Diet-induced alteration of intestinal stem cell function underlies obesity and prediabetes in mice

1Diet-induced obesity is a serious public health and economic problem. Obese people are at higher risk of developing type 2 diabetes (T2D), cardiovascular diseases and cancer, all leading causes of death worldwide (https://www.who.int). Today, bariatric surgery is the most effective treatment to achieve long-term weight loss and notably leads to diabetes remission. Surgical procedures cause profound changes in secretion of gut hormones with beneficial effects on whole body metabolism, appetite and food intake. These findings suggest gut hormones as candidates for new therapies against obesity and diabetes.

2The gut, as the body’s digestive and largest endocrine system, serves as a central regulator of energy and glucose homeostasis and quickly responds to dietary and nutritional changes–. Constant overnutrition is thought to lead to intestinal maladaptation and dysfunction and to contribute to the development of obesity and prediabetes. This is evident as two hallmarks of obesity, excessive food intake and a reduced stimulation of postprandial insulin secretion by gut hormones, are linked to impaired gut function. Moreover, differences in gut morphology and physiology have been observed between lean and obese individuals,,. Tackling gut dysfunction at an early stage of disease might therefore be a promising treatment option to fight obesity and the resulting risks and complications.

3Intestinal functions are carried out by specialized epithelial cells lining the gut: absorptive enterocytes, antimicrobial-peptide-secreting Paneth cells, hormone-secreting enteroendocrine cells (EECs), mucus-secreting goblet and chemosensory tuft cells. The cells of the intestinal epithelium are constantly generated from ISCs. ISC identity is defined by multi-lineage potential and self-renewal capacity, but also by properties that are not hard-wired, such as the proliferative, epigenetic and metabolic state,. High cell turnover and cellular plasticity contribute to the natural adaptive capacity of the gut, but the mechanisms underlying maladaptation in response to an obesogenic diet are still unclear. Specifically, we do not know whether the hormonal imbalance and increased absorptive capacity emerge at the level of early lineage commitment from ISCs,. Moreover, the functions of enterocytes and EECs differ between gut regions–. Proximal enterocytes, for instance, are specialized to absorb iron and nutrients (carbohydrates, fat and protein), whereas distal enterocytes absorb bile acids and vitamin B12. Proximal EEC types secrete serotonin and ghrelin, whereas distally located EECs preferentially secrete Glp-1 (ref. ). Gut functions are also spatially compartmentalized along the crypt–villus axis. Enterocytes shift their expression profile from an antimicrobial to a nutrient absorption to an immunomodulatory program while migrating from the bottom of the villus to its tip. The spatial–functional compartmentalization of nutrient absorption is achieved through zoned expression of nutrient transporters, with the highest expression of carbohydrate and amino acid transporters in the mid-villus region and of apolipoproteins and fatty acid transporters, such as Apoa4 and Fabp1, at the villus tip. Similarly, enteroendocrine cell types switch their hormone expression pattern along the crypt–villus axis. Regional identities are thought to be determined at the level of the ISCs by epigenetic mechanisms, and the crypt–villus compartmentalization is at least partly established by growth-factor gradients, but it is unknown whether and how the compartmentalization of gut functions is affected by an obesogenic diet,.

4Developing pharmacological approaches to counteract obesity and diabetes require in-depth understanding of the mechanisms that underlie maladaptation and endocrine dysfunction in the gut, specifically during the transition from a healthy to an obese and to a prediabetic state. In this study we combined single-cell profiling with genetic lineage labelling and tracing of ISC fate decisions and in situ metabolomics to elucidate the cellular and molecular mechanisms that underlie intestinal maladaptation to an obesogenic western-style high-fat/high-sugar diet (HFHSD) in mice.

5To study mechanisms of intestinal maladaptation to an HFHSD, we maintained male Foxa2–Venus fusion (FVF) reporter mice on a diet regimen for 12 weeks. Over this time, mice on an HFHSD gained significantly more body weight (fat and lean mass) than did control diet (CD)-fed mice, and this was accompanied by an increase in length and weight of the small intestine (SI), increased villus length and decreased crypt density (Extended Data Fig. 1a–l). Crypt depth, number of cells per crypt and cell sizes in crypts and villi were not changed (Extended Data Fig. 1k,m–p). Histological assessment of the SI mucosa showed cellular fat inclusions in the villi of HFHSD-fed mice, suggesting disturbances in fatty acid metabolism in the gut epithelium (Extended Data Fig. 1q). Metabolic assessment showed that our diet-induced obese mice developed prediabetes, which was characterized by fasting hyperglycaemia, impaired glucose tolerance and pronounced hyperinsulinaemia as well as insulin resistance (Extended Data Fig. 1r–u). Thus, an HFHSD changes mucosal morphology, which is indicative of altered ISC homeostasis and lineage recruitment.

6To dissect the metabolic impact of an HFHSD on ISC lineage recruitment, we employed single-cell RNA-sequencing (scRNA-seq) of crypt cells from the SI of CD- and HFHSD-fed FVF reporter mice (Fig. 1a). FVF-lineage labelling enabled us to flow sort and enrich for ISCs and EECs, cell types that together usually make up less than 6% of the intestinal epithelial cells (Fig. 1a and Extended Data Fig. 2a). This enrichment strategy not only allowed us to assess compositional changes on an HFHSD, but also to molecularly and functionally characterize rare crypt-cell types. We profiled 27,687 cells obtained from three biological replicates of CD- and HFHSD-fed mice and detected on average 3,500 genes per cell (Extended Data Fig. 2b). Unsupervised graph-based clustering and annotation based on known marker genes revealed all mature intestinal cell lineages as well as ISCs and distinct progenitor states for each intestinal lineage (Fig. 1b,c and Extended Data Fig. 2c,d). Lineage marker gene expression as well as the expression of known regulators that drive intestinal lineage decisions toward the absorptive (for example Notch1) and secretory lineages (Dll1, Dll4 and Atoh1) and PAGA topology were unchanged between diet conditions (Fig. 1c, Extended Data Fig. 2c–f and Supplementary Table 1). Corresponding subtypes showed high correlation in their transcriptomes, indicating that an HFHSD did not alter lineage identities (Extended Data Fig. 2g). Instead, we observed a shift of cell densities indicating that an HFHSD impacts the composition of the mucosal epithelium (Fig. 1d,e and Extended Data Fig. 2d,h). Note that we excluded the Paneth cell lineage from the scRNA-seq analysis because of reported sampling issues, which might impact cell-type ratios. To assess whether the HFHSD affects regional subtypes, we separated cell types into proximal and distal cells using recently described regional signature genes (Extended Data Fig. 3a). We found that the fractions of ISCs, enterocytes and goblet cells with proximal identity increased in mice fed an HFHSD (Fig. 1f,g and Extended Data Fig. 3b). Together, these data suggest that an HFHSD boosts formation of the absorptive and goblet cell lineage and promotes proximal cell identities, indicating that the intestine adapts cell-type composition and specific cellular functions to nutrient availability.

