Background and Aims. Embryonic stem (ES) cells have a pluripotent ability to differentiate into a variety of cell lineages in vitro. We have recently found that ES cells can give rise to a functional gut-like unit, which forms a three-dimensional dome-like structure with lumen and exhibits mechanical activity, such as spontaneous contraction and peristalsis. The aim of the present study was to investigate the electrophysiological and morphological properties of ES cell-derived contracting clusters.
Methods. Electrical activity was examined by an extracellular recording. Morphology and cellular components were investigated by immunohistochemistry and electron microscopy.
Results. Clusters with rhythmic contractions displayed electrical slow waves at a regular rhythm, and clusters with highly coordinated peristalsis showed regular slow waves and spontaneous spike action potentials. Immunoreactivity for c-Kit, a marker of interstitial cells of Cajal (ICC), was observed in dense network structures. Neuronal marker PGP9.5 immunoreactivity was observed only in clusters with peristalsis. The topographical structure of the wall was organized by an inner epithelial layer and outer smooth muscle layer. The smooth muscle layer was provided with an ICC network and innervated with enteric neurons.
Conclusions. ES cells can differentiate into a functional gut-like organ in vitro that exhibits physiological and morphological properties characteristic of the gastrointestinal (GI) tract. This ES cell-derived gut provides a powerful tool for studying GI motility and gut development in vitro, and has potential for elucidating and treating a variety of motility disorders.
The gastrointestinal (GI) tract manifests mechanical activity such as spontaneous contraction and peristalsis, and is morphologically organized by the enteric components of all three embryonic germ layers: epithelial cells (endoderm), smooth muscle and interstitial cells of Cajal (ICC) (mesoderm), and enteric neurons (ectoderm). Recent investigations have indicated that the ICC network in the musculature of the GI tract greatly contributes to the generation of electrical pacemaker activity for GI motility [1-, 5]. This pacemaker activity manifests itself as rhythmic slow waves in membrane potential, and controls the frequency and propagation characteristics of GI motility. Enteric neurons also innervate smooth muscle and are essential for peristalsis in GI motility [6-, 9]. Recently, a large number of studies have reported that loss or defects in ICC and/or enteric neurons could be related to pathophysiology in a variety of motility disorders [8-, 12]. Thus, there is great interest in understanding the mechanisms that govern the differentiation of ICC and enteric neurons, not only to understand GI motility but also to elucidate motility disorders. However, previous in vitro analyses of gut development have relied heavily on primary cultures of cells or dissociated tissues from GI sources, because of the lack of an appropriate in vitro model to reproduce the gut organization process.
Recently, pluripotent embryonic stem (ES) cells have been used to generate particular types of cell lineages in vitro. ES cells are clonal cell lines derived from the inner cell mass of developing blastocysts [13-, 14]. The distinguishing feature of these cells is their capacity to differentiate into a broad spectrum of derivatives of all three embryonic germ layers [15,, 16]. More recently, human ES cells were isolated and found to have a pluripotent ability to develop into a wide range of cell types [17-, 21]. This ability has drawn attention to ES cells as a novel source of cell populations for new therapeutic strategies such as cell transplantation and tissue regeneration. When ES cells are allowed to differentiate in a suspension culture, they form spherical multicellular aggregates, termed embryoid bodies (EBs), that have been shown to contain a variety of cell populations [15,, 16]. To date, some success has been achieved in inducing mouse ES cells to differentiate into particular types of cells, such as hematopoietic cells [22,, 23], cardiomyocytes , smooth muscle cells [25,, 26], neurons [27-, 30], within an EB environment. Although developmental programs associated with cell lineage commitment in EBs show remarkable similarities to those found in normal embryos [15,, 16], the in vitro differentiation of ES cells into particular types of organs has not yet been shown except for a recent report about pancreatic islet-like organization from ES cells .
Using an EB culture system, we have recently found that mouse ES cells can give rise to a functional gut-like unit, which forms a three-dimensional dome-like structure with lumen and exhibits mechanical activity, such as spontaneous contraction and highly coordinated peristalsis. From the view of the involvement of ICC and enteric neurons in GI motility, we have speculated that these contracting cell clusters are organized by a variety of enteric components including smooth muscle cells, ICC, and neurons. In the present study, we investigated whether ES cell-derived contracting clusters could possess electrophysiological and morphological properties characteristic of the GI tract. Our results provide a useful model for studying GI motility and gut development in vitro.
Materials and Methods
ES Cell Culture
Undifferentiated ES cells (EB3) were maintained on gelatin-coated dishes without feeder cells in Dulbecco's modified Eagle's medium (Sigma; St. Louis, MO; http://www.sigma-aldrich.com) supplemented with 10% fetal bovine serum (GIBCO/BRL; Grand Island, NY; http://www.tmc.tulane.edu/sif/tulgib.htm), 0.1 mM 2-mercaptoethanol (Sigma), 0.1 mM non-essential amino acids (GIBCO/BRL), 1 mM sodium pyruvate (Sigma), and 1,000 U/ml of leukemia inhibitory factor ([LIF] GIBCO/BRL). The EB3 cells (a kind gift from Dr. Hitoshi Niwa, Osaka University) carried the blasticidin S-resistant selection marker gene driven by the Oct-3/4 promoter (active under undifferentiated status) and were maintained in medium containing 10 μg/ml blasticidin S to eliminate differentiated cells . The EB3 cells are a subline derived from E14tg2a ES cells  and were generated by a targeted integration of Oct-3/4-IRES-BSD-pA vector into the Oct-3/4 allele . To induce EB formation, dissociated ES cells were cultured in hanging drops [15,, 16] with minor modifications. The cell density of one drop was 500 cells per 15 μl of ES cell medium in the absence of LIF. After 6 days in a hanging drop culture, the resulting EBs were plated onto plastic 100-mm gelatin-coated dishes and allowed to attach for the outgrowth culture.
The electrical activity of the contracting clusters was examined by an extracellular recording technique. Electrical activity was recorded using a glass electrode filled with culture solution, into which a platinum wire (200 μm) was inserted, along with a Dual-Channel Bioelectric Amplifier (MEG-2100) under filtration at LO CUT (0.08 Hz) and HI CUT (10 K). The inside diameter of the electrode was 0.5 mm and its length was 2 cm. This electrode was gently placed on the surface of the clusters, so as not to cause injury, under an inverted microscope.
Tissues were fixed with Zamboni's solution at room temperature for 4 hours and incubated at 4°C overnight with antibodies against α-smooth muscle actin (1:150; Sigma), smooth muscle myosin (1:100; Biomedical Technologies Inc.; Stoughton, MA; http://www.btiinc.com), and PGP9.5 (1:5,000; UltraClone Limited; Isle of Wight, UK). For c-Kit immunohistochemistry, tissues were incubated with Alexa 594 (Molecular Probes; Eugene, OR; http://www.probes.com) conjugated ACK2 (1:100) for 10 minutes and fixed with a 4% paraformaldehyde solution for 30 minutes, as previously described . Localization of fluorescence was examined with a confocal laser scanning microscope (Bio-Rad MRC-1024; Bio-Rad; Hercules, CA; http://www.bio-rad.com).
Tissues were fixed with 2.5% glutaraldehyde, 1.25 mM CaCl2, and 3% sucrose in a 0.05 M cacodylate buffer (pH 7.4) at room temperature for 3-4 hours. They were post-fixed with 1% OsO4, and then stained en bloc with uranyl acetate and embedded in epoxy resin. Ultrathin sections were double stained with uranyl acetate and lead citrate, before being examined with an electron microscope (Hitachi H-7100; Hitachi; Brisbane, CA; http:/www.hitachi.com).
ES Cell Differentiation and Mechanical Activity
ES cells were cultured for 6 days in a hanging drop culture system and allowed to form aggregates, known as EBs. The resulting EBs were attached to gelatin-coated dishes, after which various types of cells, including cardiac beating muscle cells, emerged from the outgrowths. After 5 to 7 days in culture, multiple clusters within each developing EB, different from the cardiac beating muscle cells, began to contract spontaneously with an irregular rhythm. Each contracting cluster underwent a dramatic transformation into a hemispherical dome-like structure, with a cavity that contained fluid and solids. On approximately day 14, these hemispherical clusters proliferated to form more prominent three-dimensional structures with lumens and began rhythmic contractions. On approximately day 21, the clusters showed distinct and highly coordinated contraction patterns with regular rhythms. This mechanical activity was composed of periodic contraction and relaxation, and was accompanied by a transportation of contents (Fig. 1). It was very similar to GI motility, i.e., peristalsis.
We investigated the electrical activity of the contracting clusters at various differentiation stages using an extracellular recording technique. Clusters with rhythmic contractions present on day 14 displayed spontaneous slow depolarizing waves at a regular rhythm (Fig. 2A). The slow waves averaged 23.8 ± 1.5 μV in amplitude and 13.1 ± 1.0 cycles per minute in frequency at 35°C (n = 6 clusters). The frequency was strongly dependent on the temperature and decreased to 9.4 ± 1.3 cycles per minute at 20°C (n = 6 clusters) (data not shown). A cholinergic agent, carbamylcholine (CCh) (1 to 10 μM) increased the frequency of the slow waves (22.4 ± 1.5 cycles per minute, n = 6 clusters), though an L-type calcium channel blocker, nifedipine (1 to 10 μM), did not affect the frequency of the slow waves. The slow-wave component or pacemaker activity of ICC has been proven to be insensitive to L-type calcium channel blockers . Thus, these results suggest that those clusters with rhythmic contractions consisted not only of smooth muscle cells but also ICC networks, which generate electrical pacemaker activity for the musculature. As for the clusters demonstrating highly coordinated peristalsis-like contractions on day 21, we found spontaneous spike action potentials as well as regular slow waves, both of which were synchronized with each contraction (Fig. 2B). A neuron blocker, tetrodotoxin (TTX) (3.1 to 31 μM), did not have any influence on the slow-wave component; however, it abolished the spike action potentials, suggesting that these spike action potentials were elicited by neurons. Subsequent to the application of TTX, methylene blue (MB) (100 μM) with light, previously shown to injure ICC , completely abolished the slow-wave component and worsened the regularity of the remaining electrical activity. These data provided physiological evidence that the clusters with peristalsis-like contractions contained not only enteric smooth muscle cells, but also ICC and enteric neurons. We also investigated the electrical activity of some of the contracting clusters using a conventional intracellular recording technique and found that the resting membrane potential (RMP) averaged -55.7 ± 10.2 mV (n = 5 clusters) (data not shown), which was consistent with the RMP of smooth muscle cells found in the mouse small intestine .
Immunohistochemistry was used to identify the cellular and protein components of the contracting clusters. From the early stages, immunoreactivity for both smooth muscle-specific actin and myosin was observed in the wall surrounding the lumen and spreading out from the clusters (Figs. 3A and 3B). Immunoreactivity for c-Kit, a useful marker for ICC [1,, 2,, 35,, 36], was detected in clusters that had spontaneous contractions at a regular rhythm and generated electrical slow waves. The c-Kit immunopositive (c-Kit+) cells were mostly multipolar, and formed a distinct and dense network (Fig. 3C). This network structure of c-Kit+ cells was similar to that of ICC at the level of the myenteric plexus in a murine embryo or neonate, as described previously . Neuronal marker PGP9.5 immunoreactivity was observed in clusters that showed peristalsis-like contractions. PGP9.5 immunopositive (PGP9.5+) cells were also distributed outside the clusters, where they formed ganglion-like structures (Fig. 3D). PGP9.5+ fibers from ganglions located close to the clusters were often seen projecting into the contracting clusters. PGP9.5+ cells in the cluster were distributed in a solitary manner and seldom formed a ganglion (Fig. 3E). Most of them were unipolar, with a protruding single axon-like process with varicosities, though a few were multipolar with prominent dendrites (Fig. 3F).
Ultrastructural Characteristics of the Contracting Clusters
Electronmicroscopic analysis confirmed the cellular components of the cluster. The wall was constituted of three distinct layers, as shown in Figure 4A, in a semi-thin section stained with toluidine blue. The innermost was a flat and single layer of epithelium (EP), which did not form any plica, intestinal villus, or crypt. Most of the epithelial cells were columnar type, with fewer and shorter microvilli than those in epithelia in the mouse GI tract. Well-developed goblet cells were common, and tuft cells were also scattered among the columnar cells (Figs. 4B and 4C). We considered it remarkable that the entero-endocrine cells that contained numerous secretory granules were well differentiated (Fig. 4D). All of these epithelial cells were separated from the connective tissue (CT) by a basal lamina. The outermost layer was composed of prominent smooth muscle cells (ML) (Fig. 4A), filled with filament-like structures (i.e., thick, thin, and intermediate filaments) in their cytoplasm, and had membrane-free or -bound dense bodies (Fig. 4E). These smooth muscle cells ran parallel to one another, and formed fascicles or sheets corresponding to circular or longitudinal muscle layers, respectively, though the muscularis mucosa could not be identified. We confirmed the occurrence of ICC in the muscle layer by ultrastructural features such as electron-dense cytoplasm, numerous mitochondria, and caveolae on the cell membrane (Fig. 4F). ICC were connected to each other by their long cell processes surrounding smooth muscles, suggesting a multipolar cell type. Nerve fibers were also seen in the muscle layer, though such enteric ganglia as Auerbach's or Meissner's plexuses were seldom observed in the cluster (data not shown). The third layer, between the epithelium and muscle layer, was a thin CT corresponding to the lamina propria and/or submucosa (Fig. 4A). There were many fibroblasts and collagen fibers present; however, the absence of both blood and lymphatic vessels was striking. The outer surface of the cluster was lined with a flat and thin monolayer, similar to the serosa.
Here we demonstrated that mouse ES cells can give rise to a functional gut-like organ in vitro, which exhibits mechanical activity such as spontaneous contraction and peristalsis. The ES cell-derived gut-like organ was composed of the enteric derivatives of all three embryonic germ layers: epithelial cells (endoderm), smooth muscle cells and ICC (mesoderm), and enteric neurons (ectoderm). The topographical structure of the wall was well-organized by an inner epithelial layer and outer smooth muscle layer. The smooth muscle layer was provided with an ICC network and innervated with enteric neurons.
Notably, the in vitro organized gut produced electrical activity characteristic of the GI tract at various EB differentiation stages. Clusters with rhythmic contractions present on day 14 of EB culture displayed electrical slow waves at a regular rhythm. On day 21, the contracting clusters demonstrated a highly coordinated peristalsis, and showed spontaneous spike action potentials as well as regular slow waves. The slow wave frequency of the contracting clusters increased in response to a CCh; however, it was not affected by an L-type calcium channel blocker (nifedipine). These findings suggest that the clusters consisted not only of smooth muscle cells but also of ICC, since the pacemaker activity of ICC is known to be insensitive to L-type calcium channel blockers . Furthermore, MB with light, previously shown to injure ICC , completely abolished the slow-wave component and worsened the regularity of the remaining electrical activity, providing further physiological evidence that clusters with rhythmic contractions consisted of ICC networks that generate electrical pacemaker activity for the musculature. As for the clusters with highly coordinated peristalsis, a neuron blocker (TTX) abolished the spike action potentials, suggesting that the clusters were innervated with enteric neurons. These electrophysiological findings suggest that the mechanical activity of the in vitro organized gut is regulated by ICC and enteric neurons.
To confirm the presence of ICC and neurons within the in vitro organized gut, immunohistochemical studies were performed. As several classes of mouse ICC are known to express the proto-oncogene c-kit, the antibodies against the gene product (c-Kit) are used for their identification [1,, 2,, 3,, 35,, 36]. Immunoreactivity for c-Kit was first detected within the wall of clusters that spontaneously contracted with an irregular rhythm on days 7 to 10 of EB culture, though the network of c-Kit+ cells was not completely constituted (data not shown). In contrast, contracting clusters with electrical slow waves at a regular rhythm on day 14 showed a distinct and dense network of c-Kit+ cells. This network structure was similar to that of ICC at the level of the myenteric plexus in a murine embryo or neonate, as described previously . Our results indicate that ES cells can reproduce the ICC network differentiation process in vitro. Furthermore, they are consistent with our previous observations of murine embryos, in which the development of the ICC network was correlated with the initial onset of electrical rhythmicity .
Immunoreactivity for PGP9.5, a specific antibody that recognizes neurons, was observed only in clusters that showed highly coordinated peristalsis, suggesting that innervation to the clusters governs the peristalsis-like contractions. Interestingly, PGP9.5+ cells were also distributed outside the clusters, where they formed ganglion-like structures. PGP9.5+ fibers from ganglions located close to the clusters were often seen projecting into the contracting clusters. To clarify the origin of neurons detected inside the clusters, we further investigated the immunoreactivity for p75, which has been reported to label migratory-stage enteric neuron precursors . Some cells inside the cluster as well as ganglion cells outside expressed p75 immunoreactivity at the earlier stage of EB development (on day 7) (data not shown). These findings suggest that the neurons located inside the cluster originated from enteric neuron precursors outside. This process might be similar to that by which neural crest cells migrate into the GI tract and differentiate into enteric neurons during mouse embryogenesis [8,, 9,, 37]. Further investigations are needed to explain how the enteric neurons can be differentiated within developing EBs.
Our ultrastructural observations provided more convincing evidence that the contracting clusters were organized by cell populations similar to those in the GI tract, such as epithelial cells, smooth muscle cells, ICC, and enteric neurons. The innermost portion of the wall was a flat and single layer of columnar epithelial cells, among which goblet and tuft cells were scattered. The outermost layer was composed of smooth muscle cells, which had filament-like structures in their cytoplasm. Cells harboring several characteristics of ICC, such as an electron-dense cytoplasm, numerous mitochondria, and caveolae formation on the cell membrane, were located in the muscle layer. Although the in vitro organized gut failed to form tubular structures and lacked some components, such as lymphatic vessels and vascular formation, the essential mode of topographical organization across the transmural direction was similar to that found in gut development during embryogenesis in vertebrates . The in vitro morphogenesis of an ES cell-derived gut-like organ may conform to a fundamental principle of developmental programs, such as reciprocal epithelium-mesenchymal interaction, that has also been observed in embryonal organogenesis of the GI tract [38,, 39]. Previous studies have reported that ES cells differentiate into sheet-like enteric smooth muscles within the adhesive outgrowth of EBs that are cultured in suspension for 2 to 4 days [16,, 25,, 26]. In contrast, using EBs cultured in suspension for 6 days, we have demonstrated that ES cells can give rise to a three-dimensional gut-like organ, which consists not only of smooth muscle cells but also of ICC, enteric neurons, and epithelial cells. These results suggest that the amount of time for EB formation in a suspension culture may play an important role in the capacity and pattern of ES cell differentiation. When ES cells are cultured in suspension for 5 to 6 days, they form EBs with egg-cylinder-like structures that consist of two cell layers—an outer primitive endodermal layer and inner primitive ectodermal layer [15,, 16,, 40,, 41]. In a developing embryo, primitive endoderm cells form the visceral yolk sac endoderm, whereas the primitive ectoderm cells are capable of forming all the fetal tissues, including the embryonic endoderm, mesoderm, and ectoderm . Thus, EBs at this stage may have the potential to regulate developmental programs associated with cell lineage commitment, and provide an appropriate microenvironment to differentiate ES cells into enteric derivatives of all three embryonic germ layers and reproduce the gut organization process in vitro.
Extensive studies of the ordered sequence of events leading to a particular cell lineage commitment have been performed using the EB culture system [16,, 22-, 30]. Contrary to the well-characterized cellular differentiation from ES cells, little is known about the in vitro differentiation into particular types of organ. Although difficulties in producing an organ in culture are easily predictable, our success in inducing the in vitro differentiation of gut from ES cells may encourage other challenges for ES cell-derived organs, such as the heart, liver, and pancreas, and could also facilitate an advance in organ engineering. Notably, we showed for the first time that ES cells can differentiate into ICC within an in vitro organized gut. Although ICC are now considered to be mesoderm-derived [3,, 4], their differentiation process is not fully understood. The in vitro organized gut containing an ICC network provides a useful tool for studying the relationship between ICC differentiation and GI motility. Furthermore, this model may provide an important insight in understanding the pathophysiological mechanisms of congenital motility disorders associated with the developmental abnormalities in ICC and enteric neurons. In addition, this system could provide not only a potentially limitless source for enteric components including ICC, enteric neurons, and smooth muscle cells, but also an infinite supply of well-organized gut with motor function in vitro. In this context, an ES cell-derived gut may contribute to new therapeutic strategies such as cell or tissue transplantation for a variety of motility disorders. To explore further possibilities for therapeutic application, it will be worthwhile to develop an even more efficient way to enrich for human ES cell-derived gut from developing EBs.
In conclusion, we demonstrated that ES cells can differentiate into a functional gut-like organ in vitro. The in vitro organized gut was morphologically well-organized by an inner epithelial layer and outer smooth muscle layer containing an ICC network and enteric neurons, and possessed mechanical and electrical activities characteristic of the motor function of the GI tract. This model system is reproducible and relatively easy to handle using an EB culture system. The ES cell-derived gut provides a powerful tool for studying GI motility and gut development in vitro, and has great potential for elucidating and treating a variety of motility disorders.
We thank Dr. Hitoshi Niwa for his critical review of the manuscript and Dr. Hiroshige Nakano for his helpful discussion.
This work was supported by Research Grants-in-Aid from the Ministry of Education, Science and Culture of Japan.