal-Wallis test, ***: X2 = 99.17, p = 2.3 × 10−21; Post-hoc test, p(P-PS) = 3.8 × 10−9; p(P-unPS) = 3.9 × 10−9; p(unP-PS) = 3.8 × 10−9; p(unP-unPS) = 3.8 × 10−9. (i) Spearman correlation between neural calcium responses during freezing and non-freezing periods in response to auditory (top) and visual (bottom) stimuli. (j) The Euclidean angular distance (θ) of neural population vectors during auditory and visual CS time periods when rats were not freezing for the paired (left) and unpaired (right) groups during recall. (k) The distribution of angular distances in the Unpaired group in j by bootstrap resampling. Angular distance of paired group (Red line) in j falls outside of the 95% confidence interval (dashed lines). (l-m) Same as in l-m, but for CS time periods in which animals were freezing.
Extended Data Fig. 5 Freezing behavior is encoded in dmPFC neural activity.
(a) Decoding accuracy for distinguishing freezing from non-freezing behavior averaged from each rat in the paired group during auditory (red), visual (blue) CSs and shuffled dataset (grey) in recall. (b) Median decoding accuracy of freezing from dmPFC population activity for experimental (colored bars) and shuffled data (grey bars) during auditory (left) and visual (right) CSs in Recall. Dots represent individual animal values (filled = male, open = female). Mann-Whitney U test (two-sided); ** (left): r = 57, p = 0.0022; ** (right): r = 57, p = 0.0022. (c, d) Left: Peri-event time histograms showing normalized calcium activity around freezing periods during auditory (c) and visual (d) CSs for all activated neurons during recall. SEM=shaded. Right: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium responses before and after (green shading in c, d) freezing onset. Mann-Whitney U test (two-sided), c, ***: r = 103447.5, z = 4.42, p = 1.0 × 10−5; d, **: r = 98567, z = 3.07, p = 0.0022; (e) Calcium response (normalized by subtraction of means) to auditory stimulus and SEM (shaded) around the onset of freezing behaviours for co-responsive neurons (defined by activity during recall). (f) Similar to e, but for auditory-only neurons. (g) Similar to e, but for visual-only neurons.
Extended Data Fig. 6 Decoding accuracy for differentiating auditory and visual representations following preconditioning and during recall.
(a) Left: Peri-event time histogram showing decoding accuracy for distinguishing auditory (A) from visual (V) stimuli based on dmPFC population activity (including all neurons) during preconditioning in the paired groups. Right: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) of decoding accuracy. Mann-Whitney U test (two-sided); *: z = 2.50, r = 1301, p = 0.012; (b) Similar to a, but for decoding accuracy during recall. Mann-Whitney U test (two-sided); ***: z = 6.55, r = 1617, p = 5.6 × 10−11; (c-d) Similar to a-b, but for unpaired group. Mann-Whitney U test (two-sided); ***(c): z = 5.28, r = 1517.5, p = 1.3 × 10−7; ***(d): z = 6.03, r = 1576, p = 1.7 × 10−9; (e) Box plot showing comparison of decoding accuracy (A-V) for paired group during preconditioning and recall. Mann-Whitney U test (two-sided); ***: z = 4.89, r = 1487, p = 1.0 × 10−6; (f) Similar to e, but for the unpaired group. Mann-Whitney U test (two-sided); *: z = 2.04, r = 1265, p = 0.041; (g) Similar to e, but for decoding accuracy during recall comparing paired and unpaired groups. Mann-Whitney U test (two-sided); *: z = 2.03, r = 1264, p = 0.043. (h, i) Spearman correlations between averaged calcium responses in single cells in response to auditory stimuli during recall and foot shocks during aversive conditioning for the paired (h) and unpaired (i) groups. Each dot represents one neuron. (j, k) Similar to h, i, but for the correlation between averaged Calcium responses in single neurons in response to visual stimuli during recall and foot shock during aversive conditioning for the paired (j) and unpaired (k) groups. (l) The distribution of correlation coefficients in the unpaired group between auditory responses during Recall and US responses during Aversive Conditioning by bootstrap resampling. The correlation coefficient in the paired group (red line) falls outside of the 95% confidence interval (dashed lines). (m) Similar to l, but for the correlation coefficient between visual response during recall and US response during aversive conditioning.
