Step one: Transition from squamous to columnar-lined esophagus
To understand what constitutes a columnar-lined esophagus an understanding of the anatomy and histology of the normal gastroesophageal junction is required. Unfortunately, the very definition of what is normal in this area remains controversial, with much debate centered on whether cardiac mucosa is normally present at the gastroesophageal junction. Although our understanding is gradually improving, Hayward’s remark in 1961 that “the lower end of the esophagus is a region where the pathology, the physiology, and even the anatomy are not quite clear” remains appropriate even today. In one of the first reports describing the normal gastroesophageal junction, Hayward indicated that a junctional or buffer zone of columnar mucosa is normally interposed between the acid-secreting oxyntic gastric mucosa and the acid-sensitive squamous esophageal mucosa. Although an appealing concept, Hayward provided no data in support of his theory, and did not discuss the role of the lower esophageal sphincter which had been demonstrated to exist before his publication. According to Hayward, this junctional mucosa is normally found in a length of up to 2 cm at the gastroesophageal junction. He also noted the following about this junctional mucosa: (1) it was histologically distinct from normal gastric fundic and pyloric epithelium; (2) it did not secrete acid or pepsin but was resistant to both; (3) it was not congenital but acquired; (4) it was mobile and varied in length - creeping progressively higher into the esophagus with continued gastroesophageal reflux; and (5) it was potentially reversible with correction of reflux. Furthermore, he pointed out that it was located in the esophagus, and that it developed in association with gastroesophageal reflux.
Now, over 40 years later, there is still dispute about the histology of the normal gastroesophageal junction, but it is clear that normally there is none or at most 4 mm of cardiac mucosa in the distal esophagus at the gastroesophageal junction[10-13]. Longer lengths of cardiac mucosa are acquired secondary to chronic gastroesophageal reflux[14,15]. Supporting evidence for the concept that cardiac mucosa is acquired comes from both clinical and experimental studies. Experimental evidence comes from a 1970 study by Bremner et al in which a series of dogs underwent stripping of the distal esophageal squamous mucosa, with or without creation of a cardioplasty to destroy the function of the lower esophageal sphincter. Squamous re-epithelialization occurred in those animals without gastroesophageal reflux, whereas in the animals with reflux after cardioplasty, the esophagus was re-epithelialized by a columnar epithelium that lacked parietal cells - the equivalent of cardiac mucosa in humans. There is also clinical evidence in humans that columnar mucosa can replace normal esophageal squamous epithelium in the setting of gastroesophageal reflux. Following an esophagectomy with gastric pull-up, reflux of gastric juice into the residual esophagus is common because there is no lower esophageal sphincter and a large hiatal hernia has been created. Postoperative endoscopy has revealed that many of these patients develop columnar epithelium that, on histology, is identical to cardiac mucosa proximal to the anastomosis in the residual esophagus, in what had pathologically been proven to be squamous mucosa at the time of the operation. Several series have revealed that this process is common, and occurs in ≥ 50% of patients after esophagectomy with gastric pull-up, and that the length of columnar mucosa increases with longer follow-up[8,17-20]. Furthermore, the cardiac mucosa that develops in these patients proximal to the esophagogastric anastomosis has been shown to be biochemically similar to cardiac mucosa found in non-operated patients at the native gastroesophageal junction. Additional support for the concept that cardiac mucosa is acquired comes from the fact that it is not found anywhere else in the gastrointestinal tract, and when present at the gastroesophageal junction, it is always inflamed and demonstrates reactive changes unrelated to either Helicobacter pylori infection or mucosal pathology elsewhere in the stomach. This is atypical for a normal epithelium. Lastly, the presence of cardiac mucosa can be correlated with objective markers of GERD, including an incompetent lower esophageal sphincter, increased esophageal acid exposure on 24-h pH monitoring, a hiatal hernia, and erosive esophagitis.
The earliest manifestation of GERD might in fact be the presence of microscopic foci of cardiac mucosa at the gastroesophageal junction. This leads to the question of why the finding of a microscopic length of cardiac mucosa at the gastroesophageal junction is so common even in patients without the typical reflux symptoms of heartburn or regurgitation. This is likely to be related to the pathophysiology of early reflux disease. Evidence is accumulating that reflux disease begins with gastric distention after large and particularly fatty meals. Gastric distension leads to effacement of the lower esophageal sphincter and exposure of the squamous mucosa at the distal extent of the sphincter to gastric juice. The pathophysiology of the gastroesophageal junction has been best studied by Fletcher and McColl. They have noted that the gastric distension that occurs with eating can cause the lower esophageal sphincter to unfold by almost 2 cm in normal volunteers. Moreover, they have identified an unbuffered acid pocket at the gastroesophageal junction following a meal; a phenomenon that they have attributed to gastric juice floating upon a lipid layer after ingestion of fatty food. By pulling back a pH catheter before and after a meal, they have been able to show that the pH step-up that corresponds to the functioning lower esophageal sphincter moved proximally with gastric distension, secondary to unfolding of the distal portion of the sphincter. By measuring acid exposure with a pH catheter positioned at the squamocolumnar junction, and another located 5.5 cm proximal to the squamocolumnar junction, Fletcher et al have demonstrated significantly greater acid exposure at the squamocolumnar junction (median total percentage time pH < 4 of 11.7% vs 1.8% at 5.5 cm proximal to the squamocolumnar junction). This study has confirmed the presence of significant acid exposure at the most distal intrasphincteric segment of the esophagus in patients with otherwise normal acid exposure proximally at 5.5 cm above the squamocolumnar junction. These findings were subsequently extended when it was demonstrated that salivary nitrite is rapidly converted into nitric oxide when it comes in contact with gastric acid that contains physiological levels of ascorbic acid, and this reaction has been found to be maximal at the gastroesophageal junction. The levels of nitric oxide generated at the gastroesophageal junction are potentially mutagenic, and might play a role in the pathophysiology of this region.
It is likely that continued injury to the distal esophagus and lower esophageal sphincter leads to progressive loss of the abdominal length of the sphincter. What started as transient sphincter unfolding with gastric distension gradually progresses to permanent sphincter destruction. With destruction of the sphincter, reflux disease is allowed to explode into the esophagus, and can lead to an increase in the length of cardiac mucosa, either as tongues or as a circumferential replacement of the distal esophageal squamous mucosa. This leads to progressive migration of the squamocolumnar junction proximally[25,26]. Confirmation of esophageal submucosal glands deep to areas lined by cardiac mucosa provides clear evidence that the development of cardiac mucosa is occurring in the esophagus in areas previously covered with squamous mucosa and not in the proximal stomach.
The precise details of the molecular mechanism by which squamous mucosa is transformed into cardiac mucosa remain unknown. However, there is likely to be a crucial interaction between normally sequestered esophageal stem cells and an intraluminal stimulus that drives this metaplastic process. Tobey et al have demonstrated that exposure of esophageal squamous mucosa to gastric juice produces dilated intercellular spaces that allow molecules of up to 20 kDa to permeate down to the stem cells in the basal layer. Perhaps the sensation of heartburn occurs as a consequence of diffusion of hydrochloric acid through these intercellular spaces and stimulation of sensory afferent nerves. These ultrastructural changes occur before gross or microscopic changes become apparent. Thus, one possibility is that factors present in the refluxed juice that gain access to the basal layer stem cells via these dilated intercellular spaces induce a phenotypic transformation such that cardiac columnar mucosal cells rather than squamous cells are produced.
Step two: Intestinalization of cardiac mucosa
Cardiac mucosa is thought to be an unstable epithelium, in part because of the severe inflammatory and reactive changes present on histology. It is hypothesized that cardiac mucosa progresses down one of two possible pathways, based on a combination of environmental and genetic factors. One pathway involves the expression of gastric genes and leads to the formation of parietal cells within glands below the cardiac mucosa. Gastric differentiation leads to a mucosa called oxyntocardiac mucosa, and this is thought to represent a regressive or favorable change because oxyntocardiac mucosa is not premalignant, and appears to be protected from developing intestinal metaplasia. In the second pathway, expression of intestinal genes causes the formation of goblet cells within cardiac mucosa. In contrast to gastric differentiation, intestinal differentiation represents a progressive or unfavorable change because this mucosa is premalignant. Both oxyntocardiac mucosa and Barrett’s esophagus have less inflammation than cardiac mucosa, which suggests that these mucosal types are more stable epithelia.
The development of goblet cells marks the transformation of cardiac mucosa into intestinal metaplasia. When an endoscopically visible length of this mucosa is present in the esophagus, the definition of Barrett’s esophagus has been met. While gastroesophageal reflux is known to be the primary factor responsible for the development of Barrett’s esophagus, the specific cellular events that lead to the transformation of cardiac mucosa into intestinalized cardiac mucosa are unknown. However, evidence is accumulating that intestinalization requires a specific condition or stimulus, and that Barrett’s esophagus occurs in a stepwise process. The first step, from squamous to cardiac mucosa, is likely to occur in response to acid reflux. The second step, development of intestinal metaplasia, is likely to occur in response to a different type of luminal insult. Numerous studies have demonstrated that, although isolated acid reflux can cause esophagitis, Barrett’s esophagus is associated with the presence of a mixture of acid and bile salts[30-32]. Furthermore, clinical experience dating back 30 years has suggested a role for refluxed bile in the development of intestinal metaplasia. In 1977, Hamilton and Yardley observed the development of columnar mucosa and intestinal metaplasia above the esophagogastric anastomosis in a group of patients after esophagectomy. They noted that “severe symptoms of gastroesophageal reflux and bile staining of the refluxed material were documented only in the group with Barrett’s. In addition, pyloroplasty had been performed more commonly in this group.”. Recently, in two separate analyses of patients with reflux with and without Barrett’s esophagus, we found that the factor most associated with the presence of Barrett’s esophagus in both men and women with GERD was abnormal bilirubin reflux, as determined by Bilitec monitoring[34,35].
Fitzgerald et al have reported several interesting observations on how the dynamics of mucosal exposure to luminal contents might affect columnar epithelial cell proliferation and differentiation. Using cultured human Barrett’s esophagus biopsy specimens, they have demonstrated that continuous exposure to acidic media at pH 3.5 resulted in increased villin expression (a marker for epithelial cell differentiation) and reduced cell proliferation. Villin expression was not detected when the culture medium was made more acidic (pH < 2.5). In contrast, a dramatic increase in proliferation occurred when the Barrett’s esophagus tissue was exposed to a short (1 h) pulse of acidic medium (pH 3.5) followed by a return to neutral pH. Clinically, this same group has noted that effective acid suppression results in a shift of the Barrett’s epithelium away from proliferation and toward differentiation. However, the cellular consequences of duodeno-gastroesophageal reflux in the setting of gastric alkalization with acid suppression medications were not addressed in that study.
It has been hypothesized that the mechanism by which acid and bile interact to cause Barrett’s esophagus is related to the ionized state of bile salts. It appears that in a weakly acidic environment certain bile acids are particularly toxic. At pH 3-6, these bile salts are soluble and non-ionized, and can enter mucosal cells, accumulate, and cause direct cellular injury. When the luminal pH is higher than the pKa, these same bile acids are ionized and cannot cross the phospholipid membrane. Further, when the luminal pH is lower, as normally it is in the stomach, bile acids precipitate out of solution and are harmless. Thus, it is only at this critical pH range of 3-5 that certain bile acids become non-ionized and able to cross the cell membrane. Once inside the cell, the pH is 7 and the bile acids become ionized and are trapped inside the cell where they have been shown to result in mitochondrial injury, cellular toxicity and mutagenesis[41-44]. Consequently, this mid-range gastric pH of 3-5 is a danger zone for patients with duodeno-gastroesophageal reflux.
It remains uncertain whether the transformation of cardiac mucosa to intestinalized cardiac mucosa represents a phenotypic change secondary to the induction of genes, or a mutational event within the columnar cells. Mendes de Almeida and colleagues have demonstrated biochemically that both cardiac mucosa and intestinal metaplasia express sucrase-isomaltase and crypt cell antigen - two small intestine marker proteins; however, in that study only three patients with cardiac mucosa were evaluated. Das has developed a murine monoclonal antibody (DAS-1) that reacts specifically with normal colonic epithelial cells, and subsequently he has found that it also reacts with an unknown epitope in Barrett’s mucosa. Griffel et al have reported that the DAS-1 antibody stained cardiac mucosa without intestinal metaplasia in seven patients, and that six of these patients later developed histological evidence of intestinalization on repeat biopsies. Likewise, we noted that the pattern of immunostaining with cytokeratins 7 and 20 was similar in cardiac mucosa and Barrett’s esophagus. These findings suggest that, biochemically, cardiac mucosa and intestinal metaplasia are similar, and that cardiac mucosa is the precursor of intestinalized columnar epithelium, or Barrett’s esophagus.
