Enteroendocrine cell lineages that differentially control feeding and gut motility

Enteroendocrine cells are specialized sensory cells of the gut-brain axis that are sparsely distributed along the intestinal epithelium. The functions of enteroendocrine cells have classically been inferred by the gut hormones they release. However, individual enteroendocrine cells typically produce multiple, sometimes apparently opposing, gut hormones in combination, and some gut hormones are also produced elsewhere in the body. Here, we developed approaches involving intersectional genetics to enable selective access to enteroendocrine cells in vivo in mice. We targeted FlpO expression to the endogenous Villin1 locus (in Vil1-p2a-FlpO knock-in mice) to restrict reporter expression to intestinal epithelium. Combined use of Cre and Flp alleles effectively targeted major transcriptome-defined enteroendocrine cell lineages that produce serotonin, glucagon-like peptide 1, cholecystokinin, somatostatin, or glucose-dependent insulinotropic polypeptide. Chemogenetic activation of different enteroendocrine cell types variably impacted feeding behavior and gut motility. Defining the physiological roles of different enteroendocrine cell types provides an essential framework for understanding sensory biology of the intestine.


Introduction
The gut-brain axis plays a critical role in animal physiology and behavior. Sensory pathways from the gut relay information about ingested nutrients, meal-induced tissue distension, osmolarity changes in the intestinal lumen, and cellular damage from toxins (Bai et al., 2019;Brookes et al., 2013;Prescott and Liberles, 2022;Richards et al., 2021;Williams et al., 2016). Responding neural circuits evoke sensations like satiety and nausea, coordinate digestion across organs, shift systemic metabolism and energy utilization, and provide positive and negative reinforcement signals that guide future consumption of safe, energy-rich foods (Andermann and Lowell, 2017;Sternson and Eiselt, 2017;Zimmerman and Knight, 2020). Moreover, manipulations of the gut-brain axis have been harnessed clinically through gut hormone receptor agonism or bariatric surgery to provide powerful therapeutic approaches for obesity and diabetes intervention (Richards et al., 2021;Seeley et al., 2015).
Enteroendocrine cells are first-order chemosensory cells of the gut-brain axis and are sparsely distributed along the gastrointestinal tract . Like taste cells, enteroendocrine cells are epithelial cells with neuron-like features, as they are electrically excitable, release vesicles upon elevation of intracellular calcium, and form synaptic connections with second-order neurons through specialized extrusions called neuropods (Bohórquez et al., 2015;Reimann et al., 2012). Single-cell RNA sequencing approaches revealed a diversity of enteroendocrine cell types that produce different gut hormones (Beumer et al., 2018;Gehart et al., 2019;Haber et al., 2017). Superimposing cell birthdate on the enteroendocrine cell atlas through an elegant genetically encoded fluorescent clock revealed five major enteroendocrine cell lineages defined by expression of either glucose-dependent insulinotropic polypeptide (GIP), ghrelin, serotonin (called enterochromaffin cells), somatostatin, or a combination of glucagon-like peptide 1 (GLP1), cholecystokinin (CCK), and/or neurotensin (Gehart et al., 2019).
Enteroendocrine cell-derived gut hormones evoke a variety of physiological effects (Drucker, 2016). GLP1 and CCK are satiety hormones released following nutrient intake, ghrelin is an appetitepromoting hormone whose release is suppressed by nutrients, and serotonin can be released by non-nutritive signals like irritants, force, and catecholamines. Sugar-induced release of GIP and GLP1 causes the incretin effect which rapidly promotes insulin release and lowers blood glucose (Holst et al., 2009). CCK, serotonin, and other gut hormones additionally regulate a variety of digestive functions, including gut motility, gastric emptying, gastric acidification, absorption, gallbladder contraction, and exocrine pancreas secretion.
The functions of individual enteroendocrine cell types could in some cases be inferred by summing the actions of their expressed hormones. For example, chemogenetic activation of enteroendocrine cells in the distal colon which express insulin-like peptide 5 triggers a multipronged physiological response that includes appetite suppression through a peptide YY (PYY) receptor, improved glucose tolerance through GLP1, and defecation indirectly through the serotonin receptor HTR3A (Lewis et al., 2020). However, a challenge in generalizing this approach is that some enteroendocrine cells release hormones with apparently opposing functions (Gehart et al., 2019;Haber et al., 2017), and moreover, many gut hormones are also produced by other cell types in the body (Lee and Soltesz, 2011;Okaty et al., 2019). To overcome these challenges, we developed approaches involving intersectional genetics to obtain highly selective access to major transcriptome-defined enteroendocrine cell lineages. Chemogenetic activation of each of these enteroendocrine cell types produced variable effects on gut physiology and behavior. Obtaining a holistic model for enteroendocrine cell function provides a critical framework for understanding the neuronal and cellular logic underlying gut-brain communication.