Extended Data Fig. 7 Merging of shock and inferred memory representations after aversive learning and preconditioning induced changes in baseline activity.
(a) Euclidean angular distance of auditory (A) stimulus and shock-evoked (US) neural population vectors for the paired (left) and unpaired (right) groups following preconditioning (top) and during recall (bottom). (b) Angular distance (median calculated for all rats in each group) between auditory stimulus and shock-evoked population vectors in the paired and unpaired groups following preconditioning (left) and during recall (right). Mann-Whitney U test (two-sided); n.s.: r = 36, p = 0.70; **: r = 23, p = 0.0087. Two-way ANOVA, interaction (CSs vs groups), F(1, 20) = 1.63, p = 0.22. (c) The distribution of angular distances in the unpaired group between population vectors in response to shock US during aversive conditioning and auditory CS during preconditioning by bootstrap resampling. The angular distance of the paired group (red line) is within the 95% confidence interval (dashed lines). (d) Similar to c, but for angular distance of population vectors in response to shock US during aversive conditioning and auditory CS during recall. The angular distance of the paired group (red line) falls outside of the 95% confidence interval (dashed lines). (e) The Euclidean angular distance of visual stimulus and shock-evoked neural population vectors for the paired (left) and unpaired (right) groups following preconditioning (top) and during recall (bottom). (f) Angular distance (median calculated for all rats in each group) between visual stimulus and shock-evoked population vectors in the paired and unpaired groups following preconditioning (left) and during recall (right). Mann-Whitney U test (two-sided); n.s. (left): r = 42, p = 0.70; n.s.(right): r = 35, p = 0.59. Two-way ANOVA, interaction (CSs vs groups), F(1, 20) = 0.69, p = 0.42. (g) Similar to c, but for angular distance of population vectors in response to shock US during aversive conditioning and visual CS during preconditioning. (h) Similar to c, but for angular distance of population vectors in response to shock US during aversive conditioning and visual CS during recall. The angular distance of the paired group (red line) falls outside of the 95% confidence interval (dashed lines). (i, j) Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium activity of auditory-only (‘Audit’), visual only (‘Visual’) and co-responsive (‘Co’) neurons in paired (i) and unpaired (j) groups during baseline before 1st trial of day 1 preconditioning. Paired: Kruskal-Wallis test (two-sided), n.s.: X2 (2, 86) = 0.18, p = 0.92; Unpaired: Kruskal-Wallis test (two-sided), n.s.: X2 (2, 86) = 2.22, p = 0.33; (k) Similar to i, but for baseline activity in the Paired group before day 2 preconditioning. Kruskal-Wallis test (two-sided), *: X2 (2, 116) = 8, p = 0.018; Post-hoc test, #: p = 0.046; &: p = 0.016; (l) Similar to k, but for baseline activity in the unpaired group before day 2 preconditioning. Kruskal-Wallis test (two-sided), n.s.: X2 (2, 101) = 3.94, p = 0.14; (m) Peri-event time histograms showing population averaged auditory and visual evoked calcium responses (lines, normalized by baseline subtraction) and SEM (shaded) of co-responsive neurons (defined by sensory test after day 2 preconditioning) during the first 5 trials in day 1 and day 2 preconditioning. (n) Similar to m, but for auditory and visual stimulus evoked calcium responses in sensory test (when co-responsive neurons are defined) after day 2 preconditioning. (o) Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for auditory stimulus evoked calcium responses during day1 preconditioning early 5 trials and day 2 sensory test for day 2 co-responsive neurons. Mann-Whitney U test (two-sided); *: z = 2.31, r = 450, p = 0.021; (p) Similar to o, but for visual stimulus evoked Calcium response. Mann-Whitney U test (two-sided); n.s.: z = 0.18, r = 377, p = 0.86. For b and f, filled dots denote male and open indicate female.
Extended Data Fig. 8 Enhanced responsiveness of co-responsive cells following preconditioning and stabilization of their responding into the aversive learning phase.
