A neural mechanism for learning from delayed postingestive feedback

Abstract

Animals learn the value of foods on the basis of their postingestive effects and thereby develop aversions to foods that are toxic^1-10 and preferences to those that are nutritious^11-13. However, it remains unclear how the brain is able to assign credit to flavours experienced during a meal with postingestive feedback signals that can arise after a substantial delay. Here we reveal an unexpected role for the postingestive reactivation of neural flavour representations in this temporal credit-assignment process. To begin, we leverage the fact that mice learn to associate novel^14,15, but not familiar, flavours with delayed gastrointestinal malaise signals to investigate how the brain represents flavours that support aversive postingestive learning. Analyses of brain-wide activation patterns reveal that a network of amygdala regions is unique in being preferentially activated by novel flavours across every stage of learning (consumption, delayed malaise and memory retrieval). By combining high-density recordings in the amygdala with optogenetic stimulation of malaise-coding hindbrain neurons, we show that delayed malaise signals selectively reactivate flavour representations in the amygdala from a recent meal. The degree of malaise-driven reactivation of individual neurons predicts the strengthening of flavour responses upon memory retrieval, which in turn leads to stabilization of the population-level representation of the recently consumed flavour. By contrast, flavour representations in the amygdala degrade in the absence of unexpected postingestive consequences. Thus, we demonstrate that postingestive reactivation and plasticity of neural flavour representations may support learning from delayed feedback.

Fig. 1. Novel flavours support one-shot, delayed CFA learning and preferentially activate the amygdala at every stage of learning.

a, Schematic of the CFA paradigm. i.p., intraperitoneal injection. b, Flavour preference across three consecutive daily retrieval tests for mice that consumed either a novel (top) or familiar (bottom; all statistical tests not significant (NS)) flavour and then were injected with either LiCl or saline on the conditioning day (n = 8 mice per group). The flavour (sweetened grape Kool-Aid) and amount consumed (1.2 ml) was the same for all groups. The group given a familiar flavour was pre-exposed to the flavour on four consecutive days before conditioning, whereas the group given the novel flavour was completely naive. c, Example FOS imaging data (100 µm maximum-intensity projection) and cell detection results from the brain-wide light-sheet microscopy imaging pipeline. Scale bars, 1 mm (left), 100 µm (bottom right) and 25 µm (top right). d, Description of the GLMM for the brain-wide FOS dataset (n = 12 mice per flavour condition per time point for e–i; see Methods and equation (2)). e, Novel – familiar ΔFOS effect distribution at each time point across all significantly modulated brain regions (n = 130 brain regions). Each point represents a single brain region. f, Hierarchical clustering of novel – familiar ΔFOS effects (see Extended Data Fig. 3d for an expanded version). g, Detail of the amygdala network (cluster 1 from f) that is preferentially activated by novel flavours at every stage of learning. h, Visualization of the difference in FOS⁺ cell density across flavour conditions with Allen CCF boundaries overlaid. i, Comparison of individual mice for the novel and familiar flavour conditions for the CEA at each time point. Error bars represent the mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01, ****P ≤ 0.0001. See Supplementary Table 2 for details of statistical tests and for exact P values. See Supplementary Table 1 for a list of brain-region abbreviations and for GLMM statistics. Source data

Fig. 2. CGRP neurons mediate the effects of postingestive malaise on the amygdala, and monosynaptic connections to the CEA support the acquisition of delayed CFA.

