THE AMERICAN JOURNAL OF GASTROENTEROLOGY 2001 by Am. Coll. of Gastroenterology Published by Elsevier Science Inc.

Vol. 96, No. 5, 2001 ISSN 0002-9270/01/$20.00 PII S0002-9270(01)02355-3

Cytokeratin Subsets for Distinguishing Barretts Esophagus From Intestinal Metaplasia in the Cardia Using Endoscopic Biopsy Specimens
Hala M. T. El-Zimaity, M.D., and David Y. Graham, M.D. Gastrointestinal Mucosa Pathology Laboratory and Departments of Medicine and Pathology, Veterans Affairs Medical Center and Baylor College of Medicine, Houston, Texas

OBJECTIVES: It has been suggested that Barretts epithelium and intestinal metaplasia in the gastric cardia have different cyotokeratin (CK) staining patterns and that Barretts epithelium can be distinguished by CK staining pattern. The aim of this study was to test the utility of CK staining for distinguishing Barretts esophagus from gastric intestinal metaplasia. METHODS: Topographically mapped gastric biopsy specimens were obtained from patients without Barretts esophagus, and esophageal biopsies were obtained from patients with long-segment Barretts esophagus (3 cm). Serial sections were stained with Genta or El-Zimaity triple stain, and biopsies with intestinal metaplasia were stained with antibodies against CK 4, 13, 7, and 20. RESULTS: Sections from 33 biopsies with Barretts esophagus, 23 with intestinal metaplasia of the gastric cardia, 27 with intestinal metaplasia of the gastric body, and 33 with intestinal metaplasia of the antrum were examined. CK 4 and CK 13 stained squamous epithelium only. The proposed diagnostic CK Barretts 7/20 pattern was found in only 39% of long-segment Barretts compared to 35%, 4%, and 24% in intestinal metaplasia from the gastric cardia, body, and antrum, respectively. The criteria proposed had a sensitivity of 45% and a specicity of 65%. CONCLUSIONS: These results do not support keratin phenotyping as a tool for differentiating intestinal metaplasia originating in the cardia from intestinal metaplasia of Barretts. (Am J Gastroenterol 2001;96:1378 1382. 2001 by Am. Coll. of Gastroenterology)

tion such that the distinction between H. pylorirelated intestinal metaplasia of the cardia and short-segments Barretts esophagus is important. Recently, Ormsby et al. (10), using antibodies to cytokeratin (CK) 7 and CK 20, found that the staining pattern of gastric intestinal metaplasia was entirely different from that of Barretts epithelium. The Barretts CK 7/20 pattern was dened as staining of the supercial epithelium with CK 20 and staining of both the supercial and deep metaplastic epithelium with CK 7. This pattern was present in 97% of specimens with long-segment Barretts and was not observed in gastric intestinal metaplasia (10). They hypothesized that the Barretts CK 7/20 pattern was specic for Barretts epithelium. The current study was designed to conrm their hypothesis and to extend the observation to biopsies taken at endoscopy by investigating the utility of CK 7, 20, 4, and 13 in the histological distinction of Barretts esophagus from intestinal metaplasia in the cardia.

MATERIALS AND METHODS

Patients and Histology Mucosal biopsy specimens were obtained from patients who had previous upper GI endoscopy with gastric mapping which typically involved taking 14 biopsies from specied sites (11). The anatomic cardia was dened as the mucosa immediately below the site of the junction of the mucosa of the tubular esophagus and the stomach (Z-line) and above the beginning of the rst gastric fold. Biopsies of the gastric cardia (mean and median of two biopsies) were always taken antegrade (not retrograde with retroexion of the endoscope). None of the patients had tongues of gastric type epithelium. Each biopsy was placed in a separate bottle of 10% buffered formalin. Biopsies were embedded on edge, sectioned at 5 m with six sections per slide, and stained with the Genta stain (12) or El-Zimaity triple stain (13). The median size of biopsy specimens (xed tissue measured on a glass slide) was 8 4 mm. Each specimen was reviewed by one pathologist and scored using a visual analog scale from 0 (absent/normal) to 5 (maximal intensity) for intestinal metaplasia (14).