(a, b) Left: Population averaged peri-event time histograms showing auditory evoked calcium activity (lines) and SEM (shaded) of activated auditory-only (red) and co-responsive (Co, purple) neurons during preconditioning (PC) for the paired (a) and unpaired (b) groups. Right: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium responses during auditory CSs for the cell classes in the adjacent peri-event time histograms; Mann-Whitney U test (one-sided); *: z = 2.18, r = 2005, p = 0.015; n.s.: z = 1.17, r = 939, p = 0.12. Two-way ANOVA, interaction (cell-types vs groups), F(1, 180) = 0.68, p = 0.41. (c, d) Similar to a-b but for visual responses of the paired (c) and unpaired (d) groups. Mann-Whitney U test (one-sided); *: z = 1.85, r = 2410, p = 0.032; n.s.: z = 1.52, r = 1091, p = 0.064. Two-way ANOVA, interaction (cell-types vs groups), F(1, 213) = 0.28, p = 0.59. (e, f) Spearman correlations between averaged auditory evoked calcium responses in single cells following preconditioning and during recall for the paired (e) and unpaired (f) groups. Each dot represents one neuron. (g, h) Similar to e-f, but for visual responses in the paired (g) and unpaired (h) groups. (i) The distribution of correlation coefficients in the unpaired group between auditory (A) responses during recall and preconditioning by bootstrap resampling. The correlation coefficient in the paired group (red line) falls outside of the 95% confidence interval (dashed lines). (j) Similar to i, but for the correlation coefficients between visual (V) response during recall and preconditioning. (k) Removing cells that were co-responsive during preconditioning (PC) produced a significant shift (beyond 95% confidence interval, CI) in the correlation between visual responsiveness during preconditioning and that during recall compared to a distribution generated by iterative removal of an equivalent number of randomly selected neurons. (l) Representative image of inhibitory opsin expression in dmPFC injection sites for optogenetic inhibition through bilateral optical fiber implants. (m) Schematic showing optical fiber placements (approximated AP position) and eArch3.0-EYFP expression for each rat in the opsin treated group.
Extended Data Fig. 9 Unpaired aversive conditioning after preconditioning did alter dmPFC representations.
(a) Schematic shows the unpaired aversive conditioning component of the modified sensory preconditioning paradigm. (b) Percentage of time spent freezing (‘freezing duration’, y-axis) during the CS period of each trial during aversive conditioning (AC) and recall for the paired-AC (dashed) and unpaired-AC (black) groups. (c) Median freezing duration (calculated for rats in each group) for the paired and unpaired groups during auditory (left) and visual CS periods (right) at recall. Dots represent individual animal values (filled = male, open = female). Mann-Whitney U test (two-sided); **(left): r = 16, p = 0.0043; **(right): r = 15, p = 0.0043; (d) Left: peri-event time histogram showing population averaged auditory-evoked calcium responses (lines) and SEM (shaded) of activated neurons during preconditioning (PC, grey) and recall (red) in the unpaired-AC groups. Dashed lines represent the onset and offset of sensory stimuli. Right: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for auditory stimulus evoked calcium responses from peri-event time histograms in the left panel. Mann-Whitney U test (one-sided); n.s.: z = 0.25, r = 73698, p = 0.40; (e) Similar to d, but for calcium responses during visual CSs. Mann-Whitney U test (one-sided); n.s.: z = 1.13, r = 87535, p = 0.13; (f) The Euclidean angular distance (θ) of auditory and visual stimulus evoked population vectors for the unpaired-AC group during preconditioning (left) and recall (right). (g) The distribution of angular distances between auditory (A) and visual (V) stimulus-evoked population vectors for the unpaired-AC group during preconditioning by bootstrap resampling. Angular distance during Recall is within the 95% CI. (h) Euclidean angular distance of auditory stimulus and shock-evoked neural population vectors for the AC-paired (left) and AC-unpaired (right) groups during recall. (i) The distribution of angular distances between auditory CS during Recall and shock US during AC drawn from the AC-unpaired by bootstrap resampling. Angular distance of the AC-paired group (red line) falls outside of the 95% confidence interval (dashed lines). (j, k) Similar to h, i, but for the population vectors of visual CSs and US. (l) Left: Peri-event time histogram showing auditory-evoked calcium responses at recall for all (black), auditory-only (red) and co-responsive (Co, purple) cells classified during preconditioning in the AC-unpaired groups. Right: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for sensory stimulus evoked calcium responses from cell classes. Kruskal-Wallis test (two-sided), n.s.: X2 (2,303) = 3.24, p = 0.20; (m) Similar to l, but for visual stimulus-evoked calcium responses at recall. Kruskal-Wallis test (two-sided), n.s.: X2 (2, 323) = 0.92, p = 0.63; (n) Left: Peri-event time histogram showing calcium responses to visual stimuli (20 s time-period within dotted vertical lines) and shocks (denoted by green rectangle) during unpaired aversive conditioning for cells which were identified as auditory-only, visual-only, co-responsive and others (non-significantly responsive and inhibited cells) during recall in the AC-Unpaired group. Right: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for averaged shock-evoked calcium responses (from traces in the left panel). Kruskal-Wallis test (two-sided), n.s.: X2 (3, 327) = 5.61, p = 0.13. (o) Similar to n, but for cells which were identified as auditory-only, visual-only, co-responsive and others after preconditioning. Kruskal-Wallis test (two-sided), n.s.: X2 (3, 282) = 0.83, p = 0.84.
