Introduction

The pathogenesis of gastric cancer is a multifactorial process in which both environmental and host-related factors play significant roles. The risk of developing gastric cancer is closely related to Helicobacter pylori-associated progressive gastric inflammation [2, 5, 26, 32]. The degree of damage is initially most severe in the antrum, but over time the damage progresses into the gastric corpus and can be visualized as an advancing atrophic border, which involves the lesser curve more rapidly than the greater curve [2, 8, 16, 22]. Intestinal type gastric cancers are thought to evolve through a multi-step process starting with superficial gastritis and progressing through atrophy, followed by the development of dysplasia, and finally carcinoma [1]. The role of intestinal metaplasia (IM) in the transition from superficial gastritis to cancer is unclear. There are data suggesting that cancer and IM arise from different cell lineages such that IM may not be a precursor lesion but rather is likely a marker for increased risk [6, 17, 29].

Gastric IM was classified as complete (small intestine) or incomplete (colonic) using immunohistochemical staining techniques and into three types based on the staining pattern of its mucins: I (complete), and II and III (incomplete). Type I is the most common type; Paneth cells are present and goblet cells secrete sialomucins. Types II and III are characterized by the presence of columnar cells and goblet cells secreting sialomucins and/or sulfomucins, and columnar cells secret sialomucins in type II and sulfomucins in type III [12]. Although several studies have suggested that patients with type III IM may be at a higher risk for development of gastric cancer [3, 34], the predictive value of type I and type II IM remains unclear.

Mucins are heavily glycosylated proteins that constitute the major component of the mucous protective layer above mucous surfaces. Core proteins for 12 human mucins (MUC1, 2, 3, 4, 5AC, 5B, 6, 7, 8, 9,11,12) were identified [15, 23, 30, 39]. The normal gastric mucosa shows cell type specific expression of MUC1, 5AC, and 6, and does not express MUC2. Alterations of the expression pattern of mucins were described in carcinoma and IM. Loss of expression of MUC5AC and increased mucin heterogeneity were reported in gastric cancer, whereas under-expression of MUC1, MUC5AC and MUC6 and de novo expression of MUC2 were described in IM. IM was also classified based upon the cell classification status using both gastric and intestinal phenotypic markers [19, 36–38]. In that system, IM is divided into two major types; a gastric and intestinal mixed (GI) type, and a solely intestinal (I) type. MUC5AC and MUC6, galactose oxidase-Schiff, etc. are markers of gastric phenotype, and MUC2, CD10, Villin, sucrase, etc. are intestinal phenotypic markers [19, 35–38].

In retrospective studies, type III IM was found in 75% to 90% of cases of intestinal type gastric cancer that underwent surgical resections [13, 24]. The presence of type III IM was also associated with an increased risk of developing early gastric cancer [3, 14, 28] but not with diffuse type gastric carcinoma [20].

IM is a manifestation of atrophic mucosa occurring in patients with chronic H. pylori gastritis. Type I IM is commonly found in random biopsies of patients with H. pylori infection and the prevalence increases with age. Because detection of IM in routinely obtained endoscopic biopsy materials is common, its presence alone is not considered a reliable marker for identifying patients in whom surveillance is indicated [11]. In addition, biopsies taken repeatedly from predetermined points showed that neither the presence nor the type of IM was constant [7].

There are a number of studies suggesting that the presence of type III IM may be a risk factor for the intestinal type gastric cancer. However, the majority of these studies involved surgical specimens of human stomach with gastric cancer [8, 18, 24, 25] such that while it is possible to conclude that type III IM is frequently present in patients with gastric cancer, little can be said regarding its predictive value. The same problem is present regarding the role of altered mucin expression pattern in IM as most of the studies have come from patients who already had gastric cancers. There are no data regarding the potential value of mucin expression pattern as a risk factor for gastric cancer using fixed-point biopsy samples. We previously used endoscospic biopsies from predetermined sites and found that the presence of IM at the corpus lesser curve was the prevalent finding in patients with early gastric cancer [32].

The aim of the present study was to examine the usefulness of typing of IM and the pattern of mucin expression in IM. We compared the histopathological findings in patients with a history of EMR for early gastric cancer with a suitable control group to identify markers useful to identify patients at increased risk of having or developing gastric cancer.

Materials and methods

This was a case-control study in patients who previously had endoscopic mucosal resection (EMR) for early gastric cancer. The study was performed at the Osaka Medical Center for Cancer and Cardiovascular Diseases and the Aishinkai Nakae Hospital in Japan. Patients were enrolled for the study between November 2003 and November 2005.

