Gene activity in individual hormone-producing cells from lab-grown human small intestine tissue

Enteroendocrine cells (EECs) are the source of a variety of gut hormones that control local intestinal functions such as gallbladder contraction, pancreatic enzyme secretion and intestinal motility, as well as peripheral nutrient metabolism and appetite. They make up around 1% of all intestinal epithelial cells, and are constantly regenerated from stem cells in the crypts, with individual epithelial cells having a lifespan of around 5 days in the small intestine (Gribble & Reimann, 2019). Particularly relevant to the control of nutrient metabolism and food intake are the cells producing the hormones glucagon‐like peptide‐1 (GLP‐1), which forms the basis for the new class of injectable diabetes and weight loss drugs (Drucker & Holst, 2023), cholecystokinin (CCK), peptideYY (PYY) and glucose‐dependent insulinotropic polypeptide (GIP).

Recent years have witnessed an explosion of interest in, and understanding of, EEC function, made possible by the use of transgenic mice and intestinal organoids expressing fluorescent markers driven by EEC‐specific promoters, which enable live EEC identification and purification for transcriptomic and functional analysis (Beumer et al., 2020; Reimann et al., 2008). As a result of these studies, the traditional alphabetical EEC nomenclature in which e.g. I‐cells produce CCK, K‐cells produce GIP and L‐cells produce GLP‐1 and PYY, no longer represents our improved understanding of these cell populations, as several hormones traditionally ascribed to EECs with distinct single‐letter codes actually arise from overlapping cell populations (Kaelberer & Bohorquez, 2018). Furthermore, studies on EEC nutrient‐sensing mechanisms have revealed that different EEC subtypes express the same pool of receptors, including the sodium‐coupled glucose cotransporter (SGLT1), and G‐protein‐coupled receptors (GPCRs) such as free fatty acid receptors (FFAR1–4), bile acid receptor (GPBAR1), 2‐mono‐oleoylglycerol receptor (GPR119), and amino acid receptors (CASR and GPR142) (Santos‐Hernandez et al., 2024). Nevertheless, postprandial plasma excursions of different gut hormones are not identical (Veedfald et al., 2022), and likely reflect both the subtle characteristics of the cells producing them and their location along the gastrointestinal tract.

One method to generate an unbiased comparison between different EECs is single‐cell RNA sequencing (scRNAseq), as the experimental design allows parallel analysis of large numbers of cells (e.g. ∼10,000 using 10X technology) from the same tissue preparation, collected and analysed in a single run. Although unbiased single‐cell RNAseq analysis of the entire intestinal epithelium yields relatively few EECs due to their rarity, this can be improved using EEC‐restricted fluorescent protein expression to enable EEC purification prior to analysis (Beumer et al., 2020). This method has been performed in transgenic mice using NeuroD1 or NeuroG3 as global EEC markers (Bai et al., 2022; Billing et al., 2019; Hayashi et al., 2023; Smith et al., 2024), or in human organoids in which EEC formation was enhanced by inducible NEUROG3 expression (Beumer et al., 2020). In the current study, we generated fluorescent reporter organoids from the human duodenum and ileum using the chromogranin‐A (CHGA) promoter as a pan‐EEC marker (Beumer et al., 2020), and used scRNAseq to cluster and characterise the full EEC populations in the upper and lower small intestine. Although the expected nutrient‐sensing receptors were spread across EEC clusters, differences between cell clusters were apparent using this approach. The dataset was merged with published data from native human intestine (Hickey et al., 2023) and organoids (Beumer et al., 2020), which contained fewer EECs, revealing good overlap between EEC clusters from the different studies.

CHGA‐Venus labelled human duodenal and ileal organoids were generated using CRISPR‐Cas9‐mediated HDR using the strategy shown in Fig. 1A. Organoids from both regions exhibited scattered fluorescent cells with morphology typical of EECs (Fig. 1B and C). Fluorescent CHGA‐positive cells from both intestinal regions were collected by flow cytometry for 10X droplet‐based library preparation and single‐cell RNA sequencing (Fig. 1D and E).

