Mature enteroendocrine cells contribute to basal and pathological stem cell dynamics in the small intestine

Abstract

Lgr5-expressing intestinal stem cells (ISCs) maintain continuous and rapid generation of the intestinal epithelium. Here, we present evidence that dedifferentiation of committed enteroendocrine cells (EECs) contributes to maintenance of the epithelium under both basal conditions and in response to injury. Lineage-tracing studies identified a subset of EECs that reside at +4 position for more than 2 wk, most of which were BrdU-label-retaining cells. Under basal conditions, cells derived from these EECs grow from the bottom of the crypt to generate intestinal epithelium according to neutral drift kinetics that is consistent with dedifferentiation of mature EECs to ISCs. The lineage tracing of EECs demonstrated reserve stem cell properties in response to radiation-induced injury with the generation of reparative EEC-derived epithelial patches. Finally, the enterochromaffin (EC) cell was the predominant EEC type participating in these stem cell dynamics. These results provide novel insights into the +4 reserve ISC hypothesis, stem cell dynamics of the intestinal epithelium, and in the development of EC-derived small intestinal tumors.

NEW & NOTEWORTHY The current manuscript demonstrating that a subset of mature enteroendocrine cells (EECs), predominantly enterochromaffin cells, dedifferentiates to fully functional intestinal stem cells (ISCs) is novel, timely, and important. These cells dedifferentiate to ISCs not only in response to injury but also under basal homeostatic conditions. These novel findings provide a mechanism in which a specified cell can dedifferentiate and contribute to normal tissue plasticity as well as the development of EEC-derived intestinal tumors under pathologic conditions.

INTRODUCTION

The intestinal epithelium renews continuously every 3–5 days. This rapid turnover is sustained by actively cycling leucine-rich repeat-containing G protein-coupled receptor 5 positive (Lgr5⁺) intestinal stem cells (ISCs) that reside at the base of crypt (4). However, the current model suggests when homeostatic renewal is perturbed by loss or stress of Lgr5⁺ ISCs upon inflammation or injury, restoration of the epithelium is mediated by a reserve ISC population (51, 57). The first recognized reserve ISC population is the quiescent stem cell that resides above Paneth cells at +4 position (34). This cell population can be identified by its DNA label-retaining property and thus called the label-retaining cell (LRC). A growing list of the putative reserve ISC markers, including Bmi1 (41), DCLK1 (29), HopX (49), Lgr5 ^low (6), Lrig1 (37), Sox 9^high (55), and Tert (31) are expressed by LRCs as well as some neighboring crypt cells. Although the nature of the reserve ISCs remains obscure, recent gene expression studies have identified genes of secretory cell lineage, including genes specific to enteroendocrine cells (EECs) in the cell populations isolated based on either their label-retention property (8, 24) or the expression of reserve ISC markers (6, 17, 55, 58).

The EECs, which comprise ~1% of intestinal epithelium, share a common lineage with other principle nonendocrine cells in the intestine and originate from Lgr5-expressing ISCs. The lineage fate decision is made through Notch signaling (13). Notch signaling activates the promoters of the HES transcription repressor that normally inhibits expression of proendocrine basic helix-loop-helix transcription factor genes Atoh1 (Math 1) and neurogenin 3 (Ngn3). The inactive state of Notch signaling allows cells to express Atoh1 and Ngn3 genes and differentiate into endocrine cell lineage-specific precursors (18, 23). Subsequently, through a complex network of transcription factor genes such as NeuroD1 (32), Pax4 (22), and Nkx2.2 (11), the Ngn3⁺ endocrine precursors differentiate to mature hormone-producing EECs that are classified into at least 15 different terminally differentiated subsets according to their specific hormone expression (1, 40). These EECs constitute the largest and the most complex endocrine system in the body and regulate motility and secretion in the digestive system, appetite, gut immunity, and metabolism (12, 20, 56).

Recent converging evidence now suggests a role for EECs in stem cell dynamics and plasticity in the intestine (8, 14, 43, 55, 58). Studies of LRCs identified using Cyp1a1-H2B-YFP revealed that YFP⁺-LRCs expressed Paneth and EEC genes, Mmp7 and ChgA, respectively, along with Lgr5, Tert, HopX, and Lrig1 (8). Further studies of Sox9⁺ cells showed significant enrichment of multiple EEC genes along with reserve ISC genes Bmi1 and HopX in the reserve ISC-enriched population based on Sox9 gene expression (55). Furthermore, noncycling Lgr5^low/Ki67⁻ +4 cells were found to express reserve ISC genes Bmi1, HopX, Tert, and Lrig1 as well as secretory cell markers, including EEC genes such as ChgA (6). Consistently, Mex3a⁺/Lgr5^low cells were found to be slow-cycling ISCs and to express EEC genes highly (5). More recently, chromatin immunoprecipitation sequencing and RNA sequencing studies of Bmi1-expressing cells concluded that the Bmi1⁺ population consists of mostly EEC lineage (17, 58). Taken together, these studies suggest that at least some subsets of reserve ISCs have characteristics of EEC lineage.

More direct evidence for the stem cell potential of the secretory cell lineage was obtained from lineage-tracing studies. Consistent with the findings by Buczacki et al. (8), a mouse lineage-tracing model using Cre-based recombination driven by Notch ligand Delta-like-1 showed reserve stem cell potential of Delta-like-1⁺ secretory precursors during radiation-induced injury (54). Moreover, stem cell potential was also found beyond the stage of secretory cell precursors in an earlier lineage-tracing study of Ngn3⁺ EEC precursors (43). Subsequently, Gross et al. (14) showed stem cell potential of post-Ngn3, Nkx2.2-derived EECs. More recently, lineage-tracing studies of Prox1-expressing cells showed homeostatic and injury-inducible stem cell potentials of post-Ngn3, Prox1-derived EECs (58). Therefore, these results from the lineage-tracing studies are not only consistent with gene expression characteristics of the cell populations isolated based on their label-retaining property or expression of reserve ISC markers but also suggest direct involvement of EECs in stem cell dynamics.

The degree and type of EEC possessing these reserve stem cell properties as well as the mechanism by which they contribute to stem cell dynamics remains unclear. NeuroD1, a basic helix-loop-helix transcription factor gene that regulates post-Ngn3 differentiation and maturation of EECs is expressed in most EECs (23, 32). Here, we report that lineage-tracing experiments using inducible NeuroD1-CreERT2 indicate that mature post-Ngn3 EECs have stem cell potential and contribute to stem cell dynamics under basal conditions and in response to injury using an irradiation model. Furthermore, we show that a particular terminally specified EEC, the serotonin-producing enterochromaffin (EC) cell, is the predominant EEC type that has these reserve stem cell properties using a Tryptophan hydroxylase 1 (Tph1)-CreERT2 mouse model. In summary, we provide multiple lines of evidence from in vivo and ex vivo lineage tracings, kinetics, and organoid formation from FACS-sorted single cells, which show a subset of mature EECs, including EC cells, possesses stem cell potential, and we discuss the implications for their potential role as a cell of origin for small intestinal neuroendocrine tumors (10, 44). Although not directly observed, the stem cell potential of EECs demonstrated in the present study is consistent with dedifferentiation (or reversion) as suggested by others (17) and not by transdifferentiation or reprograming (conversion; 19).

MATERIALS AND METHODS

Animal usage.