Extended Data Fig. 10 Extinction induced changes in dmPFC neural processing.
(a) Schematic showing approximate positions of GRIN lens implantations for each rat. (b) Percentage of time spent freezing (‘freezing duration’, y-axis) during auditory and visual CSs in each trial of aversive conditioning (visual-shock pairings), extinction and test for the auditory extinction group. Data are presented as mean value ± SEM. (c) Top: Peri-event time histograms showing auditory evoked calcium responses of activated neurons during preconditioning (PC, grey) and post-extinction test (red). Bottom: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for averaged calcium responses during auditory stimuli in the top panel. Mann-Whitney U test (two-sided); n.s.: z = 1.16, r = 79788, p = 0.24. (d) Similar to c, but for visual responses. Mann-Whitney U test (two-sided); ***: z = 3.51, r = 98351, p = 4.5 × 10−4. (e) Top: Peri-event time histogram showing calcium responses of activated neurons during auditory (red) and visual (blue) stimuli during recall for the paired group w/o extinction. Bottom: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for averaged calcium responses during sensory stimuli in the top panel. Mann-Whitney U test (two-sided); n.s.: z = 1.61, r = 95902, p = 0.11. (f) Top: Peri-event time histogram showing decoding accuracy for distinguishing auditory stimuli from baseline period based on dmPFC population activity (including all neurons) in response to auditory stimuli during early auditory extinction (A-EX trial 1-5, light red), auditory CS during test (dark red) and visual CS during test (blue). Bottom: Box plot showing median (red lines) and interquartile range (blue) for decoding accuracy from data in the top panel. Kruskal-Wallis test (two-sided), **: Χ2 = 13.55, p = 0.0011. Post-hoc test, *: p = 0.045; ***: p = 8.6 × 10−4. (g) The Euclidean angular distance (θ) between auditory and visual stimulus evoked neural population vectors for the auditory extinction group during preconditioning (left) and test after auditory extinction (right). (h) The distribution of angular distances between auditory (A) and visual (V) stimulus evoked population vectors in the auditory extinction group during preconditioning by bootstrap resampling. The angular distance of the population vector during test (red line) is within the 95% confidence interval (dashed lines). (i) Spearman correlation between averaged calcium responses in single cells in response to auditory stimuli during post-extinction test and foot shock during aversive conditioning. Each dot represents one neuron. (j) Similar to i, but for visual responses during test and foot shock responses during conditioning. (k) Similar to i, but for auditory responses in early extinction and foot shock responses during conditioning. (l) Pie-chart showing the percentage of shock-activated cells during aversive conditioning which were subsequently reactivated by auditory or visual stimuli during recall in the paired group w/o extinction. (m) Left: Peri-event time histograms showing calcium responses to auditory CSs at recall without extinction (solid line) or at test following extinction (dashed line) for cells which were identified as shock-activated during aversive conditioning. Right: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium responses during sensory CSs in the left panel. Mann-Whitney U test (two-sided); **: z = 2.80, r = 12168, p = 0.0051. (n) The Euclidean angular distance (θ) between shock US and auditory stimulus evoked population vectors during preconditioning, across extinction trial blocks and test. (o) the distribution of angular distances between US and auditory stimulus evoked population vectors during preconditioning by bootstrap resampling. The angular distance during test after extinction (red line) is within the 95% confidence interval (CI, dashed lines), while the angular distance during early extinction (orange line) falls outside of the 95% CI. (p) The Euclidean angular distance (θ) of shock US and visual stimulus evoked neural population vectors during US and PC (left)/Test (right) for the auditory extinction group. (q) The distribution of angular distances between the US and visual stimulus evoked population vectors during preconditioning by bootstrap resampling. The angular distance during test (red line) fall outside of the 95% confidence interval (dashed lines).
