A Single Progenitor Population Switches Behavior to Maintain and Repair Esophageal Epithelium

See allHide authors and affiliations

Science  31 Aug 2012:
Vol. 337, Issue 6098, pp. 1091-1093
DOI: 10.1126/science.1218835


Diseases of the esophageal epithelium (EE), such as reflux esophagitis and cancer, are rising in incidence. Despite this, the cellular behaviors underlying EE homeostasis and repair remain controversial. Here, we show that in mice, EE is maintained by a single population of cells that divide stochastically to generate proliferating and differentiating daughters with equal probability. In response to challenge with all-trans retinoic acid (atRA), the balance of daughter cell fate is unaltered, but the rate of cell division increases. However, after wounding, cells reversibly switch to producing an excess of proliferating daughters until the wound has closed. Such fate-switching enables a single progenitor population to both maintain and repair tissue without the need for a “reserve” slow-cycling stem cell pool.

Murine esophageal epithelium (EE) consists of layers of keratinocytes. This tissue lacks structures such as crypts or glands that form stem cell niches in other epithelia (Fig. 1, A and B) (15). Proliferation is confined to cells in the basal layer (6). On commitment to terminal differentiation, basal cells exit the cell cycle and subsequently migrate to the tissue surface from which they are shed. Early studies suggested that all proliferating cells were functionally equivalent, but recent reports propose that a discrete population of slow-cycling stem cells is responsible for both maintenance and wound healing (711). This controversy and the importance of EE in disease motivated us to resolve the proliferative cell behavior in homeostatic EE and in tissue challenged by systemic treatment with the vitamin A metabolite all-trans retinoic acid (atRA) or acute local wounding (12, 13).

Fig. 1

Esophageal epithelium contains no slow-cycling epithelial cells. (A) Microendoscopy showing esophageal lumen; scale bar, ~500 μm. (B) Section of epithelium, basal layer (b), suprabasal layers (sb), and lumen (l); scale bar, 10 μm. (C) Protocol: Adult Rosa26M2rtTA/TetO-HGFP mice treated with doxycycline (DOX) express HGFP (green). After DOX withdrawal, HGFP is diluted upon cell division, except in slow-cycling cells. (D and E) Rendered confocal z stacks, showing HGFP (green) at time 0 (D) and after 4-week chase (E). Scale bar, 10 μm. Dashed line indicates basement membrane. Inset shows CD45 (red) staining in HGFP-retaining cell at 4 weeks. 4′,6-diamidino-2-phenylindole (DAPI), blue; scale bar, 5 μm.

To investigate cell division rates in EE, we used a transgenic label-retaining cell (LRC) assay (Fig. 1C) (1, 14, 15). Doxycycline (DOX) induction of the fusion protein Histone-2B enhanced green fluorescent protein (HGFP) expression in Rosa26M2rtTA/TetO-HGFP mice resulted in nuclear fluorescent labeling throughout the EE (Fig. 1D and fig. S1A). When DOX is withdrawn, HGFP is diluted by cell division, leaving 0.4% basal layer cells (561 out of 140,000) retaining label after a 4-week chase (Fig. 1E and fig. S1B). Three-dimensional imaging showed that these LRCs had smaller nuclei than the surrounding keratinocytes and did not stain for the basal keratinocyte marker Keratin14 (0 out of 561 LRCs) (fig. S1, C and D). The stem cell markers CD34 and Lgr5 were also undetectable in LRCs or other cells (figs. S2 and S3) (2, 4, 10, 16). However, 99.9% (2457 out of 2459) of LRCs were positive for the pan leukocyte marker CD45 (Fig. 1E, inset), comprising a mixture of Langerhans cells and lymphocytes (fig. S1, E and F). These findings lead to the unexpected conclusion that, unlike tissues such as the epidermis, there are no slow-cycling or quiescent epithelial stem cells in EE (1, 17). Indeed, HGFP dilution in basal cells was strikingly homogeneous, suggesting that all cells divide at a similar rate of about twice per week (fig. S1G).