Selective access to enteroendocrine cells in vivo through intersectional genetics
We first sought to identify genetic tools that broadly and selectively mark enteroendocrine cells. Transcription factors such as Atoh1, Neurogenin3, and NeuroD1 are expressed in enteroendocrine cell progenitors and/or precursors and act in early stages of enteroendocrine cell development (Li et al., 2011). We obtained Atoh1-Cre (both knock-in and transgenic lines), Neurog3-Cre, and Neurod1-Cre mice and crossed them to mice containing a Cre-dependent tdTomato reporter (Rosa26 CAG-lsl-tdTomato herein defined as lsl-tdTomato). Neurog3-Cre and Neurod1-Cre lines labeled a sparse population of intestinal epithelial cells characteristic of enteroendocrine cells, although the Neurog3-Cre line additionally labeled other cells in intestinal crypts and in occasional mice produced broad labeling of intestinal epithelium; neither Atoh1-Cre line tested displayed selective labeling of enteroendocrine cells (Figure 1-figure supplement 1A; Schonhoff et al., 2004). Two-color analysis of tdTomato and gut hormone expression verified tdTomato localization in enteroendocrine cells of Neurod1-Cre; lsl-tdTomato mice, consistent with prior findings (Figure 1-figure supplement 1B; Li et al., 2012). Single-cell RNA sequencing of tdTomato-positive cells obtained from these mice (see below) also verified selective enteroendocrine cell labeling.
Neurod1-Cre mice provide broad, indelible, and selective marking of enteroendocrine cells within the intestine, but NeuroD1 is also expressed in a variety of other tissues, including the brain, retina, pancreas, peripheral neurons, and enteric neurons (Figure 1B and C, Figure 1-figure supplement 1D and E;Cho and Tsai, 2004;Li et al., 2011). Knockout of NeuroD1 is lethal, causing severe deficits in neuron birth and survival, as well as in the development of pancreatic islets and enteroendocrine cells (Gao et al., 2009;Naya et al., 1997). We employed an intersectional genetic strategy of combining Cre and Flp recombinases to limit effector gene expression to enteroendocrine cells. Villin1 (Vil1) is expressed with high selectivity in the lower gastrointestinal tract (el Marjou et al., 2004;Maunoury et al., 1992), so we generated a knock-in mouse allele (Vil1-p2a-FlpO) that drives FlpO recombinase expression from the endogenous Vil1 locus. Vil1-p2a-FlpO mice displayed expression of a Flp-dependent Gfp allele in epithelial cells throughout the entire length of the intestine with striking specificity ( Figure 1A, Figure 1-figure supplement 1C). Reporter expression was not observed in most other tissues examined, including most brain regions, spinal cord, peripheral ganglia, and enteric neurons; rare GFP-expressing cells were noted in taste papillae, epiglottis, pancreas, liver, and thalamus ( Figure 1C, Figure 1 -figure supplement 1C and D;Höfer and Drenckhahn, 1999;Madison et al., 2002;Rutlin et al., 2020). Combining Neurod1-Cre and Vil1-p2a-FlpO alleles (Neuro-d1 INTER ) yielded highly selective expression of an intersectional reporter gene encoding tdTomato (Rosa26 CAG-lsl-fsf-tdTomato herein defined as inter-tdTomato) in enteroendocrine cells, with only occasional cells observed in pancreas, and no detectable expression in other cell types labeled by either allele alone ( Figure 1C, Figure 1-figure supplement 1D and E).