Extended Data Fig. 11 The destabilization of co-responsive neurons after the extinction of inferred aversive memory and the neural activity during and after extinction of directly associated aversive memory.
(a) Percentage of activated cells at recall (filled) that were sensory activated during preconditioning (PC, entire rectangle) for co-responsive cells in the auditory extinction and non-extinction groups. Chi-square test (two-sided); *: X2 = 4.21, p = 0.040; (b) Left: peri-event time histogram showing population averaged auditory-evoked calcium responses (lines) and SEM (shaded) of co-responsive neurons following day 2 Preconditioning (grey) and auditory extinction (A-EX) test (red). Dashed lines represent the onset and offset of sensory stimuli. Bottom: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium responses during auditory CSs from peri-event time histograms in the left panel. Mann-Whitney U test (one-sided); **: z = 2.87, r = 266, p = 0.0020; (c) Similar to b, but for calcium responses during visual CSs. Mann-Whitney U test (one-sided); **: z = 3.06, r = 270, p = 0.0011; (d) Schematic shows the behavioral paradigm for extinguishing direct aversive memories. A, auditory, V, visual. (e) Left: Peri-event time histogram showing visual CS-evoked calcium responses of activated neurons during the 1st-4th block of extinction training for the visual extinction group. Right: Averaged (from responses in left panel) calcium responses during visual CSs across extinction training. Kruskal-Wallis test (two-sided), *: X2 (3, 1400) = 10.08, p = 0.018; Post-hoc test; #: p = 0.035; &: p = 0.036; (f) Left: Peri-event time histogram showing auditory and visual CS-evoked calcium responses of activated neurons during test for the visual extinction (V-EX) group. Right: Averaged (based on responses in left) calcium responses during auditory CSs. Mann-Whitney U test(two-sided); n.s.: z = 0.41, r = 104967, p = 0.68; (g) Pie chart showing the percentage of US-activated neurons which were subsequently reactivated by auditory and visual stimuli during early extinction (trials 1–5) and extinction Test following visual extinction. Chi-square Test (two-sided); n.s.(top): X2 = 0.36, p = 0.54; n.s.(bottom): X2 = 0.39, p = 0.53; (h) Left: Peri-event time histogram showing CS-evoked calcium responses during early extinction (grey) and auditory (red) and visual (blue) CS-evoked responses during test for cells which were shock-activated during aversive learning. Right: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for averaged (based on responses in left panel) auditory and visual CS-evoked calcium responses. Kruskal-Wallis test (two-sided), n.s.: X2 (2, 308) = 5.54, p = 0.063.
Extended Data Fig. 12 Histology, behavior and learning induced changes in response properties for experiments involving imaging from amygdala projecting dmPFC cells.