INTRODUCTION
Intestinal metaplasia is a common nding in biopsy specimens taken from just below the gastroesophageal junction in patients with current or past Helicobacter pylori infection (15). This nding has prompted considerable discussion regarding whether it is a variant of gastroesophageal reux disease (GERD) (6 9), (e.g., short-segment Barretts), a consequence of H. pylori of the stomach, or both (15). Barretts esophagus is considered a premalignant condi-

AJG May, 2001

Cytokeratin Subsets in Barretts Versus Stomach

1379

Table 1. Cytokeratin 7/20 With Barretts Esophagus and Gastric Intestinal Metaplasia, by Site Disease Barretts esophagus Gastric intestinal metaplasia Cardia Body Antrum N (biopsies) 33 6 complete 27 incomplete 23 8 complete 15 incomplete 27 12 complete 11 incomplete 33 14 complete 19 incomplete Barretts CK7/20 Pattern 13 (39%) 2 (33%) 11 (41%) 8 (35%) 2 (25%) 6 (40%) 1 (4%) 0 1 (9%) 8 (24%) 1 (7%) 7 (37%) Gastric CK7/20 Pattern 15 (15%) 3 (50%) 12 (44%) 7 (30%) 1 (13%) 6 (40%) 7 (26%) 5 (42%) 3 (27%) 5 (15%) 2 (14%) 3 (16%) Other Patterns 5 (15%) 1 (17%) 4 (15%) 8 (35%) 5 (63%) 3 (20%) 17 (63%) 7 (58%) 7 (64%) 20 (61%) 11 (79%) 9 (47%)

Cases were selected because of previously documented intestinal metaplasia. Indications for upper GI endoscopy included duodenal ulcer, gastric ulcer, and previous diagnosis of intestinal metaplasia. Cases with previously documented long-segment Barretts metaplasia were obtained from the les of the Veterans Affairs Medical Center. High IronDiamine Staining Biopsies with intestinal metaplasia were stained with high iron diamine/Alcian blue (HID/AB) to identify neutral, sialo-, and sulfomucins. Briey, slides were immersed in HID solution for more than 18 h, at 2325C. Slides were then rinsed with deionized water and stained with 1% Alcian blue, pH 2.5 for 2 min (15). Subtyping intestinal metaplasia was done according to the system used by Jass and Filipe (16, 17). Type I is classied as complete intestinal metaplasia and type II and III are grouped incomplete metaplasia. Immunohistochemical Studies For immunophenotyping, 5-m thick sections were stained using a modied streptavidin-biotin complex method with antigen retrieval as required. Briey, the following reagents were used in sequential steps at 36C: inhibitor for endogenous peroxidase, primary antibody for 12 h, biotinylated secondary antibody, avidin-biotin complex with horseradish peroxidase, 3,3-diaminobenzidine tetrahydrochloride (DAB). When indicated, slides were pretreated for antigen retrieval by steam for 15 min in a Black and Decker (Schaumburg, IL) steamer in 10 mmol/L citrate buffer (pH 6.0), followed by cooling for 20 min. Slides were counterstained with hematoxylin. The antibody panel included CK 7, 20, 4, and 13 from Dako (Carpinteria, CA). After grading slides for intensity of stain, intestinal metaplasia glands were visually divided into thirds (upper, middle, and lower). The staining pattern was considered supercial if the upper third or two thirds was positive. Slides were then analyzed for the different patterns as dened by Ormsby et al. (10), e.g., supercial staining with CK 20 and strong CK 7 staining of both supercial and deep glands was dened as a Barretts CK7/20 prole. The whole area of metaplasia was also screened for patchy or diffuse staining.

Statistical Analyses Scores were analyzed using Sigma Stat (Jandel Scientic Software, San Rafael, CA). Fishers exact test or, when appropriate, the 2 test (both two-tailed) were used for comparison of proportions.

RESULTS
Intestinal Metaplasia A total of 116 biopsies with intestinal metaplasia from 102 patients were examined (33 Barretts esophagus, 23 cardia, 27 body, and 33 antral). CK 4 and CK 13 stained squamous epithelium only; areas with intestinal metaplasia remained unstained. Intestinal metaplasia was incomplete in 79% and 65% of patients presenting with intestinal metaplasia in the esophagus and cardia, respectively. Intestinal Metaplasia in Barretts A total of 33 biopsies from 29 patients with Barretts esophagus were examined. The proposed diagnostic CK Barretts 7/20 as dened by Ormsby et al. (10) was found in only 39% of biopsies (Table 1) (Fig. 1). Incomplete intestinal metaplasia was present in 27 biopsies from 23 patients; 11 biopsies had the esophageal pattern, 12 biopsies had a gastric pattern, and four had other patterns. Two patients had multiple biopsies. The rst patient had two biopsies and both had the esophageal pattern. The second patient had four biopsies, one had an esophageal pattern, and three had a gastric pattern. Six biopsies from six patients had complete intestinal metaplasia. Only three (50%) had the diagnostic Barrett pattern. Intestinal Metaplasia in the Stomach A total of 83 gastric biopsies with intestinal metaplasia were examined (23 cardia, 27 body, and 33 antral). Intestinal Metaplasia in the Cardia In all, 23 biopsies from 18 patients were examined. The proposed diagnostic CK Barretts 7/20 pattern as dened by Ormsby et al. (10) was found in eight of 23 biopsies

1380

El-Zimaity et al.

AJG Vol. 96, No. 5, 2001

Figure 1. Representative section for cytokeratins 20 (A) and 7 (B) in the esophagus. The proposed diagnostic pattern was observed in only 39% of biopsies. Contrary to the proposed diagnostic pattern, CK 20 was expressed in the entire length of the gland in 17 biopsies (59%), and CK 7 was supercial or mixed, with areas of no staining in eight cases (28%).

(35%), (Table 1). Incomplete intestinal metaplasia was present in 15 biopsies from 11 patients. The esophageal pattern was present in six (40%) (Fig. 2). Four patients had two biopsies each. An esophageal pattern was seen in both biopsies in one patient; the second patient had a gastric pattern in both biopsies; and two patients had a mixed pattern (i.e., one esophageal and one gastric in one patient, and one gastric and one other pattern in the other patient). Complete intestinal metaplasia was present in eight biopsies from seven patients. The diagnostic pattern was present in two of eight biopsies with complete intestinal metaplasia (25%). One patient had two biopsies, one biopsy had an esophageal pattern and the other biopsy had supercial staining of the glands with CK 7 and CK 20 (other pattern). Intestinal Metaplasia in the Corpus The proposed diagnostic CK Barretts 7/20 pattern as dened by Ormsby et al. (10) was found in only one of 27 biopsies (3%) (95% C.I. 0% to 20%) (p 0.0036 com-

pared to Barretts) (Table 1). The proposed diagnostic CK Barretts 7/20 pattern was found in only one of 11 biopsies (11 patients) with incomplete intestinal metaplasia. Although CK 7 stained two cases with incomplete intestinal metaplasia in the body, the characteristic Barretts staining of the entire length of the gland was observed in one patient only. Similarly, complete intestinal metaplasia was present in 12 biopsies from 11 patients; none had the Barretts pattern. Intestinal Metaplasia of the Antrum The proposed diagnostic CK Barretts 7/20 pattern as dened by Ormsby et al. (10) was found in nine of 33 biopsies (27%) (Table 1). Incomplete intestinal metaplasia was present in 19 biopsies from 16 patients. Seven biopsies (37%) had the esophageal pattern. Three patients had two biopsies each. One patient had an esophageal pattern and a gastric pattern. The other patient had an esophageal pattern, and supercial staining with CK 20 with no staining with

Figure 2. Representative section for cytokeratins 20 (A) and 7 (B) in the cardia. The diagnostic cytokeratin Barretts 7/20 was found in 35% of biopsies. Supercial staining with CK 20 was observed in 65% of biopsies, and CK 7 stained the entire length of the gland in 52%.

AJG May, 2001

Cytokeratin Subsets in Barretts Versus Stomach

1381

CK 7. The last patient had supercial staining with CK 20 and no staining with CK 7 in both biopsies (other pattern). Complete intestinal metaplasia was present in 14 biopsies from 12 patients; only one biopsy had the esophageal pattern. Two patients had two biopsies each. One had a gastric pattern in both biopsies, and the other patient had supercial staining with CK 20 and no staining with CK 7 in both biopsies.