a, Schematic of the pathway that conveys malaise signals to the amygdala. b, CGRP neurons are activated in vivo by LiCl-induced malaise (n = 5 mice). c, Top, schematic of the slice electrophysiology experiment. Bottom, traces showing strong monosynaptic connections from CGRP neurons to the CEAc and CEAl and weaker connections to the CEAm (n = 5 neurons from 3 mice per region). Dark lines represent the average and transparent lines represent individual trials for example neurons. oEPSC, optically evoked excitatory postsynaptic current; TTX, tetrodotoxin; 4AP, 4-aminopyridine. d, Top, schematic for the CGRP neuron stimulation experiment. Bottom, example image of ChR2–YFP expression and data for the retrieval test (n = 6 mice per group). e, Left, schematic of the CGRP^CEA projection stimulation experiment. Middle, example image of ChRmine–mScarlet expression. Right, retrieval test data (n = 6 mice per group). f, Left, schematic of the CGRP^CEA projection inhibition experiment. Middle, example image of eOPN3–mScarlet expression. Right, retrieval test data (n = 11 mice for eOPN3, 9 mice for YFP). g, Schematic of the CGRP neuron stimulation FOS experiment (n = 14 mice for novel flavour, 13 mice for familiar flavour for h–k). ITI, inter-trial interval. h, Summary of FOS⁺ cell counts in the CEA for individual mice. i, Correlation between average FOS⁺ cell count for LiCl-induced malaise versus CGRP neuron stimulation (n = 12 for the amygdala network, 117 for the other regions). j, Analogous to i, but comparing the difference between flavour conditions. k, Visualization of the difference in FOS⁺ cell density across flavour conditions. l, Schematic of the FISH experiment. m,n, Comparison of marker gene expression (m) and co-expression (n) (n = 6 mice for novel flavour, 7 mice for familiar flavour; 490 ± 54 Fos⁺ neurons per mouse (mean ± s.e.m.); all statistical tests NS). Error bars represent the mean ± s.e.m. Shaded areas in b represent the mean ± s.e.m. and in i and j represent the linear fit estimate ±95% confidence intervals. Units in j are per cent per mm³. *P ≤ 0.05, **P ≤ 0.01, ****P ≤ 0.0001. See Supplementary Table 2 for details of statistical tests and for exact P values. Scale bars, 1 mm (d–f) or 100 µm (l). Source data

Fig. 3. Postingestive CGRP neuron activity preferentially reactivates the representation of a recently consumed flavour in the amygdala.

a, Hypotheses for how the amygdala associates temporally separated flavour and malaise signals. b, Schematic of the CGRP neuron-stimulation recording strategy. c, Reconstruction of recording trajectories registered to the Allen CCF. Each line represents one shank of a four-shank Neuropixels 2.0 probe (32 shanks from 8 mice). d, Average spiking of all individual neurons (n = 1,104 single units and multiunits from 8 mice). e, Average spiking of novel-flavour-preferring (n = 373), water-preferring (n = 121) and nonselective (n = 610) populations. f, Left, average spiking of individual neurons during CGRP neuron-stimulation bouts. Right, population averages (same sample sizes as e). g, Example of a multinomial logistic regression decoder session. h, Average decoder posterior time locked to CGRP neuron stimulation (mean across 6 mice). i, Average reactivation rates of novel flavour or water representations (n = 6 mice). j, PCA schematic. k, Population trajectories for novel-flavour consumption, water consumption and CGRP neuron stimulation. l, Trajectories for individual example mice. m, Schematic of the LiCl-induced-malaise recording strategy. n, Average spiking of novel-flavour-preferring (n = 280 neurons from 4 mice), water-preferring (n = 80) and nonselective (n = 218) populations. o, Average reactivation rates of novel flavour or water representations (n = 4 mice). p, Example CGRP immunoreactivity data confirming the ablation of CGRP neurons in the PB (outlined in green). Scale bar, 100 µm. q, Analogous to n, but for mice in which CGRP neurons were ablated (n = 124 novel-flavour-preferring, 20 water-preferring, 256 nonselective neurons from 4 mice; all statistical tests NS). r, Analogous to o, but for mice in which CGRP neurons were ablated (n = 4 mice). Shaded areas represent the mean ± s.e.m. Inset box plots show the 10th, 25th, 50th, 75th and 90th percentiles. **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. See Supplementary Table 2 for details of statistical tests and for exact P values. Source data

Fig. 4. Postingestive CGRP neuron activity induces plasticity to stabilize flavour representations in the amygdala upon memory retrieval.