Patients

Patients with prior medical EMR for early stage, non-cardiac intestinal type gastric cancer without lymph node metastasis were age and sex-matched with control patients who underwent endoscopy for follow-up of peptic ulcer or for health screening. Patients were excluded if they had received eradication therapy for H. pylori infection, used anti-secretory drugs or non-steroidal anti-inflammatory drugs (NSAIDs), or had predisposing factors of hemorrhagic disease or chronic diseases such as insulin-dependent diabetes mellitus, cirrhosis or renal failure. Drinking and smoking were defined as regular intake when respective consumption was more than 35 g of ethanol and 5 cigarettes per day, respectively.

The study was approved by the Osaka Medical Center for Cancer and Cardiovascular Diseases Ethical Committee, and informed consent was obtained from each patient. The biopsy samples from a subpopulation of these patients (40 controls and 40 gastric cancer patients) were also used in studies regarding loss of Sonic Hedgehog as an indicator of H. pylori-induced atrophic gastritis progressing to gastric cancer and for histological and serum risk markers for non-cardia early gastric cancer [31, 32].

Endoscopic finding of gastric cancer

EMR was done in patients with intestinal type cancers in which the depth of invasion was clinically limited to the mucosa. Patients were followed up at least once per year after EMR. Experienced endoscopists performed all endoscopies.

IM and atrophy grading

Two specimens each were obtained from the greater curvature of the antrum 3 cm proximal to the pylorus and the mid-point of the greater and lesser curvature of the gastric body. A 4-point visual analogue scale was used to grade the mucosal biopsies ranging from zero (absent/normal) to three (maximal intensity) for IM and atrophy [4]. The grading was determined twice for each slide by a single pathologist who was blinded to the previous histological scores and other experimental results. The average histological scores were for the analyses.

Immunohistochemistry

Three adjacent sections 4 μm thick were cut onto each of three polylysine-coated glass slides. The sections were deparaffinized, washed in phosphate–buffer–saline (PBS) and autoclaved for 10 min in sodium citrate buffer for antigenic retrieval. For immunostaining, sections were incubated with diluted primary mouse monoclonal antibodies against MUC2, MUC5AC, MUC6 (Novocastra Laboratories, Newcastle, UK) at a dilution of 1:200, 1:500, 1:500, respectively, at room temperature for 1 h. After washing in PBS, the sections were incubated for 1 h with an anti-mouse immunoglobulin conjugated to horseradish peroxidase dextran labeled polymer using an EnVision TM peroxidase mouse system (DAKO, Carpinteria, CA, USA). The sections were washed in PBS, stained in 0.05% diaminobenzidine hydrochrolide (DAB) solution for 5 min, and counterstained with haematoxylin. After dehydration with xylene, the sections were mounted under a glass coverslip.

Subtyping IM

Biopsies were stained with Alcian blue (AB)/high-iron diamine (HID) to categorise IM as complete type (Type I) or incomplete type (Type II and III). Slides were immersed in HID solution for 20 h at room temperature, then rinsed with deionised water, and stained with 1% AB (pH 2.5) for 2 min. IM subtyping was done according to the system used by Fillipe et al. [12] as follows: Type I, non-secretory absorptive cells and sialomucin secreting goblet cells; Type II, few absorptive cells, columnar cells secreting sialomucin, and goblet cells secreting mainly sialomucin but occasionally sulphomucin; Type III, columnar cells secreting predominantly sulphomucin and goblet cells secreting sialomucin or sulphomucin.

Assay for serum levels of pepsinogen (PG) I and PG II

Blood samples were collected just before endoscopic examination. Serum levels of PG I and PG II were measured by radioimmunoassay (RIA).

Diagnosis of H. pylori infection

H. pylori infection was diagnosed by specific IgG H. pylori antibodies analyzed by an enzyme-linked immunosorbent assay (ELISA) kit (E plate, Eiken Kagaku Inc, Tokyo).

Statistical analyses

Values were expressed as the mean±SD or median and 25–75% range whichever was appropriate. Mantel–Haenszel chi square analyses, Kruskal–Wallis and the non-parametric Mann–Whitney U test were performed to measure differences. Analysis of proportions among the three different groups was performed using chi square analyses. Comparisons were made between those with cancer and the entire control group and control subjects with IM. After comparison by Kruskal–Wallis, Dunn’s methods were used to isolate the groups that differ from the others.

Mantel–Haenszel statistics were used to assess the relationship between gastric cancer and histopathological findings showing the odds ratio (OR) and 95% confidence interval (CI). Risk factors for gastric cancer were analyzed by multivariate logistic regression analyses. A two sided p value of less than 0.05 was considered statistically significant. All statistical computations were performed using SPSS (version 11.0 for Windows, SPSS Inc, Chicago, IL).

