Article Text

other Versions


Original article
Identification and genetic manipulation of human and mouse oesophageal stem cells
  1. Youngtae Jeong1,
  2. Horace Rhee1,2,
  3. Shanique Martin1,
  4. Daniel Klass1,
  5. Yuan Lin1,
  6. Le Xuan Truong Nguyen1,
  7. Weiguo Feng1,
  8. Maximilian Diehn1,3
  1. 1Stanford Cancer Institute and Institute for Stem Cell Biology and Regenerative Medicine, Stanford University School of Medicine, Stanford, California, USA
  2. 2Division of Gastroenterology and Hepatology, Stanford University School of Medicine, Stanford, California, USA
  3. 3Department of Radiation Oncology, Stanford University School of Medicine, Stanford, California, USA
  1. Correspondence to Dr Maximilian Diehn, Institute for Stem Cell Biology & Regenerative Medicine, Stanford University, 875 Blake Wilbur Dr, Stanford, CA 94305, USA; diehn{at}


Objective Human oesophageal stem cell research is hampered by the lack of an optimal assay system to study self-renewal and differentiation. We aimed to identify and characterise human and mouse oesophageal stem/progenitor cells by establishing 3-dimensional organotypic sphere culture systems for both species.

Design Primary oesophageal epithelial cells were freshly isolated and fluorescence-activated cell sorting (FACS)-sorted from human and mouse oesophagus and 3-dimensional organotypic sphere culture systems were developed. The self-renewing potential and differentiation status of novel subpopulations were assessed by sphere-forming ability, cell cycle analysis, immunostaining, qPCR and RNA-Seq.

Results Primary human and mouse oesophageal epithelial cells clonally formed esophagospheres consisting of stratified squamous epithelium. Sphere-forming cells could self-renew and form esophagospheres for over 43 passages in vitro and generated stratified squamous epithelium when transplanted under the kidney capsule of immunodeficient mice. Sphere-forming cells were 10–15-fold enriched among human CD49fhiCD24low cells and murine CD49f+CD24lowCD71low cells compared with the most differentiated cells. Genetic elimination of p63 in mouse and human oesophageal cells dramatically decreased esophagosphere formation and basal gene expression while increasing suprabasal gene expression.

Conclusions We developed clonogenic and organotypic culture systems for the quantitative analyses of human and mouse oesophageal stem/progenitor cells and identified novel cell surface marker combinations that enrich for these cells. Using this system, we demonstrate that elimination of p63 inhibits self-renewal of human oesophageal stem/progenitor cells. We anticipate that these esophagosphere culture systems will facilitate studies of oesophageal stem cell biology and may prove useful for ex vivo expansion of human oesophageal stem cells.


Statistics from

Significance of this study

What is already known on this subject?

  • The human and mouse oesophagus is lined by stratified squamous epithelium, comprised of basal and suprabasal cells.

  • Stem cells reside within the basal layer and are able to proliferate and maintain the oesophageal epithelial integrity. However, the precise identity and hierarchy of oesophageal stem and differentiated cells are not clearly understood.

What are the new findings?

  • We present novel three-dimensional, clonogenic and organotypic sphere culture systems for human and mouse oesophageal stem/progenitor cells.

  • Oesophageal stem cells are most enriched in CD49fhiCD24low cells in human and CD49f+CD24lowCD71low cells in mouse.

  • p63 regulates the self-renewal and differentiation of human and mouse oesophageal stem cells.

How might it impact on clinical practice in the foreseeable future?

  • Identification and in vitro manipulation of human oesophageal stem/progenitor cells in this culture system will enhance our understanding of the role of these cells in normal physiological and pathological processes such as inflammation and cancer. Ultimately, these assays may allow ex vivo expansion of human oesophageal stem cells for regenerative medicine applications.


The digestive epithelia are maintained by tissue resident stem cells through the balance of self-renewal and terminal differentiation. Unique within the digestive tract, the oesophagus is lined by a stratified squamous epithelium, which is maintained by a basal layer of proliferating cells adherent to an underlying basement membrane. When committed to differentiate, these cells give rise to suprabasal cells by exiting the cell cycle, detaching from the basement membrane, and moving towards the oesophageal lumen. Suprabasal cells undergo a defined programme of terminal differentiation and provide the oesophagus with a protective epithelial barrier against ingested contents.1 ,2