NeuroD1-CreERT2 (JAX mice strain 025867), HopX-CreERT2 (JAX 017606), Lgr5-EGFP-IRES-CreERT2 (JAX 008875), and Rosa26-tdTomato (JAX007908) were obtained from the Jackson Laboratory. Tph1-CFP mice were established at the University of Massachusetts Medical School (now JAX 028366). Tph1-CreERT2 mice were established at University of Michigan Medical School using a bacterial artificial chromosome (BAC) transgene generated from the Tph1 gene. A 3,718-bp cassette containing the fusion of sequence-encoding enhanced green fluorescent protein (eGFP), Cre recombinase, and the mutant estrogen ligand-binding domain (ERT2) was inserted into a BAC clone (RP23-29A7) containing the full-length gene encoding Tph1 at the ATG start codon. This modified BAC construct was injected into C57BL/6 single-cell embryos to produce a transgenic line at the University of Michigan transgenic core. The eGFP protein fluorescence expression is not detectable. All procedures involving mice followed National Institutes of Health guidelines and were approved by the National Institute of Diabetes and Digestive and Kidney Diseases animal care and use committee.

Animal treatment.

For radiation exposure, mice received a single dose of abdominal irradiation (14 Gy, 2.25 Gy/min) using an X-Rad 320 (Precision X-Ray, North Branford, CT). Custom lead blocking was used to shield the remainder of the body from irradiation. Dose delivered was confirmed using lucite phantoms in identical experimental conditions with thermoluminescent dosimeters. For Cre induction, mice were given a single dose of tamoxifen (TAM; cat. no. T-5648, Sigma; 100 μl, 10 mg/ml in corn oil) by oral gavage. For BrdU labeling, mice were given BrdU (BD PharMingen) daily by intraperitoneal injection (2 mg in 0.1 ml saline) for 5 days.

Crypt preparation and immunohistochemistry.

Crypt-enriched preparations were obtained from the small intestine by incubation in Dulbecco’s phosphate-buffered saline (DPBS) containing 2 mM EDTA for 30 min at 4°C followed by vigorous pipetting and filtering through 70-μm mesh. The crypt-enriched materials were then fixed in 4% paraformaldehyde (PFA) in DPBS for 1 h at room temperature and treated with DPBS containing 0.1% Triton X-100 for 30 min. Crypts were incubated in DBPS containing 1% bovine serum albumin for 30 min and immunostained with primary antibody at room temperature for 1 h. The primary antibodies used were: 1:2,000 rabbit polyclonal anti-chromogranin A (ChgA); cat. no. 20085, ImmunoStar, Hudson, WI), 1:50 rabbit polyclonal anti-lysozyme (cat. no. RP-028, DBS, Inc., Pleasanton, CA), 1:50 rabbit polyclonal anti-mucin2 antibody (cat. no. sc-15334, Santa Cruz, CA). After three washes with DPBS containing 0.1% Triton X-100, crypts were incubated with Alexa Fluor 488 or Alexa Fluor 647-conjugated goat anti-rabbit IgG secondary antibody (Invitrogen, Carlesbad, CA). For BrdU staining, 4% PFA-fixed crypts were treated with heating at 99°C in 10 mM sodium citrate buffer (pH 6.0) for 15 min to denature DNA. After three washes with DPBS, the crypts treated for denaturing DNA were stained with Alexa Fluor 488- or Alexa Fluor 647-conjugated anti-BrdU antibody (BioLegend) following ChgA-staining. After three washes with DPBS containing 0.1% Triton X-100, immunostained crypts were subjected to confocal and bright-field examination with a complete Z-stack using a Zeiss LSM-510 Meta confocal microscope (Carl Zeiss USA).

Examination of tdTomato-expressing cells.

Cre-induced tdTomato-expressing cells were clearly identifiable by either epifluorescence or confocal microscopy. For analysis of position and number of Cre-derived tdTomato-expressing cells, crypts were fixed, stained for ChgA, and examined by confocal microscopy. Each crypt was scanned along the z-axis to accurately acquire the position and number of the cells at different depths. For examination of ribbon-like streaks, a continuous generation of progeny from an induced Cre-derived stem cell, a whole mount of the intestine was first screened using an Olympus IX-71 epifluorescence microscope (Olympus USA). Detailed analysis of ribbons was obtained from dissected tissue that was fixed, stained for ChgA, and examined using a Zeiss LSM-510 Meta confocal microscope (Carl Zeiss USA).

Organoid formation and lineage tracing.

Preparation and culture of organoids were performed as previously reported (42). Briefly, crypts were separated from the intestine by incubation in DPBS containing 2 mM EDTA for 30 min at 4°C followed by vigorous pipetting and passing through a 70-μm mesh. After crypts were counted, ~100 crypts were mixed with 25 μl of Matrigel (cat. no. 356231, Corning, Tewksbury, MA) and plated in 8-chamber well plates (Nunc Laboratory-Tek cat. no. 155411, ThermoFisher, Waltham, MA). After polymerization of Matrigel, the crypts were merged and cultured in 200 μl of growth medium [either Advanced DMEM/F12 (Invitrogen) containing 50 ng/ml EGF (cat. no. 315–09, Peprotech, Rocky Hill, NJ), 500 ng/ml R-spondin 1 (cat. no. 3474-RS, R&D, Minneapolis, MN), and 100 ng/ml Noggin (ca. no. 250–38, Peprotech), or IntestiCult Organoid Growth Medium (STEMCELL, Vancouver, BC, Canada)]. Organoid growth was maintained by replacing the growth medium every 2–3 days. The distribution of tdTomato⁺ cells within the developing organoids was recorded using a Zeiss LSM-510 Meta confocal microscope. Multiple images were obtained at sequential focal planes and time points (8–24 h interval over 10 days).

Organoid immunohistochemistry.

Organoids were released from Matrigel by gentle pipetting following replacement of the media with DPBS. The free-floating organoids were transferred to a microfuge tube, centrifuged at 150 g for 2 min, and fixed with 4% PFA in DPBS for 1 h at room temperature. Following treatment with DPBS containing 0.1% Triton X-100 for 30 min, the organoids were incubated in DPBS containing 1% BSA for 30 min. Subsequently, organoids were immunostained with primary antibodies to ChgA at 4°C overnight in DBPS containing 1% BSA. After three washes with DPBS containing 0.1% Triton X-100, the organoids were incubated with Alexa Fluor 647-conjugated goat anti-rabbit IgG secondary antibody for 30 min at RT. After three washes with DPBS containing 0.1% Triton X-100, immunostained organoids were gently placed in an 8-chamber well in PBS and observed using a Zeiss LSM-510 Meta confocal microscope.

FACS and single-cell organoid formation assay.

The entire mouse small intestine was collected, washed, and incubated in DPBS containing 2 mM EDTA, 1 mM Dithiothreitol, and 10 μM Y27632 (Y-0503, Sigma) for 20 min at 37°C in a shaking water bath. After washing, tissue was incubated for another 20 min with 75 U/ml type 1 collagenase (Gibco) in DMEM (Invitrogen). Cells were resuspended and filtered through a 40-μm mesh to obtain a dispersed single-cell population. Cells were resuspended in phenol-free DMEM containing 10 μM Y27632 (Y-0503, Sigma) and 50 μg/ml DNase I (DN-25, Sigma) and sorted by FACS using a BD FACSAria Fusion (Becton Dickinson). FACS-sorted cells were collected in Advanced DMEM/F12 (Invitrogen) containing 10 μM Y27632 and 50 μg/ml DNase I. The sorted cells were mixed with 25 μl of Matrigel containing 1 μM Jagged-1 (Anaspec, Fremont, CA) and cultured in the growth medium as described above except for the addition of 10 μM Y27632 during the first 3 days.

Computational simulation to predict the growth dynamics of neuroD1-derived ISCs.