Results

The study groups consisted of 80 patients with a prior EMR for non-cardiac early gastric cancer and 80 age and sex-matched controls. Demographic and clinical characteristics of the study groups are shown in Table 1. Multiple malignant lesions were found in 30 patients (asynchronous 15, synchronous 15); the remaining 50 had single lesions. Malignant lesions in 42 patients were located in the lower stomach (antrum or angulus), and 29 patients had lesions in the upper stomach (corpus). In the remaining nine patients, malignant lesions were located in the two different locations. Malignant lesions in 20 patients were depressed type and those in 50 patients were protruded type. Finally, ten patients had both types of malignant lesions.

Table 1 Patient demographic and clinical characteristics
Full size table

The IM scores for each region of the stomach were significantly higher in the cancer group than in the control group (Fig. 1). There is a significant association between atrophic scores and types of IM in the corpus lesser curve (p<0.001). The frequency of type III IM increased along with the scores of atrophy (Fig. 2). The ratio between the concentration of PG I and PG II (PG I/II) was significantly lower in the type II and type III groups compared to the group without IM (p<0.01). The PG I/II ratio was inversely related to the presence of type III IM and was significantly lower (p<0.05) in the type III group than in the type I group (Fig. 3).

Fig. 1
figure 1

Comparison of IM scores between patients with gastric cancer (▪) and control subjects (▫). The ends of the boxes define the 25th and 75th percentiles with the bold line at the median and error bars define the 10th and 90th percentiles. *p<0.001; p values by Mann–Whitney U test. AG, antrum greater curve; CG, corpus greater curve; CL, corpus lesser curve

Full size image
Fig. 2
figure 2

Correlation between types of IM and atrophic scores in the corpus lesser curve. The frequency of type III IM increased along with the scores of atrophy (p<0.001, by Kruskal–Wallis)

Full size image
Fig. 3
figure 3

Comparison of the ratio between the concentration of PG I and PG II (PG I/II). The PG I/II was significantly lower in the type II and type III groups compared to the control group (p<0.01). The PG I/II ratio was inversely related to the presence of type III IM and was significantly lower in the type III group than in the type I group (p<0.05). *p<0.05, **p<0.01 by Dunn’s methods

Full size image

In the gastric mucosa without IM, MUC5AC expression was detected in the foveolar epithelium and mucous neck cells of both the antrum and corpus. MUC6 expression was detected in the pyloric glands of the antrum and the mucous cells of the neck zone of the corpus. MUC2 was only detected in the mucosa with IM. MUC2 expression displayed a vacuolar staining in most goblet cells. In the gastric mucosa with type I IM, MUC5AC in the goblet cells was detected in 46% in the antrum and 13% in the corpus (Fig. 4). In the gastric mucosa with type II and type III IM, MUC5AC was detected in 83% and 94% of patients, respectively, in the antrum and 57% and 90%, respectively, in the corpus (Figs. 4 and 5). In contrast, expression of MUC6 in IM in the antrum was present in 46% of patients with type II IM and 50% of patients with type III (Table 2).

Fig. 4
figure 4

Immunohistochemical staining of serial sections of type I (*) and II (**) IM in the antrum greater curve (original magnification ×40). a. High-iron diamine–alcian blue (HID–AB) staining shows goblet cells stained in blue (sialomucin) of type I IM (*) and cells stained in brown (sulphomucin) of type II IM (**). b. MUC5 was detected in the mucosa without IM and type I IM (*)

Full size image
Fig. 5
figure 5

MUC2 and MUC5AC immunohistochemical staining of serial sections of type III IM in the corpus lesser curve (original magnification ×40). a. High-iron diamine–alcian blue (HID–AB) staining shows goblet cells and columnar cells stained in brown (sulphomucin). b. MUC5AC was detected in the goblet cells and in the columnar cells. c. IM does not express MUC6

Full size image
Table 2 Correlation of HID–AB staining and expression of MUC5AC and MUC6
Full size table

Gastric cancer was associated with a higher frequency of incomplete IM (types II and III) and IM expressing MUC5AC (Table 3). The sensitivity and specificity of detecting incomplete IM in either point were 86% and 65% (OR=11, 95% CI 5.1–25, p<0.001). Those of type III IM were 55% and 95% (OR=22; 95% CI 7.5–67, p<0.001), and IM with MUC5AC expression were 81% and 69%, respectively (OR=8.9, 95% CI 4.3–19, p<0.001).