Analysis revealed higher expression of particular hormones in distinct clusters corresponding to the established one‐letter code for EEC naming (Fig. 2A and B), even though most hormones were detectable in several clusters (Fig. 2C–E). In duodenal CHGA‐Venus organoids, the prominent EEC types were enterochromaffin cells (EC cells, characterised by TPH1 expression), K‐cells (expressing GIP), I‐cells (expressing CCK), M/X cells (expressing MLN and GHRL) and D‐cells (expressing SST). In ileal CHGA organoids, the major EEC clusters were EC cells, L‐cells, M/X cells and D‐cells. A subset of cells from the duodenum expressed a number of hormones at lower levels than other clusters and lacked a hormone expressed at higher levels – these are denoted as polyhormonal cells in the figures. Expression of some gut hormones showed overlap between certain EEC populations, such as the detection of CCK in K‐cells, consistent with previous reports from the mouse intestine (Bai et al., 2022; Habib et al., 2012; Hayashi et al., 2023; Smith et al., 2024). The duodenal I‐cell and K‐cell populations co‐expressed GAST, as also previously identified in human NEUROG3‐driven human duodenal organoids (Beumer et al., 2020). SCT expression was very infrequent, likely reflecting a lack of the most mature population of EECs in these organoids. In addition to the hormones defining the cell clusters, human EECs also expressed low levels of TAC3, NPW and VGF, with no apparent predominance in any EEC cluster. TAC1 expression was lower than expected for crypt‐like ECs based on data from mice (Beumer et al., 2020; Hayashi et al., 2023; Roth & Gordon, 1990), and was largely restricted to ileal ECs. Amongst the chromogranin/secretogranin family, CHGA and SCG5 exhibited higher expression in EECs from the ileum, and SCG2 was more highly expressed in duodenum (Fig. 2E).

We next sought to evaluate the expression of GPCRs and nutrient transporters potentially involved in detecting food ingestion (Fig. 3A). FFAR1 expression was highest in I‐cells and L‐cells; GPBAR1 was highest in ileal L‐cells but also detected in most other EEC populations; FFAR4 was highest in M/X cells and ileal L‐cells; GPR119 in L‐cells; OR51E1 in a number of EEC types and notably higher in cells from ileum than duodenum. Compared with other EEC types, EC cells expressed low levels of FFAR1/4, GPR119 and CASR. The intestinal glucose sensor SGLT1 (encoded by SLC5A1), was notably most highly expressed in L‐cells and K‐cells.

Figure 3B shows expression of GPCRs potentially involved in paracrine cross‐talk between different EEC populations. Notably, GIPR was expressed across all human EEC cell populations; GLP1R expression was less common, and mostly restricted to EECs from the duodenum. RXFP4, the receptor for INSL5, a hormone co‐released by colonic L‐cells (Billing et al., 2019), was broadly expressed in EC cells, and was also detectable in scattered EECs from other clusters in the duodenum. NPY1R was predominantly located in ileal EC cells, suggesting that there may be inhibitory paracrine cross‐talk between PYY and 5‐HT release (as NPY1R is Gi coupled). Amongst the SST receptor family, SSTR2 was most highly expressed, with little expression of SSTR5, contrasting with functional data in rodents showing strong activity of SSTR5 in intestinal EECs (Jepsen et al., 2021; Moss et al., 2012).

Transcription factors relevant to EEC specification in the human intestinal epithelium were examined (Fig. 3C). This confirmed the previously reported selective enrichment of LMX1A and (to a lesser extent) MNX1 in ECs (Beumer et al., 2020). HHEX was highly localised to SST‐expressing D‐cells, while the relative absence of ARX in ECs and D‐cells was also noted. PAX6 was most highly expressed in K‐ and L‐cells from duodenum and ileum, respectively.

As EECs release their hormonal contents by regulated exocytosis, we investigated the expression of machinery potentially involved in vesicle function and release. The secretory machinery itself was largely similar across EEC clusters, although with some minor variation between clusters and regions (Fig. 3D). All EEC populations expressed the core SNARE complex proteins (SNAP25, STX1A and VAMP2), accessory proteins implicated in vesicular docking and priming (STXBP1, UNC13A/B, RAB3B, CPLX1/2) and the calcium‐binding protein secretogogin (SCGN). We note, however, that the cluster of polyhormonal cells had fewer hits for secretory machinery genes, suggesting this may represent immature EECs not yet committed to a definitive hormone.

Data from CHGA‐Venus organoids were combined with publicly available single‐cell RNA sequencing data from the native human intestine (Hickey et al., 2023) and organoids with NEUROG3‐driven enhanced secretory line differentiation (Beumer et al., 2020). Overlaps between the EEC populations from the three datasets were good for both duodenum (Fig. 4A–C) and ileum (Fig. 4D and E), supporting the validity of studying EECs derived from organoids. The new data from CHGA‐Venus organoids add a substantial number of EECs to this combined cluster map.

By analysis of fluorescently labelled EECs purified from CHGA‐Venus organoids, the current study provides single‐cell expression data on organoid‐derived EECs from the human duodenum and ileum, adding to the previously published relatively small EEC dataset from human organoids driven by induced NEUROG3 overexpression (Beumer et al., 2020), and EEC populations from native human small intestine (Beumer et al., 2020; Hickey et al., 2023). Importantly, merging the datasets from these three studies revealed strong overlap of the EEC clusters. Cluster analysis of CHGA‐Venus organoid‐derived EECs identified the following major cell populations (in descending order of cluster size): duodenum – EC > M/X > I > D > K‐cells; ileum – EC > L > M/X > D‐cells. Some hormone genes were detected at low levels in the majority of cells isolated from that intestinal region, and could reflect mRNA contamination of the sheath fluid arising from lysis of a few cells expressing very high levels of the hormone during the microfluidic cell separation. While this could contribute to the appearance of the polyhormonal cell cluster, the alternative explanation that some EECs co‐express a number of hormones is supported by immunohistochemical and mass‐spectrometric studies, showing, for example, that CCK is produced by murine K‐ and L‐cells (Cho et al., 2015; Habib et al., 2012) and human K‐ and M/X cells (Guccio et al., 2024; Miedzybrodzka et al., 2021). A distinct duodenal I‐cell cluster was identified that had not been found in the previously published human organoid study (Beumer et al., 2020) but was evident in native duodenum (Hickey et al., 2023), allowing transcriptomic comparisons to be made with this important CCK‐producing cell type. SCT expression was largely undetectable in organoid‐derived EECs, as also reported previously (Beumer et al., 2020), although a distinct population of S‐cells was reported in the native human EEC dataset (Hickey et al., 2023). As SCT is predominantly expressed in mature EECs in duodenal villi (Aiken & Roth, 1992; Hysenaj et al., 2024), its relative absence in the organoid‐derided cell populations suggests that the organoids are biased towards generation of crypt over villus EECs.

Relatively little is understood about how transcription factors drive formation of the different EEC clusters, but a few clear patterns were observable in our TF analysis: (1) HHEX was largely restricted to D‐cells, as also previously observed in mouse and human intestinal organoids, and mirroring the restricted expression of Hhex in other somatostatin‐expressing cells such as pancreatic delta cells (Zhang et al., 2014) and gastric D‐cells (Adriaenssens et al., 2015); (2) ARX was particularly highly expressed in L, K, M/X and I‐cells; which set of genes it regulates in EECs is unclear but it has previously been implicated in endocrine cell specification in the endocrine pancreas (Gage et al., 2015); (3) LMX1A and MNX1 were most highly expressed in EC cells, as reported previously (Beumer et al., 2020); and (4) the TF expression profiles of K, I and L‐cells were largely indistinguishable, consistent with the strong overlaps between these cell populations found in a number of studies (Egerod et al., 2012; Haber et al., 2017; Habib et al., 2012; Kaelberer & Bohorquez, 2018; Smith et al., 2024).

Transcriptomic characteristics of human EEC clusters were broadly similar to those described previously in mice, isolated from intestinal tissue labelled by fluorescent reporter expression controlled by NeuroD1‐Cre expression (Smith et al., 2024) but with several notable differences. Whereas a number of analyses of mouse EECs have revealed strong overlap between L‐cells and I‐cells (Egerod et al., 2012; Habib et al., 2012; Smith et al., 2024), these clusters were more distinct in the human data, with relatively low co‐expression of GCG and PYY in duodenal 'I‐cells', or of CCK in ileal 'L‐cells'. Instead, the human I‐cell cluster exhibited relatively high expression of GAST. GIP was similarly co‐localised with GAST, as also reported previously in human organoids (Beumer et al., 2020). Production of GAST in the human small intestine does not seem to be an artefact of generating EECs in organoids, as our previous studies identified relatively high levels of GAST peptide in human duodenal extracts and of GAST mRNA and protein in purified native jejunal EECs (Roberts et al., 2019). In mouse, by contrast, there are very few Gast‐positive cells outside the stomach and no evident co‐localisation with Gip (Smith et al., 2024). Alongside this study, we further optimised differentiation protocols to generate mature functional K‐cells in human GIP‐Venus organoids (Guccio et al., 2024), which also expressed high levels of GAST mRNA and peptide; however, the physiological role of duodenal‐derived gastrin is currently obscure.

Although nutrient‐sensing GPCRs have been identified in many different EEC populations (mouse and human) by bulk RNA sequencing, there is less understanding of how expression levels vary between EEC cell types. Applying scRNAseq to CHGA‐labelled cells enabled direct comparisons between EEC clusters and revealed several notable differences in this study. SLC5A1 expression was highest in K‐ and L‐cells, consistent with the extensively documented ability of oral glucose to trigger elevations in plasma GIP and GLP‐1 (Nauck et al., 1993), attributed mechanistically to the action of SGLT1 as the intestinal glucose sensor (Gorboulev et al., 2012). GPBAR1 was identified in all EEC clusters but was particularly highly expressed in L‐cells and was higher in the ileum than duodenum, consistent with the ileum being the major site of bile acid reabsorption – a step previously shown to be essential for bile acids to reach GPBAR1 located on the basolateral surface of L‐cells (Brighton et al., 2015). The long‐chain free fatty acid receptor FFAR1 was most highly expressed in K‐, I‐ and L‐cells, mirroring the well‐documented FFAR1‐dependent increases in plasma GIP, CCK and GLP‐1 levels following lipid ingestion (Hauge et al., 2015; Liou et al., 2011). FFAR4 expression was highest in M/X cells: although FFAR4 has been linked to gut hormone secretion in a number of studies, the function of FFAR4 in EECs remains uncertain (Hirasawa et al., 2005; Sankoda et al., 2017) and we were previously unable to demonstrate any stimulatory contribution of FFAR4 agonists to MLN secretion in human MLN‐labelled organoids (Miedzybrodzka et al., 2021). Notable amongst the receptors for short‐ and branched‐chain fatty acids were the high expression of FFAR2 in L‐cells and of both FFAR2 and OR51E1 in ileal ECs, suggesting that small intestinal microbial fermentation products might exert their strongest effects on the secretion of GLP‐1, PYY and 5‐HT. Roles for Ffar2 and the OR51E1 homologue Olfr558 have been demonstrated previously in murine L‐cells and EC cells (Bellono et al., 2017; Tolhurst et al., 2012). Subtle differences in the nutrient‐sensing machinery between some EEC clusters likely contribute to different patterns of plasma gut hormone concentrations measured after food ingestion.

Hormone secretion by EECs occurs through calcium‐dependent vesicular exocytosis, a highly regulated process involving a cohort of conserved core SNARE complex proteins alongside a range of accessory and regulatory proteins, the composition and function of which have yet to be characterised in human EECs (Sudhof, 2013). We found that, similar to most neuronal and neuroendocrine cell types which exhibit regulated calcium‐dependent exocytosis, all populations of human ileal and duodenal EECs identified expressed the core SNARE complex proteins (SNAP25, STX1A and VAMP2) and accessory proteins implicated in vesicular docking and priming (STXBP1, UNC13A/B, RAB3B and CPLX1/2). This suggests that exocytosis in EECs likely occurs through mechanisms similar to those of other secretory cells, although this is yet to be confirmed with functional studies. We also found that all EECs showed expression of exocytotic proteins known to be particularly enriched in neuroendocrine cells such as those found in the pancreatic islets, specifically the calcium‐binding proteins SYT7 and SCGN (Gustavsson et al., 2009; Wagner et al., 2000). Expression of many of the exocytotic proteins in EECs appears to be consistent between human and mouse, where the core SNARE complex proteins STX1A, SNAP25 and VAMP2, and the accessory protein STXBP1 have been demonstrated to have functional roles in exocytosis in mouse L‐cells (Campbell et al., 2020; Li et al., 2014; Wheeler et al., 2017). Expression of exocytotic proteins was broadly similar across all EECs meaning that the terminal steps of secretion are likely regulated in a similar manner among the different EEC types. However, we did observe an enrichment of SYT1 in the ileal population of EECs, the calcium sensor responsible for fast and synchronous vesicular exocytosis in neuronal chemical synapses (Sudhof, 2013). This is interesting given that there is some evidence that EECs may be capable of fast neurotransmitter‐mediated transmission to neuronal afferents (Kaelberer et al., 2018; Lu et al., 2019).

Following the success of gut hormone‐based therapies for treating intestinal disorders, diabetes and obesity, EECs are a prime target for future drug development. Despite the widespread application of single‐cell transcriptomics to generate atlases of cell types in the human body, there are insufficient EECs in most databases to enable robust cluster analysis because the entire EEC population only makes up around 1% of the intestinal epithelium. By performing scRNAseq on EECs purified from CHGA‐Venus organoids, we have substantially increased the number of EECs for analysis. Human EECs generated in organoids replicate the transcriptomic properties of their native counterparts, making them a good model for studying EEC physiology. Supporting the relevance of the transcriptomic approach taken here, a number of studies have demonstrated that EECs in human organoids replicate physiological properties of the native enteroendocrine system, employing nutrient‐sensing machinery such as SGLT1, FFAR1 and GPBAR1 (Goldspink et al., 2020; Guccio et al., 2024; Miedzybrodzka et al., 2021). The ability to knock out candidate genes in organoids using CRISPR‐Cas9 further offers a viable and valuable approach to validate the physiological relevance of individual genes identified in transcriptomic studies, complementing more traditional pharmacological approaches that are frequently hampered by limitations in drug specificity. The lack of SCT‐expressing cells in organoids generated by current protocols suggests they have not reached full maturity, and while they are a good in vitro model system for the human enteroendocrine system, there is still a place for studies on intact perfused intestinal models, free‐living animals and human volunteers where the mucosa is mature, intact and in contact with other cell types not present in the organoid system such as inflammatory cells and neurones.

Although the data clustered broadly into cell types according to the dominantly expressed hormone, mirroring the classical single‐letter nomenclature, overlaps of hormone expression were apparent and are supported by many previous studies (Egerod et al., 2012; Haber et al., 2017; Habib et al., 2012; Kaelberer & Bohorquez, 2018) raising the question of whether it is time for the field to agree a new nomenclature for EECs that is both intuitive and representative.

F.M.G. and F.R. have received funding from AstraZeneca and Eli Lilly for projects not overlapping the material in this manuscript. They also received sponsorship to host the European Incretin Study Group meeting in Cambridge 2024 from AstraZeneca, Eli Lilly, Sun Pharma and Mercodia. D.G. is currently an employee of GlaxoSmithKline, but her work towards this manuscript was performed when an employee of the University of Cambridge and is unrelated to and independent of her position at GSK.

C.S., F.R. and F.G.: Conception or design of the work; acquisition, analysis or interpretation of data for the work; drafting the work or revising it critically for important intellectual content; final approval of the version to be published; and agreement to be accountable for all aspects of the work. V.L., R.B.B., E.M., A.D. and D.G.: Conception or design of the work; acquisition, analysis or interpretation of data for the work; drafting the work or revising it critically for important intellectual content; and final approval of the version to be published. All authors have approved the final version of the manuscript and agree to be accountable for all aspects of the work. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

This research was funded by a Wellcome joint investigator award to FR/FMG (220271/Z/20/Z) and the MRC Metabolic Diseases Unit (MRC_MC_UU_12012/3). Core support was provided by MRC [MRC_MC_UU_00014/5] and Wellcome [100574/Z/12/Z]). F.G. and F.R. received funding from AstraZeneca and Eli Lilly for non‐overlapping research on other projects. They received sponsorship from AstraZeneca, Eli Lilly, Sun Pharma and Mercodia to run the 5th European Incretin Study Group conference in Cambridge (April 2024).

V.B. Lu: Department of Physiology and Pharmacology, University of Western Ontario, London, ON, Canada N6A 5C1. R. Bany Bakar: Department of Medicine, University of Cambridge, Addenbrooke's Hospital, Cambridge, UK. E.Miedzybrodzka: Centre for Regenerative Medicine, Institute for Regeneration and Repair, The University of Edinburgh, 5 Little France Drive, Edinburgh EH16 4UU, UK. D.Goldspink: GlaxoSmithKline, Medicines Research Centre, Gunnels Wood Road, Stevenage, SG1 2NY, UK.