(a) Schematic showing the approximate positions of GRIN lens implantations for each rat in the paired (left) and unpaired (right) groups. (b) Percentage of time spent freezing (‘freezing duration’, y-axis) during auditory and visual CSs in each trial of aversive conditioning (visual-shock pairings) and recall for the paired (grey solid lines) and unpaired (black dotted lines) groups. BL: baseline. Two-way ANOVA (two-sided), ***: F(1,65) = 16.85, p = 0.00012; n.s.: F(1,65) = 3.88, p = 0.053. Data are presented as mean value ± SEM. (c) Median percentage (calculated for rats in each group) of time spent freezing during CSs for paired and unpaired groups in the recall of auditory stimulus (left) and visual stimulus (right) for the paired and unpaired groups. Dots represent individual animal values (filled = male, open = female). Mann-Whitney U test (two-sided); *: r = 74, p = 0.040; n.s.: r = 65, p = 0.34; Two-way ANOVA after log-transformation, interaction (CSs vs groups), F(1, 26) = 1.51, p = 0.23. (d) Heat-maps of calcium responses for all imaged neurons projecting to amygdala in the paired (top) and unpaired (bottom) groups during preconditioning and recall. Each row represents the averaged calcium responses of an individual neuron across each trial type. The preconditioning and recall heatmaps are aligned separately for each trial type (A, auditory and V, visual) according to the calcium responses during the visual stimulus. Dashed lines indicate onset and offset of sensory stimuli. (e, f) Top: peri-event time histograms showing population averaged auditory-evoked calcium responses (lines) and SEM (shaded) of activated neurons during preconditioning (PC, grey) and recall (red) in the paired (e) and unpaired (f) groups. Dashed lines represent the onset and offset of sensory stimuli. Bottom: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium responses during auditory CSs from peri-event time histograms in the top panel. Mann-Whitney U test (one-sided); **: z = 2.81, r = 35301, p = 0.0025; n.s.: z = 1.18, r = 29368, p = 0.12. Two-way ANOVA, interaction (stages vs groups), F(1, 698) = 9.66, p = 0.002. (g, h) Similar to e-f, but for visual responses in the paired (g) and unpaired (h) groups. Mann-Whitney U test (one-sided); n.s.(left): z = 0.65, r = 33124, p = 0.26; n.s.(right): z = 0.26, r = 37321, p = 0.40. Two-way ANOVA, interaction (stages vs groups), F(1, 718) = 0.050, p = 0.82.
Extended Data Fig. 13 Preconditioning with stimulus switching, learning induced changes in response properties of inhibited neurons and decoding from dmPFC-amygdala projecting neurons.
(a) Schematic shows the aversive inference behavioral paradigm in which the auditory CS was directly paired with shock and the visual CS was the inferred stimulus. A, auditory, V, visual. (b) Left: peri-event time histograms showing calcium responses (lines) and SEM (shaded) of activated neurons during visual CS presentations in preconditioning (PC, grey) and recall (blue). Right: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium responses during visual CSs of Preconditioning and Recall. Mann-Whitney U test (two-sided); *: z = 2.10, r = 8381, p = 0.036; (c) Similar to b, but for calcium response during auditory CSs. Mann-Whitney U test (two-sided); n.s.: z = 0.78, r = 7177, p = 0.432; (d) Top: peri-event time histograms showing population averaged auditory-evoked calcium responses (lines) and SEM (shaded) of inhibited neurons in the unpaired (grey) and paired (red) groups during preconditioning. Bottom: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium responses during auditory CSs from the peri-event time histograms in the top panels. Mann-Whitney U test (two-sided); n.s.: z = 0.0011, r = 13376, p = 0.99. (e) Similar to d, but for auditory responses during recall. Mann-Whitney U test (two-sided); n.s.: z = 0.75, r = 10903, p = 0.45. (f) Similar to d, but for visual responses during preconditioning. Mann-Whitney U test (two-sided); n.s.: z = 0.11, r = 9044, p = 0.91. (g) Similar to d, but for visual responses during recall. Mann-Whitney U test (two-sided); n.s.: z = 0.10, r = 11058, p = 0.92. (h-i) Top: Peri-event time histograms showing decoding accuracy for distinguishing auditory stimuli from baseline activity based on activity of dmPFC neurons projecting to amygdala during preconditioning (h) and recall (i) in the paired (red, solid) and unpaired (grey, dashed) groups as well as corresponding shuffled control datasets. Bottom: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) of decoding accuracy for auditory stimuli from the peri-event time histograms in the top panels. Kruskal-Wallis