Cancers arising in the oesophageal epithelium are among the most common fatal tumors in the world. Despite this, comparatively little is known about the cell biology and organization of this tissue. Recently, in vitro and in vivo techniques developed over the past 30 years for the study of the epidermis have been applied to the study of the oesophageal epithelium. This approach, combined with data from previous histochemical studies, has lead to the identification and isolation of putative oesophageal epithelial stem cells. Oesophageal epithelial stem cells demonstrate several unusual properties, and their identification may facilitate studies on oesophageal carcinogenesis.
The oesophageal epithelium
The luminal surface of the oesophagus is lined by a non-keratinising,stratified squamous epithelium that has a highly complex organization(Fig. 1)(Geboes and Desmet, 1978). Histologically, the oesophageal epithelium can be divided into two zones: the basal zone consists of several layers of small basophilic cells, whereas the differentiated zone consists of multiple layers of progressively flattened,differentiated squames. The basal zone contains a single layer of cells that adhere to the basement membrane (the basal layer) and a variable number of cell layers above this (the epibasal layers). The lamina propria invaginates the epithelium at regular intervals, producing tall papillary structures. Hence, the basal layer is further divisible anatomically into two components:one flat (the interpapillary basal layer, IBL) and one covering the papillae(the papillary basal layer, PBL) (Fig. 1). Cellular proliferation is limited to the basal zone, and cells are thought to migrate from this area towards the oesophageal lumen(Jankowski et al., 1993). Migration is associated with the initiation of differentiation and the sequential expression of differentiation markers(Jankowski et al., 1993).
The stem cell model
Data from rapidly proliferating epithelia indicate that their organization can be understood in terms of a stem/transit-amplifying-cell model. In this model, epithelia are maintained by a limited number of stem cells resident in the generative layer of the tissue (Hall and Watt, 1989; Potten and Morris, 1988). Stem cell division serves two purposes:replenishing the stem cell compartment and generating transit amplifying cells. Transit amplifying cells undergo several rounds of division, and their progeny eventually execute the differentiation pathways characteristic of the epithelium. Differentiation is associated with a reduction in proliferative capacity, and ultimately the differentiated progeny are lost. In a tissue such as the oesophageal epithelium that has a single differentiation pathway,tissue renewal is performed by the transit amplifying cell population.
In many invertebrate systems, stem cell division always gives rise to one stem cell daughter and one transit amplifying cell. Such `invariant asymmetry'is rare in adult mammalian tissues (Watt and Hogan, 2000). In most mammalian tissues, stem cell division has three possible outcomes: the generation of two stem cells, two transit amplifying cells or one daughter cell of each type. In the steady state, the balance between stem and transit amplifying cell numbers is maintained in the tissue as a whole by multiple feedback loops (`populational asymmetry')(Watt and Hogan, 2000).
Many issues concerning the nature of epithelial stem cells remain to be resolved, and there is a growing realization that there may be more plasticity in adult tissues than the stem/transit-amplifying-cell model predicts(Morrison et al., 1997;Vogel, 2000). However, it provides a useful framework for us to conceptualize cell behavior in vivo and also makes several predictions that allow identification and isolation of putative human stem cells. Firstly, relative to the transit amplifying population, stem cells should proliferate less frequently in vivo in the steady state. In addition, as the progeny of transit cells undergo terminal differentiation after several rounds of division, stem cells have a higher proliferative capacity both in vivo and in vitro. Secondly, the difference in function between stem cells and transit amplifying cells is likely to be reflected in different levels of expression of functionally relevant molecules. Identification of such molecules could allow the isolation and separation of the two cell types. Thirdly, because stem cells do not generally initiate a program of differentiation, they are `phenotypically primitive',that is, they do not express differentiation markers characteristic of the tissue.
Identification and isolation of oesophageal epithelial stem cells
Interpapillary basal layer cells proliferate infrequently and asymmetrically in vivo
Recent detailed analysis of mitotic figures in the oesophageal epithelium combined with immunohistochemical staining for proliferating cells demonstrates a highly complex and predictable pattern of cell proliferation in the basal zone (Seery and Watt,2000). Firstly, proliferating cells are far more common in the epibasal layers than in the basal layer. Secondly, even within the basal layer, mitotic cells are approximately four times more common in the PBL than in the IBL. Thirdly, the orientation of mitosis is heterogeneous in the basal layer. In the PBL, cell division is symmetrical relative to the underlying stroma, yielding two daughter cells in contact with the basement membrane. In contrast, in the IBL, cell division is inherently asymmetrical. The majority of cell divisions in the IBL occur at right angles to the underlying basement membrane, yielding one daughter that remains in the basal layer and one that enters the epibasal layers (Fig. 2).
Cells in the IBL are thus candidates for oesophageal epithelial stem cells. They proliferate relatively infrequently in vivo, and their division yields one daughter cell that remains in an area of low proliferative activity (a putative stem cell) and one that enters an area of high proliferative activity(a putative transit amplifying cell).
Interpapillary basal layer cells have a high in vitro proliferative capacity
The level of expression of adhesion molecules is not uniform in the basal layer of the epidermis (Jones and Watt,1993; Moles and Watt,1997). In the interfollicular epidermis, basal keratinocytes on the dermal papillae express high levels of β1 integrin compared with basal cells in the rete ridges (Jensen et al., 1999). Epidermal basal cells can be separated from suprabasal cells on the basis of their forward- and side-scatter characteristics by fluorescence activated cell sorting (FACS)(Jones and Watt, 1993). Jones and Watt isolated interfollicular basal epidermal keratinocytes by using this technique and further separated them on the basis of their relative levels ofβ1 integrin expression. They showed that the 20% of basal cells expressing the highest level of β1 integrin are enriched for cells capable of forming large, rapidly proliferating colonies (>32 cells after two weeks of culturing) in vitro. They argued that such colonies are likely to arise from stem cells. Therefore, the `integrin bright' region of the dermal papillae represents the stem cell compartment of the epidermal basal layer(Jones and Watt, 1993;Jensen et al., 1999).
Basal cells from the oesophageal epithelium can also be separated from suprabasal keratinocytes on the basis of their FACS profile(Jankowski et al., 1992a;Seery and Watt, 2000).
In addition, β1 integrin expression is also predictably heterogeneous in the oesophageal basal layer. Surprisingly, cells in the IBL express lower levels of this molecule compared with cells in the PBL. However, when cells from the two regions are separated by FACS on the basis of β1 integrin levels and then grown at clonal density, IBL cells are consistently two-fold enriched for cells capable of forming large actively growing colonies(Seery and Watt, 2000)(Fig. 3).
Interpapillary basal layer cells are phenotypically primitive
The expression of several cytokeratin (CK) species varies during the differentiation program of oesophageal keratinocytes(Takahashi et al., 1995). Using in situ hybridization, Viaene and Baert have shown that the pattern of CK mRNA expression is heterogeneous in the oesophageal epithelial basal layer(Viaene and Baert, 1995). CK13 protein is produced by differentiating keratinocytes in all endodermderived stratified squamous epithelia, including the oesophageal epithelium(Moll et al., 1982). CK13 mRNA is present at high levels in the cells of both the PBL and epibasal layers but is absent from the keratinocytes of the IBL. In addition, the signal intensity for CK14 and CK15 mRNA is patchy in the IBL, which contrasts with the high levels of expression of these species in the PBL and epibasal layers. Furthermore, mRNA for the differentiation marker CK4 is detectable in the papillary region from the second epibasal layer onwards but does not appear in the interpapillary region until the third epibasal layer(Viaene and Baert, 1995). Hence, in terms of differentiation markers, IBL cells are the `least differentiated' cell type in the tissue.
The data discussed above are thus consistent with the idea that keratinocytes of the IBL are stem cells and with a model of the oesophageal epithelium in which stem and transit amplifying cells are contained in distinctive anatomical compartments: the IBL and epibasal layers, respectively(Fig. 4). Why the putative IBL stem cells, in contrast to the epidermal stem cells, express low levels of integrin β1 is discussed below.
The oesophageal epithelium has unexpected properties
The stratified squamous epithelia of the oesophagus and epidermis have different functions and embryological origins. Perhaps it is not surprising,therefore, that they exhibit many differences in cellular behavior and organization. Firstly, in the epidermis, stem cells and transit amplifying cells are both present in the basal layer, and the available evidence suggests that the fate of stem cell daughters is determined by populational asymmetry(Watt and Hogan, 2000). In contrast, putative stem cells and transit amplifying cells of the oesophageal epithelium reside in separate anatomical compartments. Secondly, stem cell divisions in the IBL seem to generate transit amplifying daughters by an invariant strategy: division at right angles to the underlying basement membrane inevitably places one daughter cell in the putative epibasal transit compartment. This is analogous to the behavior of stem cells in some embryonic systems. It is probably rare in adult mammalian epithelia and is not observed in the epidermis (Watt and Hogan,2000). Thirdly, whereas the stem cells of the epidermis express high levels of β1 integrin, the putative stem cells of the IBL express low levels of this marker. These differences between epidermal and oesophageal keratinocytes may reflect differences in the role of epithelial-mesenchymal interactions in the control of stem cell function in the two cell types.
The behavior of epidermal stem cells is independent of interactions with their underlying stroma. Epidermal keratinocytes can regulate stem cell proliferative activity and number independently of interactions with the dermis (Jones et al., 1995;Watt, 2001). In contrast, the basement membrane of the oesophageal epithelium plays a central role in controlling oesophageal stem cell behavior. Interactions between oesophageal stem cells and the oesophageal basement membrane determine the asymmetric orientation of cell division and dictate the overall tissue architecture. When oesophageal keratinocytes are cultured in vitro on denuded acellular dermis,the orientation of cell division in the basal layer of the epithelium formed is random. In addition, this in vitro epithelium is flat and featureless. In contrast, when oesophageal keratinocytes are cultured on denuded acellular oesophageal submucosa with intact basement membrane, the orientation of cell division is predominantly asymmetrical, and an epithelium with prominent papillae is formed that resembles the in vivo tissue(Seery and Watt, 2000).
Recent data from the study of Drosophila embryogenesis suggest a mechanism by which basement-membrane/adhesion-molecule interactions might influence the orientation of basal cell division. Lu et al. have shown that the formation of adherens junctions dictates the orientation of cell division in Drosophila neuroblasts and neuroepithelial cells. Following disruption of adherens junctions, neuroepithelial division, which normally occurs in the plane of the neuroepithelium, becomes asymmetric in orientation(Lu et al., 2001). In addition, Le Borgne et al. have shown that cadherin-catenin function in the single organ precursor cell of Drosophila determines the orientation of cell division at certain stages in the formation of the dorsal sensory organ of the thorax (Le Borgne et al.,2002). Intracellular adhesion and cadherincatenin levels are regulated by interactions with the extracellular matrix and in particular by integrin ligation (Gimond et al.,1999). If basement-membrane/adhesion-molecule interactions can influence the orientation of basal cell division, this might offer an explanation for the observed difference in the orientation of mitotic cells in the IBL and PBL. The precise composition of the oesophageal basement membrane has not been determined, but the levels of at least one laminin isoform (theβ2 laminin chain) vary between the IBL and PBL(Seery and Watt, 2000).
Several molecules have been identified in Drosophila that induce an asymmetric accumulation of cell-fate determinants during cytokinesis in embryonic stem cells (e.g. Frizzled, Inscuteable and Bazooka)(Bellaiche et al., 2001;Orgogozo et al., 2001). These cell-fate determinants (e.g. Numb, an antagonist of Notch signaling) confer different phenotypes on each daughter cell, thus contributing to the generation of cellular diversity in the embryo(Jan and Jan, 2000). Functional homologues of these proteins have been described in vertebrates(Cayouette et al., 2001). We do not know if they, or related molecules, play a role in determining the fate of the progeny of IBL stem cell divisions, because there is no data concerning their patterns of expression in the oesophagus. In fact, such mechanisms may not be necessary in keratinocytes. In common with their epidermal equivalents,oesophageal keratinocytes differentiate in suspension culture(Adams and Watt, 1989;Dazard et al., 2000;Seery and Watt, 2000). Hence,following asymmetric mitosis in the IBL, loss of contact with the basement membrane eventually triggers terminal differentiation in the epibasal daughter cell and its progeny. Thus, by determining the orientation of cell division in the IBL, basement-membrane/adhesion-molecule interactions may also determine the fate of the daughter cells produced.
The idea that basement membrane components in the IBL somehow provide a`stem cell niche' is in keeping with the prevailing situation throughout the rest of the gastrointestinal tract. Regional variation along the crypt-villus axis in laminin components of the basement membrane plays the central role in establishing stem cell identity in the columnar-lined gastrointestinal tract(Simon-Assman et al., 1998).
Invariant stem cell division might seem inflexible for situations of increased cell need. However, cell divisions that are symmetrical relative to the underlying basement membrane are occasionally observed in the IBL(Seery and Watt, 2000). Whether the ratio of symmetrical to asymmetrical divisions in the IBL can be altered by signals from the surrounding tissue is not known. Cells in the IBL have a characteristic morphology during S phase(Seery and Watt, 2000). The cell remains attached to the basement membrane by a thin film of cytoplasm,but its nucleus is displaced into the epibasal layers. Such a morphology is reminiscent of the intermitotic nuclear migration during stem cell division in the neuroepithelium (Alvarez-Buylla et al.,1998). Jan and Jan have suggested that intermitotic nuclear migration allows stem cells to receive signals from the surrounding tissue that dictate asymmetric partitioning of molecules and consequently determine the orientation of cell division (Jan and Jan, 2000). If such signals can alter the balance of symmetrical and asymmetrical mitoses in the IBL, this should offer a novel mechanism for controlling the balance between stem cell and transit amplifying cell production. By altering stem cell numbers, such a mechanism would have marked effects on the rate of new cell production in the tissue(Potten and Morris, 1988;Morrison et al., 1997).
Many questions to answer
Although the model outlined in Fig. 4 is complex, it must be a gross oversimplification of the in vivo situation, and much basic information remains to be acquired. Every cell in the IBL is unlikely to be a stem cell. In situ hydridization studies have shown that the levels of several CK mRNAs vary within the IBL, which indicates that there are further subdivisions of cell function in this region(Viaene and Baert, 1995). We cannot yet isolate individual cells or subgroups of cells from the IBL to allow determination of their in vitro properties. Similarly, epibasal cells cannot be unequivocally designated as the transit amplifying population. No cell surface marker that allows their in vitro isolation has been described;hence, we have no data on their clonogenic capacity. However, it is reasonable to speculate that the epibasal cells are a transit amplifying population:proliferation is frequent in the epibasal layers, and loss of contact with the basement membrane eventually triggers differentiation(Seery and Watt, 2000). In addition, when oesophageal basal cells are selectively isolated from surgical specimens and cultured in vitro at clonal density, the majority of colonies formed (>85%) are large and rapidly growing (>32 cells after two weeks in culture) (Seery and Watt,2000) (Fig. 4). This contrasts with the situation in the epidermis. When epidermal basal keratinocytes are cultured in vitro under similar conditions, up to 40% of colonies formed are small (<32 cells) and abortive (exhibiting predominantly differentiated cell morphology)(Jones and Watt, 1993). Jones and Watt have argued that such abortive colonies are derived from transit amplifying cells (Jones and Watt,1993). Therefore, the absence of such colonies from cultures of PBL and IBL cells is consistent with the idea that transit amplifying cells are not present in significant numbers in the oesophageal basal layer.
The status of the cells in the PBL is difficult to define in terms of the stem/transit-amplifying-cell model. In terms of in vivo proliferation and CK mRNA expression pattern, the cells of this region appear to be intermediate between the putative stem cells of the IBL and the putative transit amplifying cells of the epibasal layers (Seery and Watt, 2000; Viaene and Baert,1995). Interestingly, the in vitro clonogenicity of basal epidermal keratinocytes varies linearly with integrin expression levels. There might thus be a continuum of cellular behavior in the proliferative zone of the epidermis rather than strictly defined populations of stem cells and transit amplifying cells. A similar situation may apply in the oesophageal epithelium, PBL cells being functionally intermediate between these two cell types. However, no data defining the lineage relationship between the cells of the IBL and the PBL are yet available, and the model outlined inFig. 4 makes no prediction about this relationship. The hypothesis that the papillae are formed by sideways expansion of cells in the IBL remains a possibility(Jankowski et al., 1992b). The progency of the rare divisions parallel to the basement membrane in this site could contribute to the PBL (Seery and Watt, 2000). In the epidermis, interactions between Delta1 expressed on the surface of cells in the `integrin bright' stem cell cluster and Notch1 on surrounding cells are important in controlling stem cell activity. Interaction between Delta1 and Notch1 promotes cellular differentiation at the edges of the stem cell clusters, thus maintaining the anatomical integrity and size of the stem cell compartment(Lowell et al., 2000). Whether, such interactions at the junction of the IBL and PBL play a role in controlling their relative size and composition is unknown. The pattern of expression of Notch and Delta in the oesophagus has not been described in detail.
The model in Fig. 4 takes no account of the glandular structures associated with the oesophageal epithelium. Tubuloalveolar glands are present in the submucosa along the length of the oesophagus. Ducts connecting these glands to the luminal surface are lined by a cuboidal epithelium that becomes stratified in its terminal part (Hopwood et al., 1986). Gillen et al. suggested that reconstitution of the surface epithelium from stem cells in the tubuloalveolar glands, following mucosal injury, is the source of Barrett's oesophagus (vide infra)(Gillen et al., 1988). Although there is no direct evidence for the presence of a stem cell population in these glands, their existence would not necessarily conflict with the model discussed here and would, in fact, allow an interesting analogy to be drawn with the epidermis. In the epidermis, stem cells appear to reside not only in the interfollicular epithelium but also in the epidermal appendages, specifically in the bulge region of the hair follicle(Cotsarelis et al., 1990;Rochat et al., 1994;Panteleyev et al., 2001). The detailed histological and functional analyses carried out on the cell biology of the hair follicle have not been applied to the study of the glandular structures of the oesophagus. This is an area of study of considerable clinical importance.
Oesophageal stem and transit amplifying cells in disease
Metaplasia of the oesophageal epithelium from the normal stratified squamous epithelium to a columnar phenotype (Barrett's oesophagus,Fig. 5) is a commonly recognized disorder in clinical practice. Barrett's oesophagus, is an acquired condition and arises secondary to longstanding gastroesophageal reflux disease(GORD) (Richter, 2001). In normal individuals, a sphincter mechanism at the junction of the oesophagus and stomach prevents the reflux of gastric contents (acid and bile salts) into the oesophageal lumen. This sphincter mechanism is defective in up to 40% of adults, and gastric contents reflux into the oesophageal lumen. Barrett's metaplasia develops in up to 10% of such individuals(Richter, 2001). Barrett's oesophagus is the precursor lesion of adenocarcinoma of the oesophagus, an almost invariably fatal malignancy. The incidence of oesophageal adenocarcinoma has doubled in the developed world in recent years (Devasa et al., 1998).
With regard to the hypothesized origin of Barrett's oesophagus from stem cells in the tubuloalveolar glands, note that duodenogastric reflux can trigger a similar transdifferentiation in the oesophagus of rats(Pera et al., 2000). Because rats do not have glandular structures in the oesophagus, this indicates that,at least in this species, the differentiation program of keratinocytes can be modified by GORD to induce columnar differentiation. Furthermore, as gastric refluxate is not genotoxic (Fein et al.,2000), it is possible that the effect of gastroesophageal reflux on oesophageal transdifferentiation represents an epigenetic effect on post-mitotic cells rather than an abnormality of stem cells. In this regard,it is of interest that bile salts exert regulatory effects on cellular differentiation in the haemopoietic system through effects on transcriptional regulatory pathways (Zimber et al.,2000). Interestingly, it has recently been shown that components of the Wnt signaling pathway play a key role controlling the balance between squamous and glandular differentiation in epidermal cells(Arnold and Watt, 2001; Merril et al., 2001; Miyoshi et al.,2002; Niemann et al.,2002). Although abnormalities in Ecadherin and catenin signaling have been implicated in the progression of Barrett's oesophagus to adenocarcinoma (Bailey et al.,1998; Seery et al.,1999), the role of these pathways in the primary pathogenesis of Barrett's metaplasia has not been determined.
That gastroesophageal refluxate can influence transit amplifying cell differentiation in the oesophagus is illustrated by the effects of GORD on oesophageal epithelial morphology. Although oesophageal keratinocytes differentiate when held in suspension(Seery and Watt, 2000), the putative transit amplifying cells of the epibasal layers seem to be capable of initiating a program of differentiation, as evidenced by expression of several differentiation markers, and to continue proliferating in vivo. Furthermore,the relative rates of the two processes seem to be affected by GORD, because the relative thickness of all the layers described inFig. 1 varies in this disorder. In the presence of GORD, an unidentified proliferative stimulus results in marked expansion of the transit amplifying population, with elongation of the papillae, thickening of the epibasal layers and thinning of the differentiated zone (Livstone et al.,1977).
Given that the methods of oesophageal keratinocyte isolation and culture described in this commentary result in little contamination with glandular elements, it may now be possible to dissect in vitro the role of genetic and epigenetic factors in controlling oesophageal keratinocyte differentiation and proliferation.
The oesophageal epithelium is worthy of further study. The combination of readily defined stem and transit amplifying components, the clinical importance of primary diseases of oesophageal keratinocytes and the ease of their in vitro culture is unique among adult epithelia.
I thank Alberto Gandarillas and Fiona Watt for discussions and advice on keratinocyte cell biology and Kieran Sheahan, Department of Pathology, St. Vincent's Hospital, Dublin, for supplying the images inFig. 5. I also thank Dr. Burke,Dr. King, Mr. Townsend and Mr. Fountain of Harefield Hospital, UK and Mr Sagor and Dr O'Reilly of Hemel Hempsted Hospital, UK.