Although oesophageal stem cells are thought to reside within the basal layer, it remains unclear whether only a subpopulation of basal cells or every basal cell can serve as a stem/progenitor cell. Differences in cell kinetics and nucleotide label retention initially suggested that only a subpopulation of oesophageal basal cells were slow-cycling, long-lived stem cells that had the capacity to self-renew in human and mouse.3–5 However, a pulse-chase study using a histone label found no quiescent label retaining cells in mouse oesophageal epithelium.6 Furthermore, fate mapping of individual basal cells revealed a pattern consistent with a single progenitor population, suggesting that all basal cells are functionally equivalent. Similar findings were also observed in another stratified epithelium, the interfollicular epidermis.7 However, clonal fate mapping found two distinct progenitors that can differentially modulate epidermal homoeostasis and repair within the basal cell layer of the skin.8 Whether such cellular heterogeneity also exists within oesophageal basal cells remains incompletely explored, particularly in the human oesophagus. This lack of clarity was recently illustrated by a study that found regional differences in mitosis and cell marker expression across the human oesophageal basal layer, but no differences in their clonogenic, self-renewal and reconstitutive potentials.9

The delineation of putative oesophageal stem cells would be facilitated by the development of clonogenic and organotypic assays that allow for the prospective identification and functional analysis of basal cell subpopulations. Classic colony formation assays performed on adherent cultures lack the normal three-dimensional (3D) tissue architecture, which can limit their interpretation. Although organotypic cultures of oesophageal epithelial cells have previously been described,10–13 most of these models assess tissue reconstitution without quantitative determination of clonogenicity.

Here we report the establishment of 3D, clonogenic and organotypic ‘esophagosphere’ culture systems for primary human and mouse oesophageal epithelial cells. We find that oesophageal epithelial cells are heterogeneous in their ability to form clonally derived spheres. Furthermore, we identify cell surface markers that enrich for sphere-forming basal cells that are capable of self-renewal and can give rise to fully differentiated suprabasal cells. Transcriptional profiling revealed that sphere-forming basal cells are enriched for stem cell gene sets. These esophagospheres can be genetically manipulated in culture to study effects on self-renewal and differentiation, and represent an ideal tool for oesophageal stem cell research.

Materials and methods

Full methods are available in online supplementary materials and methods.

Human oesophageal epithelial cell isolation, FACS and culture

Under institutional review board approved protocols, oesophagus biopsy samples were obtained from routine endoscopy and incubated in dispase for 30 min. If present, the submucosa was removed and the mucosal epithelium was trypsinised for 5 min at 37°C, prepared for culture as mentioned in online supplementary materials and methods for mouse oesophagus, and cultured in CnT-57 progenitor media and CnT-02-3DP5 sphere media (CELLnTEC). Cells were stained with antihuman CD24, CD49f and EpCAM.14


Establishment of an organotypic sphere culture system for primary human squamous oesophageal stem/progenitor cells

The identification of human oesophageal stem cells has been hampered by the lack of an ideal model system. We sought to establish an organotypic culture system of primary human squamous oesophageal cells that could recapitulate a stratified epithelium, and quantitatively analyse the clonogenicity and self-renewal of putative oesophageal stem cells. When isolated from endoscopic biopsies and seeded into Matrigel, primary human oesophageal epithelial cells formed 3D structures in the absence of stromal cells. Using defined media, we established two separate culture conditions that promote either stem/progenitor cell expansion (Progenitor media) or growth of organotypic spheres (Sphere media). In progenitor media, human oesophageal cells grew into amorphous clusters largely composed of uniform cells that expressed basal cell markers Krt14, CD49f and p63 (figure 1A, B). When switched or directly plated into sphere media, these oesophageal cells gave rise to ‘esophagospheres’ consisting of circumferential stratified squamous epithelium surrounding a central core of keratinised cells. The outermost cells expressed squamous basal markers of Krt14, CD49f and p63, while the inner cell layers expressed the suprabasal marker Krt4. From single plated cells, esophagospheres expanded in size up to 300 μm in diameter and reached full differentiation over 2–3 weeks. The overall efficiency of differentiated sphere formation from bulk-sorted human oesophageal epithelial cells was approximately 2.5% (figure 2B). Cell counting at ∼14 days after plating in progenitor media enabled more than sevenfold expansion of oesophageal cells (figure 1C). Thus, a single esophagosphere-forming cell can be expanded up to ∼300-fold under stem/progenitor condition. Furthermore, passaged esophagosphere cells continued to self-renew and resembled primary esophagospheres by morphology (figure 1D). Thus, human esophagospheres are organotypic, reflecting the morphological and histological features of human oesophageal epithelium.

Figure 1

Characterisation of esophagosphere formation by human oesophageal epithelial cells. (A) Immunostaining of human oesophageal epithelium. Basal cells and suprabasal cells express CD49f and CD24, respectively (size of bar=100 µm). (B) Primary human oesophageal epithelial cells from endoscopic biopsies cultured in two different culture media (progenitor media or sphere media) at day 15 after seeding. Images show microscopic bright field, H&E staining and immunohistochemical staining of esophagospheres. Human esophagospheres express basal markers (KRT14, CD49f, P63) and suprabasal markers (KRT4) (size of bar=100 µm). (C) Relative number of initially plated and subsequently harvested cells. After 15–20 days in culture, human esophagospheres were dissociated into single cells, and cell numbers were counted. Data are presented as mean fold-change±SEM (N=3, **p<0.01). (D) Human esophagospheres passaged into progenitor and sphere media (size of bar=100 µm).

Figure 2

Esophagosphere-forming cells are enriched in CD49fhiCD24low basal cells. (A) FACS analysis of primary human oesophageal epithelial cells using CD49f and CD24. (B) Sphere-forming efficiency of unsorted (▾), CD49fhiCD24low (•), CD49flowCD24low (▪), and CD49flowCD24hi (▴) cells. (C) Expression of basal and suprabasal cytokeratins in unsorted (white), CD49f+CD24low (black), CD49flowCD24low (dark grey) and CD49flowCD24hi (light grey) cells assessed by qPCR. (D) FACS analysis of dissociated esophagosphere cells derived from unsorted and CD49fhiCD24low cells in progenitor media and sphere media. (E) Cell cycle analysis of CD49fhiCD24low, CD49flowCD24low and CD49flowCD24hi cells. (F) Schematic model of human oesophageal stem cell differentiation. Oesophageal stem cells (SC; CD49fhiCD24low cells) differentiate through early stage suprabasal differentiated cells (Early SD; CD49flowCD24low cells) into late stage suprabasal differentiated cells (Late SD; CD49flowCD24hi cells).

Identification of human oesophageal stem/progenitor cells

Since only a fraction of isolated oesophageal epithelial cells are able to form esophagospheres, we next sought to identify cell surface markers that could be used to enrich for sphere-forming cells from human oesophagus. In immunofluorescence analysis of human oesophageal tissue sections, while CD49f was strongly expressed only within the basal layer, CD24 exclusively stained suprabasal layers, suggesting that CD49f and CD24 could differentially label basal and suprabasal cells (figure 1A). FACS analysis of viable epithelial cells reflected this heterogeneous expression of CD24 and CD49f, with three distinct subpopulations, CD49fhiCD24low, CD49flowCD24low and CD49flowCD24hi (figure 2A). Clonogenic analysis of sorted cells from these subpopulations revealed that CD49fhiCD24low cells had the highest sphere-forming capacity with a frequency of ∼8%, eightfold more than CD49flowCD24hi cells, and expressed the highest level of basal marker (Krt14) and lowest level of suprabasal marker (Krt4) relative to the other subpopulations (figure 2B, C). Furthermore, sorted CD49fhiCD24low cells cultured in sphere media gave rise to CD49flowCD24hi cells (figure 2D). In contrast, CD49flowCD24hi cells expressed the highest level of Krt4 and the lowest level of Krt14 and gave rise to few spheres. In addition, most spheres from CD49flowCD24hi cells did not fully mature and appeared abortive, suggesting they were initiated by cells without long-term proliferative potential (data not shown). CD49flowCD24low cells formed spheres at the frequency of ∼2.5% and showed intermediate levels of Krt14 and Krt4 gene expression.

To further characterise these human oesophageal epithelial subpopulations, we performed cell cycle analysis. While CD49fhiCD24low and CD49flowCD24low cells contained a significant portion of cycling cells, CD49flowCD24hi cells appear to have exited from the cell cycle, consistent with being fully differentiated cells (figure 2E and see online supplementary figure S1A). Taken together, a lineage hierarchy appears to exist in human oesophageal epithelium with CD49fhiCD24low basal cells enriched for stem/progenitor cells, CD49flowCD24low cells and CD49flowCD24hi cells likely consisting of early and late stage differentiated suprabasal cells, respectively (figure 2F).

Establishment of an organotypic sphere culture system for primary mouse squamous oesophageal stem/progenitor cells

In order to be able to perform cross-species studies of pathways important in oesophageal stem/progenitor cells and to take advantage of mouse genetic techniques, we also optimised an organotypic esophagosphere culture system for primary murine oesophageal cells. Like human oesophageal cells, primary mouse oesophageal epithelial cells gave rise to esophagospheres with the characteristic expression of basal and suprabasal cell markers found in stratified squamous epithelium (figure 3A). To demonstrate that these esophagospheres are clonally derived from single epithelial cells, we sorted equal numbers of oesophageal epithelial cells from green fluorescent protein (GFP)-expressing or red fluorescent protein (RFP)-expressing mice. When plated together in esophagosphere culture conditions, 99% of 374 spheres were monochromatic (187 GFP+, 183 RFP+). Of the four spheres (1.1%) positive for GFP and RFP, all consisted of two monochromatic regions, suggesting that they arose from two spheres that merged (figure 3B, C). Additionally, we demonstrated that single oesophageal epithelial cells plated in esophagosphere culture conditions lead to the formation of single esophagospheres. Taken together, these data indicate that esophagospheres are clonally derived from single epithelial cells.