test (two-sided), n.s.: X2 = 4.62, p = 0.20; ***: X2 = 34.9, p = 1.3 × 10−7; Post-hoc test, %%%: p = 1.3 × 10−4; & & &: p = 4.3 × 10−6; ###: p = 2.0 × 10−6; (j-k) Similar to h-i, but for visual CSs during preconditioning (j) and recall (k). Kruskal-Wallis test (two-sided), *** (j): X2 = 22.55, p = 5.0 × 10−5; *** (k): X2 = 36.27, p = 6.6 × 10−8; Post-hoc test, p(g,unP-Ps) = 0.0031; p(g,unP-unPs) = 4.0 × 10−5; p(h,P-Ps) = 3.3 × 10−5; p(h,unP-unPs) = 5.4 × 10−4; p(h,unP-Ps) = 2.0 × 10−5; p(h,P-unPs) = 8.2 × 10−4. (l) Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for decoding accuracy (auditory response vs baseline) during preconditioning and recall for the paired group. Mann-Whitney U test (two-sided); ***: z = 3.32, r = 1365, p = 9.0 × 10−4. (m) Similar to l, but for decoding accuracy in the unpaired group. Mann-Whitney U test (two-sided); n.s.: z = 1.09, r = 1191, p = 0.28. Two-way ANOVA, interaction (l-m, stages vs groups), F(1) = 12.35, p = 0.0006. (n) Decoding accuracy (visual response vs baseline) during preconditioning and recall in the paired group. Mann-Whitney U test (two-sided); ***: z = 4.14, r = 1429, p = 3.4 × 10−5. (o) Similar to n, but for decoding accuracy in the unpaired group. Mann-Whitney U test (two-sided); **: z = 2.86, r = 1329, p = 0.0042. Two-way ANOVA, interaction (n-o, stages vs groups), F(1) = 2.43, p = 0.12.
Extended Data Fig. 14 Freezing related neural activity of amygdala projecting neurons and lack of auditory and visual CS representational merging after conditioning.
(a, b) Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for auditory CS-evoked calcium responses in paired and unpaired groups taken from time periods during auditory CSs when rats were not freezing. Mann-Whitney U test (two-sided); ***: r = 35608, z = 4.73, p = 2.2 × 10−6; n.s.: r = 27464, z = 0.95, p = 0.34; (c) Decoding accuracy of freezing from non-freezing behavior averaged from each rat in the paired group during auditory (red), visual (blue) CSs and shuffled dataset (grey) in recall. (d) Median decoding accuracy (calculated for each rat) of freezing from the activity of amygdala projecting neurons for experimental and shuffled data during auditory (left) and visual (right) CSs in recall. Dots represent individual animal values (filled = male, open = female). Mann-Whitney U test (two-sided); *** (left): r = 77, p = 5.8 × 10−4; *** (right): r = 77, p = 5.8 × 10−4. (e, f) Left: Peri-event time histogram showing normalized calcium activity around freezing periods occurring during auditory (e) and visual (f) CSs for all activated neurons during recall. SEM = shaded. Right: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium responses before and after (green shading in e, f) freezing onset. Mann-Whitney U test (two-sided), e, **: r = 34850.5, z = 3.17, p = 0.0015; f, n.s.: r = 25543, z = 1.01, p = 0.31. (g, h) Similar to e, f, but for calcium responses around the offset of freezing. g: Mann-Whitney U test (two-sided); n.s.: r = 32518, z = 0.77, p = 0.44; h: Mann-Whitney U test (two-sided); n.s.: r = 25516, z = 0.98, p = 0.33. (i) The Euclidean angular distance (θ) of auditory and visual stimulus evoked neural population vectors for the paired (left) and unpaired (right) groups during preconditioning (PC, top) and recall (bottom) in amygdala projecting dmPFC neurons. (j) The distribution of angular distance between auditory (A) and visual (V) stimulus evoked population vectors in the unpaired group during preconditioning (left) and recall (right) by bootstrap resampling in amygdala projecting dmPFC neurons. Angular distances in the paired group marked as red lines is within the 95% confidence intervals (CI) at preconditioning and recall. (k) Left: Median (of all rats in each group) Spearman correlation coefficient between calcium responses during auditory stimuli in recall and during foot-shocks of aversive conditioning for each rat in the paired and unpaired groups. Dots represent individual animal values (filled = male, open = female). Right: Similar to the left panel, but for the correlation coefficient between calcium responses during visual stimuli in recall and during foot-shocks of aversive conditioning. Mann-Whitney U test (two-sided); *: r = 74, p = 0.040; n.s.: r = 68, p = 0.19. (l) Pie-chart showing the percentage of shock-activated cells during aversive conditioning which were subsequently reactivated by auditory-only, visual-only or auditory and visual (Co, co-responsive) CSs at memory recall in the paired and unpaired groups. Chi-square Test (two-sided); *: X2 = 3.98, p = 0.046; n.s.(blue): X2 = 1.30, p = 0.25; n.s.(purple): X2 = 2.54, p = 0.11.