7As our scRNA-seq experiment indicated alterations of lineage abundances in the crypt on HFHSD, we next performed short-term genetic lineage tracing of ISC fate decisions using a dual-fluorescent inducible Cre-reporter Foxa2nEGFP-CreERT2/+;Gt(ROSA)26mTmG/+ mouse (Fig. 2a), to determine how these changes impact the composition of the mature villus compartment. Foxa2 is expressed in quiescent and rapidly dividing ISCs (Supplementary Fig. 1). In this mouse model, Foxa2 expression-driven CreERT2 induces a switch from membrane-Tomato (mT) to membrane-GFP (mG) in ISCs after tamoxifen administration. At 48 h after Cre-ERT2 activation by tamoxifen, we observed labelled mG-positive ISCs next to Paneth cells at the crypt bottom and labelled single cells and small cell clusters in duodenal crypts at similar frequencies in CD- and HFHSD-fed mice (Extended Data Fig. 4a). To determine the effect of an obesogenic diet on ISC lineage recruitment, we analysed the abundance of mature enterocytes, goblet cells and EECs in Cre-reporter labelled mG-positive lineage ribbons in the villi of CD- and HFHSD-fed mice 70 h after tamoxifen induction (Fig. 2b). In mice fed an HFHSD, longer ribbons (>6 cells) appeared more frequently, suggesting higher ISC or progenitor proliferation or increased cell turnover (Fig. 2c,d and Extended Data Fig. 4b). Within the lineage ribbons, total numbers of villin+ enterocytes and Muc2+ goblet cells were increased, whereas the number of ChgA+ EECs was reduced (Fig. 2c,d). The relative abundances of enterocytes and ChgA+ EECs within lineage ribbons were diminished owing to the large increase in goblet cell numbers in mice fed an HFHSD (Fig. 2c and Extended Data Fig. 4c). In accordance, goblet cell numbers in the duodenal villi were increased and numbers of ChgA+ EECs were reduced in the duodenum and ileum of HFHSD-fed FVF mice (Fig. 2e–j and Extended Data Fig. 4d). Note that the number of ileal goblet cells was not altered, suggesting that gut regions are differently affected by an HFHSD (Extended Data Fig. 4e). As we excluded Paneth cells from our scRNA-seq analysis, we determined the Paneth cell numbers in situ and found no difference between CD- and HFHSD-fed mice (Extended Data Fig. 4f,g). To understand whether the shift in regional identity in crypt cells translates to the villi and results in altered function, we performed scRNA-seq of villus cells from the SI (Extended Data Fig. 5a–c). We focused on enterocytes because regional and spatial compartmentalization of enterocyte function is well described and classified them into proximal and distal enterocytes using reported regional transcription factors (Extended Data Fig. 5d,e),. Consistent with crypt scRNA-seq data, we also found an increased fraction of proximal-type villus enterocytes in mice fed an HFHSD (Extended Data Fig. 5f). Furthermore, proximal and distal-type enterocytes showed upregulation of functional proximal genes such as Fabp1 and Apoa4 (Extended Data Fig. 5g).

8Next, we assessed whether enterocyte zonation is affected by an HFHSD. We inferred a pseudospatial ordering of cells using previously reported zonation markers and partitioned cells into five zones along the axis from villus bottom to tip (Extended Data Fig. 6a). We found that in proximal and distal enterocytes, the spatial expression pattern of several genes associated with carbohydrate and fatty acid absorption was altered in HFHSD-fed mice (Extended Data Fig. 6b). Analysis of Fabp1 and Apoa4 immunolocalization in the ileum and duodenum confirmed the proximalization of enterocytes and altered enterocyte zonation. The expression zones of Fabp1 and Apoa4 were enlarged on an HFHSD and reached from villus tip to villus bottom (Extended Data Fig. 6c–k).

9Thus, more goblet cells and proximal-type enterocytes are generated under HFHSD conditions, which leads to morphological changes of the gut mucosa and an increase in fatty acid transport and absorption. Reduced ChgA+ EEC numbers imply changes in specific EEC subsets.

10Secreted gut hormones are critical regulators of food intake and systemic metabolism along the gut–brain–pancreas axis and hormonal imbalance is linked to the metabolic syndrome,. To elucidate the mechanisms underlying EEC dysfunction in response to an obesogenic diet, we first refined the clustering of our 2,865 EEC lineage cells into distinct subpopulations, which revealed five enteroendocrine progenitor (EEP) clusters characterized by expression of Sox4, Ngn3, Arx/Isl1, Ghrl and Pax4, respectively, and six polyhormonal EEC clusters: SILA cells (coexpress Sct, Cck, Gcg, Ghrl and Gal), SILP cells (coexpress Sct, Cck, Gcg and Pyy), SIK cells (coexpress Sct, Cck and Gip), SAKD cells (coexpress Sct, Cck, Ghrl, Gip and Sst), SIN cells (coexpress Sct, Cck, Gcg and Nts), enterchromaffin (EC) cells (coexpress Sct, Tac1, Tph1) and Reg4+ EC cells (coexpress Sct, Tac1, Tph1, Ucn3 and Reg4) (Fig. 3a,b and Extended Data Fig. 7a,b),,. We also identified a population of heterogeneous EECs coexpressing endocrine and ISC markers that we termed Lgr5+ EECs and which are reminiscent of Lgr5+ label-retaining cells (Fig. 3a,b and Extended Data Fig. 7a). Notably, we found that Lgr5+ EECs are characterized by active Bmp signalling, which has been shown to regulate the hormonal plasticity of EECs (Extended Data Fig. 7c). To determine EEC lineage hierarchy and to understand the relationship of EEP and mature clusters, we used PAGA in combination with RNA velocity (Methods). The abstracted graph represents possible differentiation paths that cells follow and RNA velocity predicts the future state of a cell based on gene expression state (gene induction or repression), thus indicating the direction of differentiation. We observed two main differentiation trajectories from the Sox4+ progenitors to mature subtypes: (1) a path via Arx/Isl1+ and Ghrl+ progenitors to peptidergic EECs (SILA, SILP, SIK and SAKD) and (2) a route via Pax4+ progenitors to serotonergic EC and mature Reg4+ EC cells (Fig. 3c). To identify transcriptional signatures and regulators of EEC lineage allocation, we extracted genes that were either only transiently expressed in a specific EEP stage or mature cell subtype (state-specific genes) or turned on with sustained expression in subsequent states (global or lineage-specific genes) (Fig. 3d, Extended Data Fig. 7a,d and Supplementary Tables 2 and 3). Transiently expressed markers in progenitor stages (for example Sfrp5 in Ngn3+ progenitor) are potentially important for specification. Global or lineage-specific markers (for example Nefm in peptidergic lineage) might regulate endocrine cell identity. Together, our data confirm EEC lineage differentiation from an early common progenitor into EC-biased (Pax4+) and non-EC-biased (Arx/Isl1+) progenitors and their respective molecular programs,,. We then compared EEC lineage allocation in CD- and HFHSD-fed animals. We found that an HFHSD reduces the number of Sox4+ early EEPs, increases the fraction of Ngn3+ EEPs, reduces the number of Lyz1−BrdU+ cells that correspond to label-retaining Lgr5+ EECs and affects mature SILA and Reg4+ EC cells, which are most abundant in the duodenum (Fig. 3e–o and Extended Data Fig. 7e–i). Serotonin (5-HT)+Reg4+ EC numbers were decreased in HFHSD crypts, whereas Ghrl+ SILA cells were more abundant in the duodenal crypts and villi (Fig. 3h–m). Numbers of ileal located SILA/SILP cells expressing Gcg, which encodes the incretin Glp-1, were also increased with an HFHSD (Fig. 3n,o). We next assessed whether an HFHSD induces transcriptome changes that might affect EEC lineage allocation or function. Single-cell messenger RNA expression levels of hormones did not differ between CD and HFHSD EEC subsets (Extended Data Fig. 7j). However, genes associated with metabolism (for example Slc5a1, Acadl in SIK cells), vesicular trafficking machinery and the secretome (for example Cplx2 in EC and Sct in SIK cells) as well as signalling (transduction) (for example ID1/ID3 in Sox4+ progenitors, Gnas in EC cells) and transcription factors (for example Cdx2 in Sox4+ progenitor, Hmgn3 in EC cells) were differentially expressed between CD and HFHSD conditions (Extended Data Fig. 7k–m and Supplementary Table 4).

11Next, we examined EEC functionality by assessing hormone secretion (Extended Data Fig. 7n). Intestinal EC cells produce over 90% of the body’s serotonin. We found that basal plasma levels of serotonin were lower in mice fed an HFHSD, which correlates with reduced numbers of 5-HT+ EECs (Reg4+ EC cells) (Extended Data Fig. 7o). By contrast, despite a higher number of Ghrl+ SILA EECs, plasma levels of ghrelin were reduced in HFHSD-fed mice (Extended Data Fig. 7p), suggesting that Ghrl+ EECs are functionally impaired upon HFHSD. Glp-1 plasma levels were also increased, which corresponds with increased numbers of ileal Gcg+ cells (Extended Data Fig. 7q). Taken together, these results show that an HFHSD impacts EEPs, alters the number of specific mature EEC subtypes and circulating gut hormone levels and affects expression of genes that are potentially important for endocrine cell differentiation and/or function.

12We next asked whether the increased number of Ngn3+ EEPs, as well as the increase in villus length and higher abundance of the enterocyte and goblet cell lineage under HFHSD conditions, resulted from enhanced proliferation. To test this, we identified proliferating cells using a cell-cycle signature score that assigns each cell to a cell-cycle state (G1, S and G2M; Methods) (Extended Data Fig. 8a). HFHSD mainly increased the proportion of cells in G2M phase, indicating that HFHSD alters cell-cycle dynamics (Fig. 4a,b). In particular, in ISCs, enterocytes, goblet cells and the early Sox4+ and Ngn3+ EEP clusters, the proportion of cycling cells was clearly increased (Fig. 4c and Extended Data Fig. 8b). Moreover, several genes, including the known cell-cycle regulators Ccnb1, Cenpa, Dut, Pbk and Smc2, which correlated with the S/G2M cell-cycle score, were differentially expressed in proliferating cells upon HFHSD, further suggesting that cell-cycle dynamics of ISCs and progenitors is accelerated (Fig. 4d and Supplementary Table 5).

13To confirm the increase in proliferative activity, we first identified 5-ethynyl-2ʹ-deoxyuridine (EdU)-positive progenitor cells 2 h after an EdU pulse. Consistent with the crypt scRNA-seq data, we found a higher number of proliferative progenitors in the transit-amplifying (TA) zone in SI crypts of HFHSD-fed FVF mice (Fig. 4e,f). To directly assess the number of proliferating ISCs we put Lgr5-ki mice on an HFHSD (Extended Data Fig. 8c). In contrast to our scRNA-seq data, we found no difference in the numbers of ISCs between CD and HFHSD by flow cytometry on the basis of high Lgr5-EGFP fluorescence intensity (Extended Data Fig. 8d and Fig. 4g). However, the numbers of EdU-positive ISCs were increased on HFHSD (Fig. 4h,i).

14Further, we determined antigen immunoreactivity to Ki67 and the migration rate of crypt cells. We observed that Ki67-labeled domains were significantly larger in mice fed an HFHSD and, in contrast to controls, extended into the villi. Also, the crypt-to-villus migration rate of pulse-labeled 5-bromo-2ʹ-deoxyuridine (BrdU)-positive cells was increased on HFHSD (Extended Data Fig. 8e–i) which altogether suggests that an HFHSD enhances cell turnover.

15In summary, these data demonstrate that an obesogenic diet increases the proliferation rate of ISCs and progenitors leading to increased villus length. Enhanced cell-cycle activity of ISCs did not result in physical expansion of the ISC pool due to accelerated differentiation and cell turnover. This is reflected by the decreased fraction of ISCs and increased fraction of progenitors in our crypt scRNA-seq data (Fig. 1e). These results further indicate that ISC homeostasis and identity is disturbed in HFHSD-fed animals.

16A central role of metabolic pathways in the regulation of stem cell maintenance and fate control has been described in several adult stem cell systems,. Both Wnt/β-catenin and Igf1/insulin signalling integrate metabolic and proliferative cues, and increased activity of these pathways has been shown to induce ISC hyperproliferation upon high-fat diet (HFD) feeding,–. We assessed Wnt/β-catenin pathway activation in bulk ISC-enriched FVFlow cells and FVFneg enterocyte progenitors, which we isolated by flow cytometry (Supplementary Fig. 2a,b). Unexpectedly, the level of nuclear β-catenin was decreased or unchanged in ISC-enriched FVFlow cells or enterocyte progenitor-enriched FVFneg cells, respectively (Supplementary Fig. 2c,d). Consistently, expression of Wnt/β-catenin target genes was downregulated in ISC-enriched FVFlow cells and HFHSD-derived single-cell ISCs (Supplementary Fig. 2e–g and Supplementary Table 1). As glycogen synthase kinase 3β (Gsk3β) is a negative regulator of Wnt/β-catenin signalling and an important modulator of cellular metabolism, we checked Gsk3β activity. Decreased levels of phosphorylated Gsk3β in HFHSD-derived SI crypts indicated increased Gsk3β activity and enhanced Gsk3β-mediated destruction of β-catenin (Supplementary Fig. 2h,i). Together, these results demonstrate that Wnt/β-catenin signalling does not drive hyperproliferation of ISCs and progenitors and enhanced progenitor recruitment in our HFHSD mouse model. Instead, decreased Wnt/β-catenin signalling indicates that differentiation of ISCs is accelerated and thus provides further evidence of disturbed ISC homeostasis,. HFHSD caused obesity and hyperinsulinaemia in our mouse model. Analysis of InsrIgf1rAkt pathway activity in SI crypt lysates showed increased phosphorylation of InsrIgf1r and Akt as well as upregulation of several genes associated with PI3K and Akt signalling, which is in line with pronounced hyperinsulinaemia and confirms previous findings that Igf1insulin signalling induces hyperproliferation in diet-induced obesity (Extended Data Figs. 1r–u and 9a–c).

17To uncover additional metabolic pathways that link HFHSD to hyperproliferation and cell fate changes, we compared metabolite profiles from SI crypt regions of CD- and HFHSD-fed FVF mice using matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI–MSI). MALDI–MSI allows analysis of metabolites directly in tissue sections without isolation bias. We identified 297 discriminative masses; of these, 257 were enriched and 40 were less abundant in SI crypts of HFHSD-fed animals (Fig. 5a). Pathway enrichment analysis revealed that HFHSD upregulated metabolite signatures related to fatty acid biosynthesis (for example, octadecanoic acid) and linoleic acid metabolism (for example, phosphatidylcholine) and downregulated metabolites linked to pathways of glucose metabolism, such as the pentose phosphate pathway and pentose glucuronate interconversions (for example, d-glyceraldehyde 3-phosphate) (Fig. 5b and Supplementary Table 6). To map the metabolic changes from tissue to cell-type level, we integrated metabolomics data with bulk transcriptomes of ISC-enriched FVFlow cells, secretory progenitor-enriched FVFhigh cells and enterocyte progenitor-enriched FVFneg cells (Extended Data Fig. 9d, Fig. 5c and Methods). The HFHSD metabolite signature overlapped with regulated genes involved in carbohydrate metabolism (for example, sucrose degradation, maturity onset diabetes of young (MODY) signalling) and pro-proliferative fatty acid biosynthesis pathways (for example, stearate synthesis and acyl-CoA-hydrolysis). Regulation of these pathways was most pronounced in enterocyte progenitor-enriched FVFneg and ISC-enriched FVFlow cells (Fig. 5d and Supplementary Table 7). Next, to determine the cell subtypes in which an HFHSD altered metabolism and to identify the molecular pathways associated with deregulated HFHSD metabolites, we compared the transcriptional profiles of CD- and HFHSD-derived cells. We found that gene signatures associated with Ppar signalling (for example Hmgcs2, Acdvl, Acaa2, Fabp1 and Fabp2) and fatty acid biosynthesis (for example Acot1, Acox1, Scd2, Srebf1, Fads1 and Me1) were upregulated in HFHSD-derived ISCs and progenitor clusters (Fig. 5e–g and Supplementary Tables 1,8 and 9). Upregulation of these gene sets was stronger in subpopulations with proximal identity; in particular, in proximal ISCs, enterocyte and goblet progenitors (Fig. 5e,f and Supplementary Table 8). Moreover, genes associated with carbohydrate metabolism (for example glycolysis, gluconeogenesis and the pentose phosphate pathway) were downregulated in these subtypes (Fig. 5e–g). We confirmed upregulation of active Srebp1 (mature, m-Srebp1), Acc, Pparγ and Scd1, the master transcriptional regulators and key enzymes of Ppar signalling and fatty acid synthesis in HFHSD crypt protein lysates and in ISCs and progenitors, by a sensitive, targeted single-cell qPCR approach (Fig. 5h,i, Extended Data Fig. 10 and Supplementary Table 10). Finally, consistent with crypt enterocytes, we also found that mature villus enterocytes increase the expression of genes associated with intracellular fat accumulation and lipid uptake, de novo lipogenesis and peroxisomal fatty acid oxidation (Supplementary Fig. 3).

18Thus, by integration of in situ metabolomics, bulk and single-cell transcriptomics and targeted protein expression analysis, we have revealed diet-induced metabolic rewiring and cell- and subtype-specific transcriptional changes that correlate with an increase in proliferation and endocrine dysfunction.

19In this study we provide a basic mechanistic explanation of diet-induced ISC dysfunction and intestinal maladaptation that underlie the development of obesity and prediabetes and increase the risk for gastrointestinal cancer.

20With our study we aimed to determine the immediate effects of a western-style HFHSD on intestinal function. We observed enlargement of the SI, longer villi and decreased crypt density and an altered cellular composition in the crypts, which was confirmed in the villi by lineage-tracing studies. We are aware that our lineage-tracing approach using the dual-fluorescent, inducible Cre-reporter Foxa2nEGFP-CreERT2/+;Gt(ROSA)26mTmG/+ mouse has its limitations; for example, inefficient labelling of stem cells, which generates fewer clonal ribbons in comparison to the Lgr5-ki reporter and analysis of only one time point. Therefore, we validated our findings from the crypt scRNA-seq experiment not only by a lineage-tracing approach but also independent of a reporter in tissue sections.

21We show that ISCs and progenitors are hyperproliferative, and that differentiation and cell turnover are accelerated by an HFHSD. Accelerated differentiation and cell turnover also explain the discrepancy between the scRNA-seq data (decrease in ISCs) and the data from the Lgr5-reporter mice (no change in the number of ISCs). ISCs in HFHSD-fed mice divide faster and newly formed cells are still Lgr5–EGFPhi, but already express markers of differentiated cells. Whole-transcriptome-based clustering in our scRNA-seq data groups Lgr5hi ISCs and Lgr5hi progenitors separately, whereas when using only Lgr5–EGFP as a marker in tissue sections and flow cytometry we cannot discriminate between an ISC and a progenitor.

22Hyperproliferation in the crypt, which can promote cancer initiation and progression, has previously been associated with an HFD; however, the molecular pathways that couple dietary cues to this cellular response are still debated and probably depend on the type of diet (for example fat and sugar source),,. Also, a role of metabolic pathways in ISC maintenance, number and fate control has been highlighted in several studies,,. Hyperproliferation upon HFHSD is not driven by Pparδ-mediated activation of the major oncogenic pathway in the gut, Wnt/β-catenin signalling, as previously reported for a lard-based chronic HFD,,. Instead, we found that a coconut oil- and sucrose-based high-lipid and -carbohydrate content diet specifically elevates pro-proliferative Pparγ signalling, Srebp1-mediated lipogenesis and InsrIgf1r–Akt signalling. Upregulation of Pparγ and Srebp1-mediated de novo lipogenesis has been associated with inflammation, increased proliferation and tumour progression in many types of cancer and thus provides a possible link between HFHSD-induced metabolic signalling, crypt-cell hyperproliferation and increased risk of gastrointestinal cancer. As tumour initiation and progression require biomass production and therefore depend on a high nutrient supply, chronic activation of these pathways as observed with an HFHSD might reduce the barrier to oncogenic transformation or tumour growth and proliferation.

23Further, we found a profound impact of an HFHSD on the enterocyte lineage. Enterocytes are metabolically rewired, increase the expression of genes linked to carbohydrate and fat uptake and show an intracellular fat accumulation, which is reminiscent of the vesicular accumulation of triglycerides in hepatic steatosis that causes liver fibrosis and cancer–. In addition, HFHSD induces a regional and spatial repatterning of enterocyte gene expression and function. An HFHSD increased the number of proximal-type enterocytes, which are specialized on carbohydrate and fatty acid absorption and altered enterocyte zonation along the crypt–villus axis (increased expression of the fatty acid transporter Fabp1). Thus, our results imply that enterocytes functionally adapt to an HFHSD, which may increase calorie intake and fat accumulation and promote obesity.

24Our in-depth molecular and functional analysis of the EEC lineage reveals mechanisms of EEC dysfunction in obesity that include (1) higher abundance of Ngn3+ EEPs owing to increased proliferation, (2) lower abundance of serotonergic Reg4+ EC cells and lower plasma serotonin levels, (3) higher abundance of peptidergic Ghrl+ SILA cells but lower ghrelin plasma levels, (4) increased numbers of ileal Gcg+ cells and increased Glp-1 levels and (5) lower abundance of Lgr5+ EECs. The physiological roles of gut-derived serotonin are broad and it regulates various processes both in the gut and systemically. Accumulating evidence indicates a link between peripheral serotonin and systemic glucose and lipid metabolism, as well as metabolic diseases. Whether blood serotonin levels are changed in obesity is controversial, due to difficulties in measuring serotonin. However, intraperitoneal injection of 5-HT to mice inhibits weight gain, hyperglycaemia and insulin resistance on an HFD. Thus, reduced numbers of serotonin-producing cells and blood serotonin levels as observed in our mouse model might promote the development of obesity. The hunger hormone ghrelin is known to increase gastric emptying and decrease insulin secretion. A negative correlation between plasma insulin and ghrelin has been reported in human obesity, which is in line with our data. Notably, we found that lower plasma ghrelin levels are not due to a reduced number of Ghrl+ SILA cells, suggesting that an impaired secretory machinery might affect ghrelin levels. However, given that ghrelin-secreting cells are also present in the stomach, changes in plasma ghrelin levels may not be solely attributed to duodenal Ghrl+ EECs. Lower postprandial GLP-1 levels are reported in obesity and incretin hormone secretion and activity are impaired in individuals with prediabetes, although findings are contradictory,. In our obesity model, however, the number of Gcg-expressing cells and Glp-1 plasma levels were increased. Glp-1 stimulates insulin secretion from pancreatic β-cells, so the higher levels of circulating Glp-1 in HFHSD-fed mice may be a compensatory response to insulin resistance at the prediabetic state, which leads to hyperinsulinaemia.

25Finally, in contrast to other studies, Paneth cell numbers were not affected in our obesity model.

26T he discrepancies between our results and those of previous studies support the emerging evidence that the intestine fine-tunes its response to environmental stimuli,,. Different factors, such as the selected mouse model (for example, diet versus genetically induced obesity), type of diet (for example HFD versus HFHSD), fat source (for example lard versus coconut oil) and/or duration of diet (short-term versus chronic) can influence intestinal remodelling and response. For instance, in contrast to the coconut oil, high-sucrose diet for 12–14 weeks used in this study, a lard-based low-sugar chronic diet for 9–14 months caused shorter villi, increased the number of ISCs, decreased the number of Paneth cells and did not affect goblet cells and EECs. Hence, careful selection and reporting of dietary information in animal studies is crucial to interpret and contextualize results.

27In summary, our study reveals that functional maladaptation of the gut in response to an HFHSD is caused by disturbed ISC identity, changes in the regional identity of cells and an altered mature cell-type composition. Further, we describe targetable pathways that are induced by an HFHSD and potentially underlie the pathogenesis of the metabolic syndrome and gastrointestinal cancer. This new understanding of the mechanisms of disease is crucial to develop non-invasive therapeutic options to resolve obesity and insulin-dependent diabetes, for example by counteracting enteroendocrine dysregulation (for example, by elevating peripheral serotonin levels) and increased nutrient absorption (for example, by inducing distal enterocyte phenotypes).

28Animal experiments were carried out in compliance with the German Animal Protection Act and with the approved guidelines of the Society of Laboratory Animals and of the Federation of Laboratory Animal Science Associations. This study was approved by the institutional Animal Welfare Officer (Helmholtz Center Munich) and by the Government of Upper Bavaria, Germany. Homozygous FVF mice were generated as previously described and backcrossed to C57BL/6 background. Foxa2nEGFP-CreERT2 mice (CD1 background) were crossed with Gt(ROSA)26mTmG mice (mixed 129/SvJ, C57BL/6J background) to obtain heterozygous Foxa2nEGFP-CreERT2/+;Gt(ROSA)26mTmG/+ animals and bred in our own facilities. Other mouse lines were Lgr5-EGFP-IRES-creERT2 (ref. ) (Lgr5-ki, C57BL/6J background) and wild-type C57BL/6N (bred in our own facilities).

29Mice were housed in groups of two to four animals and maintained at 23 ± 1 °C and 45–65% humidity on a 12-h dark/light cycle with ad libitum access to diet and water unless otherwise indicated. All experiments were performed using male animals at 3 to 7 months of age.

30For dietary interventions, 10–12-week-old male mice were randomized into test groups matched for body weight, with similar variance, and given ad libitum access to either an obesogenic HFHSD (58% kcal from fat, 25% kcal from carbohydrates, 17% kcal from protein (Research Diets, no. D12331)) or CD (11% kcal from fat, 64% kcal from carbohydrates, 25% kcal from protein (ssniff Spezialdiäten, E15051-04)) for a period of 11–13 weeks. Body weights were measured every second week.

31Lean and fat masses were measured in FVF mice 12 weeks after the start of the diet using quantitative nuclear magnetic resonance technology (EchoMRI).

32Glucose tolerance was assessed by an oral glucose tolerance test (oGTT) in FVF mice, maintained for 12 weeks on CD or HFHSD. After a 6-h fast, mice received an oral glucose bolus (1.5 mg g−1 body weight of 20 % (wt/v) d-(+)-glucose (Sigma-Aldrich) in PBS). Tail blood glucose concentrations were measured with a handheld glucometer (Abbott) before (0 min) and 15, 30, 60 and 120 min after the glucose bolus. To measure the insulin secretion, tail vein blood samples were collected into EDTA-coated microvette tubes (SARSTEDT) at time points 0, 15 and 30 min of the oGTT. Plasma was extracted by centrifugation (3,500 r.p.m., 15 min, 4 °C) and insulin concentration was determined using the Ultra-Sensitive Mouse Insulin ELISA Kit (Crystal Chem, 90080) according to the manufacturer’s instructions.

33The homeostasis model assessment of insulin resistance (HOMA-IR) and HOMA-β were used to assess insulin resistance and beta-cell function, respectively, in FVF mice at 12 weeks after diet start. HOMA indices were calculated from basal blood glucose and plasma insulin levels after a 6-h fast based on the conventional formulas: HOMA-IR = fasting blood glucose (mg per 100 ml) × fasting insulin (µU per ml)/405 and HOMA-β = fasting insulin (µU per ml) × 360/fasting glucose (mg per 100 ml) − 63 (ref. ).

34Circulating hormones, serotonin, ghrelin and Glp-1 were assessed in FVF mice maintained for 13 weeks on a CD or an HFHSD. For basal levels (ghrelin and serotonin), mice were fasted for 6 h and tail vein blood was sampled into EDTA-coated microvette tubes (SARSTEDT). To compare postprandial plasma hormone levels (Glp-1, ghrelin), fasted mice were gavaged with 250 µl of mixed-meal-containing liquid diet (Osmolite HiCal, Abbott) supplemented with dextrose at 20% (wt/v) (Sigma-Aldrich). Blood was collected 10 min after the mixed-meal bolus either from the tail vein or, under terminal anaesthesia with isoflurane, from the vena cava–. For Glp-1 measurement, blood samples were immediately mixed with 0.1 mM Diprotin A (Abcam, 145599) and 500 KIU ml−1 aprotinin (Sigma, A-1153). Plasma was extracted by centrifugation (13,000 r.p.m., 2 min, 4 °C). Total ghrelin concentrations were determined using an ELISA kit from Millipore-Merck (EZRGRT-91K), total plasma Glp-1 was measured using a mouse Glp-1 ELISA kit (Crystal Chem, 81508) and serotonin concentration was determined using a serotonin ELISA kit from Enzo (ADI-900-175).

35To assess the epithelial replication rate in the SI, EdU (Thermo Fisher Scientific, A10044) or BrdU (Sigma, no. B5002) was administered as an intraperitoneal injection at 100 μg g−1 body weight or at 50 μg g−1 body weight, respectively, each from a 10 mg ml−1 stock and in sterile PBS. Mice were killed 2 h post-EdU or 24 h post-BrdU administration. For the assessment of BrdU label retention, FVF mice maintained for 10 weeks on a diet were given BrdU in drinking water at 1 mg ml−1 supplemented with 1% sucrose for 14 d. BrdU was then withdrawn and mice were further maintained on a diet with ad libitum drinking water for a chase period of 21 d. A group of mice was killed after 14 d of continuous BrdU labelling or after a 21d period of chase. For short-term genetic lineage studies, Foxa2nEGFP-CreERT2/+;Gt(ROSA)26mTmG/+ mice were fasted for 3 h and Cre-recombinase activity was induced by a titrated single dose of tamoxifen administered orally by gavage (Sigma-Aldrich, T5648) at 0.25 mg g−1 body weight in sunflower oil. Mice were killed 70 h after the tamoxifen gavage.

36Isolation of small intestinal crypts was carried out as previously reported. In brief, SIs were removed and washed with cold PBS. Villi were scraped off with a glass slide. The remaining tissue was cut into 2-cm pieces, washed several times with cold PBS and incubated in 2 mM EDTA/PBS for 35 min at 4 °C on a tube roller. Subsequently, crypts were collected by rigorous shaking and filtered through a 70-µm mesh to remove villous fragments. For single-cell preparation, isolated crypts were incubated with TrypLE (Life Technologies, no. 12605) for 5 min on ice and then 5 min at 37 °C and treated with 10 µg ml−1 DNase in crypt complete medium (DMEM/F-12 containing 10% FCS) for 5 min at 37 °C. Single-cell suspension was achieved by gentle repeated pipetting. Cells were washed twice with 2% FCS in PBS and pelleted at 300g for 5 min at 4 °C. For flow cytometry, cells were collected in 1–2 ml FACS buffer (2% FCS, 2 mM EDTA in PBS) (Sigma-Aldrich, no. Y0503) and passed through 40-µm cell strainer caps of FACS tubes.

37To obtain a single-cell suspension of villi cells, villi were scraped off and incubated with TrypLE as described for crypt cells.

38For gene expression measurement (microarray, single-cell transcriptomics) and western blotting, small intestinal crypt cells were sorted using FACS-Aria III (FACSDiva software v.6.1.3, BD Bioscience) with a 100-µm nozzle. For all experiments, single cells were gated according to their FSC-A (front scatter area) and SSC-A (side scatter area). Singlets were gated dependent on the FSC-W (front scatter width) and FSC-H (front scatter height) and dead cells were excluded using 7-AAD (eBioscience, no. 00-6993-50). For quantitative PCR with reverse transcription (qRT–PCR), cells were sorted directly into Qiazol lysis reagent (QIAGEN, no. 79306). To obtain FVF-enriched small intestinal crypt-cell samples for scRNA-seq, we sorted 30,000 FVF+ (FVFlow and FVFhigh) cells followed by sorting 30,000 live crypt cells per sample. Cells were sorted into modified FACS buffer (2% FCS, 0.02 mM EDTA in PBS).

39For bulk gene profiling studies (qPCR, microarray), RNA isolation from crypts or flow-sorted cells was performed using the RNA isolation kit miRNeasy Mini (QIAGEN, no. 217004) or miRNeasy Micro kit (QIAGEN, no. 217084) depending on the amount of the sample. Complementary DNA was synthesized using the SuperScript VILO cDNA synthesis kit (Invitrogen, no. 11754). RNA was reverse transcribed and amplified with the Ovation PicoSL WTA System V2 kit (NuGEN, no. 331248).

40For gene profiling of flow-sorted FVFlow, FVFhigh and FVFneg cells from CD- and HFHSD-fed FVF mice, total RNA was isolated as described above and RNA integrity was assessed using an Agilent 2100 Bioanalyzer (Agilent RNA 6000 Pico Kit). RNA was amplified with the Ovation PicoSL WTA System V2 in combination with the Encore Biotin Module (Nugen). Amplified cDNA was hybridized on Affymetrix Mouse Gene 1.0 ST arrays. Staining (Fluidics script FS450_0007) and scanning of the microarray were performed according to the Affymetrix expression protocol including minor modifications as suggested in the Encore Biotin kit protocol. Expression Console (v.1.3.0.187, Affymetrix) was used for quality control and annotation of the normalized robust microarray analysis gene-level data, and standard settings, including median polish and sketch-quantile normalization, were used.

41For gene expression analysis, real-time qRT–PCR was performed using TaqMan probes (Life Technologies), TaqMan Fast Advanced Master Mix (Applied Biosystems, no. 4444557) or TaqMan Universal Master Mix II (Applied Biosystems, no. 4440040) for amplified cDNA and the ViiA 7 Real-Time PCR System (Thermo Fisher Scientific).

42The following probes were used: Mm00782745_s1 for Rpl37, Mm00438890_m1 for Lgr5, Mm01320260_m1 for Olfm4, Mm01268891_g1 for Ascl2, Mm00443610_m1 for Axin2, Mm03928990_g1 for RN18S, Mm02524776_s1 for Fzd2, and Mm00433409_s1 for Fzd7.

43FVF-enriched single-cell samples of crypts isolated from the small intestinal epithelium (duodenum, jejunum and ileum) and villus samples from the SI of C57BL/6N mice were prepared as described above. Dead cells were excluded by flow cytometry after 7AAD labelling. Dead cell exclusion was controlled by trypan blue staining and sorted cells were counted. Single-cell libraries were generated using the Chromium Single cell 3′ library and gel bead kit v2 (10X Genomics, no. 120237) according to the manufacturer’s instructions. Libraries were sequenced on a HiSeq4000 (Illumina) with 150-bp paired-end sequencing of read 2.

44Fresh-frozen samples were cut into 12-μm sections using a cryo-microtome at −20 °C (Leica CM1950, Leica Microsystems) and mounted onto precooled conductive indium-tin-oxide-coated MALDI target glass slides (Bruker Daltonics). Sections were coated with 9-aminoacridine hydrochloride monohydrate matrix (Sigma-Aldrich) at 10 mg ml−1 in water/methanol 30:70 (v/v) by a SunCollect automatic sprayer (Sunchrom). The matrix application was performed at flow rates of 10, 20 and 30, respectively, for the first three layers. The other five layers were performed at 40 μl min−1. MALDI–MSI measurement was performed on a Bruker Solarix 7T FT-ICR-MS (Bruker Daltonics). MALDI–MSI data were acquired over a mass range of m/z 50–1,000 in negative ionization mode with 30-μm spatial resolution using 50 laser shots at a frequency of 500 Hz. The acquired data underwent spectrum processing in FlexImaging v.4.2 (Bruker Daltonics). Following MALDI imaging experiments, the matrix was removed with 70% ethanol. Tissue sections were stained with haematoxylin and eosin. Slides were scanned with a MIRAX DESK digital slide-scanning system (Carl Zeiss MicroImaging).

45For protein expression analyses, whole-cell lysates from isolated crypts or flow-sorted cells were prepared using the RIPA buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Igepal, 0.1% SDS, 0.5% sodium deoxycholate) containing phosphatase (Sigma-Aldrich, P5726, P0044) and proteinase inhibitors (Sigma-Aldrich, P8340). Nuclear and cytosolic extracts from isolated crypts were prepared using the NE-PER Nuclear and Cytoplasmic Extraction Reagents kit (Thermo Fisher Scientific, no. 78833) according to the manufacturer’s instructions. Cell lysates were mixed with Laemmli sample buffer, resolved by SDS–PAGE and blotted onto a PVDF membrane (Bio-Rad). Membranes were blocked with 5% milk in Tris-buffered saline containing 0.2% Tween-20, then incubated overnight with primary antibodies in blocking solution at 4 °C, followed by a 1-h incubation with horseradish peroxidase (HRP)-conjugated IgG secondary antibodies. Protein bands were visualized using a chemiluminiscence reagent (Bio-Rad, no. 170-5061) and quantified using ImageJ software. For quantification, expression of proteins was normalized to α-tubulin or lamin in cytoplasmic or nuclear fractions, respectively.

46Primary antibodies used were mouse anti-Srebp1 (1:1,000 dilution, Novus Biologicals, NB600-582SS); rabbit anti-Acc (1:1,000 dilution, Cell Signaling Technology, 3676); rabbit anti-Pparγ (1:1,000 dilution, Cell Signaling Technology, 2435); goat anti-lamin (1:1,000 dilution, Santa Cruz, sc-6217); mouse anti-α-tubulin (1:1,000 dilution, Sigma-Aldrich, T6199); rabbit anti-Fasn (1:1,000 dilution, Cell Signaling Technology, 3180); rabbit anti-Scd1 (1:1,000 dilution, Cell Signaling Technology, 2794); mouse anti-β-catenin (1:1,000 dilution, BD, 610154), rabbit anti-Gsk3β (1:5,000 dilution, Cell Signaling Technology, 12456); and rabbit anti-phospho Gsk3β (1:5,000 dilution, Cell Signaling Technology, 5558). Secondary antibodies used were goat anti-mouse HRP (1:15,000 dilution, Dianova, 115036062); goat anti-rabbit HRP (1:15,000 dilution, Dianova, 111036045) or rabbit anti-goat HRP (1:15,000 dilution, Dianova, 305035045).

47For tissue histology, intestines were flushed and fixed in 4% paraformaldehyde (PFA) overnight, paraffin embedded according to standard procedures and sectioned at 6 µm. Sections were dried, dehydrated through a graded ethanol series and cleared in xylene. Standard haematoxylin and eosin staining was performed and images were acquired using the Zeiss AXIO Scope A1 microscope (Carl Zeiss AG).

48SIs were isolated, rinsed with ice-cold PBS and fixed with 4% PFA for 3 h at 4 °C. Tissue was cryopreserved through a progressive sucrose gradient (7.5% for 1 h, 15% for 1 h, 30% sucrose overnight), embedded in a tissue-freezing medium (Leica Biosystems, no. 14020108926) and sectioned at 14 µm. For whole-mount staining, isolated small intestinal crypts were fixed in 4% PFA for 30 min at room temperature (RT) and subsequently washed in PBS. For immunofluorescence staining, sections or crypts were permeabilized with 0.5% Triton X-100 in PBS for 30 min at RT, blocked (10% FCS, 0.1% BSA and 3% donkey serum in PBS/0.1% Tween-20) for 1 h and incubated with primary antibodies overnight at 4 °C. Sections or crypts were washed in PBS/0.1% Tween-20 and incubated with secondary antibodies in blocking solution for 1 h at RT, stained with DAPI (ROTH, 6335.1) to visualize the nuclei and mounted with the Elvanol antifade reagent.

49To assess proliferation, EdU staining was performed using the Click-iT Staining kit (Invitrogen, no. C10340) according to the manufacturer’s instructions.

50For BrdU staining, sections were incubated with 3.3 N HCl for 10 min on ice, followed by incubation for 50 min at 37 °C and two wash steps with borate buffer, pH 8.5, to neutralize the reaction (each wash 15 min at RT).

51Ki67 immunoreactivity was assessed on paraffin tissue sections after heat-induced antigen retrieval (at ~90 °C, 10 min) using the Antigen Unmasking Solution, Citric Acid-Based (Vector Laboratories, no. H-3300-250).

52Fluorescent images were obtained with a Leica SP5 confocal microscope (Leica Microsystems) and analysed using LAS AF software (LAS AF software v.2.6.0-7266).

53The primary antibodies used were chicken anti-GFP (1:600 dilution, Aves Labs, GFP-1020); goat anti-ChgA (1:200 dilution, Santa Cruz, sc-1488); rabbit anti-Lyz1 (1:1,000 dilution, DAKO, M0776); rabbit anti-Muc2 (1:1,000 dilution, Santa Cruz, sc-7314); rat anti-BrdU (1:200 dilution, Abcam, ab6326); rabbit anti-5-HT (1:1,000 dilution, Neuromics, RA20080); anti-rabbit Ngn3 (1:100 dilution, a kind gift from H. Edlund); goat anti-villin (1:200 dilution, Santa Cruz, sc-7672); goat anti-ghrelin (1:200 dilution, Santa Cruz, sc-10368); rabbit anti-Ki67 (1:200 dilution, Abcam, ab15580); rabbit anti-E-cadherin (extracellular domain) (1:1,000 dilution, a gift from D. Vestweber); rabbit anti-Fabp1 (1:300 dilution, Abcam, ab222517); and rabbit anti-Apoa4 (1:300 dilution, Abcam, ab231660). The secondary antibodies used were donkey anti-chicken Alexa Fluor 488 (1:800 dilution, Dianova, 703225155); donkey anti-mouse Cy5 (1:800 dilution, Dianova, 715175151); donkey anti-goat Alexa Fluor 555 (1:800 dilution, Invitrogen, A21432); donkey anti-rabbit Alexa Fluor 555 (1:800 dilution, Invitrogen, A31572); and donkey anti-rabbit Alexa Fluor 649 (1:800 dilution, Dianova, 711605152).