Table 3 Comparison of histological findings between the cancer group and the control
Full size table

Among the patients with IM, the presence of type III IM in the antrum and corpus lesser curve, and incomplete and GI mixed type MUC5AC expressing IM in each part were significant risk factors for gastric cancer (Table 4). However, in a multivariate analysis among those factors, gastric cancer was significantly associated with a higher frequency of incomplete IM (types II or III) in the corpus lesser curve (OR=6.4; 95% CI, 2.0–21, p=0.002).

Table 4 Comparison of the prevalence of type of intestinal metaplasia among the subjects with intestinal metaplasia
Full size table

Asynchronous multiple lesions were associated with incomplete IM in the corpus greater curve (OR=4.8; 95% CI, 1.4–16, p=0.01). However, there was no significant association between types of IM and locations of malignant lesions.

Discussion

In this study, we confirmed that MUC5AC expression was observed most often in type III IM. Previous studies examined biopsy samples or surgical specimens taken from the gastric mucosa adjacent to carcinomas, and MUC1 and MUC5AC together with MUC2 were expressed in every case with IM types II and III [27, 33]. A recent study from Japan examined the pyloric regions of stomachs resected for gastric cancer and showed that MUC5AC was expressed in 38% of type I glands, 78% of type II, and 91% of type III IM [35]. These results are similar to those reported here. The most incomplete IM (type II and III) preserving gastric mucin is the GI mixed type, whereas the complete type is the type I IM especially in the corpus lesser curve.

Gastric cancer risk increases with age and is highest in those with severe atrophic gastritis. We found that incomplete IM in the corpus lesser curve is common in patients with gastric cancer. The presence of incomplete IM at this location is indicative of the presence of corpus atrophy, which is a known risk factor for an increased risk of the intestinal type gastric cancer [31, 32]. The fact that multivariate analysis showed a better association of the type of IM with the presence of gastric cancer than the score of atrophy suggests that, in an examination using a limited number of biopsy specimens, the type of IM may better reflect the severity and extent of atrophy. This notion is consistent with our results showing an inverse correlation between the type of IM, the PG I/II and the severity of corpus atrophy.

Using a systematic endoscopic protocol with fixed-point biopsy samples, we previously showed that the scores for IM in the corpus greater curve and the scores for atrophy in the corpus lesser curve were significantly higher in patients with multiple malignant lesions compared to those with a single lesion [32]. This study extended those findings showing that the presence of incomplete IM in the corpus greater curve was an additional histological risk marker for recurrent new malignant lesions. Kabashima et al. reported that the incidence of IM expressing gastric phenotype in non-neoplastic mucosa within 5 mm from the margins of carcinoma cells was higher in the patients with multiple early gastric cancer lesions than in the patients with a single lesion [21].

This study suggests that the type and location of IM provides important additional information regarding the extent and severity of gastric atrophy and thus the risk of developing gastric cancer. Changes in the type of IM could also reflect the amount of genetic damage that has taken place and/or changes in the intragastric environment (e.g., development of low acid secretion). For example, hypoacidity might allow stem cell differentiation in a particular direction to IM and independently to cancer [10, 31]. Recent studies investigating relationship between genetic alterations and cellular phenotypes in cancer and IM have provided evidence that the phenotypic expression in cancer and precancerous lesions of the stomach is strongly dependent on genetic changes [9], and that cancer and IM may arise from different cell lineages [6, 17, 29, 36]. Tatematsu and colleagues used the classification of IM using gastric and intestinal cell phenotypic markers, and hypothesized that IM phenotype shift from GI mixed type to I type [35–37]. Intestinalization progresses from GI mixed type to I type in non-cancerous and cancerous tissue independently. However, our results do not support the concept of IM phenotype shift from mixed type to intestinal type as incomplete IM appeared to follow complete IM and was correlated with the PG I/II ratio and with the severity of corpus atrophy. The pathogenesis of IM and actual relationship to gastric cancer remains unclear and additional studies are warranted to explore the relationship.

Our study using targeted biopsies showed that the most incomplete IM preserving gastric mucin is the GI type, whereas the complete type is the type I IM especially in the corpus lesser curve, and that IM type was associated with the presence of early gastric cancer. These data suggest that patients with incomplete IM in the corpus lesser curve are likely to have severe atrophic pangastritis and such patients are at increased risk of developing gastric cancer. Moreover, the presence of incomplete IM in the corpus greater curve was an additional histological risk marker for recurrent new malignant lesions. In conclusion, the classification of IM by histochemical detection using biopsy samples obtained from fixed-points may enhance the ability of surveillance programs to detect patients at increased risk of either having or developing gastric cancer.