A recent computational model of lineage-tracing studies indicates that the homeostasis of ISCs is maintained by neutral drift of equipotent ISC cell populations (27, 47). To test whether reverted NeuroD1-derived stem cells follow stochastic neutral drift dynamics after their reversion to actively cycling ISCs, we designed a simple simulation based upon a population asymmetry-based neutral drift model (27, 47). Because it was not experimentally feasible to measure clonal size of rare labeled ISCs, we chose to simulate the growth of NeuroD1-derived ribbons as a surrogate for the number of ISCs. The simulation incorporating the assumptions described below were programmed in Matlab.

Assumptions.

First, contraction and expansion of labeled stem cell clones follows a stochastic behavior in a cell-extrinsic self-renewal manner (contact process) and a pattern of neutral drift dynamics (21). Using previously described notations (27), this process is described as appears below,

$$S^{\left(l\right)}{S}^{\left(u\right)}\stackrel{\lambda }{\to }\{\begin{array}{cc}{S}^{\left(l\right)}{S}^{\left(l\right)}& \text{Prob}\text{.\hspace{0.17em}1/2}\\ S^{\left(u\right)}{S}^{\left(u\right)}& \text{Prob}\text{.\hspace{0.17em}1/2}\end{array}$$ (1)

where s^(l) and s^(u) denote the labeled and unlabeled stem cells, respectively, Prob denotes probability, and λ denotes the average rate of stem cell turnover (λ = 1/day).

Second, within a crypt, each labeled NeuroD1-derived EEC that dedifferentiates to a labeled ISC becomes subject to neutral drift. In accordance with assumption 1, fate and growth of the labeled clone (n = 1 at time 0) follow a simple random walk (50) between lower (n = 0) and upper [n = upper boundary number (Nub)] boundaries. The lower boundary condition is absorbing, but the upper boundary condition can be simulated to be either absorbing or reflecting. In this stochastic simulation, clones were allowed to expand temporarily up to Nub = 20 instead of strictly making absorption at Nub = 16, which is the estimated average number of stem cells per crypt. This was because we have observed fluctuations in the number of Lgr5-eGFP⁺ ISCs at least up to 20. In previously published models (27, 47), it was elucidated from master equations that the average number of stem cells per surviving clone (Nave) increased over time according to a square-root power law, thus $\text{N}ave(t,\hspace{0.17em}\text{time})\approx \sqrt{\lambda t}$ (27), or $\approx \sqrt{\pi \lambda t}$ (47), λ = 1 for short-term clonal evolution. The present simulation using a simple random walk model yielded in silico growth of surviving ISC clones that was consistent with the predicted growth curves, $\sqrt{t}$ and $\sqrt{\pi t}$, and with the growth of ISC clones observed in the lineage-tracing studies (27, 47). This prior validation of the simulation supported its present use in our studies and formed the basis to apply assumptions 3–5 in simulating the growth of NeuroD1-derived ribbons.

Third, the stem cell population in the stem cell compartment of the crypt maintains homeostasis and a constant basal number of transit amplifying cells that divide every 12 h up to six cycles before differentiating into various functional cells, the majority of which are enterocytes (33).

Fourth, the renewal of the epithelium is maintained by continuous production and upward migration of epithelial cells. The average lifetime of an epithelial cell in the small intestine is ~60 h (53).

Fifth, the majority of crypts produce two ribbons per crypt. This labeled, visible structure consists of ~800 cells on average [250–300 cells in the crypt (33) and 250–300 cells in each of two villi] by our counting.

RESULTS

Distribution and growth dynamics of neuroD1-derived cells in the small intestine.

To study the distribution and growth dynamics of NeuroD1-derived cells, NeuroD1-CreERT2;Rosa26-tdTomato mice were induced with a single TAM administration and examined at 1, 2, 5, and 14 days postinduction. Crypt-enriched preparations from the small intestine were immunostained for the mature EEC marker ChgA and examined under the confocal microscope for the position and number of NeuroD1-tdTomato and/or ChgA-expressing cells (Fig. 1). The three cell populations, NeuroD1⁺/ChgA⁺ cells (yellow triangle), NeuroD1⁻/ChgA⁺ cells (green triangle), and NeuroD1⁺/ChgA⁻ cells (red triangle) were identified and analyzed (Fig. 1). NeuroD1⁺/ChgA⁺ cells were mature ChgA-expressing NeuroD1-derived EECs. NeuroD1⁺/ChgA⁻ cells were either NeuroD1-derived immature EEC or nonendocrine cells that were present as a rare population following their reversion into stem cells (as evidenced below in the present study). NeuroD1⁻/ChgA⁺ cells were either mature EECs that were either incompletely labeled or mature EECs that differentiated from precursors that arose after the Cre-induction phase. Because the proportion of NeuroD1-negative cells among the ChgA-positive cells declined from 34% at day 1 to 15.7% at day 2 and increased thereafter (Fig. 1B), the majority of the NeuroD1⁻/ChgA⁺ cells were most likely incompletely labeled at day 1 whereas the increase observed after day 2 most likely represented cells derived from precursors that arose after the Cre induction ended. Therefore, it was considered that NeuroD1-Cre induction by a single TAM administration effectively labeled mature ChgA-expressing EECs, which were identified as NeuroD1⁺/ChgA⁺, by day 2.

Fig. 1.

Dynamics of NeuroD1-derived cells in small intestine of NeuroD1-CreERT2;Rosa26-tdTomato mice. Distribution of NeuroD1⁺/ChgA⁺ cells (A), NeuroD1⁻/ChgA⁺ cells (B), and NeuroD1⁺/ChgA⁻ cells (C) along the crypt axis at 1, 2, 5, and 14 days post-TAM induction. The x- and y-axes indicate cell position along the crypt axis and number of cells per 100 crypts, respectively. Note that the count for positions higher than p10 (>p10) included from p11 to the top edge of isolated crypts. Representative images of crypts containing NeuroD1⁺/ChgA⁺ cells (yellow triangle), NeuroD1⁻/ChgA⁺ cells (green triangle), and NeuroD1⁺/ChgA⁻ cells (red triangle) at days 2, 5, and 14 post-TAM induction (D). NeuroD1-tdTomato, ChgA (Alexa-488) and DAPI are shown in red, green, and blue, respectively. *NeuroD1⁺/ChgA⁻ cell found at position 1 at 14 days, and red triangle in C. TAM, tamoxifen.

A distribution resulting from counts of NeuroD1 and/or ChgA expressing cells contained in a crypt at different time points reflected dynamics of migration and differentiation of NeuroD1-derived cells. We suggest that the distribution of the NeuroD1⁺/ChgA⁺ cells at day 2 reflects the average distribution of NeuroD1-expressing cells in normal crypts of the small intestine (Fig. 1A). The NeuroD1⁺/ChgA⁺ cells were most frequently found at +4 position and above position 10. The time-dependent changes in distribution of the NeuroD1⁺/ChgA⁺ cells from day 2 to day 14 reflect the migration of these cells. Fourteen days post-Cre induction, the number of NeuroD1⁺/ChgA⁺ cells was dramatically reduced because of upward villus migration. The reduction was most prominent above p10, suggesting migration from crypts to villi (counts include only the cells up to the top edge of isolated crypts). Notably, the greatest proportion of NeuroD1⁺/ChgA⁺ cells within the crypt remained at the crypt base, with a peak at +4 position where LRCs reside (Fig. 1A). This result was consistent with our previous finding of a subset of the EECs that either remained at +4 position or migrated down to the crypt bottom instead of migrating up toward villi (45). Interestingly, similar to the 6.5% LRC frequency reported by Potten et al. (36), this subset was found in 6.7% of small intestinal crypts after a single dose TAM induction (Fig. 1A).

