This conversion process has been the focus of several studies over the past 15 years, beginning with an observation by Patapoutian et al. that smooth muscle cells directly transdifferentiated into striated muscle cells.¹ Using mice as a model of esophageal myogenesis, where the conversion from one cell type to the other is virtually 100%, these authors showed coexpression of smooth muscle and striated muscle markers in muscularis externae myocytes by immunohistochemical labeling, and suggested that smooth muscle cells were able to switch phenotype well after commitment and differentiation. These findings were both supported by other investigators and challenged by other investigators.² Thus, the mechanism of esophageal striated myogenesis remained unresolved. We performed a series of experiments to determine whether transdifferentiation accounts for esophageal muscle phenotype switching.^3–5

In the first experiment, we fate mapped smooth muscle cells by crossing a mouse expressing eGFP and Cre-recombinase under the control of the smooth muscle myosin heavy chain (SmMHC) promoter (so-called SMCG2 mice) with R26R mice.³ The expression of eGFP and Cre-recombinase is strictly confined to smooth muscle cells in the SMCG2 mouse. R26R mice (also known as Rosa26R mice) have a constitutively expressed β-lactamase gene (under the control of the Rosa promoter), which is inhibited from expression by a lox-P flanked stop codon. Removal of the stop codon by Cre-recombinase results in constitutive β-lactamase expression. Thus, by crossing the SMCG2 and R26R mice, we could selectively activate β-lactamase expression in any cell that has expressed SmMHC (i.e., smooth muscle cells). Examination of E15 mouse embryos showed robust eGFP and Cre-recombinase expression (and β-lactamase expression) along the full length of the esophagus. However, P1 and P15 esophagi showed expression of these smooth muscle elements to more caudal portions of the esophagus. Importantly, no striated muscles in the muscularis externae showed β-lactamase expression (as evidenced by X-gal staining). This would suggest that these cells had never activated the SmMHC promoter to express Cre-recombinase and, therefore, did not derive from a smooth muscle cell precursor cell.

We then used the SMCG2 mouse to examine coexpression of eGFP or Cre-recombinase (which is localized to nuclei in smooth muscle cells) and the striated muscle transcription factor, myogenin.⁴ We could find no instances of coexpression of these two markers (one smooth, one striated) in any cells within the developing muscularis externae.

Finally, we used a selective deletion strategy to determine whether smooth muscle cells transdifferentiate into striated muscles.⁵ We used the SMCG2 mouse to selectively delete the myogenin gene from smooth muscle cells by crossing it with a mouse that had a lox-P-flanked myogenin gene. Since myogenin is essential for striated myogenesis, we hypothesized that, if smooth muscle cells transdifferentiated into striated muscle cells within the esophagus, deletion of myogenin in these cells prior to the transdifferentiation process would prevent this event, resulting in a myodysplasia, characterized by persistence of smooth muscle along the length of the esophagus. However, despite robust Cre-recombinase expression (and, presumably, efficient myogenin deletion), esophageal myogenesis progressed in these mice in a normal manner.

We first used a BMP reporter mouse line harboring a BRE-lacZ allele along with in situ hybridization to localize transcripts for BMP signaling components, including different antagonists. We demonstrated that when epithelium stratifies between embryonic day (E) 10.5 and E14.5, BMP signaling is not activated in the epithelium of the esophagus and forestomach, correlating with the presence of the BMP antagonist Noggin. However, activation of BMP signaling after E14.5 at the top layers of the epithelium is correlated with differentiation of the epithelium (Fig. 1). We further showed that deletion of the Noggin gene stalls the conversion of simple columnar into stratified epithelium. We then exploited a Shh-Cre allele that drives recombination in the embryonic foregut epithelium to generate either gain-or-loss-of-function models for the Bmpr1a (Alk3) receptor.

In gain-of-function embryos (Shh-Cre; Rosa26^{CAG–loxstoploxp–caBmpRIa}), high levels of ectopic BMP signaling also prevents the transition from simple columnar to multilayered undifferentiated epithelium in the esophagus and forestomach. In loss-of-function experiments, conditional deletion of the BMP receptor in Shh-Cre;Bmpr1a^flox/flox embryos allows the formation of a multilayered squamous epithelium, but which fails to differentiate, as shown by the absence of expression of the suprabasal markers, loricrin, and involucrin.

The replacement of one epithelium with another during esophageal development has been documented in both mouse and human, and in 1993, Thorey et al. proposed that the basal layer of the stratified squamous epithelium is derived from the columnar epithelium.^10–12 However, direct evidence for this conversion was lacking due to the availability of suitable model systems.

We have developed an in vitro model for studying esophageal development based on the culture of esophageal explants from embryonic day 11.5 (E11.5) mouse embryos. This culture model recapitulates the in vivo development of the esophagus.⁷ Using immunohistochemical analysis of sections from developing mouse embryos, we initially demonstrated that the transition from columnar to stratified squamous epithelium is accompanied by a change in expression of the intermediate filament proteins cytokeratin 8 (K8, a marker of columnar epithelium) and 14 (K14, a marker for the basal epithelial layer). At E11.5 all the cells of the esophageal epithelium express K8 but not K14. Gradually, between E15.5 and E17.5, K8 expression is lost in the basal layer but is retained in the suprabasal layers until postnatal day 3 when it is completely lost. In contrast, K14 is absent in the esophageal epithelium at E11.5 but begins to be expressed around E17.5 in the basal cells, which are no longer expressing K8.

The explant culture model was established by placing esophageal tissue dissected from E11.5 embryos onto fibronectin-coated glass coverslips. Within 24 h, the explants attached and flattened onto the fibronectin substratum at which stage they are composed of a tube of epithelium surrounded by mesenchyme. The explants remain viable in culture for up to 20 days and differentiate as stratified squamous epithelium, exhibiting the same switch in expression of K8 to K14 as observed in vivo.

To determine the origin of the stratified squamous epithelium, we used our explant model to explore the possibility that basal cells arise through programmed cell death of the columnar cells, selective overgrowth of the squamous epithelium, or though a direct conversion of one cell type to another.

To distinguish the first of these possibilities, we separated the cells after 7, 11, or 15 days of culture by flow cytometry and then immunostained for either K8 or K14 and propidium iodide to identify dead cells. If the switch in cell type was due to a preferential death of the columnar cell population, the number of dead cells would be higher in the K8-expressing population. However, there was no observable difference in the proportion of dead cells in the K8- and K14-positive populations indicating that programmed cell death does not play a role in the conversion from columnar to stratified squamous epithelium.

To determine whether the switch from columnar to stratified squamous epithelium was due to selective overgrowth of the squamous epithelium we immunostained for the proliferation marker Ki-67. The embryonic esophageal epithelium in culture is highly proliferative up until the culture differentiates into multilayers after which time the majority of the proliferating cells become localized to the K14-positive cells of the basal layer. In order to determine whether cell division was required for conversion, we treated esophageal explants with mitomycin C, a DNA crosslinking agent that prevents the cells from entering the cell cycle. If the cell conversion is dependent on cell division, then we predicted it would not be able to occur in the presence of a cell cycle inhibitor. However, the transition from K8-positive columnar layer to the K14-expressing cells continued to occur in the presence of mitomycin C indicating that cell proliferation does not play a role in the epithelial cell conversion.

The possibility that the cells of the basal layer arise directly from the columnar epithelium was also investigated. Upon costaining for K8 and K14, we found some cells coexpressing both markers suggesting that K14-positive cells arose directly from K8-positive cells. To trace the lineage of the stratified squamous basal cells, we electroporated a reporter plasmid, the K14 promoter driving GFP expression, into esophageal explants obtained from E15.5 embryos. This timepoint was chosen since the basal phenotype starts to appear between E15.5 and E17.5. Costaining for GFP and K8 revealed that at least some of the cells that have an activate K14 promoter (GFP-positive) also express K8, indicating that the basal cells of the squamous epithelium arise through a direct conversion from K8-positive columnar cells.

Barrett's esophagus (BE) is a premalignant lesion of esophageal adenocarcinoma in which the normal squamous epithelium is replaced by intestinalized columnar epithelium. Although it is known to develop as a consequence of chronic gastroesophageal reflux, its molecular aspect is not fully understood. Two lines of evidence suggested that research on esophageal development may shed light on the molecular mechanism of BE. (a) BE may originate from stem cells residing in the esophagus (e.g., interpapillary basal cells, ductal cells of submucosal glands). (b) Human embryonic esophagus is initially lined with a ciliated pseudostratified columnar epithelium containing goblet cells, which is later replaced by a stratified non-keratinized squamous epithelium containing submucosal glands. A question becomes obvious, Is BE an adaptive reversal of embryonic development?