a, Spike waveforms, autocorrelograms (Autocorr.) and flavour response rasters for an example neuron tracked across conditioning and retrieval days. b, Average spiking of all tracked neurons during the consumption, delay and CGRP neuron-stimulation periods on the conditioning day and during consumption on the retrieval day (n = 939 neurons from 8 mice). c, Left, average spiking of the novel-flavour-preferring population (n = 265 neurons) during flavour consumption on conditioning and retrieval days. Middle and right, average spiking of novel-flavour-preferring neurons with the highest 10% CGRP response magnitudes and of the remaining novel-flavour-preferring neurons. d, Correlation between the change (retrieval – conditioning) in flavour response or selectivity for each neuron during consumption to its average response during the CGRP neuron-stimulation period. The novel-flavour-preferring (n = 265 neurons), water-preferring (n = 123 neurons) and nonselective (n = 551 neurons) populations are shown separately. e, Analogous to d, but for mice with CGRP^CEA projection stimulation (n = 286 novel-flavour-preferring neurons from 8 mice). f, Left, schematic of the flavour-familiarization experiment. Right, average spiking of the initially flavour-preferring population (n = 201 neurons from 7 mice; classified on novel-flavour day) during flavour consumption on the novel day and the familiar day. g, Illustration of the neural mechanism for learning from delayed postingestive feedback using malaise-driven reactivation and stabilization of flavour representations in the amygdala. Shaded areas in c and f represent the mean ± s.e.m. and in d and e represent the linear fit estimate ±95% confidence intervals. Inset box plot shows the 10th, 25th, 50th, 75th and 90th percentiles. *P ≤ 0.05, ***P ≤ 0.001, ****P ≤ 0.0001. See Supplementary Table 2 for details of statistical tests and for exact P values. Source data

Fig. 5. Novel-flavour consumption triggers PKA activity in the amygdala, which provides a potential biochemical eligibility trace for reactivation by postingestive CGRP neuron activity.

a, Simplified schematic of the biochemical pathway that has been proposed to allocate a subset of amygdala neurons to the CFA memory-retrieval ensemble (or ‘memory engram’)^–. b, Schematic of the strategy for recording PKA activity in the CEA across flavour-familiarization using the AKAR2 sensor. F, familiar; N, novel. c, Example image of AKAR2 expression in the CEA. Scale bar, 1 mm. d, PKA activity in the CEA in response to consumption of a novel or familiar flavour (port A) and to water (port B) across four consecutive days (n = 13 mice). Points and error bars at the top of each plot indicate the timing of the next reward consumption. e, PKA activity for individual mice in response to a novel or familiar flavour and to water consumption. Mice were sorted by novel-flavour response (day 1). f, Summary of PKA activity in response to a novel or familiar flavour and to water consumption (n = 13 mice). g, PKA activity in response to novel flavour (left) and water (right) consumption on day 1 for individual mice aligned to the Allen CCF (n = 13 mice). h, Schematic of a putative hypothesis linking biochemistry with neural activity during CFA, with novelty-dependent increases in PKA in novel-flavour-coding amygdala neurons leading to increased reactivation of these cells by delayed CGRP neuron inputs and recruitment to the CFA memory-retrieval ensemble. Error bars in d represent the 25th, 50th and 75th percentiles and in f represent the mean ± s.e.m. Shaded areas represent the mean ± s.e.m. *P ≤ 0.05, ****P ≤ 0.0001. See Supplementary Table 2 for details of statistical tests and for exact P values. Source data

Extended Data Fig. 1. Brain-wide novel versus familiar flavour activation patterns at each stage of one-shot, delayed CFA learning.