As stated above, NeuroD1^-/ChgA⁺ cells after day 2 were mature EECs that likely arose following completion of the Cre-induction phase and are therefore tdTomato-unlabeled. Thus, this population gradually increased throughout the crypt during the 14-day postinduction period (Fig. 1 B). As expected, their distribution at day 14 closely resembled the distribution of NeuroD1⁺/ChgA⁺ cells at day 2. Thus, these cells were most frequently found at +4 and above p10.

Finally, we identified a small number of NeuroD1⁺/ChgA^- cells that may represent immature EECs. Time-dependent decreases in this population were seen at +4 position where immature EECs normally appear and subsequently decrease with maturation. However, it is possible that some of the NeuroD1⁺/ChgA⁻ cells remaining at the crypt bottom at 14 days, (* marked cell in Fig. 1 C and D), are former EECs that reverted to ISCs with clonal expansion potential as described below and found by chance in this preparation. In summary, this lineage-tracing study of NeuroD1-derived cells demonstrates the distribution, migration, and differentiation of NeuroD1-derived cells that are consistent with our previous observations of EECs in the murine small intestine (45).

Many of the EECs that remained at +4 positions for over 2 wk are typical LRCs.

To determine whether a subset of the EECs that remained at the +4 position for more than 2 wk were LRCs, we performed a combined experiment of NeuroD1 lineage tracing and BrdU labeling. Mice were given BrdU daily by intraperitoneal injection for 5 days and a single dose of TAM on the middle day of the BrdU-labeling period (Fig. 2A). Crypts were harvested 2 wk after TAM induction and immunostained for ChgA and BrdU. We screened over 1,500 crypts and examined a total of 109 NeuroD1-derived tdTomato-positive cells for their positions in the crypt, status of ChgA expression and BrdU labeling. We found that the NeuroD1-derived tdTomato-positive cells were all ChgA positive EECs except one ChgA-negative Paneth cell. Consistent with the results from the time-course study (Fig. 1A), the majority of NeuroD1-derived tdTomato-positive, ChgA⁺ EECs remained around the crypt base, with a peak at the +4 position (Fig. 2B). Importantly, about one-half of NeuroD1-derived tdTomato-positive ChgA⁺ EECs retained BrdU. We found the highest BrdU labeling (66%) in the tdTomato-positive ChgA⁺ EECs at +4 position (Fig. 2, A and B). We also reverse-screened for BrdU-positive cells to examine expression of tdTomato and ChgA. The results from screening 337 crypts showed that 10 cells were ChgA⁺ EECs (4 cells at +4 position), in total 18 BrdU-positive cells whereas the rest were Paneth cells. Therefore, these results suggest that many of the EECs that remained at +4 positions for over 2 wk were typical LRCs. This label-retaining subset is consistent with the short-term LRC reported by Li et al. (24).

Fig. 2.

Retention of BrdU in the subset of NeuroD1-derived EECs. A: BrdU was intraperitoneally injected daily for 5 days. Single dose of TAM was given on the middle day of the BrdU-labeling period. Crypts were collected 2 wk after TAM induction and examined following immunostaining for ChgA and BrdU. B: distribution and BrdU-labeling status of 108 NeuroD1-derived tdTomato-positive ChgA-positive EECs 2 wk after TAM induction. C: representative images of BrdU-retaining NeuroD1-derived EECs. Top: images of crypt containing two ChgA⁺ cells at +4 position. Left: ChgA+ cell was tdTomato-positive BrdU-labeld EEC that remained for 2 wk. Bottom: images of crypt containing two ChgA⁺ cells at +4 position (right) and +5 position (left). Right: ChgA⁺ cell was tdTomato-positive BrdU-labeld EEC that remained for 2 wk. NeuroD1-tdTomato, ChgA (Alexa-488), BrdU (Alexa-647) and DAPI were shown in yellow, green, red, and blue, respectively. EEC, enteroendocrine cell; TAM, tamoxifen.

A subset of neuroD1-derived cells revert to ISCs under basal conditions.

To determine whether a subset of the NeuroD1-derived cell population could revert to the ISCs, we performed lineage tracing using NeuroD1-CreERT2;Rosa26-tdTomato mice induced with a single TAM administration and examined for the presence of NeuroD1-derived tdTomato-labeled ribbons. Screening the entire intestine using epifluorescent microscopy, we found ribbons of differentiated epithelial cells in the small intestine as early as 1 to 2 wk after induction with a single TAM administration (Fig. 3A). We found no ribbon in the large intestine. This is consistent with lack of LRC in the large intestine reported by Hughes et al. (15). Low-power confocal examination of histological orientation indicated typical ribbons originate from the crypt bottom and extended to the tips of villi (Fig. 3B). Confocal examination of ChgA-immunostained, dissected mucosa indicated that the ribbons started from columnar epithelial cells between Paneth cells at the crypt base and extended to the tips of villi (Fig. 3C). The cells forming ribbons were ChgA-negative, indicating a nonendocrine cell of origin (Fig. 3C). These results suggest that there is a subset of EECs among the NeuroD1-derived cell population that dedifferentiates to an active ISC and contributes to generation of the epithelium under normal physiological conditions.

Fig. 3.

NeuroD1-derived cells reverted to ISCs and gave rise to intestinal epithelial cells in mice in vivo. A: representative epifluorescent images of NeuroD1-derived tdTomato-positive ribbons in unfixed small intestinal tissue: ×40 (top) and ×100 (bottom) magnification; scale bars, 200 μm. B: low-power confocal image of a tissue section containing NeuroD1-derived ribbons extending from a tdTomato-positive crypt to a top of villi. scale bar, 100 μm. C, left: low-power confocal image of a tissue section containing NeuroD1-derived ribbons extending from a tdTomato-positive crypt immunostained for ChgA. NeuroD1-tdTomato, ChgA (Alexa-647) and nucleus (DAPI) are shown in yellow, red, and blue, respectively. Top, middle: high-power image (inset, left) of a tdTomato-positive ribbon and nearby ChgA-positive cell near the top of the villus. Top, right: high-power image of the ChgA-positive cell (inset, middle). Bottom, middle and right: high-power image of the crypt (inset, bottom, left) and merged with the transmitted light image, respectively, illustrating the position of the NeuroD1-derived tdTomato-expressing cells between Paneth cells (p). D: graph of the experimentally derived average number of NeuroD1-derived tdTomato-expressing ribbons in the small intestine 1, 2, 4, 6, 8, and 10 wk (closed circles) after a single TAM administration (n = 3 at 4 and 7 wk, n = 4 at 10 wk). No ribbons were observed in the large intestine. The blue, red and yellow lines indicate means ± SD from 6 simulation experiments that were designed based upon a stochastic neutral drift model and performed assuming an initial 400 (blue), 600 (red), and 800 (yellow) clones. ISC, intestinal stem cell; TAM, tamoxifen.

To determine whether these ISCs derived from NeuroD1-reserve ISCs follow stochastic neutral drift dynamics, we examined the entire small intestine at 1, 2, 4, 6, 8, and 10 wk following the induction with TAM. Results showed a time-dependent increase in the number of ribbons following induction (Fig. 3D, closed circles). Subsequently, based on the neutral drift model (27, 47), we designed a computational simulation of neutral drift dynamics for ISC generation of ribbons (see material and methods) to determine if our results were consistent with this model (Fig. 3D, colored lines). Our experimental results align closely with our simulation that was based upon an assumption of an initial 600 clones and therefore, suggest that ISCs derived from reversion of NeuroD1-expressing EECs follow the previously found neutral drift dynamics of ISCs. These results also suggest that the drift is unbiased and indicate that upon reversion to actively cycling ISCs, the EEC-derived ISCs grow and expand without any advantage or disadvantage compared with preexisting ISCs.

