Upon a nutrient challenge, L cells produce glucagon-like peptide 1 (GLP-1), a powerful stimulant of insulin release. Strategies to augment endogenous GLP-1 production include promoting L-cell differentiation and increasing L-cell number. Here we present a novel in vitro platform to generate functional L cells from three-dimensional cultures of mouse and human intestinal crypts. We show that short-chain fatty acids selectively increase the number of L cells, resulting in an elevation of GLP-1 release. This is accompanied by the upregulation of transcription factors associated with the endocrine lineage of intestinal stem cell development. Thus, our platform allows us to study and modulate the development of L cells in mouse and human crypts as a potential basis for novel therapeutic strategies in patients with type 2 diabetes.
Impairment of insulin secretion is a hallmark of type 2 diabetes. New strategies in the treatment of type 2 diabetes are based on the glucose-lowering effects of the intestinally produced hormone glucagon-like peptide 1 (GLP-1), which augments glucose-dependent insulin release, improves β-cell survival, and promotes satiety (1–3). GLP-1–producing L cells are scattered in the intestinal epithelium among enterocytes and other secretory cells. They also produce GLP-2 and peptide YY. GLP-1 is released in response to ingested nutrients and is rapidly degraded by the enzyme dipeptidyl peptidase 4. Current antihyperglycemic agents include inhibitors of dipeptidyl peptidase 4, which enhance the bioavailability of endogenously secreted GLP-1 and GLP-1 receptor agonists.
Alternatively, increasing the L-cell number to augment GLP-1 secretion can be a useful therapeutic strategy. L cells are generated from stem cells at the base of intestinal crypts. The intestinal stem cells proliferate and give rise to transit-amplifying progenitor cells that subsequently differentiate (4). Enteroendocrine cells and cells from other secretory cell lineages, such as goblet and Paneth cells, originate from a common progenitor cell (5–7). Later in differentiation, endocrine cell progenitors express neurogenin 3 (Ngn3) (8). Insight into the development of L cells, and the determination of factors and downstream signaling pathways that drive L-cell differentiation are hampered by the lack of an in vitro system that allows the study of L cells in their regular cell environment. Therefore, we applied a three-dimensional intestinal crypt culture system that was developed recently at the Hubrecht Institute (9). In this system, intestinal crypts are grown as self-renewing organoids that continuously produce differentiated epithelial cells, including chromogranin (Chg)A–positive cells, similar to intestinal crypts in vivo (4,9,10). So far, it has not been established whether these ChgA-positive cells in organoids are representative of L cells in vivo.
Here we show a continuous generation of mouse and human L cells in vitro, with a possibility to observe L-cell development in real time and selectively modulate the generation of L cells by nutrient stimulation with short-chain fatty acids (SCFAs).
Research Design and Methods
Animal experiments were conducted under the approval of the animal care committee of the Royal Netherlands Academy of Arts and Sciences (permit HI 11.2503). Twelve-week-old C57BL/6 mice were purchased from Charles River (L'Arbresle, France). Mice expressing yellow fluorescent protein (YFP) under transcriptional control of the proglucagon promotor (GLU-Venus mice) (11) were bred on a C57BL/6 background in our animal facility. Mice were fed regular chow ad libitum.
Surgically resected intestinal tissues or endoscopic biopsy samples were obtained from the University Medical Centre Utrecht, which was approved by the ethics committee of the University Medical Centre Utrecht. Informed consent was provided by all subjects.
Mouse and human intestinal crypts were isolated from mouse and human jejunum, as described previously (9,12,13), and were seeded into Matrigel (BD Biosciences), in which they grew into organoids (9,13). Crypts were cultured in advanced Dulbecco’s modified Eagle’s medium/F12 containing 100 units/mL penicillin/streptomycin, 10 mmol/L HEPES, 2 mmol/L Glutamax, supplements N2 (1×) and B27 (1×), and 50 ng/mL murine or human epidermal growth factor (for mouse and human crypt culture, respectively) (all from Life Technologies); and 1 μmol/L N-acetylcysteine (Sigma-Aldrich) and murine Noggin and human R-spondin-1, both as 10% conditioned medium (13). The culture medium for human crypts was additionally supplemented with 50% Wnt-3A conditioned medium (13), 10 nmol/L gastrin (Sigma-Aldrich), 10 mmol/L nicotinamide (Sigma-Aldrich), 500 nmol/L inhibitor of transforming growth factor-β type I receptor ALK5 kinase (A83-01) (Tocris Bioscience), and 10 μmol/L p38 mitogen–activated protein kinase inhibitor (SB202190) (Sigma-Aldrich) (13). The medium was refreshed every 2–3 days. Mouse and human colon crypts were cultured using the same medium composition as human jejunum. Analysis of human and mouse organoids was performed on 1.5- or 2-month-old cultures. For passage, organoids were removed from the Matrigel, mechanically dissociated using a glass Pasteur pipette, pelleted, and replated in fresh Matrigel in 24-well plates. Mouse organoids were passaged every 5 days with a 1:4 splitting ratio. Human organoids were passaged every 10–14 days with a 1:8 split ratio. To test the effect of SCFAs on L-cell differentiation, a combination of acetate, propionate, and butyrate (5, 1, and 1 mmol/L, respectively) was added to the culture medium. These SCFA concentrations were chosen based on earlier in vitro studies (14) and the ratios of these SCFAs in plasma and intestinal lumen (15). For control mouse organoids, regular medium without SCFAs was used. For dose testing in Supplementary Fig. 2F, different concentrations of SCFA combination were used with a constant ratio of 5:1:1 for acetate/butyrate/propionate, respectively.
To improve differentiation of human organoids during SCFA testing, Wnt-3A, nicotinamide, A-83-01, and SB202190 inhibitors were omitted (13). Human and mouse organoids were collected for analysis 48 h after SCFA addition.
Immunostaining and 5-Ethynyl-2′-Deoxyuridine Labeling
For immunostaining, organoids were fixed in 4% paraformaldehyde, permeabilized with 0.3% Triton X-100, and blocked with 3% donkey serum. Organoids were incubated overnight with primary antibodies against GLP-1 (Phoenix Pharmaceuticals), mucin (sc-15334; Santa Cruz Biotechnology), lysozyme (Lyz1) (A0099; Dako), ChgA (sc-1488; Santa Cruz Biotechnology), or ChgC (sc-1491; Santa Cruz) at 4°C. Alexa Fluor 568 donkey anti-goat and Alexa Fluor 488 donkey anti-rabbit (Invitrogen) were used as secondary antibodies. Images were acquired on a confocal laser-scanning microscope (SP5; Leica) using LAS software. The percentage of L cells in organoids was determined based on the number of L cells and DAPI-positive cells in 3-µm optical slices from z-stacks with a distance of 3 µm between the slices. For 5-ethynyl-2′-deoxyuridine (EdU) labeling, mouse organoids were incubated in 10 μmol/L EdU (Click-it; Invitrogen) for 30 min and human organoids 2 h before fixation. The detection was performed according to the manufacturer’s protocol.
Quantitative PCR Analysis
Total RNA was extracted from organoids using Trizol (Invitrogen) and reverse-transcribed with a Fermentas kit. Quantitative real-time PCR was performed on a real-time PCR System (Bio-Rad) using SYBR green assays. We tested GAPDH, HPRT, and beta-2 microglobulin (B2M) as endogenous control gene and found B2M to be most stable during organoid culture and passaging (data not shown). Gene expression of L-cell–specific functional markers (11) was analyzed in sorted L cells from organoids, fresh villi, and crypts. Markers of intestinal cell types were analyzed in whole organoids (Table 1). Transcription factors associated with L-cell development were tested in whole organoids and sorted L cells.
Villus and Crypt L-Cell Isolation and Fluorescence-Activated Cell Sorting
For comparison of primary and organoid-derived L cells, freshly isolated small intestine crypts from GLU-Venus mouse and small intestines of GLU-Venus mouse organoids after five to eight passages were dissociated with 0.05% trypsin-EDTA (Life Technologies) at 37°C into single cells, centrifuged in 4% FBS in PBS at 300g, and immediately sorted by flow cytometry. Villi were collected from intestinal fragments prior to crypt isolation, washed in PBS containing 10% FBS, and dissociated into single cells, as described for crypt cell isolation. YFP-positive cells were separated by flow cytometry as described previously (11).