54In situ detection of FOXA2, LGR5 and OLFM4 mRNA was performed using the RNAscope Intro Pack for Multiplex Fluorescent Reagent Kit v2-Mm (ACD; no. 323136) according to the manufacturer’s protocol. RNAscope 3-Plex Negative Control Probe (dapB; ACD; no. 320871) and RNAscope 3-Plex Positive Control Probe (ACD; no. 320881) were used as internal controls. Probes used were Mm-Foxa2 (ACD; no. 409111), Mm-Olfm4-C2 (ACD; no. 311831-C2) and Mm-Lgr5-C3 (ACD; no. 312171-C3) with respective fluorescent dyes Opal 520 Reagent (AKOYA; no. FP1487001KT) for channel 1, Opal 570 Reagent (AKOYA; no. FP1488001KT) for channel 2 and Opal 690 Reagent (AKOYA; no. FP1497001KT) for channel 3, diluted 1:800 in RNAscope Multiplex TSA Buffer (ACD; no. 322809). The assay was performed on FFPE mouse intestinal jejunum sections prepared as described above with a thickness of 7 µm and standard pretreatment conditions with protease III (ACD; no. 322340) as recommended by the manufacturer.

55After completion of the RNAscope assay, sections were stained using DAPI (ROTH, 6335.1) to visualize the nuclei, washed with PBS three times and mounted with ProLong Diamond Antifade Mountant (Life Technologies; no. P36970). Samples were visualized using a Leica SP5 confocal microscope (LAS AF software v.2.6.0-7266).

56Statistical bulk transcriptome analyses were performed using the statistical programming environment R implemented in CARMAweb,. Gene-wise testing for differential expression was carried out employing the limma t-test. Sets of regulated genes were defined by raw P < 0.01, FC > 1.3× and average expression in at least one group >32. Enriched canonical pathways were analysed through the use of QIAGEN’s ingenuity pathway analysis (https://www.qiagen.com/ingenuity). Microarray data are available at Gene Expression Omnibus (GEO).

57MATLAB R2014b (v.7.10.0, Mathworks) was used as MALDI spectral pre-processing tool for the subsequent data bioinformatics analysis. Peak picking was performed using an adapted version of the LIMPIC algorithm. In brief, the parameters of peak picking included m/z 0.0005 minimal peak width, signal-to-noise threshold of 4 and intensity threshold of 0.01%. Isotopes were automatically identified and excluded. Statistical comparisons were performed with a Student’s t-test (two-tailed). Metabolites were considered to be significant if they had an intensity FC ≥ 2 and a P value ≤0.05. Metabolite annotation was performed by matching accurate mass with databases (mass accuracy ≤4 ppm, METLIN, http://metlin.scripps.edu/; Human Metabolome Database, http://www.hmdb.ca/; MassTRIX, http://masstrix3.helmholtz-muenchen.de/masstrix3/; METASPACE, http://annotate.metaspace2020.eu/). Heat-map-based clustering and enrichment analysis of metabolic pathways were performed with MetaboAnalyst v.3.0 (http://www.metaboanalyst.ca).

58Lists of genes participating in the candidate pathways from the ingenuity canonical pathways analysis of the microarray data of FVFlow, FVFhigh and FVFneg cells were compiled. For each of the genes on the lists, information on the reaction it is involved in and the participating metabolites were extracted from the mouse specific BiGG databank (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2874806/, https://www.ncbi.nlm.nih.gov/pubmed/20959003). If genes and metabolites within a reaction were also significant in their analyses as described above, the reaction and its pathway were said to be affected at both the metabolomics and the transcriptomics levels.

59To assess the transcriptional profiles of single FVFlow and FVFhigh cells, a nested single-cell qPCR design was used. FVFlow or FVFhigh cells isolated from three CD- and HFHSD-fed FVF mice were sorted as described above directly into single wells of 96-well plates containing 5 μl of a pre-amplification solution composed of 1.2 μl 5× VILO reaction mix (Invitrogen, no. 11754-050), 0.3 μl 20 U μl−1 SUPERase-In (Ambion, no. AM2694), 0.25 μl 10 % NP40 (Thermo Fisher Scientific, no. 28324), 0.25 μl RNA spikes mix (Fluidigm, no. 100-5582) and 3 μl of nuclease-free water (Promega, no. P119C). Cells were lysed by incubation at 65 °C for 90 s and cDNA transcription from RNA was performed by reverse transcription cycling (25 °C for 5 min, 50 °C for 30 min, 55 °C for 25 min, 60 °C for 5 min and 70 °C for 10 min) with 1 μl reverse transcription mix solution containing 0.15 μl 10× SuperScript enzyme mix (Invitrogen, no. 11754-050), 0.12 μl T4 Gene 32 Protein (New England BioLabs, no. M0300S) and 0.73 μl nuclease-free water. The efficiency and specificity of outer and inner primer pairs for target-specific cDNA amplification were tested in advance. Primers showing single peaks and single bands by melt curve analysis and by separation of qPCR products on a 2.5% agarose gel, respectively, were considered specific. Primer efficiency was analysed over a range of tenfold cDNA dilutions (1:1 to 1:100) and primers with 100 ± 15 % efficiency were qualified for further proceedings. Specific target amplification was performed with 9 μl reaction mix containing 7.5 μl TaqMan PreAmp Master Mix (Applied Biosystems, no. 4391128), 0.075 μl 0.5 M EDTA, pH 8.0 (Invitrogen, no. Am9260G), 1.5 μl 10× outer primer mix (500 nM) under the following cycling conditions: enzyme activation step at 95 °C for 10 min, 20 cycles of denaturation for 5 s at 96 °C and 4 min annealing/extension at 60 °C. Amplified cDNA samples were then treated with 6 μl Exonuclease I reaction mix containing 0.6 μl reaction buffer, 1.2 μl Exonuclease I (New England BioLabs, no. M0293S) and 4.2 μl nuclease-free water. To increase target specificity, amplified single-cell cDNA samples were analysed with gene-specific inner primer pairs and SsoFast EvaGreen Supermix with Low ROX (Bio-Rad Laboratories, no. 172-5210) using the 96.96 Dynamic Array on the BioMark System (Fluidigm). BioMark Real-Time PCR Analysis software (Fluidigm) was used to calculate Ct values.

60A detailed description of the computational analyses of single-cell data is provided in the Supplementary information.

61Data collection was performed using Microsoft office Excel 2016–2018 and statistical analysis was performed using GraphPad Prism 6 Software (GraphPad Software). All data are shown as mean ± s.e.m. unless otherwise specified. In box-and-whiskers plots, data are represented as minimum and maximum with centre lines indicating the median. All samples represent biological replicates. For statistical significance testing of two independent groups, an unpaired two-tailed Student’s t-test was used. For statistical comparison of longitudinal data (body weight curves, GTT and IST), two-way analysis of variance corrected by Sidak’s multiple comparison test was used. P values of < 0.05 % were considered statistically significant. For metabolic studies (GTT and IST), sample size was statistically determined; otherwise sample size estimates were not used. Studies were not blinded and investigators were not blinded during analysis.

62Further information on research design is available in the Nature Research Reporting Summary linked to this article.