Charting enteroendocrine cell diversity and gene expression
Our general goal was to use intersectional genetics to access subtypes of enteroendocrine cells that express different gut hormones. We first used single-cell RNA sequencing approaches to measure the extent of enteroendocrine cell diversity, compare findings with existing enteroendocrine cell atlases, and establish a foundation for genetic experiments. Enteroendocrine cells represent <1% of gut epithelial cells, so we used genetic markers for enrichment. NeuroD1 is expressed early in the enteroendocrine cell lineage, and we observed by two-color expression analysis that Neurod1-Cre mice target at least several enteroendocrine cell types (Figure 1-figure supplement 1B). Since prior enteroendocrine cell atlases were derived from cells expressing an earlier developmental marker, Neurog3 (Gehart et al., 2019), we sought to compare the repertoire of enteroendocrine cells captured by Neurod1-Cre and Neurog3-Cre mice.
tdTomato-positive cells were separately obtained from the intestines (duodenum to ileum) of Neurod1-Cre; lsl-tdTomato mice and Neurog3-Cre; lsl-tdTomato mice by fluorescence-activated cell sorting (Figure 2-figure supplement 1A). Using the 10X Genomics platform, mRNA was captured from individual cells, and barcoded single-cell cDNA was generated. Single-cell cDNA was then sequenced and unsupervised clustering analysis was performed using the Seurat pipeline Stuart et al., 2019). Transcriptome data was obtained for 5,856 tdTomato-positive cells from Neurog3-Cre; lsl-tdTomato mice and 1841 tdTomato-positive cells from Neurod1-Cre; lsl-tdTomato mice. Twenty-five percent of Neurog3-lineage cells (1454/5856) and 87% of NeuroD1-lineage cells (1595/1841) expressed classical markers for enteroendocrine cells (Figure 2figure supplement 1B and C). Moreover, the full diversity of known enteroendocrine cell types was similarly captured by both Cre lines, with Neurog3-Cre mice additionally labeling many other cells, including paneth cells, goblet cells, enterocytes, and progenitors ( Figure 2-figure supplement 1C). These findings are consistent with NeuroD1 acting later than Neurogenin3 in the enteroendocrine cell lineage, but prior to cell fate decisions leading to enteroendocrine cell specialization (Jenny et al., 2002).
Since Neurog3-Cre and Neurod1-Cre mice similarly labeled all known enteroendocrine cell lineages, transcriptome data was computationally integrated for analysis of enteroendocrine cell subtypes. Selective clustering analysis of 3049 enteroendocrine cells from both mouse lines revealed 10 distinct cell clusters, with one cluster representing putative progenitors ( Figure 2A, Figure 2-source data 1). Cell clusters were compared with previously described enteroendocrine cell types based on expression of signature genes encoding hormones and transcriptional regulators (Figure 2A-C; Gehart et al., 2019). We observed three classes of enterochromaffin cells that similarly express serotonin biosynthesis enzymes (Tph1) and associated transcription factors (Lmx1a), but differentially produce Tac1, Cartpt, Pyy, Ucn3, and Gad2 ( Figure 2B). Six other cell types preferentially express either Gip (K cells), Cck (I cells), Gcg (GLP1 precursor, L cells), Nts (N cells), Sst (D cells), and Ghrl (X cells), with L, I,   and N cells thought to be derived from a common cell lineage (Beumer et al., 2020;Gehart et al., 2019). Strong segregation was observed for some signature genes, such as Tph1 in enterochromaffin cells and Sst in D cells. In other cases, signature hormone genes like Cck and Ghrl were enriched in particular cell clusters but expression was not absolutely restricted and also observed at lower levels in other cell clusters ( Figure 2B). We note that glutamate transporters were not readily detected in root ganglion) or wholemounts (tongue) of fixed tissues indicated from Neurod1-Cre; lsl-tdTomato mice (left), Vil1-p2a-FlpO; fsf-Gfp mice (middle), and Neurod1 INTER ; inter-tdTomato mice (right). Scale bars: 100 μm for all except 500 μm for tongue. Intestine sections from duodenum (middle) or jejunum (left, right). See Figure 1-figure supplement 1.
The online version of this article includes the following figure supplement(s) for figure 1:     The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Signature genes with differential expression across enteroendocrine cell types.  Thus, each enteroendocrine cell subtype expresses a hormone repertoire with distinct patterns of enrichment but also sometimes partial overlap.
Enteroendocrine cells also express various cell surface receptors to detect nutrients, toxins, and other stimuli. For example, enteroendocrine cells detect sugars through the sodium-glucose cotransporter SGLT1 (encoded by the gene Slc5a1), with sodium co-transport thought to lead directly to cell depolarization (Gorboulev et al., 2012;Reimann et al., 2008). This mechanism is distinct from sugar detection by taste cells or pancreatic beta cells. Gustatory sensations of sweet (and savory/umami) involve taste cell-mediated detection of sugars (and amino acids) through heterodimeric G protein-coupled receptors termed T1Rs (Yarmolinsky et al., 2009), while pancreatic beta cells respond to sugar through increased metabolic flux, ATP-gated potassium channel closure, and depolarization. Expression of Slc5a1 was observed in multiple enteroendocrine cell subtypes, and highest in K, L, D, and N cells, while abundant expression of T1Rs was not detected in any enteroendocrine cell type ( Figure 2B). These findings are consistent with the ability of taste blind mice lacking T1Rs to develop a preference for sugar-rich foods through SGLT1-mediated post-ingestive signals of the gut-brain axis (Sclafani et al., 2016;Tan et al., 2020). In addition, free fatty acid receptor genes Ffar1 and Ffar4 were broadly expressed in several enteroendocrine cell lineages, but largely excluded from enterochromaffin cells ( Figure 2B). Orthogonally, the toxin receptor gene Trpa1 was enriched in enterochromaffin cells (Bellono et al., 2017), but not abundantly expressed in other enteroendocrine cells ( Figure 2B and D). Enterochromaffin cells also reportedly sense force through the mechanosensory ion channel PIEZO2 (Alcaino et al., 2018); Piezo2 transcript was not readily detected in our transcriptomic data, but is enriched in enteroendocrine cells from colon that we did not analyze (Billing et al., 2019;Treichel et al., 2022; Figure 2B). Thus, enteroendocrine cells often express multiple cell surface receptors, suggesting polymodal response properties, and some receptors are expressed by multiple enteroendocrine cell types.