a, Comparison of individual familiar and novel flavour condition mice for every brain region that was significantly novel flavour-activated during consumption (n = 12 mice per flavour condition). b, Analogous to a, but for brain regions that were significantly familiar flavour-activated during consumption (n = 12 mice per flavour condition). c, Map of average FOS⁺ cell density across all mice for the consumption time point (n = 24 mice). The Allen CCF is overlaid. Coronal sections are spaced by 0.5 mm, the section corresponding to Bregma is marked with a *, and key brain regions are labeled. d, Map of the difference in average FOS⁺ cell density across novel versus familiar flavour condition mice for the consumption time point (n = 12 mice per flavour condition). e, Map of average FOS⁺ cell density across all mice for the malaise time point (n = 24 mice). f, Map of the difference in average FOS⁺ cell density across novel versus familiar flavour condition mice for the malaise time point (n = 12 mice per flavour condition). g, Map of average FOS⁺ cell density across all mice for the retrieval time point (n = 24 mice). h, Map of the difference in average FOS⁺ cell density across novel versus familiar flavour condition mice for the retrieval time point (n = 12 mice per flavour condition). An interactive visualization of these FOS⁺ cell density maps is available at https://www.brainsharer.org/ng/?id=872. Error bars represent mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. See Supplementary Table 2 for details of statistical tests and for exact P values. See Supplementary Table 1 for list of brain region abbreviations. Source data

Extended Data Fig. 2. Activation of the LS during novel flavour consumption blocks malaise-driven amygdala activation and interferes with CFA acquisition.

a, Schematic and example hM3D-mCherry expression data for the bilateral chemogenetic LS activation experiment. CNO was delivered 45-min before the experiment began to ensure that the LS was activated throughout consumption. b, Retrieval test flavour preference for the experiment described in a (n = 18 hM3D mice, 12 YFP mice). c, Schematic of the LS activation FOS time point (n = 12 mice per group for d–g). As in a, CNO was delivered 45-min before the experiment began, and the flavour was novel for both groups. d, Comparison of LS FOS (including the entire ‘Lateral septal complex’ in the Allen CCF) for individual YFP and hM3D mice, confirming strong activation by hM3D. e, Comparison of CEA FOS for individual YFP and hM3D mice, showing reduced malaise-driven activation in hM3D mice. f, Correlation between the average FOS⁺ cell count of each brain region for hM3D versus YFP mice. The amygdala network (from Fig. 1f, g; n = 12 regions), septal complex (n = 4 regions), and all other regions (n = 114 regions) are shown separately. g, Visualization of the difference in FOS⁺ cell density across YFP versus hM3D mice with Allen CCF boundaries overlaid. Error bars represent mean ± s.e.m. Shaded areas represent linear fit estimate ± 95% confidence interval. **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. See Supplementary Table 2 for details of statistical tests and for exact P values. See Supplementary Table 1 for list of brain region abbreviations. Source data

Extended Data Fig. 3. Hierarchical clustering of brain regions based on novel versus familiar flavour activation patterns.

Panels a–c show that the brain-wide shift towards activation by the novel flavour is primarily localized to subcortical regions; outlines represent kernel-density estimates of the empirical distributions. a, Novel – familiar ΔFOS effect distribution of all cortical regions (cerebral cortex in the Allen CCF) at each time point (n = 38 brain regions; all statistical tests not significant). b, Novel – familiar ΔFOS effect distribution of all subcortical forebrain regions (cerebral nuclei, thalamus and hypothalamus in the Allen CCF) at each time point (n = 54 brain regions). c, Novel – familiar ΔFOS effect distribution of all midbrain and hindbrain regions (midbrain, pons and medulla in the Allen CCF) at each time point (n = 38 brain regions). d, Hierarchical clustering of novel – familiar ΔFOS effects. This is an expanded version Fig. 1f showing all brain region names. e–n, Left, Illustration of the brain regions comprising each cluster from the hierarchical clustering analysis. Right, Summary of the novel – familiar ΔFOS effect for each cluster at each time point, showing each brain region as an individual point. Error bars represent mean ± s.e.m. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. See Supplementary Table 2 for details of statistical tests and for exact P values. No statistical tests were performed for e–n. See Supplementary Table 1 for list of brain region abbreviations and for GLMM statistics. Source data

Extended Data Fig. 4. The amygdala cluster forms a functional network.

a, Correlation matrices showing the animal-by-animal pairwise FOS correlation for every pair of brain regions during consumption (left), delayed malaise (middle), or memory retrieval (right). Brain regions are sorted using the hierarchical clustermap obtained from the novel – familiar ΔFOS effects in Fig. 1f. b, Summary of the average within-cluster FOS correlation for individual amygdala network regions (Cluster 1 from Fig. 1f, g) by time point (n = 12 regions). The high animal-by-animal correlation among all of the regions in cluster 1 suggest that these regions form a functional network. Panels c,d show that activation of other clusters of brain regions is more correlated with amygdala network activation at experimental time points when those clusters are more strongly novel flavour-selective, including for clusters that were specifically engaged during the initial flavour consumption (comprising sensory cortices; cluster 2) or during retrieval (including the BST; cluster 6). c, Summary of the average across-cluster FOS correlation between the amygdala network and every other cluster at each time point as a function of the other cluster’s standardized novel – familiar effect at that time point (n = 9 clusters × 3 time points). d, Scatter plots showing the pairwise correlation between AIp (top; example cluster 2 region) or BST (bottom; example cluster 6 region) and the CEA (n = 24 mice per time point) at each experimental time point. Error bars represent mean ± s.e.m. Shaded areas represent linear fit estimate ± 95% confidence interval. NS, not significant, ***P ≤ 0.001, ****P ≤ 0.0001. See Supplementary Table 2 for details of statistical tests and for exact P values. See Supplementary Table 1 for list of brain region abbreviations. Source data

Extended Data Fig. 5. An electrophysiological atlas of the effects of CGRP neuron stimulation on amygdala activity in vivo.

a, Schematic of the acute Neuropixels recording experiment (created using BioRender.com). b, Reconstruction of recording trajectories registered to the Allen CCF. Each line represents one insertion of a single-shank Neuropixels 1.0 probe (n = 24 insertions from 4 mice). c, Left, PETHs of neural activity time-locked to CGRP neuron stimulation trains (n = 3,524 amygdala neurons from 24 insertions). Neurons were divided into four response types using a GMM (see Methods): two CGRP neuron stimulation-activated response types (7.3% strongly activated, dark green; 22.4% weakly activated, light green), one CGRP neuron stimulation-inhibited response type (24.6%, purple), and one unmodulated response type (45.8%, gray). Right, Average PETHs for each GMM response type. d, Percentage of recorded neurons that were CGRP neuron stimulation-activated based on the GMM across amygdala subregions (n = 339 CEAc, 272 CEAl, 717 CEAm, 526 BMAa, 129 COAa, 133 IA, 41 BLAp, 30 PAA, 182 MEA, 354 BLAa, 54 Other (AAA, LA, PA), 44 BLAv, 250 BMAp, and 58 COAp neurons). Regions in the CFA amygdala network (cluster 1 from Fig. 1f, g) are shown in red, and other amygdala regions are shown in black. e, Anatomical distribution of all CGRP neuron stimulation-activated (green), CGRP neuron stimulation-inhibited (purple), and unmodulated (gray) neurons projected onto a single coronal or sagittal section of the Allen CCF. f, Left, Light-sheet microscopy data for each animal showing Neuropixels probe trajectories aligned to the Allen CCF with amygdala subregions overlaid. Right, Reconstructions of recording trajectories. For each animal, a single sagittal section corresponding to the center-of-mass of the active recording sites is shown. The colormap for amygdala regions in b is also used in e,f. See Supplementary Table 1 for list of brain region abbreviations. Source data

Extended Data Fig. 6. Brain-wide novel versus familiar flavour activation pattern during CGRP neuron stimulation.

a, Example brain-wide FOS imaging data (200-µm maximum-intensity projections) for four example CGRP neuron stimulation animals. b, Summary of FOS⁺ cell counts in the PB at each time point (n = 24 consumption, 24 malaise, 27 CGRP neuron stimulation, 24 retrieval mice). c, Analysis analogous to Fig. 2i but using a GLMM, showing the correlation among the standardized coefficients for the main LiCl-induced malaise and CGRP neuron stimulation effects from Equation 3 (n = 12 amygdala, 117 other regions). d, Analysis analogous to Fig. 2j but using a GLMM, showing the correlation among the average marginal effects of flavour from Equation 4 (n = 12 amygdala, 117 other regions). e, Map of average FOS⁺ cell density across all mice for the CGRP neuron stimulation time point (n = 27 mice). f, Map of the difference in average FOS⁺ cell density across novel versus familiar flavour condition mice for the CGRP neuron stimulation time point (n = 14 novel flavour mice, 13 familiar flavour mice). g, Top, Schematic of the CGRP^CEA projection stimulation RNAscope FISH experiment. Bottom, Example slide scanner image of FOS expression with the Allen CCF overlaid. h, Example confocal image showing Fos, Sst, Prkcd, and Calcrl expression. This is an expanded version of Fig. 2l. The top row shows the full field-of-view, and the bottom row is magnified with Fos⁺ cell outlines overlaid in black. Error bars represent mean ± s.e.m. Shaded areas represent linear fit estimate ± 95% confidence interval. NS, not significant, *P ≤ 0.05, **P ≤ 0.01, ****P ≤ 0.0001. See Supplementary Table 2 for details of statistical tests and for exact P values. No statistical tests were performed for b. See Supplementary Table 1 for list of brain region abbreviations and for GLMM statistics. Source data

Extended Data Fig. 7. Chronic Neuropixels electrophysiology in the amygdala of freely moving mice.

a, Left, Schematic illustration of the chronic Neuropixels 2.0 implant assembly at progressive stages of construction from top left to bottom right. Right, Schematic illustration of the chronic Neuropixels 1.0 implant assembly upon which the 2.0 implant design is based, shown for size comparison. b, Test recordings used to select recording sites (black bars) properly targeting the CEA (red bars, based on postmortem reconstruction) for two example animals. The shanks are arranged from anterior (1, left) to posterior (4, right) and span 750-µm total. We found that we could distinguish the CEA from nearby brain regions along the vertical axis of the probe shanks as a band of dense neural activity. c, Left, Light-sheet microscopy data for each animal showing Neuropixels shank trajectories aligned to the Allen CCF with CEA subregions overlaid. Right, Reconstruction of trajectories in the CEA. For each animal, a single sagittal section corresponding to the center-of-mass of all active recording sites is shown. See Methods for a description of which animals were included in each experiment.

Extended Data Fig. 8. The amygdala is strongly activated by a novel flavour and near-perfectly discriminates flavour.

a, Cumulative intake of the novel flavour and water in the two-reward CFA paradigm in Fig. 3b (n = 8 mice). We used a randomized, trial-based structure to ensure that mice consumed the two options at an equal rate (see Methods). b, Left, PETHs to novel flavour delivery for the novel flavour-preferring (n = 373 neurons from 8 mice), water-preferring (n = 121 neurons), and nonselective (n = 610 neurons) CEA neurons in Fig. 3c–l. Middle, PETHs to water delivery for the same novel flavour-preferring, water-preferring, and nonselective CEA neurons. Right, Pie chart visualizing the proportion of novel flavour-preferring, water-preferring, and nonselective CEA neurons. Panels c–e provide additional characterization of the multinomial logistic regression decoder using CEA population activity in Fig. 4g–i. c, Cross-validated log-likelihood for the decoder classifying periods of novel flavour consumption versus water consumption versus baseline activity across a range of regularization parameter (λ) values (n = 6 mice). d, Decoder output time-locked to novel flavour delivery (top) and water delivery (bottom) in the initial consumption period (mean across 6 mice). The decoder’s predicted probability for novel flavour consumption and water consumption are both shown. e, Confusion matrix summarizing the cross-validated decoder performance across all mice (n = 6 mice). The overall misclassification rate was 0.56% (4 out of 720). Error bars and shaded areas represent mean ± s.e.m. No statistical tests were performed for c.

Extended Data Fig. 9. CGRP^CEA projection stimulation reactivates flavour representations in the amygdala, and CGRP neuron ablation impairs delayed CFA learning.