DISCUSSION
Intestinal metaplasia of the gastric cardia is common and is thought to be primarily related to H. pylori pathology (15). The question is whether a biopsy taken near the esophagogastric junction that shows intestinal metaplasia represents Barretts metaplastic epithelium or is it related to H. pylori gastritis. The differential expression of intermediate lament proteins is closely linked with specic programs of differentiation. Different epithelia have different cyotokeratin proles; CK 7 and CK 20 are markers for glandular differentiation; CK 4 and CK 13 are markers for squamous differentiation (18). Ormsby et al. studied surgical specimens of Barretts epithelium and hypothesized the presence of a unique CK prole, designated the Barretts CK 7/20 pattern, which they observed only in Barretts metaplastic epithelium (10). They suggested that this CK prole might be useful in differentiating intestinal metaplasia in the cardia from Barretts intestinal metaplasia. We were unable to conrm that hypothesis using endoscopically obtained biopsies, because we found the esophageal pattern as dened by Ormsby et al. (10) in only 45% of cases with Barretts and in 35% of patients with intestinal metaplasia in the cardia (p 0.75). The sensitivity and specicity of the Barretts CK 7/20 pattern was 45% and 65%, respectively. The expression of CKs differed among regions of the stomach with CK 7 being the most distinctive. Cytokeratin 7 was expressed in 100%, 59%, 20%, and 46% in the esophagus, cardia, corpus, and antrum, respectively. Absence of CK 7 expression might be useful to distinguish Barretts metaplastic epithelium from intestinal metaplasia in the cardia but it would have low sensitivity as CK 7 was expressed in approximately one half of the biopsies with intestinal metaplasia in the cardia. The distinction between Barretts and intestinal metaplasia in the cardia can be suggested on morphological grounds. Criteria (19 21) include the absence of enterochromafn cells and pancreatic metaplasia in Barretts metaplastic epithelium and the presence of submucosal esophageal glands, which are generally not present in the cardia (21). The muscularis mucosae in the esophagus are separated from the basal epithelium in the esophagus by an epithelium-free layer of lamina propria (21). Intestinal metaplasia in Barretts typically has a villiform surface, a pattern rarely seen in the cardia (21). Mucin histochemistry has been investigated in Barretts esophagus; however, incom-

plete intestinal metaplasia can be seen in both places (22) as was conrmed in this study. Pathologists depend on gastroenterologists for knowing the location of endoscopically obtained biopsies. One difference between our study and the original study could be differences in sampling (site, size of specimens, etc.). The cardia is a very small area and it would be very difcult to reliably sample in a retroexed manner especially if a hiatal hernia is present (D. Y. Graham, personal observation). We speculate that much of the confusion regarding short-segment Barretts and intestinal metaplasia of the cardia (19) is related to sampling technique. Of interest, the proposed Barretts CK 7/20 pattern (10) was found in only one of 27 biopsies with intestinal metaplasia in the corpus such that, if biopsies taken from the corpus were incorrectly labeled cardia, a statistically signicant difference would have been achieved between Barretts intestinal metaplasia and cardia intestinal metaplasia. These data and the increasing incidence of carcinoma in the cardia (2326) suggest that gastroenterologists should standardize methods for obtaining biopsies from the cardia for both diagnostic and research purposes. The cardia mucosa is located between the esophageal and fundi mucosa. Previous anatomic and endoscopic studies have shown the cardia to be a small structure varying from about 0 to 10 mm in length (mean, 3 mm) (20, 2729) beneath the gastroesophageal junction. As such, sampling can best be reliably achieved with an anterograde approach. Retroexion to obtain biopsies would more likely miss the cardia and sample the corpus than those taken directly adjacent to the Z line would. In addition, to the anatomic landmark of the Z line, in the experience of the senior endoscopist (D.Y.G.) it is generally possible to target the cardia mucosa directly based on the combination of the location and the appearance of a pinkish white mucosa with a delicate vascular pattern immediately distal to the typical appearance of esophageal squamous mucosa and proximal to the beginning of the rst gastric fold.

ACKNOWLEDGMENTS
This work was supported by the Department of Veterans Affairs and by NIH grant DK53659 as well as the generous support of Hilda Schwartz.
Reprint requetsts and correspondence: Hala M.T. El-Zimaity, M.D., Room 3A352, VA Medical Center (111-D), 2002 Holcombe Boulevard, Houston, TX 77030. Received July 5, 2000; accepted Jan. 31, 2001.

REFERENCES
1. Morales TG, Sampliner RE, Bhattacharyya A. Intestinal metaplasia of the gastric cardia. Am J Gastroenterol 1997;92: 414 8. 2. Hackelsberger A, Gunther T, Schultze V, et al. Prevalence and

1382

El-Zimaity et al.

AJG Vol. 96, No. 5, 2001

3.

4.

5. 6.

7.

8.

9.

10.

11.

12.

13.

14.