Genetic access to subtypes of enteroendocrine cells
Next, we obtained genetic tools for selective access to each major enteroendocrine cell lineage. We chose several combinations of Cre and FlpO lines to achieve intersectional genetic access to different enteroendocrine cells based on the cell atlas. Mice of each intersectional allele combination were crossed to inter-tdTomato mice, and reporter expression was analyzed across tissues, including in the brain, tongue, airways, pancreas, stomach, and intestine (duodenum to rectum) (Figure 3-figure supplements 1 and 2). Each of these seven intersectional combinations produced sparse labeling of intestinal epithelial cells, as expected for labeling of enteroendocrine cell subtypes ( Figure 3-figure supplement 1). Striking selectivity for enteroendocrine cells was observed across analyzed tissues for intersectional combinations targeting D, K, L, and I cells; sparse labeling was rarely observed in gastric endocrine cells and pancreatic islets, and absent from all other tissues examined. For example, Cck-ires-Cre alone (without intersectional genetics) drove reporter (lsl-tdTomato) expression in many tissues, including the brain, spinal cord, and muscle, and within the intestine, in enteroendocrine cells as well as enteric neurons, extrinsic neurons, and cells in the lamina propria; however, in Cck INTER ; inter-tdTomato mice, expression was not observed in the brain, spinal cord, or muscle, and within the intestine, was highly restricted to a subset of enteroendocrine cells, and not observed in other intestinal cell types (Figure 3-figure supplement 1). Similarly restrictive reporter expression was observed in Sst INTER ; inter-tdTomato, Gip INTER ; inter-tdTomato, and Gcg INTER ; inter-tdTomato mice. We did note that Tac1 INTER and Npy1r INTER alleles more broadly labeled rectal epithelium, and Npy1r INTER additionally labeled taste cells as well as rare cells in the airways and epiglottis (Figure 3-figure supplements 2 and 3B). We also note that other genetic tools were inefficient at targeting enteroendocrine cells, including Nts-ires-Cre and Mc4r-t2a-Cre mice (Figure 3-figure supplement 3A).
Next, we assessed the spatial distribution of each enteroendocrine cell lineage along the proximaldistal axis in the duodenum, jejunum, ileum, colon, and rectum by quantifying the number of reporterpositive cells (Figure 3). Pet INTER and Sst INTER cells were most enriched in the duodenum and colon ( Figure 3). Sst INTER cells were the sparsest of enteroendocrine cell types, consistent with observations from scRNA-seq data (Figures 2A and 3). Gip INTER cells and Gcg INTER cells displayed strikingly distinct spatial patterns. Gip INTER cells were enriched proximally, with almost no tdTomato+ cells observed in distal intestine. In contrast, Gcg INTER cells were present along the entire proximal-distal axis and were enriched in colon and rectum. Thus, various enteroendocrine cell subtypes display distinct spatial distributions along the gastrointestinal tract.