Panels a–e show that CGRP^CEA projection stimulation reactivates flavour representations in the amygdala. a, Reconstruction of recording trajectories registered to the Allen CCF for mice with CGRP^CEA projection stimulation (n = 32 shanks from 8 mice). b, Average spiking of individual neurons during the CGRP^CEA projection stimulation conditioning experiment (n = 1,221 neurons from 8 mice). c, Average spiking of the novel flavour-preferring (n = 354 neurons), water-preferring (n = 129 neurons), and nonselective (n = 738 neurons) populations across the entire experiment. d, Average spiking of the novel flavour-preferring, water-preferring, and nonselective populations during individual bouts of CGRP^CEA projection stimulation (same sample sizes as c). e, Neural trajectories in PC-space for novel flavour consumption, water consumption, and CGRP^CEA projection stimulation. Panels f,g show that genetic ablation of CGRP neurons by taCasp3-TEVp impairs delayed CFA learning. f, Example CGRP immunoreactivity data confirming genetic ablation of CGRP neurons by taCasp3-TEVp. This is an expanded version of Fig. 3p. g, CGRP neuron ablation mice show significantly higher acceptance of the conditioned flavour following LiCl-induced CFA when compared to wild type controls (n = 6 taCasp3 mice, 7 control mice). Panels h–j show that CGRP neuron stimulation-activated CEA neurons are also activated by LiCl injection. h, Schematic. i, Average spiking during CGRP neuron stimulation and then during LiCl-induced malaise (n = 821 neurons from 4 mice). j, Average spiking of the CGRP neuron stimulation-activated neurons (n = 189 neurons) and other neurons (n = 632 neurons) during LiCl-induced malaise. Shaded areas represent mean ± s.e.m. Inset box plots show the 10^th, 25^th, 50^th, 75^th, and 90^th percentiles. *P ≤ 0.05, **P ≤ 0.01, ****P ≤ 0.0001. See Supplementary Table 2 for details of statistical tests and for exact P values. Source data

Extended Data Fig. 10. Postingestive CGRP neuron activity is necessary and sufficient to stabilize flavour representations in the amygdala upon memory retrieval.

Panels a,b relate to Fig. 4b–d. a, Proportion of flavour-preferring neurons classified separately on conditioning or retrieval day (n = 8 mice). b, Population trajectories for flavour consumption, water consumption, and CGRP neuron stimulation. Panels c–e relate to Fig. 4e. c, Average spiking of all individual neurons for the CGRP^CEA projection stimulation experiment (n = 1,042 neurons from 8 mice). d, Analogous to a, but for mice with CGRP^CEA projection stimulation (n = 8 mice). e, Average spiking of the novel flavour-preferring neurons with the highest 10% CGRP^CEA response magnitudes and of the remaining novel flavour-preferring neurons. Panels f,g show that LiCl-induced malaise stabilizes the flavour representation upon retrieval, and that this is impaired by CGRP neuron ablation. f, For control mice, average spiking of the novel flavour-preferring population (n = 279 neurons from 4 mice). g, Analogous to f, but for mice with CGRP neuron ablation (n = 109 neurons from 4 mice). Panels h–k relate to Fig. 4f. h, Average spiking of all individual neurons (n = 924 neurons from 7 mice). i, Proportion of flavour-preferring neurons classified separately on novel or familiar day (n = 7 mice). j, Average spiking of the initially water-preferring population (n = 160 neurons from 7 mice; classified on novel day) during flavour consumption. k, Population trajectories for flavour and water consumption. l, Time-courses along the PC2 axis during consumption following CGRP neuron stimulation conditioning (from b) and familiarization (from k). Error bars and shaded areas represent mean ± s.e.m. Inset box plots show the 10^th, 25^th, 50^th, 75^th, and 90^th percentiles. NS, not significant, *P ≤ 0.05, **P ≤ 0.01. See Supplementary Table 2 for details of statistical tests and for exact P values. Source data