Application of a Novel Method of Double APAAP Staining With Subsequent Quantitative Image Analysis to the Examination of Integrin Expression in Undifferentiated-type Gastric Carcinomas

Abstract

In undifferentiated-type gastric carcinoma (UGC), recognition of cancer cells is not easy, which has hampered its precise phenotypic analysis. To examine alterations of the integrin phenotype during the progression of UGC, we used double alkaline phosphatase anti-alkaline phosphatase staining and computer-aided image analyses for the expression of α1, α2, α3, α5, α6, αV, β1, and β4 integrin subunits and αVβ3, αVβ5, and αVβ6 integrins in cytokeratin-positive cells in the mucosal, the submucosal, and the deeper parts of 10 early and 17 advanced UGCs, their non-neoplastic counterparts, and 9 lymph node (LN) metastases. We revealed declining expression of epithelial integrin subunits (α2, α3, α6, β4) and increasing expression of mesenchymal integrin subunits (α1, α5) as the tumor invaded deeper, reflecting gradual epithelial-to-mesenchymal transition of the integrin phenotype during tumor invasion. Enhanced expression of the αV integrin subunit and αVβ3 and αVβ5 integrins correlated with tumor invasion, and that of αVβ6 integrins with LN metastasis. Our results have demonstrated that the method we introduced is suitable for analysis of dynamic alterations of the integrin repertoire in UGC progression. (J Histochem Cytochem 57:1183–1193, 2009)

Gastric carcinoma has been commonly classified based on the development of a tubular component. According to the Japanese classification of gastric carcinoma, undifferentiated-type gastric carcinoma (UGC) is a type that poorly develops a tubular component. Studies of UGC are especially important at present, when the incidence of differentiated-type but not undifferentiated-type gastric cancers is declining worldwide (Crew and Neugut 2006), partially due to a reduction of Helicobacter pylori infection. In UGC, genetic factors may be more important than environmental factors. Despite the remarkable advances of molecular technology, however, the etiology and histogenetic pathways of diffuse gastric carcinomas are still less clear than those in the differentiated type.

This study is focused on the expression of integrins to clarify the role of epithelial–mesenchymal interactions in tumor progression. For this purpose, UGC may be suitable material because the tumor cells of UCG are dissociative and have a greater proportion of the tumor–cell stroma interface, and are expected to be regulated greatly by epithelial–mesenchymal interactions. The growth pattern of UGC varies remarkably from superficially spreading, dormant tumor to highly malignant, diffusely infiltrative carcinoma. Genetic studies have demonstrated that the latter can emerge from the former through stepwise accumulation of genomic alterations and clonal evolution in a subtype of UGC (Tamura et al. 2001; Peng et al. 2003; Yoshimura et al. 2006). This process of tumor progression may be associated with remarkable alteration in the expression of integrins.

Studies of UGC, especially of non-solid type (Japanese Gastric Cancer Association 1998), is often linked with some problems; in sections stained for immunohistochemistry (IHC) (particularly frozen sections), scattering cancerous cells could simulate inflammatory cells and active fibroblasts that show general loss of epithelial-specific proteins or gain of abnormal proteins. Hence, tumor stroma development and lymphocyte infiltration could mask the real picture. This problem becomes especially significant in studies of integrins. It was proven that invasive cells underwent dramatic alterations in levels of integrin expression and integrin affinity for extracellular matrix (ECM) substrates, which could influence tumor cell behavior and metastasis formation (Hood and Cheresh 2002) and could reflect tumor stage (Koretz et al. 1991). Therefore, while assessing integrin expression in each UGC, a researcher should differentiate cancerous cells that have lost their normal integrins and acquired mesenchymal integrins as an epithelial-to-mesenchymal transition (EMT) from stromal cells. Probably due to the above-mentioned difficulties, an overall study of all integrin repertoire changes during tumor progression of UGC from the early to the advanced stage is apparently not performed.