Reversion of neuroD1-derived cells to ISCs in ex vivo organoids.

The generation of NeuroD1-derived migration streams was also observed in ex vivo organoids (Fig. 4). For the organoid study, crypts were isolated from NeuroD1-CreERT2;Rosa26-tdTomato mice, cultured to generate organoids and induced by TAM at day 1 after the initiation of the culture. By tracing the generation and migration of NeuroD1-derived EECs, we observed formation of tdTomato-labeled ribbons (Fig. 4A, top) as well as almost entirely tdTomato-labeled budding crypts (Fig. 4A, bottom). These NeuroD1-derived ribbons were ChgA-negative and indicate that NeuroD1-derived EECs gave rise to nonendocrine cell lineages (Fig. 4B). Similar formations of tdTomato-labeled ribbons were found in ~2% to 7.5% of the organoids examined in two independent sets of experiments. These results are consistent with reversion of NeuroD1-derived mature EECs to ISCs in ex vivo organoids.

Fig. 4.

Representative images of NeuroD1-derived epithelial cells in ex vivo organoids. Organoids were established from NeuroD1-CreERT2;Rosa26-tdTomato mice and induced with TAM 1 day after the initiation of the culture. A: merged fluorescent and transmitted light images of organoids. NeuroD1-derived tdTomato (yellow)-expressing cells appear throughout the organoids. Top: organoid containing a tdTomato-positive ribbon (red line with *) 7 days post-TAM induction. Bottom: organoid with a nearly complete tdTomato-positive budding crypt (red arrow) 7 days post-TAM induction. B: whole mount ChgA-immunostained organoid. Top: transmitted light (left) and immunofluorescent images (right). NeuroD1-tdTomato, ChgA (Alexa-647) and nucleus (DAPI) appear in yellow, red, and blue, respectively. A NeuroD1-derived tdTomato-positive ribbon is marked by a red line + *. Bottom: high-power immunofluorescent images (inset, top, right) for ChgA (Alexa-647) (left), NeuroD1 (tdTomato) (middle) and merged ChgA, NeuroD1 and nucleus (DAPI) (right). ChgA⁺ cells (right) are noted with red arrows. Cells forming a NeuroD1-derived tdTomato-positive ribbon (marked by a red line + *) are all ChgA-negative and thus nonendocrine lineage.

Effect of irradiation on neuroD1-derived reserve ISC function.

A primary role of reserve ISCs is to restore actively cycling Lgr5⁺ ISCs and thus the intestinal epithelium upon their loss because of inflammation or injury (51, 57). To observe this function, we examined the effect of abdominal irradiation on NeuroD1-derived ISCs. NeuroD1-CreERT2;Rosa26-tdTomato mice were induced by a single TAM administration and given a single dose of abdominal irradiation (14 Gy) 4 days after induction (Fig. 5A). Seven weeks following irradiation, we examined the entire small intestine for the presence of NeuroD1-derived ribbons (Fig. 5A). We observed a decrease in the number of NeuroD1-derived ribbons that originate from a single crypt (rS) and an increase in the number of ribbons that originate from a patch of multiple crypts (rP) in the irradiated animals as summarized in the illustration (average 4 patches per animal; Fig. 5B). Although irradiation effects could be enhanced by prior treatment with TAM (59), there was a ~50% decrease in rS in irradiated animals compared with controls (n = 3 per group, P < 0.01; Fig. 4B). This result suggested that approximately one-half of the tdTomato-labeled cells belonged to a radiosensitive subset of reserve ISC population, possibly because of their cell cycle phase at the time of irradiation as suggested for some cell cycle-based subsets of +4 ISCs by Potten et al. (35) and/or radiosensitive actively cycling ISCs already reverted before the time of irradiation. However, there was a subset of the radio-resistant population that functioned to maintain the intestinal epithelium (Fig. 5, A and B). Although the formation of the rS was reduced, there were patches of tdTomato-positive ribbons (rP) only in irradiated mice (Fig. 5, A–C), resulting in a significant increase in the total number of ribbons (rS and rP) in the irradiated mice (n = 3 per group, P < 0.05) (Fig. 5B). These results suggest that these patches were formed in response to radiation injury. At least two mechanisms may explain the origin of these patches. Either the patch originated from a single clone, that expanded by crypt duplication (a 2.8–5.6 cycle process necessary to produce 7–42 crypts within 7 wk) or the patches formed from multiple independent clones perhaps by an unexplained local field effect. Regardless of origin, the presence of these patches strongly suggests a role for dedifferentiation of mature NeuroD1-derived EECs to ISCs in the process of small intestinal repair in response to injury.

Fig. 5.

Effect of radiation on reversion of NeuroD1-derived cells to ISCs in vivo. A: schematic schedule of TAM induction, irradiation and examination. B: cartoons illustrating basal formation of NeuroD1-derived cell ribbons in nonirradiated controls (top) and irradiated mice (bottom). While the basal formation of the ribbons that originated from a single crypt (rS; shown as V-like structures) was significantly reduced, ribbons that originated from patches of multiple crypts (rP) were found only in irradiated mice (14 Gy). C: graph of the number of rS, rP, and total (rS + rP) in irradiated vs. nonirradiated control mice. *P < 0.05 **P < 0.0005 (t-test, n = 3 per a group). D: representative images of NeuroD1-derived cell patch. Mucosal (left) and serosal (right) views of a patch are shown. Note that the patches of tdTomato-positive cell ribbons appear to have originated from 42 tdTomato-positive crypts (bars, 100 μm). ISC, intestinal stem cell; TAM, tamoxifen.

EC cells revert to ISCs under homeostatic conditions.

Multiple lines of evidence suggest a causal relationship between stem cell potential and tumorigenesis (3, 7, 30, 39, 52). The serotonin-producing EC cell is the most predominant EEC type, widely distributed throughout the gastrointestinal tract and recently shown to be the cell of origin for small intestinal neuroendocrine tumors (SI-NETs) (28, 44). This would suggest that EC cells could revert to stemness. To test whether terminally differentiated EC cells can dedifferentiate to become ISCs, we first crossed Tph1-CFP BAC transgenic mice with NeuroD1-CreERT2;Rosa26-tdTomato mice (NeuroD1⁺/Tph1⁺ mice). Two days after TAM induction, isolated crypts were fluorescently stained for ChgA and examined using a confocal microscope. As expected, more than half of NeuroD1-expressing cells in the small intestine were Tph1⁺ and all Tph1⁺ cells were NeuroD1-derived. Analysis of the frequency and distribution of NeuroD1⁺/Tph1⁺ cells in crypts indicated that NeuroD1⁺/Tph1⁺ cells were most frequently located at +4 position and above position 10 (Fig. 6A). Having established that EC cells were among the NeuroD1-derived EECs, the ability of EC cells to dedifferentiate to ISCs was assessed by lineage-tracing experiments using Tph1-CreERT2;Rosa26-tdTomato mice. Crypts from these mice were examined 2 days after TAM induction and ChgA immunofluorescence to first assess the validity of this system. Similar to the NeuroD1⁺/Tph1⁺ cells described earlier, the frequency and distribution of Tph1-tdTomato-labeled EC cells were most frequently located at +4 position and above position 10 (Fig. 6B). With this validation, lineage tracing of Tph1-derived cells was assessed under conditions previously described for NeuroD1-CreERT2;Rosa26-tdTomato mice in which we found time-dependent increases in formation of ribbons at 2, 4, and 10 wk following TAM induction (24, 35, and 51 ribbons per small intestine, respectively). Confocal examination of the ribbons indicated that the Tph1-derived tdTomato-labeled ribbons started from columnar epithelial cells between Paneth cells at the crypt base and extended to two neighboring villi (Fig. 7A). We also confirmed that Tph1-derived tdTomato-labeled epithelial layers consist of nonendocrine cells such as Mucin 2-positive goblet cells and lysozyme-positive Paneth cells (Fig. 7B). Furthermore, abdominal irradiation induced multiple patches of Tph1-derived crypts, suggesting a role for Tph1-derived EC cells in intestinal epithelial regeneration (Fig. 7C). To further validate these findings, we performed ex vivo organoid studies using isolated crypts from the Tph1-CreERT2; Rosa26-tdTomato mice and identified a tdTomato-labeled crypt generating a typical ribbon formation (Fig. 7D). The majority of the tdTomato-labeled cells were ChgA-negative and thus of nonendocrine-cell origin (Fig. 7D). Importantly, in primary crypts isolated from two mice, 2 days after TAM induction, we found three instances of tdTomato-labeled cells at +4 position dividing asymmetrically along the apical-basal axis at frequencies of 0.025% (1 in 3,888 crypts) to 0.07% (2 in 2,706 crypts) as shown (Fig. 8, A–C). In these crypts, we identified tdTomato-expressing cell doublets at +4 position in which only the cell at the basal side within the doublets was ChgA-positive. This observation suggests that the Tph1-derived cell doublets represent a terminal phase of asymmetric cell division in which the apically positioned ChgA-negative cell represents a dedifferentiated cell with potential reversion toward an ISC (Fig. 8). These findings ensure that the lineage tracing is specific to EC cell origin as intended and not nonspecific expression from cycling ISCs. Although rare, the observed frequency of +4 tdTomato-labeled cells dividing asymmetrically is consistent with the computationally estimated proportion of the EECs expected to undergo reversion to cycling ISCs following a single TAM induction. A simulation based on a stochastic neutral drift model estimates that ~600 clones in the entire small intestine are expected to initially result from a single TAM activation of EECs (Fig. 3D). Providing that the area of the entire small intestine in adult mice is ~30 cm² (≈40 cm length, ≈1 cm width at pyloroduodenal junction tapering to 0.5 cm at the terminal ileum) and the density of crypts are ≈30,000 crypts per cm² mouse small intestine (average from animals used in the study), the calculated expected frequency of the EECs reversion to cycling ISCs would be ≈20 EECs (0.07%) in 30,000 crypts per cm² following a single TAM induction. Because the observable period for the asymmetric division process is limited, the frequency of EECs undergoing dedifferentiation via the asymmetric cell division process is predicted to exceed no more than 0.07%.

Fig. 6.

Representative images and distribution of Tph1⁺ cells in the small intestine. A: NeuroD1-derived Tph1⁺ EC cells in the small intestine of NeuroD1-CreERT2; Rosa26-tdTomato; Tph1-CFP mice. Left. Crypts isolated from the small intestine were immuostained for ChgA. Representative images of crypt containing NeuroD1⁺/ChgA⁺/Tph1⁺ cells at position +4. NeuroD1-tdTomato, ChgA (Alexa-647) and Tph1-CFP are shown in yellow, red and blue, respectively. Right: distribution of NeuroD1⁺/ChgA⁺/Tph1⁺, NeuroD1⁺/ChgA⁺/Tph1⁻, NeuroD1⁺/ChgA⁻, and NeuroD1⁻/ChgA⁺ cells, along the crypt axis 2 days post-TAM induction. The x- and y-axes indicate cell position along the crypt axis and number of cells per 100 crypts, respectively. Note that the count for positions higher than p10 (>p10) was the count from p11 to the top edge of isolated crypts. B: Tph1-derived EC cells in the small intestine from Tph1-CreERT2; Rosa26-tdTomato mice. Left: isolated crypts from the small intestine were immuostained for ChgA. Representative images of crypt containing Tph1⁺/ChgA⁺ cells at position +4. Tph1-tdTomato, ChgA (Alexa-647) and DAPI were shown in yellow, red, and blue, respectively. Right: distribution of Tph1⁺/ChgA⁺, Tph1⁺/ChgA⁻, and Tph1⁻/ChgA⁺ cells along the crypt axis 2 days post TAM induction. The x- and y-axis indicates cell position along the crypt axis and number of cells per 100 crypts, respectively. Note that the count for positions higher than p10 (>p10) was the count from p11 to the top edge of isolated crypts. EC, enterochromaffin cell; TAM, tamoxifen.

Fig. 7.

Tph1-expressing EC cells revert to ISCs and gave rise to intestinal epithelial cells in mice. A: Tph1-CreERT2;Rosa26-tdTomato mouse unfixed small intestine 2 wk after TAM induction. Top: representative image of Tph1-derived tdTomato-positive cell ribbon (merged fluorescent and transmitted light images; ×40 magnification; scale bars, 100 μm). Bottom: high-power serosal view (inset, left) demonstrating tdTomato-positive crypt cells. tdTomato-negative areas in the crypts indicate long-lived Paneth cells that are not descendants of Tph1-derived ISCs (×100, scale bar, 20 μm). B, top: merged fluorescent and transmitted light images of Tph1-tdTomato (yellow) and Muc2 (Alexa-647, red). Bottom: merged fluorescent and transmitted light images of Tph1-tdTomato (yellow) and lysozyme (Alexa-647, red). Arrows point to cells expressing immunostained tdTomato. C: representative image of radiation-induced patch (mucosal villi-side view). D: reversion of Tph1-expressing EC cells to ISCs in ex vivo organoids. Left: representative merged fluorescent and transmitted light images of live organoids derived from a Tph1-CreERT2; Rosa26-tdTomato mouse and subsequently induced with TAM in vitro. First left: low-power image (×40, scale bar, 200 μm) of a Tph1-derived tdTomato -expressing ribbon formation that originated from a tdTomato-positive budding crypt. Second left: high-power image (×100, scale bar 50 μm) of the crypt in the boxed area (top) shows a continuous extension of Tph1-derived tdTomato (yellow)-expressing ribbon (red arrow). Right: whole mount ChgA-immunostained organoid shown in the live images (left). Transmitted light and immunofluorescent images of the crypt in the boxed area in A. Tph1-tdTomato, ChgA (Alexa-647) and nucleus (DAPI) appear in yellow, red and blue, respectively. tdTomato-labeld ChgA⁺ cells are noted with white arrows. EC, enterochromaffin cell; ISC, intestinal stem cell; TAM, tamoxifen.

Fig. 8.

Asymmetric cell division of Tph1-derived tdTomato-labeled cell at position 4. Images from 3 representative crypts isolated from two Tph1-CreERT2;Rosa26-tdTomato mice 2 days after TAM induction. Red and white arrows point to ChgA⁺ cells on the basal side and ChgA⁻ cells on the apical side of tdTomato-labeled asymmetrically dividing (or divided) cells, respectively. A, left, top: low-power view of a Tph1-derived cell doublet (yellow, within boxed area). Left, bottom: higher power views from 2 different confocal Z-positions of the boxed area resolves the doublets and demonstrates their +4 position next to Paneth cells (green arrows). Right: high-power merged fluorescent and transmitted light images of the area (left) Tph1-tdTomato (yellow), ChgA (Alexa-647, red) and nucleus (DAPI, blue). B, left, top: Tph1-derived cell doublets at +4 position (yellow, within boxed area). Left, bottom: high-power views from different confocal Z-positions of the boxed area that resolve the doublets and demonstrate their +4 position next to Paneth cells. Right, top: merged fluorescent image of the boxed area (left) for Tph1-tdTomato (yellow) and ChgA (Alexa-647, red). Right, bottom: fluorescent image of the boxed area (left) for nucleus (DAPI, blue). C, left, top: Tph1-derived cell doublets at +4 position (yellow, within boxed area). Left, bottom: high-power views from different confocal Z-positions of the boxed area that resolve the doublets and demonstrate their +4 position next to Paneth cells. Right, top: merged fluorescent image of the boxed area (left) for Tph1-tdTomato (yellow), ChgA (Alexa-647, red) and nucleus (DAPI, blue). Right, bottom: fluorescent image of the boxed area (left) for nucleus (DAPI, blue). D: model illustrating the timing and position of labeling during asymmetric division and dedifferentiation of +4 reserve ECs. EC, enterochromaffin cell; TAM, tamoxifen.

Taken together, these results from the Tph1 lineage-tracing studies suggest there is a subset of EC cells among the Tph1-derived cell population that can dedifferentiate to become a fully functional ISC capable of generating small intestinal epithelia under both basal physiological conditions as well in response to injury such as irradiation.

HopX-expressing EC cells have reserve stem cell potential.

Previously, we identified a population of ISC marker expressing EECs residing at and below the +4 position in murine small intestine that were distinct from the majority of ISC-marker negative EECs (45). Similarly, in human ileal mucosa, we identified a subset of EC cells that express reserve ISC genes HopX, Bmi1, and Lgr5^low (44). In addition, this subset of human EC cells was also located mostly at and below the +4 position and highly expressed NeuroD1 and Tph1 in addition to reserve ISC marker genes HopX, Bmi1, and Lgr5^low (44). Clusters of EC cells expressing +4 reserve stem cell markers HopX, Bmi1, and Lgr5^low form microtumors in the crypts of patients with SI-NETs (44). HopX-CreERT2;Rosa26-tdTomato mice mark predominantly +4 reserve ISCs early after TAM induction (typically less than 24 h; 25). To investigate further the potential of this population of EC cells to dedifferentiate to ISCs, we studied HopX expressing EC cells in Tph1-CFP, HopX-CreERT2;Rosa26-tdTomato mice 17 h after TAM induction and found a similar distribution of ChgA⁺/HopX⁺ cells at the +4 position (Fig. 9A). In fact, the majority of these +4 cells were HopX ⁺/Tph1⁺ EC cells (Fig. 9B). It is noteworthy that 57% of EC cells (ChgA⁺/Tph1⁺) at +4 position expressed HopX whereas only 16% of non-EC, EEC cells (ChgA⁺/Tph1^-) expressed HopX (Fig. 9B). To determine the potential of HopX-expressing EC cells to dedifferentiate to ISCs under basal conditions, FACS-sorted HopX ⁺/Tph1⁺ EC cells were assayed for organoid formation. Results from the organoid formation assay revealed entirely tdTomato-positive organoids, indicating that the organoids were derived from a single induced HopX-Cre-tomato cell and that HopX⁺/Tph1⁺ cells had the potential to dedifferentiate to fully functional ISCs (Fig. 9C). Moreover, organoid formation efficiency of HopX⁺/Tph1⁺ cells was significantly higher than that of HopX⁻/Tph1⁺ cells (Fig. 9D), suggesting that the HopX⁺ subset of EC cells has enhanced ability to dedifferentiate to ISCs among the EC cell population. To determine the reserve stem cell potential of the HopX⁺/Tph1⁺ cells, the same organoid formation assay was performed on FACS-sorted HopX ⁺/Tph1⁺ and HopX ⁻/Tph1⁺ ECs from mice exposed to abdominal irradiation (14 Gy). Results show that FACS-sorted HopX ⁺/Tph1⁺ cells from irradiated mice formed organoids significantly more efficiently than HopX ⁻/Tph1⁺ cells or HopX ^-/Tph1⁺ cells from untreated control mice (Fig. 9D). These results suggest that the HopX⁺ subset of EC cells, which reside predominantly at +4 position, possess reserve stem cell potential as demonstrated by their ability to dedifferentiate to ISCs and that injury further enhances this potential.

Fig. 9.

Distribution and reserve stem cell potential of HopX-expressing Tph1⁺ EC cells. A: representative images of HopX-expressing EC cells. HopX-tdTomato, ChgA (Alexa-647) and Tph1-CFP appear in yellow, red, and blue, respectively. A non-EC-enteroendocrine (HopX⁻/Tph1⁻/ChgA⁺) cell and a HopX⁺ EC cell (HopX ⁺/TPH1⁺/ChgA⁺, white arrow) were found at p2’ and p4, respectively. B: distribution of HopX ⁺/ChgA⁺/Tph1⁺, HopX ⁺/ChgA⁺/Tph1^-, HopX ⁻/ChgA⁺/Tph1⁺ and HopX⁻/ChgA⁺/Tph1⁻ cells along the crypt axis at 17-h post-TAM induction. The x- and y-axes indicate cell position along the crypt axis and number of cells per 100 crypts, respectively. Note that the count for positions higher than p10 (>p10) was from p11 to the top edge of isolated crypts. C: organoid formation from FACS-sorted HopX ⁺/Tph1⁺ EC cells. Left: representative images of a FACS-sorted HopX ⁺/Tph1⁺ EC cell that was identified by expression of HopX-tdTomato (yellow) and Tph1-CFP (blue). Right: representative images of an organoid that developed from a single HopX ⁺/Tph1⁺ EC cell and thus appears entirely tdTomato-positive (yellow). D: organoid formation potential of EC cells in relation to HopX-expression and radiation exposure. The graph shows percent organoid formation per seeded FACS-sorted cell type for HopX ⁺/Tph1⁺ and HopX⁻/Tph1⁺ cells from control (white bars) and irradiated mice (black bars) on day 4 after a single 14 Gy exposure (mean ±, *P < 0.05, **P < 0.01). EC, enterochromaffin cell.

DISCUSSION

Using both NeuroD1- and Tph1-based lineage-tracing mouse models, along with ChgA immunofluorescence, we present evidence for an important role of EECs in small intestinal stem cell dynamics. Early phase lineage tracing shows that the majority of NeuroD1-derived EECs are mature ChgA⁺ EECs most frequently located at +4 position and above position 10 in the crypts. A subset of the NeuroD1-derived EECs either remains at +4 position or migrates down to the crypt bottom rather than up the villus. These findings are consistent with our previous observations of the distribution and migration of EECs (45) and therefore support the validity of this lineage-based approach. Particularly noteworthy is a large proportion of the subset of NeuroD1-derived EECs that remained at +4 position and retained BrDU for more than 2 wk, thus satisfying the expected position and duration to be identified as +4 LRCs. We previously showed that this subset of EECs expresses ISC markers possibly because it is located within the stem cell niche (45). Longer-term (at least up to 10 wk) lineage-tracing studies showing migration streams of NeuroD1-derived differentiated epithelial cells, observed as ribbon-like streaks in vivo as well as in ex vivo organoids, indicate that NeuroD1-derived EECs can dedifferentiate to fully functional ISCs. The number of observed NeuroD1-derived ribbons increased with time after a single dose of TAM-induced labeling and fit well with a neutral drift model for ISCs. These findings indicate that NeuroD1-derived ISCs appear at a constant frequency and enter into neutral drift dynamics without either advantage or disadvantage compared with existing ISCs under basal conditions. Therefore, one may conclude that the intestinal epithelium is maintained in part by dedifferentiation of EECs, likely +4 EECs that express reserve ISC genes such as Bmi1 and HopX (44), to ISCs at a regular rate for basal maintenance of the intestinal epithelium.

Strictly speaking, the “reserve” nature of EECs should become apparent in response to injury or loss of actively cycling Lgr5⁺ ISCs as evidenced in +4 LRC or Bmi1⁺ cells (8, 51, 57). In fact, this reserve function of EECs was observed in response to abdominal irradiation. Although the number of labeled basal ribbons that originate from a single crypt (rS) was reduced, radiation induced the generation of NeuroD1-derived ribbons that originated from patches (rP). Examination of the patches from both the luminal and serosal surfaces revealed that the patches of tdTomato-labeled ribbons originated from 7 to 42 crypts. At least two possibilities may explain their origin. If each patch originated from a single labeled clone, then the labeled clone may have spread through crypt replication. If so, then a patch of 7–42 crypts would have undergone 2.8–5.6 cycles of replication within 7 wk (~9–17.5 days per crypt cycle). However, this 9-day cycle time is far shorter than previously reported for normal controls [107 days (9), 3.6 yr (26), and 14 yr (46)] and 3.6-fold shorter than observed following small intestinal resection (33 days; 9). Alternatively, patches may have formed from multiple clones because of a local field effect of unknown mechanism. Under this scenario, however, simulation of stem cell dynamics assuming a neutral drift model for 7–42 clones distributed in 7–42 crypts (one clone/crypt) would predict only 0–2 ribbons. To account for the observed >14–84 ribbons per patch within 7 wk from 7 to 42 clones per patch, simulation requires a stem cell compartment fully occupied with 16 labeled ISCs at least one week before observation. This would require sequential symmetric division with clonal expansion similar to that observed during small intestinal development using a model based upon optimal control theory and described as “bang-bang control” (16). In addition, this would assume complete loss or damage of Lgr5⁺ ISCs to allow converted labeled NeuroD1-derived clones to expand without (or greatly reduced) competition. Regardless of origin, this observed generation of multiple “nonphysiologic” NeuroD1-derived patches observed only in the irradiated mice suggests radiation-induced dedifferentiation of the radioresistant subset of NeuroD1-derived EECs to actively cycling ISCs and an active contribution of the subset to the intestinal epithelial repair process.

Recent accumulating evidence suggests that ISCs have tumor-initiating potential and thus play a critical role in tumorigenesis of intestinal tumors as reported for colorectal cancers (7, 30, 39, 52). Evidence from studies using combined mouse models of genetic lineage tracing and transformation revealed that aberrant Wnt activation can transform Lgr5⁺, Lrig1⁺, or Bmi1⁺ ISCs into tumor-initiating cells of origin for intestinal adenomas (3, 37, 41). Several recent studies also support this hypothesis in human colorectal cancers (2, 60). In our recent studies of normal human small intestine, we identified a subset of reserve ISC gene-expressing EC cells (44). Based upon the finding that early tumor formation in patients with familial SI-NETs was composed of similar EC cells expressing reserve ISC genes, we hypothesized that familial (and possibly sporadic) SI-NETs originate from this subset of cells (44). This same subset of EC cells also highly express NeuroD1 and Tph1 in addition to reserve ISC genes HopX, Bmi1, and Lgr5^low (44). In the present study, Tph1-lineage tracing demonstrated that Tph1-expressing EC cells can dedifferentiate to fully functional ISCs in vivo as well as in ex vivo organoids. In addition, FACS-sorted single HopX⁺/Tph1⁺ EC cells formed de novo organoids indicating their stem cell potential further increased following irradiation. Similarly, we found that Lgr5^low expressing EC cells can also dedifferentiate to ISC and form organoids (unpublished observation). These findings taken together support the stem cell potential through dedifferentiation of reserve ISC gene-expressing EC cells and their role(s) in small intestinal stem cell dynamics. This demonstrated potential also strengthens our assertion that the subset of the reserve ISC gene-expressing EC cells is the cell of origin for familial SI-NETs (10, 44).

The dedifferentiation of EECs, mostly +4 EECs, to ISCs is at least in part mediated by asymmetric cell division as shown in the primary crypts (Fig. 8). In support, the asymmetric division process generating actively cycling ISCs from quiescent ISCs has been previously reported (36, 38, 48). In 2002, Potten et al. (36) first captured a histological image of asymmetrically dividing +4 LRCs, which are now characterized to be secretory precursors (8) or EECs (24). Subsequently the spindle-orientation studies by Quyn et al. (38) showed asymmetrically dividing +4 LRCs along the apical-basal axis as we observed in the current study (Fig. 8). More recently, interconversion between Lgr5- and Bmi1-expressing ISCs through asymmetric cell division were proposed (48). The dividing cells evidenced in this report were asymmetrically expressing Lgr5 and Bmi1 (48). By taking into account recent findings that Bmi1 expressing cells in the crypts were mostly EEC lineage (11, 17, 58), it is highly likely that at least some of the observed Lgr5⁺/Bmi1⁺ pairs were Lgr5⁺/EEC doublets. The capturing of the dedifferentiaton of a subset of EECs as observed through the process of asymmetric division (Fig. 8, A–C) indicates that the present lineage tracing is of EEC cell-origin and not the result of nonspecific “leaky” expression from cycling ISCs. How this dedifferentiation process is coupled with the cell division mode remains an important biological question.

In summary, our results extend the recent findings by Gross et al. (14) and Yan et al. (58) beyond the Ngn3 stage of EECs to demonstrate that previously considered terminally differentiated EECs contribute to the ISC dynamics. The present study provides multiple lines of evidence from in vivo and ex vivo lineage tracings, kinetics, and organoid formation from FACS-sorted single cells, which show a subset of further differentiated EECs, including a particular subset of more specified mature EC cells, previously considered terminally differentiated, are capable of dedifferentiating to fully functional ISCs. This is achieved at least in part through asymmetric cell division. In fact, our data indicate that the EC cell is the predominate EEC with reserve stem cell capacity. The presence of a subset of these BrdU-retaining EECs residing at +4 position for more than 2 wk is consistent with their identity with previously described LRCs (24, 36). This subset of EECs contributes to basal homeostatic stem cell dynamics with a set reversion rate under normal conditions as well as in response to injury with a rate necessary to restore the epithelium in the small intestine. The exact cellular-molecular mechanisms of basal contribution to the ISC dynamics as well as patch formations in response to injury are important and require further investigation. Our findings provide novel insights into the so-called “+4 reserve ISC” hypothesis, the stem cell dynamics of the intestinal epithelium, and the development of EEC-derived tumors.

GRANTS

This work was supported by the Intramural Research Program of the National Institutes of Health and in part by National Institutes of Health Grants R01 DK-055732 (to J. Merchant).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Y.S. and S.A.W. conceived and designed research; Y.S., J.F., L.S., A.O.W., X.Z., S.Y., J.P.M., S.S., and M.M.H. performed experiments; Y.S., L.S., X.Z., S.Y., J.P.M., and S.A.W. analyzed data; Y.S. and S.A.W. interpreted results of experiments; Y.S. and S.A.W. prepared figures; Y.S. and S.A.W. drafted manuscript; Y.S., D.C., J.P.M., J.L.M., A.B.L., and S.A.W. edited and revised manuscript; Y.S. and S.A.W. approved final version of manuscript.