YFP fluorescence and bright-field images of GLU-Venus mouse organoids in a glass-bottom 24-well plate were made every 3 h during 3 days using a wide-field Leica AF7000 microscope equipped with a thermostatic chamber with humidity control and 5% CO2.
GLP-1 Secretion Assay
Basic medium for static incubations contained Hanks’ balanced salt solution (Life Technologies) supplemented with 10 mmol/L HEPES, 0.1% fatty acid–free BSA, and no glucose, pH 7.4 (NaOH). Organoids from 24-well plates were collected in 1.5-mL Eppendorf tubes (1 well per tube) and incubated in the basic medium for 2 h in a thermomixer at 300 rotations per minute. Then organoids were washed and incubated in 50 µL of basic medium containing 1 mg/mL diprotin A (Sigma-Aldrich) for 1 h. The supernatant was collected, and the organoids were incubated in 50 μL of 10 mmol/L glucose in basic medium with diprotin A for 1 h, followed by collection of the supernatant. Organoids were then lysed in CelLytic M buffer (Sigma-Aldrich). GLP-1 concentrations in the organoid cell lysates and supernatants were determined by Multi-Species GLP-1 total ELISA (Millipore) and Human Total GLP-1 multi-array (Meso Scale Diagnostics). DNA was extracted from the organoid cell lysate (16) and quantified using the PicoGreen kit (Invitrogen). GLP-1 content was normalized to the DNA content of organoids.
All data are expressed as the mean ± SEM. One-way ANOVA with post hoc Tukey tests was used for the comparison of dose testing for the SCFA combination and the expression of genes associated with GLP-1 secretion in sorted L cells. For comparison of basal and stimulated GLP-1 secretion data, the Kolmogorov-Smirnov test was applied. A Student nonpaired t test was used for all other comparisons. Significance level was set at P < 0.05.
L Cells Are Generated in Organoids Derived From Murine and Human Intestinal Crypts
Small intestinal crypts formed organoids that consisted of several outwardly protruding crypt domains and a central villus domain (Fig. 1A and B and Supplementary Fig. 1A and B). In accordance with the cell composition of crypts in vivo, the crypt domain in organoids consisted of a crypt base, containing stem cells and Paneth cells; a zone of dividing multipotent transit-amplifying cells; and differentiating cells, such as goblet cells and enteroendocrine cells (16) (Supplementary Figs. 1A–F and 2B). The villus domain mostly consisted of absorptive and secretory cells (17,18) and showed a spheroid or oval-shaped structure in mouse organoids (Fig. 1A) containing mostly nondividing cells (Supplementary Fig. 1A). Human organoids formed fold-like extensions (Fig. 1B), and the central part of the organoid often had many single dividing cells (Supplementary Fig. 1B). Human organoids showed mucin- and Lyz1-positive cells but no clearly defined Lyz1 granules or mucin droplets, characteristic for Paneth and goblet cells (Supplementary Fig. 1D and F). Mouse organoids had 5.2 ± 0.3 crypt domains per organoid 4 days after passage and human organoids had 11.1 ± 0.4 crypt domains per organoid 10 days after passage. L cells, identified as GLP-1–immunoreactive cells, were mostly found in crypt regions as single cells that were polarized toward the exterior basal side of the organoids (Fig. 1C–E). In mouse organoids, L cells constituted 0.5 ± 0.05% of all organoid cells, which corresponds to 2.1 ± 0.1 cells per average-sized organoid (ranging from 0 to 15 cells, depending on organoid size). The organoids generated L cells at a relatively constant rate for 48–96 h after splitting (Supplementary Fig. 2F). In human organoids, GLP-1–positive cells constituted 0.1 ± 0.001% of all organoid cells, or 1.5 ± 0.03 cells per organoid. After a 30-min EdU pulse, no double-positive cells for GLP-1 and EdU were found in a total of 300 human or mouse organoids (data not shown), indicating that mature L cells do not divide. Free fatty acid receptor 2 (FFAR2) appeared exclusively on GLP-1–positive cells (Supplementary Fig. 2A). GLP-1–immunoreactive cells also expressed vesicle markers ChgA and ChgC (Supplementary Fig. 2B and C, respectively). These characteristics were still present in L cells from 6-month-old cultures of mouse organoids (data not shown).
GLU-Venus mice express YFP in intestinal L cells under the control of the proglucagon promoter (11). Intestinal organoid cultures from these mice provided a unique opportunity to monitor real-time L-cell development (Fig. 2A). The majority of L cells appeared near the crypt base and moved away from the crypt base concomitantly with a proliferative expansion of the transit-amplifying compartment. The fluorescence disappeared 3–4 days (Fig. 2A) after first becoming detectable, indicating the loss of glucagon promoter activity or L-cell death. The disappearance of the fluorescent signal is unlikely to reflect photobleaching, as the cells were only excited for 70 ms every 3 h. As >90% of labeled cells were double-positive when Venus-expressing organoids were immunostained for GLP-1 (data not shown), the fluorescent time course is likely to reflect the turnover of GLP-1–expressing cells.
To test whether in vitro–generated L cells are functionally mature, we used GLU-Venus mice to compare fluorescence-activated cell (FAC)-sorted primary L cells from the small intestine and L cells from organoids after 6 passages. Estimated by FAC sorting, the percentage of L cells in the organoids was similar to that observed in fresh small intestine crypts (Supplementary Fig. 2H) and was in line with our calculations based on microscopy. We compared gene expression of specific functional markers in L cells isolated from organoids and from freshly prepared villi and crypts (Fig. 2B). Proglucagon gene expression was higher in L cells from villi compared with L cells from crypts and organoids (Fig. 2B). We found that Sglt1, Glut5, Gck, Ffar2, Kcnq1, Scn3a, Kir6.2, and Cacna1a expression was maintained in organoid-derived L cells, indicating that L cells produced in our cultures retain the molecular profile of their counterparts in vivo and express stimulus-secretion–coupling components for GLP-1 release.
We further analyzed the expression of transcription factors associated with L-cell development (8,19–23) in L cells from organoids and from freshly prepared crypts (Supplementary Fig. 3A–D). Organoid L cells had a similar expression pattern to that of crypt-derived L cells and showed reduced expression of Ngn3 compared with whole organoids, consistent with its role as early endocrine marker (8).
SCFAs Increase the Number of L Cells and GLP-1 Secretion in Mouse and Human Small Intestine Organoids
SCFAs are known to stimulate GLP-1 secretion, and we tested their effect on L-cell development. The addition of SCFAs as a combination of 5 mmol/L acetate, 1 mmol/L propionate, and 1 mmol/L butyrate almost doubled the number of L cells in mouse (Fig. 3A and B) and human organoids (Fig. 3C) within 48 h. We tested several concentrations of the combination of the three SCFAs (Supplementary Fig. 2F) and found that lower concentrations of SCFAs did not increase L-cell number. Concentrations of SCFAs higher than those used above did not result in a further increase in L-cell number (not significant, by one-way ANOVA). Propionate and butyrate (but not acetate) in concentrations >1 mmol/L had a toxic effect, resulting in changes in organoid morphology and reduction of growth (data not shown). Increasing the acetate levels alone did not have an additional effect on L-cell numbers (data not shown).
The increase in L-cell number was accompanied by a higher GLP-1 content in mouse and human organoids (Fig. 3D and E). To test whether SCFA treatment generates nutrient-responsive L cells, we measured glucose-induced GLP-1 secretion. In SCFA-treated mouse organoids, both basal and stimulated GLP-1 secretion were higher than in the control (Fig. 3F). Both control and SCFA-treated mouse organoids showed a twofold increase in GLP-1 release during stimulation with 10 mmol/L glucose (Fig. 3F), indicating that this effect was due to the increased number of L cells but not to the ability to increase GLP-1 release during the nutrient challenge. Human organoids showed very low basal secretion (Fig. 3G), consistent with fewer L cells. There was a twofold increase in GLP-1 secretion from control organoids upon glucose stimulation (Fig. 3G). After the SCFA treatment, both basal and stimulated GLP-1 secretion from human organoids were increased (Fig. 3G).
To test the effect of SCFA on differentiation of the other cell lineages, we performed marker gene expression analysis (Fig. 4A and C and Table 1). In mouse and human organoids, SCFA treatment had no effect on expression of Lgr5 and CD133 (markers for stem cells and early progenitors respectively) (18), ITF (goblet cells), Lyz1 (Paneth cells), or I-FABP (enterocytes), but the pan-endocrine cell marker ChgA was increased during SCFA treatment. In accordance with our data on GLP-1 content, Gcg expression was elevated in mouse and human organoids compared with the control. Gene expression of the enterochromaffin cell marker Tph or the K-cell marker Gip did not change in mouse organoids after SCFA treatment, indicating a specific induction of L-cell differentiation (Fig. 4A). In human organoids, the expression of SCT and TPH was increased (Fig. 4C). Next, to evaluate the potential of organoids to generate L cells, we analyzed gene expression of the transcription factors Ngn3 (19,20), Neurod1 (21), Foxa1/2 (22), and Arx (23) associated with early and late L-cell development in organoids. In mouse and human organoids, gene expression of Neurod1 and Foxa1/2 was upregulated, but the expression of Ngn3 or Arx was not (Fig. 4B and D). Some transcription factors that regulate cell development are also expressed after cell maturation and may be involved in the maintenance of the mature state of the cell. Therefore, we also compared the expression of transcription factors in the pure L-cell fraction from control and SCFA-treated mouse organoids and found no differences (Fig. 4E).
The main findings of our study are that L cells can be continuously generated from murine and human intestine in a three-dimensional culture system that allows real-time studies of L-cell development. Using organoid culture of GLU-Venus mouse crypts, we were able to observe the “birth” of L cells within a crypt environment in vitro and study gene expression from in vitro–derived L cells. The number of L cells can be selectively modulated by nutrient factors (SCFAs). This illustrates the use of this novel platform to investigate factors that modulate the L-cell number and function in order to increase endogenous GLP-1 production for the treatment of patients with type 2 diabetes.
The number of L cells in organoids was close to that observed in freshly isolated mouse small intestinal crypts and villi. The majority of L cells were in crypt domains of the organoids, similar to reports in freshly isolated tissue, where L cells are also found mainly in crypts and lower villi (24). It is assumed that in vivo the cells from the crypt region are being “pushed” toward the villus by dividing transit-amplifying cells and then shed from the surface of the villus. In the organoids, the “villus” space is smaller and is “shared” among several crypts (9), so the migration patterns of cells in organoids may not mimic the situation in intact intestine. While we did not observe major expression differences for most L-cell markers in cells derived from murine organoids or fresh tissue, proglucagon was expressed at significantly higher levels in villus L cells. A possible explanation is an increase of proglucagon expression as L cells mature during the migration from the crypts to the villus top. Other factors predominantly present in the crypt environment, such as Wnt and Notch signaling, may suppress cell differentiation. This might be of special importance in the human organoid cultures, where differences in culture medium may interfere with the establishment of a distinct crypt-villus axis and may also directly affect enteroendocrine cell maturation consistent with the observed lower L-cell density.
L cells from our mouse and human organoid cultures coexpress neuroendocrine granule markers ChgA and ChgC and the FFAR2, and are able to secrete GLP-1 in response to nutrient stimulation. The increase in GLP-1 secretion during glucose stimulation was comparable to that reported in previous studies on mouse primary L cells (11). Similar expression of genes associated with stimulus secretion coupling for GLP-1 release in primary L cells and L cells from organoids further indicates that the organoid system is an excellent model for the generation of functioning L cells. It has some advantages over the established systems because L cells in organoids are retained in a polarized environment and can be monitored over their entire life span. In addition, organoids can be established from human tissue. Organoids thus have substantial potential for translational studies and drug testing, particularly for drugs targeting L-cell differentiation.
Earlier in vivo studies on animals reported that dietary fiber increases GLP-1 levels (25–28) and colonic L-cell numbers (29,30). This effect was attributed to the production of SCFAs by intestinal microflora during the fermentation of fiber (30). SCFAs have been shown to elevate the proglucagon gene in rat intestine (28,31). In our in vitro system, SCFAs enhanced L-cell differentiation in mouse and human small intestinal crypts. Increased proglucagon gene expression, elevated GLP-1 content, and glucose-stimulated increase in GLP-1 secretion in mouse and human organoids indicate that L cells generated during SCFA treatment are functional. Importantly, SCFAs did not change the organoid growth pattern or expression of stem cell, goblet cell, and Paneth cell markers. The effect of SCFAs was selective to L cells in mouse organoids because pan-endocrine marker ChgA gene levels increased, but markers of other endocrine cell types (Sct, Tph, and Gip) did not change. Human SCFA-treated organoids showed an upregulation of TPH and SCT, indicating an additional effect on enterochromaffin cells and S cells, respectively. While some overlap in secretin and GLP-1 expression has been observed in mice (32,33), enterochromaffin cells are thought to derive through a lineage distinctive from L cells, and we did not observe any colocalization of serotonin and GLP-1 in human organoids (data not shown). It is thus possible that, at least in human organoids, SCFAs increase maturation to different enteroendocrine cell types. It should be remembered that L cells coexpress a number of different hormones, including cholecystokinin, GIP, and peptide YY, with GIP being more prevalent in the proximal small intestine and peptide YY increasing toward the distal intestine (32,34). Therefore, although an increase in L-cell secretion may be beneficial for glucose control in type 2 diabetes, the concomitant stimulation of a range of gut peptides could lead to additive anorexia but also to unwanted side effects, for example, in overstimulation of the exocrine pancreas and the gallbladder. Dissection of the pathway by which SCFAs enhance L-cell development could provide valuable knowledge on how to enhance naturally regulated GLP-1 production from L cells in duodenum, jejunum, ileum, and colon. After EdU staining, we found no dividing L cells, regardless of whether organoids were treated with SCFAs (data not shown). This indicates that existing L cells do not contribute to increased L-cell numbers and that SCFAs likely affect endocrine progenitors in the organoids. Ngn3 defines the endocrine commitment of a secretory cell progenitor (8,19,20), and, consistent with that, we found that the expression of Ngn3 is lower in mature L cells than in whole organoids that also contain a population of early endocrine-committed cells. SCFAs did not change the gene expression of Ngn3 in whole organoids, but increased expression of transcription factors Neurod1 and Foxa1/2, which are downstream of Ngn3 and have been specifically associated with L-cell development (21,22). Based on this, we suggest that SCFAs act on late enteroendocrine precursors of L cells, in which the expression of Ngn3 is already fading (20). SCFA treatment did not have additional effects on the expression of transcription factors in sorted L cells. FFAR2 mediates the stimulatory effect of SCFAs on GLP-1 secretion (14); however, it remains to be established whether it is involved in L-cell differentiation. Because SCFAs are GLP-1 secretagogues, it is possible that persistent activation of existing L cells may stimulate the development of L cells in surrounding areas.
In conclusion, the self-renewing three-dimensional intestinal crypt culture system allows the production of functional L cells, and can be used to study the development of L cells and their changes in (patho)physiological conditions that can be modeled in vitro. It is a promising platform for compound screening and studies on L-cell function. We demonstrate that L-cell differentiation can be selectively increased by SCFAs. Further identification of the signaling mechanisms induced by SCFAs may be a useful tool in our pursuit for better treatments for type 2 diabetes and obesity.
Acknowledgments. The authors thank the following colleagues from the Hubrecht Institute: Stefan van der Elst for performing fluorescence-activated cell sorting; Jori Tip-Leenders for technical assistance; Benaissa El Haddouti for animal care; Marc de Wetering for a sample of human colon organoids; Anko de Graaff and the Hubrecht Imaging Center for supporting the imaging; and Harry Begthel for assistance with immunostaining. The authors also thank Chris van der Bent of the Department of Endocrinology, Leiden University Medical Center, for performing the Mesoscale GLP-1 assay.
Funding. This study was partly funded by the Bontius Foundation.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. N.P. designed the study, researched the data, and wrote the manuscript. F.R. contributed to the study design and discussion. S.B. contributed a method for human organoid maintenance. H.F.F., R.G.J.V., H.C., and F.M.G. contributed to the discussion. F.C.R. and S.v.d.B. contributed to cell culture optimization. E.J.P.d.K. designed the study and wrote the manuscript. All authors reviewed/edited the manuscript. N.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.