Fig 2: Different excitability does not correlate with previously known diversity of CCK+ cells.(A) Axonal (black) and dendritic (blue) reconstructions of four CCK+INs in the CA3 region representing two major morphological types, the mossy-fiber associated cells and basket cells. The firing of cell#1 and #2 are shown in detail in Figure 1A. The firing from hyperpolarized MP is shown for each cell. Immuno-labelling for CCK and the nuclear protein Satb1 are shown next to each recorded cell. (B) Prevalence of TOR and RS cells within three major morphological types of CA3 CCK+. Numbers of identified cells are indicated on each bar (BC: CCK+ basket cells, MFA: mossy-fiber associated cells, SCA: Schaffer-collateral associated cells). (C) Single cell RNAseq characterization of recorded TOR (left, n = 8) and RS (right, n = 8) cells for known GABAergic cell markers. Each column corresponds to single identified CCK+INs. The bars on the right shows the logarithm of p values of the statistical comparison of TOR and RS cells for each genes. Significantly different mRNA content between TOR and RS cells was found for Satb1 and Lypd1 (p=0.0095 and p=0.0006, Mann-Whitney Test). (D) Comparison of immunohistochemical and single cell RNAseq data. Bar plots show the fraction of the recorded cells with detectable RNA content (threshold: 0.2) for the selected markers and immunopositivity for the same proteins (for examples of the immunolabelling see Figure 2—figure supplement 3). The number of tested cells are shown on each bar.

Fig 3: Similar Kv4.3-expression and different auxiliary subunit, KChIP and DPP-content in CCK+INs.(A) Normalized gene count of primary and auxiliary subunits of voltage-gated potassium channels from single cell RNAseq data of TOR (n = 8) and RS (n = 8) cells. (B) Percentage of recorded cells with detectable levels of Kv4.3 mRNA (left bars, n = 8 and 9 cells) and protein (right bars, n = 20 and 23 tested cells). (C–D) Immunofluorescent co-localization of CCK and Kv4.3 in a TOR (C), and a RS cell (D) in CA3 stratum lucidum. (E) Percentage of recorded cells with detectable levels of KChIP1 mRNA (n = 8 and 8 cells) and proteins (n = 20 and 27 cells). (F–H) Immunofluorescent co-localization of CCK with CB1 (green) and KChIP1 in a TOR (G), and a RS cell (H) from the same slice shown in low magnification image (F). (I) Percentage of recorded cells with detectable levels of KChIP4 mRNA (left bars, n = 8 and 8 cells) and protein (right bars, n = 29 and 30 tested cells). (J) Major KChIP4 splicing isoforms consist of different exons in the N-terminal region (represented as red boxes). Each row represents a single cell in the color-mapped data and columns correspond to individual exons aligned to the schematic illustration of isoforms above. Red and blue colors code high and low mRNA levels, respectively. The average exon counts from the two types of CCK+INs (n = 8 and 8) are shown at the bottom using the same color code scheme. (K) Percentage of recorded cells with detectable levels of DDP6 and DPP10 mRNA. (L) Assembly of major DPP6 isoforms and exon levels in individual CCK+INs are shown as above (J).

Fig 4: Parallel neuronal pathways convey EEC signals to the brain.(A) Experimental design. Triple-transgenic mice expressing hM3D in CCK or TAC1 EECs, or littermate controls, were subjected to either surgical vagotomy to remove subdiaphragmatic vagal afferents, or treatment with intrathecal RTX to ablate spinal afferents, or sham surgery. They were then treated with CNO or vehicle and the effect on feeding behavior and learned preferences was measured. (B,D) Tac1Cre; Vil-Flp; R26Dual-hM3D mice and littermate controls were subjected to either subdiaphragmatic vagotomy or sham surgery (B) or IT-RTX or sham surgery (D). Animals were then food deprived overnight, injected with CNO or vehicle, and one-hour food intake was recorded. Right: ratio of feeding inhibition after treatment with CNO versus vehicle. Statistics for (B, left): Two-way repeated measures ANOVA. Drug: F(1, 14)=15.03, p=0.0017. Subject: F(2, 14)=0.7452, p=0.4925. Interaction: F(2, 14)=6.480, p=0.0102. (B, right): Ordinary one-way ANOVA. F(2, 14)=6.926, p=0.0081.(D, left): Two-way repeated measures ANOVA. Drug: F(1, 16)=20.25, p=0.0004. Subject: F(2, 16)=4.673, p=0.0252. Interaction: F(2, 16)=13.38, P=0.0004. (D, right): Ordinary one-way ANOVA. F(2, 16)=13.40, p=0.0004. (C,E) CckCre; Vil-Flp; R26Dual-hM3D mice and littermate controls were subjected to either subdiaphragmatic vagotomy or sham surgery (C) or IT-RTX or sham surgery (E). Animals were then food deprived overnight, injected with CNO or vehicle, and one-hour food intake was recorded. Right: ratio of feeding inhibition after treatment with CNO versus vehicle. Statistics for (C, left): Two-way repeated measures ANOVA. Drug: F(1, 12)=32.50, p<0.0001. Subject: F(2, 12)=0.6638, p=0.5328. Interaction: F(2, 12)=15.33, p=0.0005. (C, right): Ordinary one-way ANOVA. F(2, 12)=7.267, p=0.0086. (E, left): Two-way repeated measures ANOVA. Drug: F(1, 16)=34.54, p<0.0001. Subject: F(2, 16)=1.036, p=0.3774. Interaction: F(2, 16)=7.123, p=0.0061. (E, right): Ordinary one-way ANOVA. F(2, 16)=11.03, p=0.0010. (F) Tac1Cre; Vil-Flp; R26Dual-hM3D mice (red) and littermate controls (gray) underwent sham treatment, vagotomy, or IT-RTX before CTA training as in Figure 4. All mice received CNO during CTA training. Left: Preference for sucrose in the two-bottle test after CTA training. Ordinary two-way ANOVA. Genotype: F(1, 24)=154.1, p<0.0001. Treatment: F(2, 24)=12.83, p=0.0002. Interaction: F(2, 24)=7.792, p=0.0025. Right: Ratio of total sucrose intake in the second sucrose training session (day 3, after pairing with CNO for one trial) versus first sucrose session (day 1, naive mice). Ordinary two-way ANOVA. Genotype: F(1, 30)=35.61, p<0.0001. Treatment: F(2, 30)=4.673, p=0.0171. Interaction: F(2, 30)=2.398, p=0.1081. Of note, the apparent increased preference for sucrose after vagotomy is largely due to the low ingestion of water; see also Figure 6—figure supplement 1I. (G) CckCre; Vil-Flp; R26Dual-hM3D mice and littermate controls underwent sham treatment, vagotomy, or IT-RTX and then were subjected to CFP training as in Figure 4. Shown is the preference for the CNO paired flavor in the two-bottle test before training (gray bars, day 1) and after training (red bars, day 6). Control: Two-way repeated measure ANOVA. Genotype: F(1, 10)=16.95, p=0.0021. Time: F(1, 10)=9.746, p=0.0108. Interaction: F(1, 10)=5.771, p=0.0372.Vagotomy: Two-way repeated measure ANOVA. Genotype: F(1, 12)=1.866, p=0.1970. Time: F(1, 12)=1.213, p=0.2923. Interaction: F(1, 12)=0.4892, p=0.4976. IT-RTX: Two-way repeated measure ANOVA. Genotype: F(1, 13)=0.08803, p=0.7714. Time: F(1, 13)=5.795, 0.0316. Interaction: F(1, 13)=6.671, p=0.0227. (H) Summary of neuronal pathways required for Tac1Cre EECs or CckCre EECs to regulate food intake and drive food learning. Values are reported as mean ± SEM. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001; # or ns above bars indicate p-value comparing with the vehicle control treatment of the same genotype (#P<0.05); two-way ANOVA (left of B-E; F, G), or ordinary one-way ANOVA (right of B-E). See also Figure 6—figure supplement 1, Figure 6—source data 1, Source data 1, Figure 6—source data 2, Figure 6—source data 3, Figure 6—source data 4, Figure 6—source data 5, . Figure 6—source data 1.Raw data for Figure 6B. Figure 6—source data 2.Raw data for Figure 6C. Figure 6—source data 3.Raw data for Figure 6D. Figure 6—source data 4.Raw data for Figure 6E. Figure 6—source data 5.Raw data for Figure 6F. Figure 6—source data 6.Raw data for Figure 6G.

Fig 5: Selective activation of CCK-GABA neurons enhances performance in the puzzle box test. A, Cartoon of the puzzle box test showing all trials in which a new obstacle is introduced (trial 2: underpass; trial 5: dig; trial 8: plug). In these trials, the mouse must acquire a new strategy to escape from the start area into the goal area. B, left, Escape latency of CNO- (n = 11) and vehicle-treated (n = 14) CCK-GABA/hM3Dq mice in all nine trials of the puzzle box test. Lower escape latencies indicate better performance. CNO-treated mice show significantly lower escape latency only in trial 8 (plug task). B, right, Escape latencies for CNO-treated and Veh-treated mice in all novel obstacle trials (underpass, dig, and plug task). All figures present data as mean ± SEM. Behavioral data in hM3Dq- mice is given in Extended Data Figure 3-1. The asterisk indicates statistical significance at the p < 0.05 level.