Figure 3

Characterisation of esophagosphere formation by mouse oesophageal epithelial cells. (A) Mouse esophagospheres at day 15 after seeding (size of bar=100 µm). Images show microscopic bright field, H&E staining and immunohistochemical staining of esophagospheres. Esophagospheres were stained for basal markers (P63, CD49f, Krt14) and the suprabasal marker (Krt4). (B) Schematic illustration of potential outcomes from mixing experiment. (C) Outcome of mixing experiment. (Lt.) Number of spheres with green, red and mixed colour. (Rt.) Fluorescence and bright field images of esophagospheres derived from equal numbers of co-cultured GFP+ and RFP+ oesophageal epithelial cells. (D) Schematic illustration of serial passaging of esophagosphere-forming cells. (E) Number of esophagospheres formed per number of oesophageal epithelial cells that were seeded. (F) Subrenal transplantation of oesophageal epithelial cells from esophagospheres. GFP+ cells were cultured under esophagosphere conditions, dissociated, and transplanted under the kidney capsule. Kidneys were harvested after 8 weeks and examined histologically. H&E staining and immunostaining for GFP and Krt14 are shown (size of bar=100 µm).

Murine esophagosphere forming cells self-renew and give rise to epithelial structures in vivo

To explore whether esophagosphere forming cells can self-renew, cultured esophagospheres were enzymatically dissociated into single cells and serially passaged (figure 3D). Passaged esophagospheres resembled primary esophagospheres by morphology and histological expression of basal and suprabasal markers (data not shown). The overall efficiency of sphere formation from FACS-purified primary murine oesophageal epithelial cells was approximately 20% (figure 3E) and was similarly maintained in bulk cells purified from cultured esophagospheres (see online supplementary figure S2A). Furthermore, our culture system resulted in an approximately 65-fold expansion of seeded cell number after ∼15 days (see online supplementary figure S2B). Considering about 20% sphere-forming efficiency, estimated expansion of a single oesophageal cell is about 325-fold per passage. To date, murine esophagospheres have been serially propagated for more than 43 passages, with preservation of morphological and histological appearance of esophagospheres (figure 3D and data not shown).

In order to evaluate if cells dissociated from passaged esophagospheres can generate epithelial structures in vivo, we generated esophagospheres from GFP-expressing mice and transplanted single cells dissociated from esophagospheres under the kidney capsule of congenic mice. After 8 weeks, GFP+ stratified squamous epithelia were present as sheet-like or spherical structures at the transplantation site, indicating that passaged esophagosphere cells retained the capacity to reconstitute a stratified squamous epithelium (figure 3F and see online supplementary figure S2C). Collectively, these data demonstrate that esophagosphere forming cells have self-renewal capacity and maintain differential potential of stem cells.

Identification of a murine oesophageal stem/progenitor cell hierarchy

A previous study suggested that mouse oesophageal stem cells are enriched in the CD49f+CD71low cell fraction while CD49f+CD71hi cells consist of differentiated suprabasal cells.15 We therefore tested whether CD49f+CD71low cells were enriched for esophagosphere-forming cells. Sorted CD49f+CD71low cells formed higher numbers of esophagospheres than CD49f+CD71mid or CD49f+CD71hi cells (see online supplementary figure S3A, B). However, histological analysis of CD71 expression revealed that CD71 was not expressed by a large fraction of suprabasal cells and that a fraction of basal cells expressed CD71 (see online supplementary figure S3C). This suggested that some CD71low cells observed by flow cytometry were likely differentiated cells and that some CD71+ cells were likely basal cells.

We therefore attempted to identify an alternative surface marker that was more uniformly expressed by suprabasal but not basal cells and found that, similar to the human oesophageal epithelium, CD24 was exclusively expressed in murine suprabasal layers (figure 4A). CD49f+CD24low cells formed higher numbers of esophagospheres and displayed higher levels of basal marker expression (Krt5 and Krt14) and lower levels of suprabasal markers (Krt4 and Krt13) than CD49f+CD24mid or CD49f+CD24hi cells (figure 4B and see online supplementary figure S4A, B). To further validate the utility of CD24 as a marker, we next asked whether CD24 expression could subfractionate the previously reported CD49f+CD71low subpopulation (see online supplementary figure S4C) and found that CD49f+CD71lowCD24low cells formed higher numbers of esophagospheres and displayed lower levels of suprabasal markers than CD49f+CD71lowCD24mid/hi cells (figure 4C and see online supplementary figure S4D). In order to determine the fold enrichment of esophagosphere-forming cells within the various populations, we compared the esophagosphere-forming ability of CD49f+CD24low, CD49f+CD71low, CD49f+CD24lowCD71low and bulk Lin-CD49f+ epithelial cells. Esophagosphere-forming cells were approximately 2.1-fold, 1.6-fold and 1.4-fold enriched in CD49f+CD24lowCD71low cells compared with the bulk Lin-CD49f+, CD49f+CD24low and CD49f+CD71low populations, respectively (figure 4D). Additionally, single cell sorting of CD49f+CD24lowCD71low cells led to the formation of 29 single esophagospheres from 96 single CD49f+CD24lowCD71low cells (see online supplementary figure S4E). Taken together, these data indicate that CD49f+CD24lowCD71low cells are most enriched for esophagosphere-forming cells with approximately ∼30% of this population exhibiting sphere-forming capacity.

Figure 4

Mouse oesophageal stem cells are enriched in CD49f+CD24lowCD71low basal cells. (A) Immunostaining of murine oesophageal epithelium. CD24 (red) and Krt14 (green). (B) CD49f+CD24low cells are enriched for esophagosphere-forming cells. Number of esophagospheres formed by CD49f+CD24low (black), CD49f+CD24mid (grey) and CD49f+CD24hi cells (white) relative to CD49f+CD24low cells, scaled to 100. Data are presented as mean±SEM (N=3, *p<0.05, **p<0.01). (C) Number of esophagospheres formed from CD71lowCD24low and CD71lowCD24mid/hi cells (N=3). (D) Comparison of sphere-formation between bulk sorted, CD49+CD24low, CD49+CD71low, CD49+CD24lowCD71low CD49+CD24midCD71mid and CD49+CD24hiCD71hi cells. Number of esophagospheres formed relative to bulk sorted epithelial cells, scaled to 100 (N=4). (E) Relative number of esophagospheres from CD49f+CD71hiCD24low and CD49f+CD71hiCD24hi cells (N=3). (F) Cell cycle analysis of CD49f+CD24lowCD71low, CD49f+CD24midCD71mid and CD49f+CD24hiCD71hi cells. Data are presented as mean±SEM (N=3, **p<0.01).

In contrast, the most differentiated cells in our analysis were found in the CD49f+CD24hiCD71hi population. CD49f+CD24hiCD71hi cells formed four times fewer esophagospheres and showed higher levels of Krt4 and Krt13 and lower levels of Krt5 and Krt14 expression than CD49f+CD24low/midCD71hi cells (figure 4E and see online supplementary figure S5A). When compared with the subpopulation of most enriched stem/progenitor cells (CD49f+CD24lowCD71low cells), CD49f+CD24hiCD71hi cells formed 15-fold fewer esophagospheres (figure 4D). These data suggest that CD49f+CD24hiCD71hi cells are the most differentiated oesophageal cells that can be obtained by FACS sorting and that CD24 is an informative marker for enriching for murine oesophageal stem/progenitor cells and differentiated cells.

Despite significant differences in esophagosphere formation between the sorted cell populations, they all express basal epithelial cell markers as analysed by immunofluorescence on cytospin cells. Notably, there was no difference between CD49f+CD24lowCD71low and CD49f+CD24midCD71mid cells with respect to the number of cells positively stained by the basal makers K14 and CD104 (see online supplementary figure S6A, B). However, CD49f+CD24lowCD71low cells were fourfold less proliferative than CD49f+CD24midCD71mid cells, and could give rise to differentiated CD24hi and CD71hi cells, when cultured for esophagospheres (figure 4F and see online supplementary figure S7). Thus, our data suggest that functional differences are present within basal oesophageal epithelial cells and support a model in which the oesophageal epithelium is maintained by CD49f+CD24lowCD71low slow cycling stem cells, which differentiate into CD49f+CD24midCD71mid progenitor cells, and ultimately, CD49f+CD24hiCD71hi differentiating cells (see online supplementary figure S7C).

Transcriptional profiling of murine oesophageal stem cells

To identify genes that may regulate the self-renewal and differentiation of oesophageal stem cells, we performed RNA-sequencing on our sorted cell populations and compared their gene expression profiles. Unsupervised hierarchical clustering of global gene expression levels between CD24lowCD71low, CD24midCD71mid and CD24hiCD71hi cell populations showed concordance between paired replicates (figure 5A). As expected from our prior marker analysis, CD24lowCD71low cells clustered more closely with CD24midCD71mid cells compared with CD24hiCD71hi cells. To identify differentially expressed genes, we used the R DESeq package.16 At a twofold cut-off and corrected p value threshold of <0.01, 290 genes were upregulated and 240 genes were downregulated in CD49f+CD24lowCD71low cells compared with CD49f+CD24hiCD71hi cells (see online supplementary tables S1 and S2).

Figure 5

Transcriptional profiling of murine oesophageal stem cells. (A) Comparative heatmap of CD49f+CD24lowCD71low, CD49f+CD24midCD71mid and CD49f+CD24hiCD71hi cells. (B) Representative data sets derived from gene set enrichment analysis of CD49f+CD24lowCD71low, CD49f+CD24midCD71mid and CD49f+CD24hiCD71hi cells. (C) Hierarchical clustering of representative differential gene sets. (D) Confirmation qPCR of representative upregulated or downregulated genes in CD49f+CD24lowCD71low cells. (E) Immunoblot confirmation of representative upregulated or downregulated gene (Pdpn and Aqp5) in CD49f+CD24lowCD71low cells.

To gain further functional insight into the differentially expressed genes between CD49f+CD24lowCD71low cells and CD49f+CD24hiCD71hi cells, we performed gene set enrichment analysis.17 Consistent with their stem cell features in our esophagosphere assays, CD24lowCD71low cells displayed overexpression of gene sets involved in the Wnt pathway, mammary stem cells and epithelial mesenchymal transition, while gene sets of epithelial differentiation and mammary luminal mature cell differentiation were overexpressed in CD49f+CD24hiCD71hi cells (figure 5B, C, see online supplementary tables S3 and S4). Quantitative RT-PCR and western blot confirmed these results on selected upregulated (Pdpn and Lrig) and downregulated (Aqp5 and Erbb2) genes (figure 5D, E, and see online supplementary figure S8A). The RNA-seq profiles of oesophageal stem/progenitor cells will serve as a useful research tool for future studies examining pathways important for self-renewal and differentiation of these cells.

Role of p63 in human and mouse oesophageal stem cells’ self-renewal

The functional characterisation of candidate genes identified in gene expression data is often limited by lack of suitable model systems, particularly in humans. Thus, we next wished to demonstrate that our organotypic esophagosphere system could be used to evaluate genes and modulate pathways identified in our gene expression analyses. The p63 transcription factor was found to have a high rank metric score in the ‘mammary stem cell up’ gene set, and enriched in CD24lowCD71low stem cells (figure 5B). p63, a critical regulator of epithelial stem cells, modulates proliferative capacity and differentiation into terminal lineages.18–22 Deletion of p63 during embryonic morphogenesis inhibits epithelial stratification, and the oesophagus acquires a columnar phenotype losing traditional squamous markers.23 However, the role of p63 in human and mouse adult oesophageal stem cells has not been reported, likely due to lack of a suitable assay for the former and lack of promoters specific for oesophageal basal progenitor cells for the latter.

We therefore analysed oesophageal epithelial cells from conditional p63 knockout mice regulated by an inducible Krt5-CreER driver, along with a Rosa26-YFP reporter to track recombined cells. The oesophageal stem/progenitor cells isolated from conditional p63 knockout mice and treated with tamoxifen in culture had fourfold fewer YFP-positive recombined cells, compared with their heterozygous or wild type littermates (figure 6A). Likewise, when YFP-positive recombined cells were FACS-sorted and plated in sphere-forming conditions, 6.6-fold fewer spheres were formed by p63 knockout cells compared with heterozygous controls (figure 6B), suggesting decreased self-renewal of sphere-forming cells. In addition, recombined p63 knockout cells were unable to form structurally intact esophagospheres in differentiating culture conditions, instead generating clusters of cells with loss of intercellular contacts that failed to mature into a stratified epithelium (figure 6C).

Figure 6

Role of p63 in human and mouse oesophageal stem cells. (A) Percentage of YFP-positive recombined CD49f+ cells from p63 conditional knockout and control esophagospheres. Esophagospheres from Krt5-CreER; Rosa26-YFP; p63floxed and Krt5-CreER; Rosa26-YFP; p63control were treated with tamoxifen and analysed by FACS 14 days later. (B) Secondary sphere formation from sorted YFP-positive recombined cells dissociated from p63 conditional knockout and control esophagospheres relative to YFP-negative non-recombined cells. Results from two independent sets of experiments were averaged. (C) Microscopic and histological appearance of tamoxifen-treated p63 conditional knockout and control esophagospheres in primary culture (size of bar=100 µm). (D) Relative number of control-silenced (black) or p63-silenced (white) human esophagospheres with the number of control esophagospheres scaled to 100. Data are presented as mean±SEM (N=4, ***p<0.001). (E) Relative expression of basal and suprabasal cytokeratins in control (black) and p63-silenced human esophagosphere cells assessed by quantitative PCR.

To analyse the importance of p63 in primary human oesophageal stem cells, we combined lentiviral shRNA knockdown with our esophagosphere culture system. Efficient silencing of p63 in primary human oesophageal cells was confirmed by RT-PCR (see online supplementary figure S9A). Upon puromycin selection, silencing of p63 led to fewer esophagospheres (figure 6D). Furthermore, p63 silencing decreased basal marker (Krt5 and Krt14) and increased suprabasal marker (Krt4 and Krt13) expression (figure 6E). Taken together, these data show that p63 regulates the self-renewal and differentiation of oesophageal stem cells in humans and mice. Furthermore, these data demonstrate that esophagosphere culture system is a powerful assay system for the study of human and mouse oesophageal stem cell biology.


In this study, we report novel, 3D, organotypic culture systems for human and mouse primary oesophageal epithelial cells. In addition to recapitulating a stratified squamous architecture, esophagospheres were clonally derived from single progenitor cells and were capable of self-renewal. The esophagosphere culture systems enabled us to identify cell surface markers for putative human and mouse oesophageal stem cells as well as to perform deeper characterisation using cell cycle analysis and RNA-seq. Finally we demonstrated that the p63 regulates the self-renewal and differentiation of human and mouse oesophageal stem cells.

In our culture systems, esophagosphere-forming cells displayed many characteristics of bona fide stem cells. Through mixing studies and single cell sorting, we demonstrated that esophagospheres are clonally derived, and give rise to immature and differentiated cells that recapitulated the diversity of the original epithelial cell populations. We further showed that murine esophagosphere-forming cells have long-lived potential to self-renew for greater than 43 passages. Since a robust, orthotopic transplantation assay for oesophageal stem cells is lacking, we also investigated the ability of esophagosphere-forming cells to regenerate in vivo epithelial structures after transplantation under the renal capsule. Passaged self-renewing, esophagosphere-forming cells could reconstitute a stratified squamous epithelium in vivo without the addition of stromal cells, suggesting that at least a fraction of these cells are oesophageal stem cells. These data suggest that the esophagosphere system can serve as in vitro surrogate of in vivo stem cell activity, as has previously been reported for neural, mammary and prostate stem cells.24–26

Recent studies have come to opposing conclusions regarding the cell subpopulation that maintains the oesophageal epithelium. Several studies have pointed to the existence of a discrete population of slow-cycling stem cells,3–5 ,9 ,27 while another study provided evidence for the existence of a single population of oesophageal cells that divide stochastically to generate proliferating and differentiating daughters with equal probability.6 While our study was not designed to address this controversy, our findings are consistent with the existence of heterogeneity within murine oesophageal basal cells. In our analysis of sorted murine oesophageal basal cells, we find a unique subpopulation (CD49f+CD24lowCD71low) that is most enriched for esophagosphere formation and contains a lower fraction of cycling cells, indicating the existence of heterogeneity in sphere-forming ability and cell cycle status among oesophageal basal subpopulations.

Since sphere formation was highly enriched in a subpopulation of human and mouse oesophageal cells and the most mature oesophageal cells could not give rise to spheres, our findings are consistent with a hierarchical model in which oesophageal stem cells give rise to more differentiated cells. However, we cannot rule out the possibility that more differentiated cells could, under certain conditions, revert back to a stem cell phenotype. This question would ideally be further explored for murine oesophageal cells using in vivo lineage tracing experiments. However, such experiments are not feasible until a promoter is identified that is exclusively active within CD49f+CD24lowCD71low cells. Separately, we observed that CD49f+CD24midCD71mid cells, which likely represent early progenitors on their way towards differentiation, can also give rise to some esophagospheres, albeit with a significantly lower efficiency than CD49f+CD24lowCD71low cells. Thus, it is possible that there are two murine oesophageal stem cell populations that differ in their cell cycle status. However, since expression of CD24 and CD71 varies on a continuum without clear separation between subpopulations, the stem cell population likely overlaps on FACS plots with the early progenitor population and the low level of sphere formation seen among the CD24midCD71mid cells may simply be a reflection of this overlap. Future lineage tracing studies employing specific promoters in oesophageal subpopulations could help address these issues.

The esophagosphere culture systems we report here are ideally suited for characterisation of self-renewal and differentiation of human and mouse oesophageal stem cells. By combining clonogenicity and organotypic differentiation, they offer distinct advantages over previously reported organotypic culture methods, which assay primarily for tissue reconstitution without assessing clonogenicity in a quantitative fashion.10–13 As far as we are aware, we demonstrate for the first time the generation of esophagospheres from primary human oesophageal cells obtained from endoscopic biopsies. Additionally, we have identified culture conditions that can either maintain human oesophageal stem/progenitor cells or facilitate their differentiation into organotypic spheres. While a murine sphere culture system for oesophageal basal cells has been previously reported, this was not used to analyse clonogenicity or in vivo self-renewal.28

The human esophagosphere system allowed us to identify novel cell surface markers for human oesophageal stem/progenitor cells. nerve growth factor receptor (NGFR) (p75) was previously reported as a marker for human oesophageal progenitors.29 However, this study used passaged cells rather than primary cells freshly isolated from oesophageal tissue and employed a 2D culture assay. We and others.30 have observed that oesophageal and lung epithelial cells can change marker expression as early as the first in vitro passage (data not shown), underscoring the importance of analysing cells immediately after removal from patients. Previous studies have suggested the presence of CD34 expressing quiescent basal stem cells in papilla tips of the human oesophageal epithelium.5 ,9 However, these cells did not show greater clonogenicity compared with other epithelial subpopulations including suprabasal cells, suggesting they may not be enriched for stem cells. In our analyses, putative human oesophageal stem cells were most enriched in the CD49fhiCD24low population, forming 10-fold more and larger esophagospheres than and giving rise to suprabasal CD49flowCD24hi cells. No differences in proliferation were found between CD49fhiCD24low and CD49flowCD24low subpopulations. Considering that sphere-forming efficiency of human CD49fhiCD24low cells is significantly lower than their murine counterparts, the existence of a subpopulation of slow-cycling, high esophagosphere-forming cells within CD49fhiCD24low cells remains a possibility. Our findings are consistent with a model in which human CD49fhiCD24low cells contain stem/progenitor cells that give rise to CD49flowCD24low and most differentiated CD49flowCD24hi suprabasal cells (figure 2F).

The ability to genetically manipulate esophagospheres can facilitate functional mechanistic studies of oesophageal stem cell self-renewal and differentiation. Given the lack of oesophageal specific promoters targeting basal progenitor cells, in vitro organotypic sphere culture systems are particularly useful for analysis of mutations that have lethal effects in other tissues, as is the case for p63. Although the role of p63 in epithelial tissue homoeostasis has been extensively documented,18–22 its role in mouse and human oesophageal stem cells has not been previously reported to our knowledge. We found that conditional deletion of p63 in murine oesophageal cells results in depletion of esophagosphere-initiating cells. Additionally, we show for the first time that silencing of p63 in human oesophageal cells leads to dramatically fewer esophagospheres, indicating decreased stem/progenitor cell activity. Thus, p63 plays a critical regulatory role in the maintenance of human and mouse oesophageal stem cells.

In conclusion, we believe that the human esophagosphere system will be a powerful tool to model oesophageal diseases. The combination of these assays will facilitate the study of molecular and cellular mechanisms of oesophageal stem cell self-renewal, oesophageal maintenance and repair. Additionally, we envision that the esophagosphere culture system may allow amplification of oesophageal stem cells in vitro prior to regenerative medicine applications.


The authors thank Russell Fletcher and John Ngai of UC Berkeley for the kind gift of p63 conditional knockout mice. The authors also thank their Stanford gastroenterologists for their efforts in obtaining endoscopic biopsies.


View Abstract

Review history and Supplementary material

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • YJ and HR contributed equally.

  • Contributors YJ, HR and MD: study concept and design. YJ, HR and MD: writing manuscript. YJ, HR, SM, DK, YL, LXTN, WF and MD: acquisition of data, analysis and interpretation of data.

  • Funding California Institute for Regenerative Medicine Training Grant TG2-01159 (YJ), NIH T32DK007056 (HR), P30CA147933 and P01CA139490 (MD), the CRK Faculty Scholar Fund (MD) and the Virginia and DK Ludwig Foundation (MD). MD is supported by the US National Institutes of Health Director's New Innovator Award Program (1-DP2- CA186569) and is a Doris Duke Charitable Foundation Clinical Investigator.

  • Competing interests None declared.

  • Ethics approval This study was conducted with the approval of the Stanford University.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.