Currently, the length of Barrett’s esophagus is divided into short (< 3 cm) and long (≥ 3 cm) segments based on the endoscopically determined length of the columnar streak or column in the distal esophagus. Clinically, patients with long-segment Barrett’s esophagus tend to have more severe reflux disease than those with short-segment disease. Patients with long-segment Barrett’s esophagus have a higher prevalence of hiatal hernia, more commonly have a defective lower esophageal sphincter, and demonstrate greater esophageal acid and bilirubin exposure on 24-h pH and Bilitec monitoring[30,49]. Despite the differences in length, there is evidence that short and long-segment Barrett’s esophagus are biochemically similar[48,50]. This is supported by the clinical observation that the risk of malignancy is similar for both short and long segments of Barrett’s esophagus.
The presence of goblet cells is the sine qua non of Barrett’s esophagus. The likelihood of finding intestinalization correlates with the length of the columnar segment. Once 4 cm of cardiac mucosa are present in the distal esophagus, nearly all patients will be found to have intestinal metaplasia on biopsy[49,52]. However, the location of goblet cells in a columnar-lined segment is not uniform, and often the entire length of columnar esophagus does not demonstrate intestinal metaplasia. Goblet cell density is greatest near the squamocolumnar junction and becomes more variable distally. In other words, if intestinal metaplasia is present within a columnar-lined segment of the esophagus, it will always be present proximally at the squamocolumnar junction. Goblet cells might extend throughout the entire length of the columnar segment. The length of Barrett’s esophagus is determined by the endoscopic length of columnar mucosa and not by the length of mucosa showing intestinal metaplasia. In other words, a 6-cm segment of columnar mucosa with intestinal metaplasia only at the proximal 1 cm is still considered long-segment Barrett’s esophagus, but the clinical behavior of this long-segment Barrett’s might differ substantially from a 6-cm segment of columnar mucosa with intestinal metaplasia throughout the entire length. The current definition of Barrett’s esophagus does not take this into account.
The time course to develop goblet cells is uncertain, but it appears to take a minimum of 5-10 years[38,53]. Studies involving esophagectomy patients have indicated that cardiac mucosa develops rapidly, often within 1-2 years. Intestinalization of the columnar segment in these patients occurs significantly later, typically after another 3-5 years[18-20,33,54]. These findings might reflect an accelerated course of events because these patients often have significantly greater reflux of acid and bile than the typical patient with GERD. However, this clinically relevant human model does demonstrate the two-step process of Barrett’s esophagus, starting with columnarization followed by intestinalization in some patients.
The molecular mechanisms by which cardiac mucosa acquires goblet cells remain to be elucidated. However, there is increasing evidence that expression of the homeobox gene Cdx-2 plays a pivotal role. The expression of this gene increases with progression from squamous mucosa with esophagitis to cardiac mucosa, and is maximal in the setting of intestinal metaplasia[55-57]. Experimental work has suggested that Cdx-2 expression can be modulated by the pH of luminal material. Furthermore, an individual’s response to an inflammatory stimulus might also participate in the mucosal adaptation to reflux disease. Fitzgerald et al have demonstrated that esophagitis and Barrett’s esophagus have distinct cytokine profiles that reflect different inflammatory responses to reflux-induced injury. Moreover, even within a given Barrett’s esophagus segment, the inflammatory response is more severe at the proximal end near the squamocolumnar junction, which could explain the greater tendency for intestinalization to occur at this location. Furthermore, the specific cytokine polymorphism of a given individual might also influence the development of Barrett’s esophagus. Preliminary work from Gough et al, for example, has demonstrated that specific polymorphisms of interleukin (IL)-1 receptor antagonist and IL-10 are more common in patients with Barrett’s esophagus than those with esophagitis. Thus, a genetically determined inflammatory response to reflux might influence the pathway of disease in each individual patient.