15.

pattern of Helicobacter pylori gastritis in the gastric cardia. Am J Gastroenterol 1997;92:2220 4. Goldblum JR, Vicari JJ, Falk GW, et al. Inammation and intestinal metaplasia of the gastric cardia: The role of gastroesophageal reux and H. pylori infection. Gastroenterology 1998;114:6339. el-Serag HB, Sonnenberg A, Jamal MM, et al. Characteristics of intestinal metaplasia in the gastric cardia. Am J Gastroenterol 1999;94:6227. Craanen ME, Blok P, Dekker W, Tytgat GN. Helicobacter pylori and early gastric cancer. Gut 1994;35:1372 4. Voutilainen M, Farkkila M, Juhola M, et al. Specialized columnar epitheilium of the esophagogastric junction: Prevalence and associations. Am J Gastroenterol 1999;94:913 8. Pereira AD, Suspiro A, Chaves P, et al. Short segments of Barretts epithelium and intestinal metaplasia in normal appearing oesophagogastric junctions: The same or two different entities? Gut 1998;42:659 62. Oberg S, Peters JH, DeMeester TR, et al. Inammation and specialized intestinal metaplasia of cardiac mucosa is a manifestation of gastroesophageal reux disease. Ann Surg 1997; 226:52232. Spechler SJ, Zeroogian JM, Antonioli D, et al. Prevalence of metaplasia at the gastro-oesophageal junction. Lancet 1994; 344:1533 6. Ormsby AH, Goldblum JR, Rice TW, et al. Cytokeratin subsets can reliably distinguish Barretts esophagus from intestinal metaplasia of the stomach. Hum Pathol 1999;30:288 94. El-Zimaity HM, Al-Assi MT, Genta RM, Graham DY. Conrmation of successful therapy of Helicobacter pylori infection: Number and site of biopsies or a rapid urease test. Am J Gastroenterol 1995;90:1962 4. Genta RM, Robason GO, Graham DY. Simultaneous visualization of Helicobacter pylori and gastric morphology: A new stain. Hum Pathol 1994;25:221 6. El-Zimaity HMT, Ota H, Scott S, et al. A new triple stain for Helicobacter pylori suitable for the autostainer. Arch Pathol Lab Med 1998;122:732 6. El-Zimaity HMT, Graham DY, Al-Assi MT, et al. Interobserver variation in the histopathological assessment of Helicobacter pylori gastritis. Hum Pathol 1996;27:35 41. Spicer SS. Diamine methods for differentiating mucosub-

16.

17. 18. 19. 20. 21.

22. 23. 24. 25. 26. 27. 28. 29.

stances histochemically. J Histochem Cytochem 1965;13:211 34. Rokkas T, Filipe MI, Sladen GE. Detection of an increased incidence of early gastric cancer in patients with intestinal metaplasia type III who are closely followed up. Gut 1991; 32:1110 3. Silva S, Filipe MI, Pinho A. Variants of intestinal metaplasia in the evolution of chronic atrophic gastritis and gastric ulcer. A follow up study. Gut 1990;31:1097104. Moll R, Franke WW, Schiller DL, et al. The catalog of human cytokeratins: Patterns of expression in normal epithelia, tumors and cultured cells. Cell 1982;31:1124. Petras RE, Sivak MV, Rice TW. Barretts esophagus: A review of the pathologists role in diagnosis and management. Pathol Annu 1991;26:132. Krause WJ, lvey KJ, Baskin WN, et al. Morphological observations on the normal human cardiac glands. Anat Rec 1978; 192:59 72. Appelman HD, Kalish RJ, Clancy PE, et al. Distinguishing features of adenocarcinoma in Barretts esophagus and in the gastric cardia. In: Spechler SJ, Goyal RK, eds. Barretts esophagus: Pathophysiology, diagnosis, and management, 3rd ed. New York: Elsevier, 1985:167 87. Ellison E, Hassall E, Dimmick JE. Mucin histochemistry of the developing gastroesophageal junction. Pediatr Pathol Lab Med 1996;16:195206. Antonioli DA, Goldman H. Changes in the location and type of gastric adenocarcinoma. Cancer 1982;50:775 81. Blot WJ, Devesa SS, Kneller RW, et al. Rising incidence of adenocarcinoma of the esophagus and gastric cardia. JAMA 1991;265:12879. Pera M, Cameron AJ, Trastek VF, et al. Increasing incidence of adenocarcinoma of the esophagus and esophagogastric junction. Gastroenterology 1993;104:510 3. Craanen ME, Dekker W, Blok P, et al. Time trends in gastric carcinoma: Changing patterns of type and location. Am J Gastroenterol 1992;87:5729. Bensley RR. The cardiac glands of mammals. Am J Anat 1902;2:10556. Miyagawa Y. The exact distribution of the gastric glands in man and in certain animals. J Anat 1921;55:56 67. Takubo K, Mafune K, Tanaka Y, et al. Pathology of the cardia. Nippon Geka Gakkai Zasshi 1998;9:54751.