To discriminate cancerous cells from non-cancerous cells, we used double staining for integrins as well as for cell lineage markers such as cytokeratins. For this purpose, however, immunofluorescence (IF) staining, which is often applied to reveal antigens that coexist in the same compartment, could not be used, because some integrins (e.g., α5, αV group) are expressed in normal stomach epithelium and cancerous cells too weakly to be revealed by IF. We thus used the more-sensitive alkaline phosphatase anti-alkaline phosphatase (APAAP) method (De Jong et al. 1985; Roberts et al. 1991; Gregg et al. 1995). However, a limitation of simultaneous double APAAP staining is that spatial overlapping of the studied antigens can mask some reaction products with other reaction products. We therefore developed consecutive double staining, adopting the idea of an intermediate photographic step (Wang and Larsson 1985).

The above-mentioned double staining inevitably encounters the problem of crossreactivity when two antibodies of the same species (primarily mice) are used. There are at least two ways to overcome this problem: masking with diaminobenzidine (DAB) precipitate (Hsu and Soban 1982), and blocking of the antibody by microwave boiling (Lan et al. 1995; Tornehave et al. 2000). Because the DAB-based horseradish peroxidase (HRP) method in frozen sections causes insufficient quenching of endogenous peroxidase and denaturation of certain antigens (including some intermediate filament proteins) (Hittmair and Schmid 1989), we adopted the latter, which is the easiest and the most reliable.

The IHC data were analyzed with computer-based standardization and quantification instead of subjective plus/minus scale–based analysis. Two automated methods for computer-based quantitative analysis in IHC are assessment of overall chromogen staining intensity and point counting (Gross and Rothfeld 1985; Matkowskyj et al. 2000,2003). Although those methods quickly provide great amounts of quantitative data, they could not be used under our conditions: A cancerous tissue comprises a mixture of heterogeneous cells; only specific antigen distribution (e.g., membranous or cytoplasmic) is considered positive; and high contrast between the staining product and the surrounding background could not always be achieved (particularly in αV integrin subunit expression). Therefore, we have adopted a manual method of point counting, although it is somewhat time-consuming.

In this study, we thus adopted consecutive double APAAP staining and a novel manual digital image analysis to examine intratumoral heterogeneity in the expression of such fragile targets as integrins in such an object as UGC, where cancer cells are mixed with stromal and inflammatory cells.

Materials and Methods

Tissue Material and Sample Preparation

We used 27 cases with fresh resection specimens of UGC (10 early and 17 advanced). In each case, two to ten tissue samples were taken from the tumor, including (if present) fronts of superficial spreading in the mucosa and the deepest part of extramucosal invasion. Normal tissues were taken from the antrum or the distal part of the stomach body. The tissue samples were snap-frozen in liquid nitrogen and then stored at −80C until sectioning. Serial 4-μm-thick sections were cut at −20C with a cryostat (Leica CM1850; Nussloch, Germany), mounted on Superfrost-coated glass slides (Matsunami Glass Ind., Ltd; Osaka, Japan), and placed in plastic boxes for storage at −80C until further processing. Before staining, the slides were fixed as shown in Table 1. The methods of sample collection and anonymization were approved by the Institutional Review Board on Medical Ethics, Shiga University of Medical Science.

Table 1.

Processing of the slides, including immunohistochemical procedure, MW boiling, and image capturing