Extended Data Fig. 15 Shock and auditory-inferred representations become selectively merged in dmPFC-amygdala neurons following aversive learning.
(a) Left: Peri-event time histograms showing calcium responses to auditory CSs at recall for cells which were identified as shock-activated during aversive conditioning in the paired (red) and unpaired (grey) groups. Right: Box plots showing median (center), 25%/75% interquartile range (bounds of box) and extreme data points (whiskers) for calcium responses during auditory-CSs in the left panel. Mann-Whitney U test (two-sided); *: z = 2.0692, r = 4227, p = 0.039. (b) Similar to a, but for visual CS-evoked responses during recall in shock US-activated cells identified during aversive conditioning in the paired (blue) and unpaired (grey) groups. Mann-Whitney U test (two-sided); n.s.: z = 0.32, r = 3877, p = 0.75. (c) After preconditioning (PC), the Euclidean angular distance (θ) of shock US and PC auditory (top)/visual (bottom) stimulus evoked neural population vectors for the paired and unpaired groups. (d) The distribution of angular distances between US and PC auditory (A, left) /visual (V, right) stimulus evoked population vectors from the unpaired group by bootstrap resampling. Angular distances for the paired group marked as red lines are inside the 95% confidence intervals (CI). (e, f) Similar to c, d, but for angular distance between US and recall auditory (top)/visual (bottom) neural population vectors. For f, the angular distance between the US and auditory CS evoked population vectors falls outside of the 95% CI, but for the US and visual CS it does not. (g) Schematic showing approximate optical fiber placements and eArch3.0-EYFP expression for each rat in the opsin treated group. (h) Left: Enlarged image from Fig. 5iright showing dmPFC axonal expression in amygdala. Right: Merged results from all rats dmPFC axonal expression of Arch3.0 in amygdala. (i) Schematic shows the aversive inference behavioral paradigm in which the auditory CS was directly paired with shock and the visual CS was the inferred stimulus. (j) Percentage of time spent freezing (‘freezing duration’, y-axis) during the CS period of each trial during aversive conditioning and visual and auditory CS-evoked memory recall for the optogenetic (yellow) and control (black) groups. Two-way ANOVA (two-sided); ***: F(1, 70) = 13.54, p = 0.0005; n.s.: F(1, 70) = 1.93, p = 0.17. Data are presented as mean value ± SEM. (k) Median freezing duration (calculated for rats in each group) for the optogenetic and control groups during visual (left) and auditory CS periods (right) at recall. Dots represent individual animal values (filled = male, open = female). Mann-Whitney U test (two-sided); *: r = 46, p = 0.021; n.s.: r = 61.5, p = 0.52. Two-way ANOVA after log-transformation, interaction (CSs vs groups), F(1, 28) = 4.64, p = 0.040.
Supplementary information
Supplementary Information
Supplementary Discussion, Supplementary Figures and Supplementary References.
Source data
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Gu, X., Johansen, J.P. Prefrontal encoding of an internal model for emotional inference. Nature 643, 1044–1056 (2025). https://doi.org/10.1038/s41586-025-09001-2
Received:
Accepted:
Published:
Issue date:
DOI: https://doi.org/10.1038/s41586-025-09001-2