Physiological responses to enteroendocrine cell activation
Direct study of enteroendocrine cell function has been challenging due to a lack of specific genetic tools. Hints come from Neurogenin3 point mutations in human infants or intestine-targeted Neurog3 knockout, which cause loss of enteroendocrine cells, severe malabsorptive diarrhea, and increased mortality (Mellitzer et al., 2010;Wang et al., 2006). We sought to develop cell type-specific genetic tools for enteroendocrine cell manipulation, reasoning that they might provide a specific approach to define the repertoire of evoked physiological and behavioral responses.
We first developed chemogenetic approaches for acute stimulation of all enteroendocrine cell types in freely behaving mice. Chemogenetic strategies involved designer G protein-coupled receptors (so-called DREADDs) that respond to the synthetic ligand clozapine-N-oxide (CNO) (Roth, 2016). Neurod1 INTER mice were crossed to contain an intersectional reporter allele (Rosa26 CAG-fsf-eGFP-FLEX-hM3Dq-mCherry herein defined as inter-hM3Dq-mCherry) that enables expression of a Gα q -coupled DREADD  (Sciolino et al., 2016). Since this approach yielded rare reporter expression in pancreatic islets, we used an additional control mouse line, Ptf1a-Cre; Vil1-p2a-FlpO (Ptf1a INTER ), which targets sparse Vil1-expressing pancreatic cells but not intestinal cells (Figure 4-figure supplement 1; Kawaguchi et al., 2002). First, we examined the effect of global enteroendocrine cell activation on gut motility as assessed by movement of charcoal dye following oral gavage. Neurod1 INTER ; inter-hM3Dq-mCherry mice, Ptf1a INTER ; inter-hM3Dq-mCherry mice, and control Cre-negative Vil1-p2a-FlpO; inter-hM3Dq-mCherry littermates were injected intraperitoneally (IP) with CNO (fed ad libitum, daytime). After 15 min, charcoal dye was administered, and after an additional 20 min, the gastrointestinal tract was harvested. Charcoal transit distance was calculated by genotype-blinded measurement of the charcoal dye leading edge. In control animals lacking DREADD expression, the leading edge of charcoal dye traversed part of the intestine (littermate controls lacking Neurod1-Cre: 22.6 ± 1.2 cm; littermate controls lacking Ptf1a-Cre: 22.8 ± 2.0 cm) (Figure 4, Figure 4-source data 1). Chemogenetic activation of all enteroendocrine cells in Neurod1 INTER ; inter-hM3Dq-mCherry mice accelerated gut transit, with the charcoal leading edge traversing 30.8 ± 1.5 cm of the intestine. When DREADD signaling was instead activated in all epithelial cells using Vil1-Cre; lsl-hM3Dq mice, gavaged dye failed to enter the intestine at all (Figure 4-figure supplement 1A). CNO-accelerated gut transit was not observed Ptf1a INTER ; inter-hM3Dq-mCherry mice (22.6 ± 2.6 cm) containing DREADD expression only in pancreatic cells (Figure 4, Figure 4-figure supplement 1B and C). Based on these observations, the observed effects in Neurod1 INTER ; inter-hM3Dq-mCherry mice are due to enteroendocrine cells rather than pancreatic cells, and the net effect of activating all enteroendocrine cells is to promote gut transit.

Figure 4 continued on next page
Gcg INTER cells. These findings suggest a hierarchy where neural circuits that mediate toxin responses may achieve priority over those that mediate nutrient responses, at least under conditions of equal and maximal activation. Altogether, we characterize enteroendocrine cell subtypes that have different and sometimes opposing effects on digestive system physiology.

Enteroendocrine cells that regulate feeding behavior
Next, we examined the effect of global enteroendocrine cell activation on feeding behavior. Fasted mice expressing DREADDs in all enteroendocrine cells (Neurod1 INTER ; inter-hM3Dq-mCherry) or in sparse pancreatic cells (Ptf1a INTER ; inter-hM3Dq-mCherry), and their control littermates lacking Cre recombinase, were injected (IP) with CNO and given access to food for 2 hr at dark onset ( Figure 5A). Animals lacking DREADD expression, or with sparse DREADD expression only in pancreas, ate robustly (~1 g of food over a 2 hr period). In contrast, CNO-induced activation of enteroendocrine cells caused a 26% reduction in food intake ( Figure 5B, Figure 5-source data 1).
To interrogate the roles of different enteroendocrine cell subtypes in feeding regulation, similar experiments were then performed in (1) (7) Gcg INTER ; inter-hM3Dq-mCherry mice, with Cre-negative littermates again serving as controls. Chemogenetic activation of enterochromaffin cells reduced feeding behavior ( Figure 5B, 52.1% reduction). Similar results were seen upon chemogenetic activation of Tac1 and Npy1r cells ( Figure 5-figure supplement 1A, Tac1-ires2-Cre: 48.5% reduction, Npy1r-Cre: 79.6% reduction), but we note that these intersectional allele combinations also drove expression in taste cells and rectal epithelium, cell types that could also potentially drive changes in feeding behavior. In contrast, activation of Sst INTER and Gip INTER cells did not change feeding behavior ( Figure 5B) food intake over 2 hr (g) Figure 5. Enteroendocrine cell types that reduce feeding. (A) Timeline for behavioral assay. (B) Mice of genotypes indicated were fasted overnight, injected with CNO (IP, 3 mg/kg), and total food intake was measured during 2 hr ad libitum food access, circles: individual mice, n: 8-13 mice, mean ± sem, *p<0.05 by a Mann-Whitney test with Holm-Šídák correction. See Figure 5-figure supplement 1.
The online version of this article includes the following source data and figure supplement(s) for figure 5: Source data 1. Quantification of feeding behavior.      (single CNO injection) caused a durable reduction of feeding for several hours, with total food intake normalizing by 11 hr, and also evoked a decrease in water intake and the respiratory exchange ratio, but not locomotion ( Figure 5-figure supplement 1C). For comparison, activating somatostatin cells reduced the respiratory exchange ratio but did not change feeding, water intake, or locomotion. Altogether, we find that some but not all enteroendocrine cells can regulate food intake, and can do so with varying efficacy.

Conclusion
Here we developed a toolkit involving intersectional genetics for systematic access to each major enteroendocrine cell lineage ( Figure 6A). We then used chemogenetic approaches to delineate major response pathways of the gut-brain axis ( Figure 6B). Serotonin-producing enterochromaffin cells express the irritant receptor TRPA1 (Bellono et al., 2017) and chemogenetic activation blocks feeding behavior and promotes gut transit, presumably for toxin clearance. Furthermore, different enterochromaffin cell subtypes can have different effects on gut motility, suggesting at least partially nonoverlapping communication pathways with downstream neurons. These findings are consistent with a role for enterochromaffin cells in toxin-induced illness responses, and interestingly, pharmacological blockade of the serotonin receptor HTR3A is a clinical mainstay for nausea treatment (Freeman et al., 1992). Other enteroendocrine cell types, including those that produce CCK, GIP, GLP1, neurotensin, and somatostatin, express nutrient receptors yet elicit different physiological and behavioral responses. For example, GLP1 cells slow gut motility, presumably to promote nutrient absorption and decrease feeding behavior . Additional studies are needed to define gut-brain pathways that mediate nutrient reward, and why receptors for specific nutrients are expressed across a dispersed ensemble of enteroendocrine cells. Together, these experiments provide a highly selective method for accessing enteroendocrine cells in vivo and a direct measure of their various roles in behavior and digestive physiology.