Main steps	Temperature	Time	Reagent
1. Fixation
Integrin staining	−20C	35 min	Acetone (5 min) + air drying (30 min, at room temperature)
Other protein staining	−20C	10 min	Buffered formalin, 1%
2. Rehydration	Room temperature	3 min × 3	PBS
First staining cycle
3. Protein blocking	Room temperature	10 min	Goat serum (Histofine; Tokyo, Japan), 5%
4. Incubation with the first primary antibody	37C	60 min	Antibody details are mentioned in text
5. Incubation with the first secondary antibodya	37C	20 min	Goat anti-mouse AP-labeled polymer for human tissue N-Histofine Simple Stain AP(M) (Histofine)
6. Application of AP substrate and endogenous AP blockera	Room temperature	5–7 min	New fuchsin (Histofine) + Levamisole (DAKO A/S; Glostrup, Denmark)
7. Counterstainingb	Room temperature	10 s	Mayer's hematoxylin solution (Histofine)
8. Covering with coverslips and temporary storageb	4 and [deg]C	2–5 days	PBS or aqueous permanent mounting solution (Histofine)
9. Taking the first digital image	Room temperature		–
10. Removing the coverslips	Room temperature	5–10 min	Slides were dipped in PBS and the coverslips were removed by a microtome blade
11. Microwave treatment (Hitachi, MR-M33, 500 W)	100C	5 min × 2	Sodium citrate buffer: 200 ml 10 mM, pH 6
12. Cooling by tap water	–	20 min
Second staining cycle
13. Protein blockingb	Room temperature	10 min	Same as in step 3
14. Incubation with the second primary antibody	37C	60 min	Anti-cytokeratin prediluted AE1, AE3 (Histofine)
15. Incubation with the second secondary antibodya	37C	20 min	Same as in step 5
16. Application of AP substratea	Room temperature	2 min	BCIP/NBT (blue) (DAKO A/S)
17. Mounting for permanent storage	Room temperature	–	Aqueous permanent mounting solution (Histofine)
18. Taking the second digital image	–	–	–

^aSlides were rinsed in TBS three times for 2 min each.

^bSlides were rinsed in PBS three times for 2 min each.

Alkaline phosphatase (AP)-labeled polymer means AP-labeled, amino acid polymer–attached secondary antibody. Microwave treatment was performed for denaturation of the products of the first staining and to prevent their unwanted reactions with subsequently applied reagents (Lan et al. 1995). Levamisole was not necessary in step 16, because microwaving totally destroyed endogenous AP (Lan et al. 1995). Nuclei were totally removed by microwave treatment; therefore, no counterstaining was performed after the second cycle. BCIP/NBT, 5-bromo-4-chloro-3-indolyl-phosphate/nitro blue tetrazolium.

Immunohistochemistry

The mouse monoclonal antibodies used in the study included antibodies against cytokeratins 1–8,10, 14–16, 19 (prediluted, AE1/AE3; Histofine, Tokyo, Japan) and integrin/integrin subunits: α1 (1:50; TS2/7) and β1 (1:50; K-20), both from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA); α2 (1:200; P1E6), α5 (1:10; SAM-1), α6 (1:20; 4F10), αV (1:1000; LM142 and 1:20; 13C2), β4 (1:2500; 3E1), αVβ3 (1:1000; P1F6), αVβ5 (1:200; LM609), αVβ6 (1:200; E7P6), all from Chemicon International, Inc. (Temecula, CA); α3 (1:40; J143) from GeneTex, Inc. (San Antonio, TX); β1 (1:100; DF5) from BIOMOL International, L.P. (Plymouth Meeting, PA).

The method of double staining is outlined in Table 1. Because integrins do not withstand heat treatment, they should be stained in the first staining cycle. Non-cancerous counterparts of UGC underwent only the first cycle of staining. We used alkaline phosphatase–labeled, amino acid polymer–attached secondary antibody (see step 5 in Table 1). This secondary antibody provided satisfactory nuclear preservation in the first staining cycle for frozen tissues.

For a positive control, we used normal tissue of the same stomach sample (for proteins that normally occur in stomach tissue) or specimens from our collection proven to be positive for this protein (for abnormal proteins). For negative control, every processed slide contained one serial section of the same sample, which was stained with the omission of the primary antibody.

Image Capturing and Processing

The entire hematoxylin and eosin–stained slice area was scanned using the Nikon Super COOLSCAN 5000 ED film scanner at a resolution of 4000 dpi, and three to ten areas of interest were chosen for further study.

Sections were viewed with a standard light microscope (Nikon Eclipse 80i; Nikon Corporation, Tokyo, Japan) and captured with a Nikon DXM200 digital camera at a resolution of 3840 × 3072 pixels using Nikon ACT-1, Version 2.63. The first series of images (at ×7, ×14, ×28 magnifications) of each area was taken after the first APAAP staining, and the second series of images was taken after the second APAAP staining (see Table 1, steps 9 and 18). Processing was performed with Adobe Photoshop. For qualitative assessment of protein distribution, IF-resembling images were created as follows (Figure 1): first-step images, second-step images, and merged images panels.

Figure 1.

Computer-aided processing of digital photomicroscopic images. Panels “first-step images” and “second-step images” contain the images obtained from the same area after the first and the second cycles of alkaline phosphatase anti-alkaline phosphatase (APAAP) staining (for details, see Table 1). The panel “merged images” shows immunofluorescence-resembling images resulting from computer-aided merging of two APAAP images that are located at its left side (for details see Algorithm 1). Red and blue in merged images correspond to magenta and violet, respectively, in the first-step image; green in merged images corresponds to dark-blue in the second-step image. Yellow represents coexpression of integrins and cytokeratin (CK). First row: E1, E2, and E1+2 series, coexpression of epithelial integrins (namely α6) with CK in cancer cells within MP. Second row: M1, M2, and M1+2 series, distinctive expression of mesenchymal α5 integrin and CK in cancer cells invading through perimysium. Third row: U1, U2, and U1+2 series, partial overlapping of a ubiquitous integrin (β1) and CK expression in the superficial part of the mucosa in early undifferentiated-type gastric carcinoma (UGC). Panel ‘COUNT’ in the fourth row contains two images depicting the result of the fraction of positive cells calculation. COUNT1: results of the first cycle lymph node (LN) metastasis staining, α1 integrin subunit stained magenta; COUNT2: the same area after the second staining, CK stained dark blue. For details see Algorithm 2 and Materials and Methods.

Algorithm 1: Algorithm for Producing IF-resembling Images

1. Open two images (first-step image and second-step image of APAAP staining) of the same area (Figure 1).

2. Copy-paste the second-step image as the second layer upon the first-step image and decrease its opacity to ∼70% (because at this stage, the first-step image and second-step image are called the first layer and second layer).

3. Move/rotate the second layer to match the first layer (“free transform” tool).

4. Make the second layer invisible.

5. Create a merged image of the same size as the first-step image (“new image” function) and stain it in black (“paint bucket” tool).

6. Within the first layer, select magenta color range (produced by fuchsin) with either “magic wand” (“tolerance” should be adjusted for every image, and “contiguous” option should be turned off) or command “select color range” (option “sampled colors”).

7. Copy selected magenta color and paste it in a red channel of the merged image.

8. In the same way as the magenta color range, select and copy violet color range (produced by hematoxylin) within the first layer, then paste it into a blue channel of the merged image.

9. Follow the same procedure with the dark-blue color range (produced by 5-bromo-4-chloro-3-indolyl-phosphate/nitro blue tetrazolium) within the second layer and paste it into a green channel of the merged image.

10. Visualize “IF-resembling” picture by turning-on/off channels of merged image simultaneously (“Channels” → “RGB”) (Figure 1, merged images panel).

For the estimation of the fraction of positive cells, cancer cells positive for integrin (or other protein of interest) were counted within 100 cytokeratin-positive cancer cells, as follows: (Figure 1, “COUNT1” and “COUNT2” panels).

Algorithm 2: Algorithm for the Fraction-of-positive-cells Calculation

1. For creating a mesh image, make a new image of the same size as the first-step image, add a new layer (mesh layer), and draw lines in it. We used mesh with a 50-μm cell.

2. Create one more new layer (calculation layer) in mesh image.

3. Open two images (first-step image and second-step image of APAAP staining) of the same area.

4. Copy-paste the first-step image as the first layer upon the background of the mesh image.

5. Copy-paste the second-step image as the second layer upon the first layer. Change second-layer opacity (∼70%).

6. Move/rotate the second layer to match the first layer (“free transform” tool).

7. Make the second layer visible/invisible, and mark cells with the pencil tool in the calculation layer. Positive cells were marked by green dots and negative by red dots (Figure 1, images COUNT1 and COUNT2).

8. While drawing green and red dots, calculate their number by manual calculation or by leukocyte calculator (either the mechanical device or the computer program).

9. When the total number of cells reaches 100, stop calculation, assess the fraction of positive cells.

10. Record your results in the calculation layer (Figure 1, images COUNT1 and COUNT2, upper-left corner).

11. Merge all layers when the second layer is invisible and save as first-step image count (Figure 1, COUNT1). Do the same when the second layer is visible and save as second-step image count.

Note: steps 10 and 11 can be omitted.

Positivity Assessment

A cell was judged as integrin positive if it had moderate membrane staining of at least 30% of the cell membrane or continuous thin membrane staining for all integrins or pronounced cytoplasmic staining (only for integrins of the αV family).

Statistical Analysis

Descriptive statistical parameters, including means and standard errors, were calculated according to the standard method (descriptive statistics package, Microsoft Excel).

To compare two subsequent stages of tumor progression, the statistical significance of differences between means was calculated using Student's unpaired t-test. To prove the significance of trends of integrin expression alterations, linear correlation with tumor progression stages and regression analysis were used. Values throughout this report are expressed as means ± standard error. A probability (P value) less than 0.05 was considered significant in all methods.

EM_i Coefficient

To estimate the relationship between strictly mesenchymal (α1, α5) and strictly epithelial (α2, α6, β4) integrin subunits, the epithelial–mesenchymal integrin index (EM_i) was designed:

where α1, α5, α2, α6, and β4 designate the fraction of positive cells of the corresponding integrin subunits.

Results

Integrin Expression in Non-neoplastic Stomach Mucosa

The cancer-bearing stomach cannot be considered normal, because it often shows chronic atrophic gastritis with intestinal and pyloric metaplasia (Lauren 1965). Hence, the term “non-neoplastic” (NN) is applied.

The integrin subunits studied were classified into three groups (Table 2) based on their distribution patterns in NN mucosa in our data (Figures 2 and 3A–3C, upper row) and those of other researchers (Breuss et al. 1993; Virtanen et al. 1995; Tani et al. 1996; Chénard et al. 2000; Ishii et al. 2000; Kawashima et al. 2003). Among the integrin subunits examined, α2, α6, and β4 showed a uniformly positive pattern (Group 1A), whereas α3 and β4 (Group 1B) showed vertical gradients of distribution (Figure 2) that means that α3 and β4 were expressed more strongly in the foveolar than in the glandular portion of gastric glands. Such polarization was remarkable only in the stomach body: non-neoplastic fundic glands and intestinal metaplasia. In the pyloric region, as well as in most areas of pyloric metaplasia, α3 and β4 integrins were expressed uniformly, like the integrins of Group 1A. There was no expression of mesenchymal integrin subunits (Group 2) in NN gastric glands. Integrin subunits of reactive changes (Figure 2, Group3) were observed in areas of metaplasia, inflammation, and regeneration. Intracellular distribution of integrins was commonly polarized on the basal and the lateral sides.

Table 2.

Classification of integrin subunits, studied in normal stomach epithelium and UGC tissue

β1 family			αV family
Integrins of mature tissues. Markers of epithelial and mesenchymal differentiation			Possible embryo-carcinogenic markers
Epithelial (α2 α6, β4) integrins*	Ubiquitous (α3, β1)*	Mesenchymal (α1, α5)**	Mesenchymal (αVβ3)**	Ubiquitous (αVβ5)***	Epithelial (αVβ6, αVβ8)***

Distribution in non-neoplastic stomach epithelium (see Figure 2): *Group 1; **Group 2; ***Group 3. UGC, undifferentiated-type gastric carcinoma.

Figure 2.

Distribution of integrins and integrin subunits in non-neoplastic stomach epithelium. The integrins studied were divided into three groups according to the intensity of their expression, which is expressed as density of each column.

Figure 3.

Distribution of integrin and integrin subunits in non-neoplastic stomach epithelium and different areas of tumors. CK, result of the second cycles of staining for cytokeratin in the corresponding area. (A–C) Alterations of integrin expression in one case of an advanced UGC with scirrhous stroma. Uppermost row: glandular (NG) and surface covering/pit (NS) portions of non-neoplastic epithelium. Second to fourth rows: expression of integrin and integrin subunits in varying depths of stomach wall (depicted in the leftmost scheme): M, the mucosal; MP, the muscularis propria; LN, lymph node. (D) Varying degrees of expression of α5 in different areas (MUSC, smooth-muscle bundles; CONN, connective tissue; VAS, vascular lumen) of the muscularis propria in an advanced UGC with scirrhous stroma (the same case as A–C).

Epithelial and Ubiquitous Integrin Subunits at Different Depths in the Tumor

The vertical gradient of distribution of α3 and β4 characteristic of the stomach body was lost even in early signet-ring cell carcinoma, i.e., α3 and β4 were diffusely expressed in mucosal cancer cells. Basal polarization in intracellular localization was lost in all integrin subunits examined, except α6 and β4 in some areas of primary and metastatic tumors (Figure 3A: LN).

The expression of epithelial and ubiquitous integrin subunits showed significant trends toward a decrease during deeper invasion, based on statistical analysis (Figure 4), despite the visual impression of the absence of their loss (Figure 3A).

Figure 4.

Quantitative changes of expression of integrins and integrin subunits associated with increasing depth of tumor invasion. Each group of bars represents fractions of the cells positive for a specified integrin subunit or integrin in non-neoplastic glands (NN), mucosal part of early cancer (M, early), mucosal (M, advanced), submucosal (SM, advanced), and muscularis propria (MP) parts of advanced cancer and metastatic tumor in LN. Gray lines, significant trends of alterations; gray curved lines, significant differences between subsequent stages.

Mesenchymal Integrin Subunits at Different Depths in the Tumor

Here we present alterations of α1 and α5 integrin subunits (Figure 2, Group 2); αVβ3 expression is discussed within Group 3. The expression of mesenchymal integrin subunits (α1, α5) was enhanced significantly during deeper invasion (Figures 3B and 4). The EMi index rose in parallel with enhanced expression of mesenchymal integrin subunits (Figure 4).

However, the stage of progression at which the expression level altered was different between α1 and α5; α5 expression was increased at early stages in cancer mucosa, in comparison with NN epithelium (p<0.01), whereas mucosal expression of α1 remained as weak as in NN epithelium at early stages and significantly increased (p<0.05) at advanced stages (Figure 4).

In the scirrhous-type invasion, in the proper muscle layer, the expression of α1 and α5 was significantly higher in the cells invading through the muscle (Figure 3D: MUSC) than in those invading through the perimysium (Figure 3D: CONN, VAS): α1 in muscle = 0.48 ± 0.06; α1 in perimysium = 0.24 ± 0.04; α5 in muscle = 0.71 ± 0.09; α5 in perimysium = 0.13 ± 0.03; p<0.01).

αV Family Integrins

Overall expression of αV family integrins, except αVβ6, was enhanced significantly during tumor progression (Figures 3C and 4). Despite the absence of a significant trend between the expression level of αVβ6 integrin and the depth of invasion, αVβ6 integrin was significantly upregulated in LN metastasis [in comparison with NN, submucosal (SM), and muscularis propria (MP), Figure 4].

The expression of αVβ3 in the muscularis propria in a scirrhous (non-solid) pattern was significantly (p<0.01) higher in muscular tissue than in perimysial connective tissue, similar to that of α1 and α5 mesenchymal integrins.

Discussion

Our present results concerning integrin expression in NN stomach epithelia have reconfirmed and expanded most of the previous data that were based on IF- and HRP-based studies (Breuss et al. 1993; Virtanen et al. 1995; Tani et al. 1996; Chénard et al. 2000; Ishii et al. 2000; Kawashima et al. 2003). But there were some differences between the normal and cancer-bearing NN parts of stomachs. An example was the expression of the α2 integrin subunit; whereas Virtanen et al. (1995) and Chénard et al. (2000) reported the absence of its expression in parietal cells, we observed positive expression of the α2 subunit in all stomach epithelial cells, including parietal cells (Figure 4, NN), which may reflect hypoacidity in the cancer-bearing stomach (Fossmark et al. 2008). Another example was the distribution of α3 and β4 integrins (Figure 2, Group 1B); the vertical gradient of distribution in normal fundic mucosa in adult and fetal stomachs (Virtanen et al. 1995; Chénard et al. 2000) was largely lost because pyloric metaplasia was common in fundic areas of cancer-bearing stomachs.

Here we do not mention the relationship between integrin expression and functions of tumor cells, because the expression level of integrin does not always reflect its activity (Korhonen et al. 1992; Miranti and Brugge 2002); the activity of expressed integrins is regulated by phosphorylation, the presence of bivalent cations, interaction with other receptors, etc. (Busk et al. 1992; Flug and Kopf-Maier 1995; Beaulieu 1999; Hood and Cheresh 2002; Miranti and Brugge 2002; Kim et al. 2004). In current reports, most authors trace not only an integrin, but also its ECM substrate and proteins of intracellular signaling pathways that are activated by integrin–substrate binding (Su et al. 2002; Zhang et al. 2005;2008) to demonstrate the actual contribution of each integrin in tumor invasion and metastasis. In the present study, we have adopted another approach to analyzing the significance of integrin expression alterations in tumor progression using the correlation of the integrin expression pattern with the position in the tumor, because it was postulated that alterations in integrin expression in varying depths of tumor invasion [mucosa of intramucosal early signet-ring cell carcinoma (ME), mucosa of advanced UGC (MA), SM, MP, and lymph node (LN)] could reflect varying degrees of tumor progression (Figure 5). Our results have demonstrated the trend that the deeper the tumor invaded in the stomach wall, the less epithelial integrin subunits and more mesenchymal integrin subunits of β1 family and more αV family integrins were expressed. This trend of β1 family integrin alterations may reflect the notion that integrins of this family are cell lineage (epithelial or mesenchymal) markers rather than markers for the aggressiveness of tumor cells (Miettinen et al. 1993) and that gradual changes in the integrin repertoire could reflect EMT, which is common in tumor progression (Thiery 2002). Such changes are in parallel with downregulation of E-cadherin expression (Nakamura et al. 2005) and acquisition of such non-epithelial proteins as N-cadherin (Rosivatz et al. 2004; Wang et al. 2006) and vimentin (Utsunomiya et al. 1996,2002).

Figure 5.

Postulated events of tumor progression. *Peng et al. 2003; Humar et al. 2007. **Brabletz et al. 2004. NN, non-neoplastic mucosa; ME, mucosa of intramucosal early signet-ring cell carcinoma; MA, mucosa of advanced undifferentiated-type carcinoma; SM, submucosa of advanced undifferentiated-type carcinoma; MP, muscularis propria of advanced undifferentiated-type carcinoma; LN, lymph mode metastasis.

Integrin EMT was most prominent in cancer cells migrating through the smooth-muscle cells of the muscularis propria, suggesting that an acquisition of mesenchymal integrins and, probably, the fibroblast motility mechanism (Flug and Kopf-Maier 1995) and interactions with fibronectin (α5β1, αVβ3) and collagen (α1β1, αVβ3) are necessary for this type of movement. However, the remarkable preservation of epithelial integrins and E-cadherins (unpublished data), and the presence of laminin chains around cancer cells moving through muscles (Tani et al. 1996) could indicate that a cancer cell employs both epithelial- and mesenchymal-type receptors/interactions in such movements.

The trend of gradual enhancement of αV family integrin expression as the tumor invades deeper may reflect the notion that integrins of this family are linked with morphogenetic events, remodeling, and inflammation, and are considered to be related probably not only to carcinogenesis (Korhonen et al. 1992; Breuss et al. 1995; Kawahara et al. 1995; Kawashima et al. 2003) but also to tumor progression and histological pattern formation. It has been reported that αVβ3 integrin takes part in matrix-degrading protease activation and localizing (Schramm et al. 2000; Hood and Cheresh 2002), acts as a proliferation integrin, linked to the Ras-ERK signaling pathway (Beaulieu 1992; