P63 deficiency and CDX2 overexpression lead to Barrett’s-like metaplasia in mouse esophageal epithelium

Abstract

Background

The cellular origin and molecular mechanisms of Barrett’s esophagus (BE) are still controversial. Transdifferentiation is a mechanism characterized by activation of the intestinal differentiation program and inactivation of the squamous differentiation program.

Aims

Renal capsule grafting (RCG) was used to elucidate whether CDX2 overexpression on the basis of P63 deficiency in the esophageal epithelium may generate intestinal metaplasia.

Methods

P63^−/−;Villin-Cdx2 embryos were generated by crossing P63^+/− mice with Villin-Cdx2 mice. E18.5 esophagus was xenografted in a renal capsule grafting (RCG) model. At 1, 2, or 4 weeks after RCG, the mouse esophagus was immunostained for a proliferation marker (BrdU), squamous transcription factors (SOX2, PAX9), squamous differentiation markers (CK5, CK4, and CK1), intestinal transcription factors (CDX1, HNF1α, HNF4α, GATA4, and GATA6), intestinal columnar epithelial cell markers (A33, CK8), goblet cell marker (MUC2, TFF3), Paneth cell markers (LYZ and SOX9), enteroendocrine cell marker (CHA), and Tuft cell marker (DCAMKL1).

Results

The P63^−/−;Villin-Cdx2 RCG esophagus was lined with proliferating PAS/AB+ cuboidal cells and formed an intestinal crypt-like structure. The goblet cell markers (TFF3 and MUC2) and intestinal transcription factors (CDX1, HNF1α, HNF4α, GATA4, and GATA6) were expressed although no typical morphology of goblet cells was observed. Other intestinal cell markers including enteroendocrine cell marker (CHA), Paneth cell markers (LYZ and Sox9), and intestinal secretory cell marker (UEA/WGA) were also expressed in the P63^−/−;Villin-Cdx2 RCG esophagus. Squamous cell markers (PAX9 and SOX2) were also expressed, suggesting a transitional phenotype.

Conclusion

CDX2 overexpression on the basis of P63 deficiency in esophageal epithelial cells induces Barrett’s-like metaplasia in vivo. Additional factors may be needed to drive this transitional phenotype into full-blown BE.

Introduction

As a premalignant lesion of esophageal adenocarcinoma (EAC), Barrett’s esophagus (BE) is characterized by the replacement of stratified squamous epithelium by a specialized or intestinalized columnar epithelium in the distal esophagus. BE is believed to develop as a result of chronic gastroesophageal reflux disease (GERD) and inflammation in the esophagus [1]. The baseline risk for EAC is low in the general population, whereas it is 30–125 times higher in patients with BE [2, 3]; The yearly risk for EAC in non-dysplastic BE is approximately between 0.27–0.5% per person-year [4–6]. Once EAC is developed, its prognosis is poor with a 5-year survival rate of less than 15% [7]. Thus, it is important to elucidate the molecular mechanisms of BE.

Several theories of the cellular origin of BE have been supported by experimental evidence. Intestinal metaplasia may stem from trans-differentiation of mature squamous epithelial cells, circulating bone marrow stem cells, neck cells residing in the esophageal submucosal gland duct, gastric cardia, or basal cells located at the gastroesophageal junction [8–10]. Among these possible origins, basal cells at the gastroesophageal junction have recently been favored although this does not necessarily exclude other possibilities. Several pieces of evidence support trans-differentiation as a possible origin of BE: 1) After esophagectomy some patients developed BE in the cervical esophagus which is far from the gastroesophageal junction [11]; 2) A multilayered epithelium (MLE) at the squamous-Barrett’s junction was featured by the co-occurrence of microvilli and intercellular ridge on the surface of the cells and co-expression of squamous and columnar cytokeratins in the cytoplasm [12, 13]; 3) Far away from the stomach and the duodenum, MLE was observed in the mid-esophagus in a surgical model of gastroesophageal reflux in rats [14].

Our previous study has suggested that, when normal esophageal epithelium was stimulated by gastroesophageal reflux, the squamous differentiation program may be inactivated through the loss of expression of squamous transcription factors and differentiation markers (e.g., P63, SOX2, CK1, CK4, etc.). Meanwhile, the columnar differentiation program may be activated through the expression of intestinal transcription factors and differentiation markers (e.g., CDX2, CDX1, MUC2, TFF3, etc.) [10]. P63 is a critical transcription factor of squamous stratification and a key regulator of cell adhesion and survival of progenitor cells in the squamous epithelium [15], while CDX2 regulates intestinal development [16]. P63 is constitutively expressed in the basal cell layer of esophageal squamous epithelium, but lost in BE [17, 14]. When esophageal squamous cells were exposed to bile acids and acid, P63 became downregulated [18]. CDX2 was strongly expressed in the normal intestinal epithelium and human BE, but not in esophageal squamous epithelial cells [16]. When squamous epithelial cells were stimulated by gastroesophageal refluxate, CDX2 mRNA was detected by RT-PCR [19]. Treatment of HET1A cells (an immortalized human esophageal squamous epithelial cell line) with acid, bile acids, or both resulted in the expression of CDX2 [20]. Therefore, P63 silencing and CDX2 overexpression in esophageal squamous epithelial cells may play a key role in the pathogenesis of BE. However, when P63 was silenced in esophageal squamous epithelial cells in P63^−/− mice, or when Cdx2 was overexpressed in the basal cells of K14-Cdx2 transgenic mouse esophagus [8, 21], goblet cell which is diagnostic of BE did not appear in the esophagus of both mouse models. These data clearly demonstrate that P63 deficiency or CDX2 overexpression alone in the esophagus is insufficient to induce BE.

Villin (VIL) is expressed in the esophageal epithelium of P63^−/− mice [8]. We then ask whether CDX2 overexpression on the basis of P63 deficiency in the esophageal epithelium of P63^−/−;Villin-Cdx2 mice may generate intestinal metaplasia. Since P63^−/− mice die after birth [15, 22], we established a renal capsule grafting (RCG) model to test our hypothesis.

Methods

Animals and surgical procedures

Wild-type C57BL/6J mice and P63^+/− mice were purchased from Jackson Laboratory (Bar Harbor, ME). Villin-Cdx2 mice were gifted from Dr. John Lynch’s lab, University of Pennsylvania [23]. Genetically modified mice were bred and genotyped according to the original developers. All animal experiments were approved by the Institution Animal Care and Use Committee of North Carolina Central University and Ethics Committee of the First Affiliated Hospital, Chongqing Medical University. Mice were housed five per cage, given laboratory chow and water ad libitum, and maintained on a 12:12h light-dark cycle.

Female wild-type, P63^+/− and P63^+/−; Villin-Cdx2 mice were crossed with male mice of the same genotypes, respectively, to generate wild-type, P63^−/− and P63^−/−;Villin-Cdx2 embryos. E18.5 esophagi were harvested from pregnant mice, cut into a size of 5mm, and temporarily stored in ice-cold PBS. P63^−/− genotype was estimated by the outer appearance of the embryo during harvest and later on confirmed by PCR genotyping after surgery. P63^−/− embryos are known to be smaller in size, show truncated limbs, and lack mature stratified skin epidermis and appendages[24].

Syngeneic host mice received anesthetics premixed in normal saline (80mg/kg ketamine and 12mg/kg xylazine, i.p.). Renal capsule grafting (RCG) was performed in the following steps (Supplementary Figure 1): (1) Make a 1-cm incision on the skin along the dorsal midline parallel to the spine. Gently push the kidney out through an incision on the muscle. (2) Gently grip the renal capsule and pierce the capsule with the tip of a sharp forceps, and push the blade in underneath the capsule for about 1cm without damaging the kidney, and create a “nest” between the capsule and renal parenchyma. (3) Hold the opening of the “nest” with forceps, and gently insert the harvested E18.5 esophagus into the “nest”. (4) Gently guide the kidney to slide back into the peritoneal cavity. Close the muscle incision with absorbable sutures. Close the skin incision with wound clips. Five wild-type esophagus, five p63^−/− esophagus, and five p63^−/−;Villin-Cdx2 esophagus were used for RCG each at 1 week, 2 weeks, and 4 weeks.

2. Tissue samples and pathological evaluation

At 1, 2, or 4 weeks after RCG, the host mice were sacrificed by CO2 asphyxiation. The RCG esophagus along with the whole kidney was harvested and fixed in 10% buffered formalin, processed, and embedded in paraffin. The tissue samples were cut into 5-μm sections for histochemical and immunohistochemical staining.

3. Histochemical and immunohistochemical staining

Tissue sections were deparaffined and rehydrated for both histochemical and immunohistochemical staining. Hematoxylin/eosin staining (H&E), Periodic acid Schiff and Alcian blue staining (PAS/AB), and UEA1/WGA (Vector Labs, Burlingame, CA), were performed according to well-established procedures in the lab.

To validate the genotype, we performed immunohistochemical staining of P63, VIL, and CDX2. In addition, nine groups of markers were stained as well: 1) Proliferation marker (BrdU); 2) Squamous transcription factors (SOX2, PAX9); 3) Squamous differentiation markers (CK5, CK4, and CK1); 4) Intestinal transcription factors (CDX1, HNF1α, HNF4α, GATA4, and GATA6); 5) Intestinal columnar epithelial cell markers (A33, CK8); 6) goblet cell marker (MUC2, TFF3); 7) Paneth cell markers (LYZ and SOX9); 8) Enteroendocrine cell marker (CHA); 9) Tuft cell marker (DCAMKL1).

Briefly, paraffin-embedded tissue sections were deparaffinized, rehydrated, and pretreated by heating the slides for 5–10 min in 10μm citrate buffer. Immunohistochemical staining was performed with ABC kit (Vector Labs) according to the manufacturer’s instructions. The sources of the primary antibodies, catalog numbers, and working concentrations were listed in Table 1. Normal serum or phosphate buffered saline was used instead of the primary antibodies, as negative controls. Both negative and positive control slides were processed in parallel.

Table 1.

Transcription factors and differentiation markers detected by immunohistochemical staining

Antigen	Description	Expression Pattern	Antibody Source	Catalog No.	Concentration
A33	Intestinal differentiation marker	Cytoplasm of gut, rectum and colon	R&D	AF3080	1:100
BrdU (5-bromo-2’-deoxyuridine)	Thymidine analog, naturally incorporates into proliferating cells	Incorporated in proliferative cells at the S phase	Oxford Biotechnology	OBT0030	1:400
CDX1 (Caudal-related homeobox 1)	Intestinal transcription factor	Nuclei of columnar and goblet cells	Gift from Dr. John Lynch	-	1:50
CDX2 (Caudal-related homeobox 2)	Intestinal transcription factor	Nuclei of columnar and goblet cells	Cell Marque	EPR2764Y	1:200
CHA (Chromogranin A)	Intestinal differentiation marker	Cytoplasm of enteroendocrine cells	Immunostar	20085	1:4000
CK1 (Cytokeratin 1)	Squamous differentiation marker	Cytoplasm of superficial squamous cells	Abcam	Ab93652	1:50
CK4 (Cytokeratin 4)	Squamous differentiation marker	Cytoplasm of parabasal squamous cells	Sigma	C-5176	1:100
CK5 (Cytokeratin 5)	Squamous differentiation marker	Cytoplasm of basal squamous cells	ThermoFisher	EP1601Y	1:100
CK8 (Cytokeratin 8)	Intestinal differentiation marker	Cytoplasm of gut, rectum and colon	DSHB	-	1:800
DCAMKL1 (Doublecortin Like Kinase 1)	Tuft cell marker	Nuclei of Tuft cells	Abcam	Ab31704	1:100
GATA 4 (GATA-binding protein 4)	Intestinal transcription factor	Nuclei of columnar and goblet cells	Santa Cruz	SC-1237	1:50
GATA 6 (GATA-binding protein 6)	Intestinal transcription factor	Nuclei of columnar and goblet cells	R&D	AF1700	1:50
HNF1α (Hepatocyte nuclear factor 1α)	Intestinal transcription factor	Nuclei of columnar and goblet cells	Santa Cruz	SC-6507	1:50
HNF4α (Hepatocyte nuclear factor 4α)	Intestinal transcription factor	Nuclei of columnar and goblet cells	Santa Cruz	SC-6556	1:400
LYZ (Lysozyme)	Intestinal differentiation marker	Cytoplasm of Paneth cells	Leica	NCL-MVRAM	Ready to use
MUC2 (Mucin 2)	Intestinal differentiation marker	Cytoplasm of goblet cells	Santa Cruz	SC-15334	1:100
P63	Transcription factor essential for squamous epithelium	Nuclei of basal and parabasal cells	Santa Cruz	SC-8431	1:50
PAX9 (Paired Box 9)	Transcription factor essential for normal development of thymus, parathyroid, ultimobranchial bodies, teeth, skeletal elements of skull and larynx as well as distal limbs	Nuclei of esophagus	Cell signaling	8739	1:50
SOX2 (SYR-related HMG box gene 2)	Transcription factor essential for upper gastrointestinal epithelium	Nuclei of epithelial cells in oral cavity, esophagus and stomach	Abcam	Ab92494	1:100
SOX9 (SRY-box Transcription factor 9)	Intestinal transcription factor	Nuclei of Paneth cells	Chemicon	Ab5535	1:1,600
TFF3 (Trefoil factor 3)	Intestinal differentiation marker	Cytoplasm of goblet cells	Cloud clone	PAB656Mu01	1:10
VIL (Villin)	Intestinal differentiation marker	Cytoplasm of columnar and goblet cells	NeoMarkers	MS-1499-P	1:10

Immunohistochemical staining for each marker was evaluated by two separate pathologists as either positive or negative in the epithelium of interest. The staining of BrdU, SOX2, PAX9, P63, CDX1, CDX2, HNF1α, HNF4α, GATA4, GATA6, SOX9 were all nuclear. Cytoplasmic staining was present for CK5, CK4, CK1, A33, CK8, VIL, MUC2, TFF3, UEA1/WGA, LYZ, and CHA.

4. Statistical analysis

The proliferative activity of basal cells of the RCG esophageal epithelium and normal esophageal epithelium in Week 1, 2, and 4 were compared with the Chi-square test, respectively. The proliferative activity was represented as the number of BrdU⁺ cells per 1,000 basal cells. P<0.05 indicated statistical significance.

Results

RCG esophageal epithelium showed a growth pattern comparable with its normal counterpart.

We first intended to understand whether RCG would allow the esophagus to differentiate and proliferate properly within a certain time window. In gross view, the diameter and length of the RCG esophagus at Week 4 were larger than those at Week 1 and 2. Over time its color turned from transparent (Week 1), to semi-transparent (Week 2) and white (Week 4) (Figure 1A–C). The basal cells at Week 2 were more densely arranged than those at Week 1 and 4. There were approximately 3 to 5 layers of suprabasal cells at Week 2, while 2–3 layers at Week 1 and 1–2 layers at Week 4. Massive keratinization appeared in the esophageal lumen at Week 4 as compared with Week 1 and 2 (Figure1D–F). The overall organization of basal cells was similar in the RCG esophagus and normal esophagus at these three time points. More layers of suprabasal cells were observed in the esophageal epithelium at Week 2, while more keratinized epithelial cells were present at Week 4 (Supplementary Figure 1A–C).

Figure 1.

Morphology (gross view and H&E staining), proliferation (BrdU) and expression of differentiation markers (CK5, CK4, and CK1) in wild-type esophageal epithelium at 1, 2 and 4 weeks after RCG. Scale bar=1mm (A to C), or 50μm (D to R).

Proliferative activity of basal cells in the RCG esophagus was similar to that in the normal esophagus at Week 1 and Week 2 (P>0.05). At Week 4, basal cells in the RCG esophagus became less proliferative while those in the normal esophagus maintained a stable proliferative activity (P<0.05) (Figure 1G–I, Supplementary Figure 1D–F, Supplementary Figure 2). Differentiation markers of normal esophagus, CK5 for basal cells, CK4 for parabasal cells, and CK1 for keratinized cells, was also expressed in patterns similar to those in the RCG esophagus (Figure 1J–R, Supplementary Figure 1G–O).

P63 deficiency and CDX2 overexpression induced Barrett’s-like metaplasia in RCG esophageal epithelium

At E18.5, P63^−/− and P63^−/−;Villin-Cdx2 esophageal epithelium showed a single-layer of cuboidal cells lining the inner surface of the esophagus, which was different from the stratified squamous epithelium of wild-type mice. As expected, P63 was absent in P63^−/− and P63^−/−;Villin-Cdx2 esophageal epithelium, but strongly expressed in the basal cells of wild-type esophageal epithelium (Figure 2A–C). VIL was expressed in the cytoplasm of P63^−/− esophageal epithelium, but not in wild-type esophageal epithelium (Figure 2D–F). CDX2 was expressed in the nuclei of scattered basal cells of P63^−/−;Villin-Cdx2 esophageal epithelium, but not in wild-type and P63^−/− esophageal epithelium (Figure 2G–I).

Figure 2.

Expression of P63, CDX2, and VIL in the esophageal epithelial cells of wild-type, P63^−/− and P63^−/−;Villin-Cdx2 mice (E18.5). Scale=50μm.

The gross views of P63^−/− and P63^−/−;Villin-Cdx2 RCG esophagus at Week 1, 2, and 4 were different from those of wild-type RCG esophagus. Over time its color turned from transparent (Week 1 and 2) to semi-transparent (Week 4) (Figure 3A, D, G, J, L, Q). It seemed that the RCG esophagus was bloated up by mucus inside the lumen at Week 2 and 4. P63^−/− RCG esophagus at Week 1, 2, and 4 were lined by a single layer of cuboidal cells. Intermittent islands of stratified squamous epithelium were also observed in some samples at Week 2 (Figure 3B, B’, E, E’, H, H’). P63^−/−;Villin-Cdx2 RCG esophagus at Week 1 and 2 was lined by a single layer of columnar cells with massive mucus in the cytoplasm (Figure 3K, K’, O, O’). Interestingly, at Week 4, P63^−/−;Villin-Cdx2 RCG esophagus developed a single layer of columnar cells with intestinal crypt-like structures. The columnar cells in the crypts were filled with mucus in the cytoplasm (Figure 3R and R’).

Figure 3.

Morphology (gross view and H&E staining) and PAS/AB staining of P63^−/− and P63^−/−;Villin-Cdx2 RCG esophagus at 1, 2 or 4 weeks. Red arrows indicate the esophagus under renal capsule. Panel B’, E’, H’, K’, O’, R’ are enlarged parts of Panel B, E, H, K, O, R. Panel C’, F’, I’ M’, P’, S’ are enlarged parts of Panel C, F, I, M, P, S. Scale bar=50μm.

Only a small proportion of cuboidal cells containing acidic mucus (PAS/AB+) in the cytoplasm was observed in P63^−/− RCG esophageal epithelium at Week 1 and 2. At Week 4, more PAS/AB+ cuboidal cells appeared in P63^−/− RCG esophageal epithelium (Figure 3C, C’, F, F’, I, I’). Nearly all the columnar cells lining the P63^−/−;Villin-Cdx2 RCG esophagus at Week 1, 2, and 4 contained acidic mucus in the cytoplasm (Figure 3M, M’, P, P’, S, S’). Besides, acidic mucus tended to accumulate in the lumen over time.

Expression of proliferation marker, goblet cell markers, and intestinal transcription factors in P63^−/−;Villin-Cdx2 RCG esophagus at Week 4

P63^−/−;Villin-Cdx2 RCG esophagus at Week 4 was lined with a single-layer columnar epithelium with the cytoplasmic expression of VIL but not the nuclear expression of P63 (Figure 4A–B). CDX2 was strongly expressed in the nuclei of scattered columnar cells of the RCG esophageal epithelium and weakly expressed in most other cells (Figure 4C). BrdU was positive in some columnar cells in the crypt (Figure 4D). These data demonstrated that P63^−/−;Villin-Cdx2 RCG esophagus maintained its genetic manipulations and the epithelial cells remained proliferative.

Figure 4.

Expression of P63, VIL, CDX2, proliferation marker (BrdU), goblet cell markers (TFF3, MUC2), and intestinal transcription factors (CDX1, HNF1α, HNF4α, GATA4, GATA6) in P63^−/−;Villin-Cdx2 RCG esophagus at Week 4. Scale bar=50μm.

Since we did not observe typical morphology of goblet cells (barrel-shaped cells with small nuclei and around or piriform theca) on H&E stained sections, we examined the expression of two goblet cell markers and five intestinal transcription factors, TFF3 was weakly expressed in the cytoplasm of columnar cells, while MUC2 was strongly expressed in the crypt (Figure 4E–F). CDX1, HNF1α, HNF4α, and GATA6 were strongly expressed in the nuclei of columnar cells, while GATA4 was moderately expressed (Figure 4G–K).

Expression of secretory cell marker, enteroendocrine cell maker, Paneth cell markers, Tuft cell marker, intestinal columnar cell markers, and squamous epithelial cell marker in P63^−/− and P63^−/−; Villin-Cdx2 RCG esophagi at Week 4

UEA1/WGA, a key marker of intestinal secretory cells, was expressed in the villi of secretory columnar cells lining the P63^−/− and P63^−/−;Villin-Cdx2 RCG esophagi (Figure 5A–B). CHA, an enteroendocrine cell marker, was expressed in the cytoplasm of P63^−/−;Villin-Cdx2 esophageal epithelium, but not in the P63^−/− esophagus (Figure 5C–D). Paneth cell marker, LYZ, was expressed in the cytoplasm of P63^−/−;Villin-Cdx2 esophageal epithelium, but not in the P63^−/− esophagus (Figure 5E–F). SOX9, a key transcription factor modulating intestinal hemostasis and an intestinal Paneth cell marker [25], was strongly expressed in both P63^−/− and P63^−/−;Villin-Cdx2 esophageal epithelial cells (Figure 5E–H). DCAMKL1, a Tuft cell marker, was expressed weakly in the nuclei of P63^−/− esophageal epithelium, but moderately in the P63^−/−;Villin-Cdx2 esophageal epithelium (Figure 5I–J). Intestinal columnar cell markers, CK8 and A33, were expressed strongly in the P63^−/−;Villin-Cdx2 esophageal epithelium, although some CK8+ cells were also present in the P63^−/− esophageal epithelium (Figure 5K–N). Interestingly, esophageal squamous epithelial cell markers, Sox2 and PAX9, remained positive in both P63^−/−;Villin-Cdx2 and P63^−/− esophageal epithelium (Figure 5O–R).

Figure 5.

Expression of secretory cell marker (UEA1/WGA), enteroendocrine cell maker (CHA), Paneth cell markers (SOX9, LYZ), Tuft cell marker (DCAMKL1), intestinal columnar cell markers (A33, CK8), and squamous epithelial cell marker (PAX9, SOX2) in P63^−/− and P63^−/−;Villin-Cdx2 RCG esophagus at Week 4. Scale bar=50μm.

Discussion

In this study, Barrett’s-like metaplasia was induced in the esophageal epithelium of P63^−/−; Villin-Cdx2 mice in our RCG model. Our data clearly demonstrated that the combination of P63 silencing and Cdx2 overexpression in mouse esophageal epithelium created a transitional phenotype leading to BE.

As a critical initiator of epithelial stratification and a key regulator of cell adhesion and survival progenitor cells in squamous epithelium, P63 dictates squamous differentiation of esophageal epithelium[15]. In P63^−/− mice, stratified epithelial tissues are lost several days after stratification due to the loss of regenerative cell populations, subsequently leading to lethality after birth[22]. Although P63^−/− esophagus showed a remarkably well-developed columnar epithelium and scattered staining of PAS/AB(+) mucin, goblet cells as well as intestinal crypt-like structure are not observed. Meanwhile, goblet cell differentiation markers (MUC2 and TFF3) were not expressed[8]. When esophageal squamous cells were exposed to gastroesophageal reflux, the expression of P63 became down-regulated or silenced [18, 14]. Thus, P63 deficiency alone cannot induce full-blown BE. In this study, when P63^−/− esophagus at E18.5 was grafted beneath the renal capsule and allowed for further growth up to 4 weeks, several intestinal markers (PAS/AB, UEA1/WGA, SOX9, DCAMKL1, CK8) became positive (Figure 3, 5). However, other intestinal markers were not expressed and the intestinal crypt-like structure was not observed. These data support the idea that although P63 deficiency alone was not enough to induce BE, it does inactivate the squamous differentiation program in esophageal epithelial cells and lay down a foundation leading to BE [10].

CDX2 plays an important regulatory role in intestinal development and differentiation[26, 27]. Stomach-specific Cdx2 transgenic overexpression induced intestinal metaplasia in the mouse stomach within weeks after birth [23, 28]. Heterozygous Cdx2 deficiency produced colonic harmatomas with squamous epithelium appearing in the colon [29]. Treatment of human and rodent esophageal squamous epithelial cells with either acid or bile acids, which mimics gastroesophageal reflux, induced expression of CDX2 [10, 20]. CDX2 overexpression in esophageal squamous epithelial cells in vitro induced the expression of intestinal genes, e.g., MUC2, LYZ, and CDX1[20]. Transgenic overexpression of Cdx2 in esophageal squamous epithelial cells in vivo also induced the expression of intestinal genes [21, 30]. However, histological phenotype of BE was not observed in these in vivo models, suggesting that CDX2 overexpression alone cannot drive intestinal metaplasia of esophageal squamous epithelium in vivo. In this study, we, therefore, generated P63^−/−;Villin-Cdx2 mouse and grafted its esophagus in vivo for up to 4 weeks using the RCG model. The purpose was to find out whether CDX2 overexpression on the basis of P63 deficiency may generate intestinal metaplasia in mouse esophageal epithelium. By comparing P63^−/−;Villin-Cdx2 esophagus and P63^−/− esophagus, it was clear that CDX2 overexpression led to the expression of additional intestinal markers such as CHA, LYZ, A33, CDX1, and further enhanced the expression of SOX9, DCAMKL1, CK8 (Figure 5). More importantly, CDX2 overexpression generated a crypt-like structure expressing intestinal markers in mouse esophageal epithelium (Figure 3, 4). This Barrett’s-like phenotype is obviously different from the phenotype of P63 deficiency or CDX2 overexpression as described above. These data suggest that CDX2 overexpression on the basis of P63 deficiency pushed the esophageal epithelial cells to further differentiate into BE.

When we further looked at gene expression in P63^−/−;Villin-Cdx2 esophageal epithelium, BrdU was positive indicating that this epithelium, in particular cells in the crypts, were proliferating. Several intestinal transcription factors (CDX1, HNF1α, HNF4α, GATA4, GATA6) and goblet cell markers (TFF3 and MUC2) were expressed as well (Figure 4). It is known that HNF1α/4α regulate the differentiation of intestinal columnar cells and goblet cell maturation[31, 32]. Exposure of esophageal epithelial cells to bile induced MUC4 expression through HNF1α[33]. GATA4/6 are expressed in the human intestine and BE, and also regulate the differentiation of intestinal epithelial cells[34–36]. GATAs and HNFs played cooperative roles in regulating the expression of marker genes intestinal epithelium and human BE[37–40]. Nevertheless, we failed to observe the typical morphology of goblet cells in P63^−/−;Villin-Cdx2 esophagus at 4 weeks after RCG. Several reasons may explain why typical goblet cell morphology was not observed in our study: 1) It may take a time longer than 4 weeks for P63^−/−;Villin-Cdx2 esophagus develop goblet cells. The dilemma is that the RCG esophagus grows over time; mucin accumulation in the esophageal lumen may impact the differentiation of cells including future goblet cells (Figure 3). 2) Additional drivers, e.g., Notch inhibition, NFκB activation, may be needed to completely inactivate the squamous differentiation program and drive the whole process of intestinal metaplasia. Gastroesophageal reflux is known to inhibit Notch signaling and activate NFκB signaling [41, 42]. Notch signaling promotes and coordinates esophageal squamous cell differentiation, especially interacting with P63[43, 44]. Disruption of Notch signaling resulted in an increased number of goblet cells and enhanced the development of goblet cells in the intestine[45, 46]. NFκB activation up-regulated the expression of key differentiation markers of goblet cells (CDX2 and MUC2) [47]. In Figure 5, one transcription factor (PAX9) which is critical for esophageal squamous epithelial cells but absent in BE[36], remained expressed in both P63^−/− and P63^−/−;Villin-Cdx2 esophagus. It should be noted that PAX9 is a downstream effector of Notch signaling in esophageal epithelial cells (our unpublished data).

Although our data support the basal cells in esophageal epithelium as an origin of BE, this study cannot exclude other cells in the esophagus as possible origins of BE. A recent study has shown that the esophageal submucosal gland cells, seen in both human and porcine esophagus, may provide a source of cells to repopulate damaged epithelium in a normal manner (squamous) or abnormally (columnar epithelium)[48]. In fact, the lack of submucosal glands in rodent esophagus is a limitation of this study and any other studies using rodents. We and others have reported MLE and BE in multiple surgical models in rats and mice[49, 14, 50]. Although these models help us understand the progression of BE in the presence of various settings of gastroesophageal reflux, their contribution to understanding the molecular mechanisms is overall limited due to complex molecular alterations after surgery.

The clinical relevance of our findings remains a question. Human patients after esophagectomy or gastrectomy can be regarded as “human” models for GERD and BE. It has been reported that these patients are highly susceptible to BE[51, 52]. A longitudinal follow-up study of these patients will answer the question of whether P63 silencing and CDX2 overexpression contribute to BE.

In summary, the novel finding of our study is that CDX2 overexpression on the basis of P63 deficiency in mouse esophageal epithelium led to Barrett’s-like metaplasia in vivo, which is characterized by columnar epithelium with an intestinal crypt-like structure expressing intestinal markers, within a short period of time when the esophagus was grafted under the renal capsule. This transitional phenotype appears one step ahead of another transitional phenotype, MLE, which appears in the neo-squamocolumnar junction and within columnar mucosa in patients with BE, and shares the morphologic, ultrastructural, and molecular features of both squamous and columnar epithelium [53, 14, 17]. Further studies are warranted to determine whether additional factors may drive this Barrett’s-like transitional phenotype into full-blown BE.

Supplementary Material

Suppl Figure1

Supplementary Figure 1. Renal capsule grafting of mouse esophagus.

Suppl Figure2

Supplementary Figure 2. Morphology (H&E staining), proliferation (BrdU) and expression of differentiation markers (CK5, CK4, and CK1) in the esophageal epithelium of 1-week-, 2-week- and 4-week-old wild-type mice. Scale bar=50μm

Abbreviations:

BE

Barrett’s esophagus

EAC

esophageal adenocarcinoma

GERD

gastroesophageal reflux disease

H&E

hematoxylin and eosin

MLE

multilayered epithelium

PAS/AB

periodic acid Schiff and Alcian blue

RCG

renal capsule grafting

VIL

villin