Continued on next page
Key resources

Single-cell RNA sequencing
Enteroendocrine cells were acutely harvested using a protocol modified from previous publications (Haber et al., 2017;Sato et al., 2009 ). Collected cells were then loaded into the 10X Genomics Chromium Controller, and cDNA prepared and amplified according to manufacturer's protocol (10X Genomics, Chromium single-cell 3′ reagent kit v3, 12 cycles per amplification step). The resulting cDNA was sequenced on a NextSeq 500 at the Harvard Medical School Biopolymers Facility. Sequence reads were aligned to the mm39 mouse transcriptome reference, and feature barcode matrices were generated using 10X Genomics CellRanger. Unique transcript (UMI) count matrices were analyzed in R v4.1.1 using Seurat v4.0.5 (Beutler et al., 2017;Satija et al., 2015). The cell barcodes were filtered, removing cells with a high number of UMIs (>125,000) or high percentage of mitochondrial genes (>25%). The filtered UMI count matrix was transformed using SCTransform . Transformed matrices from Neurog3 and Neurod1 samples were integrated (nFeature = 3000), and integrated matrices used for cluster identification and UMAP projections. Additional clusters of low-quality cells (defined by low-average UMI counts and low-average feature counts across the cluster) were removed. To examine the diversity among enteroendocrine cells, cell barcodes belonging to enteroendocrine cells from Neurog3 and Neurod1 samples were identified and reanalyzed separately. Matrices of enteroendocrine cells from Neurog3 and Neurod1 samples were transformed and integrated (nFeature = 3000). Differential gene expression (Wilcoxon ranked-sum test) was conducted on UMI counts matrices that were log normalized and scaled. Seurat's BuildClusterTree function was used to spatially arrange clusters based on relative similarity in gene expression. Two serotonergic clusters were merged post hoc (to become cluster EC_3) due to the absence of any single signature gene that effectively distinguished them. Gene expression data in all UMAP plots is shown as a natural log of normalized UMI counts. Further details and full parameters of analysis will be provided on GitHub upon publication: https://github. com/jakaye/EEC_scRNA, copy archived at (Hayashi, 2023).

Gut transit measurements
DREADD-expressing and control animals (ad libitum fed) were injected with CNO (3 mg/kg, IP). After 15 min, charcoal dye (200 μl, 10% activated charcoal, 10% gum Arabic in water), or for Figure 4figure supplement 1, carmen red dye (200 μl, 6% carmen red, 0.5% methyl cellulose in water), was gavaged orally, and 20 min later, mice were euthanized and the gastrointestinal tract was harvested. The distance between the pyloric sphincter and the charcoal dye leading edge was measured by an observer blind to animal genotype. All animals were naive to CNO exposure, except for some Gip-Cre mice due to limited availability of mice.

Feeding measurements
Experimental mice were individually housed for 3 days and habituated to feeding from a ceramic bowl. Animals were either fed ad libitum or fasted for the last 20-22 hr in a new clean cage with some bedding material from the previous cage. CNO was injected (3 mg/kg, IP), and food pellets presented 15 min later at the onset of darkness. Food intake was measured over the course of 2 hr by weighing the amount of residual food, with genotypes revealed post hoc to achieve a genotype-blinded analysis. Studies involved fasted mice that were naive to prior CNO exposure or fed mice that were either naive to CNO or acclimated for at least a week after prior CNO exposure.

Body composition and indirect calorimetry
Body composition (lean mass and fat mass) was first analyzed for each experimental group with a 3-in-1 Echo MRI Composition Analyzer (Echo Medical Systems, Houston, TX), and no significant differences were observed. Animals were then placed in a Sable Systems Promethion indirect calorimeter maintained at 23°C ± 0.2°C. Mice were singly housed in metabolic cages with corn cob bedding and ad libitum access to Labdiet 5008 chow (56.8/16.5/26.6 carbohydrate/fat/protein). After 18 hr, all mice were injected with PBS (IP) for acclimatation to handling and mild injection stress. The following day, mice were injected with CNO (3 mg/kg, IP) approximately 30 min before dark onset. Animals were then analyzed for food and water consumption, body weight, distance traveled, and respiratory exchange ratio. Statistical analysis was performed with CalR (Mina et al., 2018).
Statistical significance was measured using a Mann-Whitney test with Holm-Šídák correction on Prism 9 (GraphPad) for  (Mina et al., 2018).

Source data
The source data Excel file contains raw numerical data used for all bar graphs and statistical analyses. Single-cell transcriptome data is available with a GEO GSE accession number GSE224223.

Materials availability statement
Vil1-p2a-FlpO mice will be deposited in Jackson Laboratory and made generally available upon reasonable request.

Ethics
All animal husbandry and procedures were performed in compliance with institutional animal care and use committee guidelines. All animal husbandry and procedures followed the ethical guidelines outlined in the NIH Guide for the Care and Use of Laboratory Animals (https:// grants. nih. gov/ grants/ olaw/ guide-for-the-care-and-use-of-laboratory-animals. pdf), and all protocols were approved by the institutional animal care and use committee (IACUC) at Harvard Medical School.

Data availability
The source data excel file contains raw numerical data used for all bar graphs and statistical analyses. Single-cell transcriptome data are available at NCBI Gene Expression Omnibus with accession GSE224223.
The following dataset was generated: