Fig. 6: CeM^Htr2a and CeM^Sst neurons are differentially activated by multiple rewarding stimuli.

a Behavioral paradigm. b Maximum projection images of calcium responses from CeM^Htr2a neurons during habituation and drinking. c Representative traces from CeM^Htr2a neurons during reward consumption. Consumption bouts in gray. d Average z-score comparisons of the general activity of CeM^Htr2a and CeM^Sst neurons during habituation and reward (Mann-Whitney two-tailed U test. CeM^Htr2a: water, p = 0.1771, U = 7322, n = 112–145; food, p = 0.0098, U = 9440, n = 135–169; Fresubin, p = 0.0001, U = 4901, n = 111–124. CeM^Sst: water, p < 0.0001, U = 8130, n = 143–176; food, p < 0.0001, U = 7320, n = 128–176; Fresubin p < 0.0001, U = 8119, n = 138–168). e 3D Visualization of CEBRA-generated embeddings of CeM^Htr2a neurons considering consumption behavior and neural dynamics over time. Shuffled: drinking data. The three dimensions represent the three CEBRA principle components. Each dot represents a time point. Yellow and purple dots mark consumption and resting times, respectively. Comparison of behavioral decoding accuracy using a Random Forest (RF) algorithm across animals for CeM^Htr2a (f) and CeM^Sst neurons (g) (drinking, feeding, Fresubin: two-tailed Wilcoxon matched-pairs signed-rank test p < 0.0001). (Prediction scores of accuracy of behavior compared to shuffled data: Actual-Shuffled Htr2a, n = 25; Sst, n = 36). h CEBRA-generated embeddings of CeM^Htr2a neurons that were longitudinally detected during drinking and feeding, and during drinking and Fresubin consumption. Accuracy of behavioral decoding through RF, employing CEBRA embeddings of CeM^Htr2a (i) and CeM^Sst (j) neurons longitudinally detected during two behaviors (paired two-tailed t test. CeM^Htr2a: drinking-feeding: p = 0.0158, t = 4.956; feeding-drinking: p = 0.0175, t = 4.768; drinking-Fresubin: p = 0.0292, t = 3.935; Fresubin-drinking: p = 0.0337, t = 3.724. CeM^Sst: drinking-feeding: p = 0.0003, t = 6.068; feeding-drinking: p = 0.0007, t = 5.372; drinking-Fresubin: p = 0.0011, t = 4.955; Fresubin-drinking: p = 0.0012, t = 4.867). (Prediction scores of accuracy of behavior compared to shuffled data: Actual-Shuffled Htr2a, n = 4; Sst, n = 9). k Chord diagrams and bar graphs depicting stable or unstable (gray) correlations of the same CeM^Htr2a and CeM^Sst neurons detected during drinking and feeding (Htr2a: stable v.s. unstable: unpaired two-tailed t test p = 0.5418, t = 0.6371, n = 5 mice. Sst: p = 0.0193, t = 2.602, n = 9 mice). l Similar analysis for CeM^Htr2a neurons detected during feeding and Fresubin sessions (stable v.s. unstable: unpaired t test p = 0.0017, t = 4.627, n = 5 mice). Values = Mean ± SEM.