Although epithelial cells have the same rate of division, they may still differ in their ability to generate cycling and differentiated progeny. We therefore used inducible cre-lox–based genetic marking to investigate whether the proliferating cell population is heterogeneous and to quantify cell behavior (18, 19). The fate of single-cell-derived clones was tracked in cohorts of adult AhcreERT R26flEYFP/wt mice at multiple time points over a year after induction, during which period EE was homeostatic (Fig. 2A and fig. S4). Crucially, analysis of the composition of clones at 1 year showed that they were representative of unlabeled cells (fig. S5). Over the time course, clone number decreased through differentiation, whereas the size of the remaining clones progressively increased (Fig. 2, B and C). Although variation in labeling efficiency limits the accuracy with which the proportion of labeled cells can be estimated, within statistical error, this proportion remains constant, which is consistent with the labeled population being in homeostasis (Fig. 2D).

Fig. 2

Proliferating cell fate in esophageal epithelium. (A) Protocol: Clonal labeling was induced in AhcreERTR26flEYFP/wt mice and analyzed at intervals from 3 days to 1 year (triangles). Images are rendered confocal z stacks of the basal layer showing typical clones at times indicated. Enhanced yellow fluorescent protein (EYFP), yellow; DAPI, blue. Scale bars, 10 μm. (B to D) Clone quantification. (B and C) Clone density and average clone size (basal cells). Observed values (orange) with error bars (mean ± SEM); green curves show predictions of model (E). (D) Average percentage of labeled basal cells at indicated time points (orange); error bars indicate mean ± SEM. Green line and shading show average and SEM across all time points. (E) Cell fate in EE. Basal layer comprises 65% functionally equivalent EP (green, dividing at a rate of 1.9/week, consistent with the rate of dilution of HGFP) (fig. S1G) and 35% postmitotic cells (pink), which stratify (arrow) at a rate of 3.5/week. Ten percent of EP divisions generate two EP daughters, 10% two differentiated daughters, and 80% one of each fate. Values are optimal fits with 95% plausible intervals.

Notably, the average size of persisting clones increased linearly with time, and their size distribution acquired long-term scaling behavior, a hallmark of a single functionally equivalent population of cells dividing at the same rate (Fig. 2C and fig. S6, A, B, and E) (1821). Studies of interfollicular epidermis (IFE) revealed that this pattern of clonal evolution was consistent with progenitors dividing stochastically to generate differentiated and cycling daughters with equal probability (18, 19). By implementing a Bayesian inference analysis, we showed that the entire data set conforms to the IFE paradigm (Fig. 2E; fig. S6, C to E; and supplementary theory). We conclude that esophageal progenitors (EP) are functionally equivalent.

The observation of similar progenitor behavior in EE and epidermis, derived from endoderm and ectoderm, respectively, argues that squamous epithelia share a common mechanism of homeostasis irrespective of their developmental origin. However, EP behavior differs from that of crypt stem cells in the endoderm-derived intestinal epithelium, where stochastic fate is a result of competition for limited niche space (16, 22).

Unlike progenitors in other tissues, such as the epidermis, EP are not supported by a discrete slow-cycling stem cell population (1). This raises the intriguing question of how the tissue responds to stress or injury. To investigate this issue, we subjected EE to a tissue-wide challenge in the form of atRA treatment and to acute local excisional wounding.

To determine the effects of atRA, we selected a dose that induced a “hyperproliferative” response (fig. S7A) and then used quantitative lineage tracing to define the changes in cell behavior (23, 24). Mice were treated for 9 days, clonal labeling was induced, and treatment then continued for a further 21 days, when clone size was scored (Fig. 3A). Bayesian analysis revealed that the rates of EP proliferation and differentiated cell stratification had approximately doubled. There was a small but statistically significant decrease in the proportion of proliferative cells but, critically, no significant change in the proportions of symmetric and asymmetric divisions, which indicated that the treated tissue was homeostatic (Fig. 3, B to D). To evaluate this finding, we used a second experimental schedule in which clonal labeling was induced before atRA treatment. The values of parameters determined in the first experiment accurately predicted the number of basal cells per clone on completion of the second protocol (fig. S7, B to D). We conclude that during atRA treatment, EP establish a new homeostatic state.

Fig. 3

All-trans retinoic acid (atRA) treatment of EE. (A) Protocol (see text). (B and C) Size distribution of multicellular clones containing at least one basal cell in control [(B), 307 clones] and atRA-treated [(C), 300 clones] EE. Green bars indicate 95% plausible fit to models in Figs. 2E (control) and 3D (atRA). (D) Optimal fit during atRA treatment; proliferation and differentiation rates (red) increase compared with control.

To investigate the repair of EP after wounding, we developed microendoscopic biopsy of mouse esophagus (fig. S8F). Biopsy produced a typical epithelial wound response (25, 26). Cells immediately next to the defect formed a migrating front (mf) in which there was minimal proliferation, surrounded by a proliferative zone (pz) in which cell division was dramatically increased (Fig. 4B and fig. S8D). We used three different protocols to analyze cell behavior (Fig. 4). First, we examined clonally labeled EP in AhcreERT R26flEYFP/wt mice induced 1 week before biopsy (Fig. 4, A to C). Twenty-four hours after wounding, fragmented clones of labeled cells were seen, aligned toward the wound and spanning the pz and mf (Fig. 4, A and B, and fig. S8D). By 10 days, clones were evident in and around the repaired defect (Fig. 4C and fig. S8, A and E). These findings indicate that EP participate in tissue regeneration after wounding, a behavior recapitulated in explant culture, suggesting that active recruitment of immune cells is not essential for the switch in EP fate (fig. S9) (27, 28). To investigate the proportion of EP that participate in regeneration, we biopsied DOX-treated Rosa26M2rtTA/TetO-HGFP mice (Fig. 4, D to F). HGFP was substantially and evenly diluted within the pz at 2 and 5 days after biopsy compared with controls but was retained outside the mf (Fig. 4, D to F; fig. S8B; and fig. S10, A and B). We conclude that there is widespread mobilization of EP within the pz and that the recruited cells proliferate to a similar extent. In a complementary experiment, animals were injected with EdU (5-ethynyl-2′-deoxyuridine) 24 hours before culling, revealing extensive recruitment of cells into cycle in the pz at 2 days, which reverted to control levels at 5 days when the epithelial defect had closed (Fig. 4, G to I; fig. S8C; and fig. S10, C to F). This indicates that the switch in EP fate after wounding is reversible.

Fig. 4

Response of EP to wounding. Cartoons show protocols; blue triangles indicate sampling. (A to C) Wounding of clonally labeled mice. Confocal z stacks, 1 (B) or 10 (C) days after biopsy. Solid line shows pz-mf boundary; dashed line shows wound margin. (A) Day 1 unwounded control. EYFP is yellow, keratin 14 (Krt14) red, and EdU grayscale. Scale bars, 50 μm. (D to F) Dilution of HGFP. Confocal z stacks from unwounded control day 2 (D) and wounded mice at 2 (E) and 5 (F) days after biopsy, showing HGFP (green). Arrow indicates HGFP bright cell (overexposed to reveal remaining cells); such cells stain for CD45 (red, inset). Scale bars, 10 μm. (G to I) Cell proliferation. Confocal z stacks from unwounded control at 2 days (G) and experimental mice at 2 (H) and 5 (I) days after biopsy stained for EdU (grayscale). Scale bars, 10 μm.

In summary, these results show that EE is both maintained and repaired by a single progenitor cell population capable of reversibly switching between homeostatic and regenerative behavior in response to injury. These findings may be reconciled with the reported proliferative heterogeneity of EE cells in vitro if only some EP cells switch into “wound mode” when placed into culture (10). The widespread participation of progenitors in tissue repair provides a rapid and robust mechanism of wound healing without an underpinning stem cell pool.

Supplementary Materials

Materials and Methods

Figs. S1 to S13

Table S1

References (29, 30)

References and Notes

  1. Acknowledgments: We thank E. Choolun and the staff at ARES and CBS Cambridge for technical assistance; D. Winton and D. Adams (Cambridge) for mice; and M. Gonzalez (London) for the Geminin antibody. We acknowledge the support of the MRC, EPSRC (Engineering and Physical Sciences Research Council), the NC3Rs (National Centre for the Replacement, Refinement and Reduction of Animals in Research), the Wellcome Trust, Sidney Sussex College, Cambridge (D.P.D.), European Union Marie Curie Fellowship PIEF-LIF-2007-220016 (M.P.A.), the Royal College of Surgeons of England (A.R.), and Cambridge Cancer Centre (A.R.). This work uses methods included in the patent WO2009010725 (A2), a method of detecting altered behavior in a population of cells; inventors were P.H.J., B.D.S., and A.M.K.
View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article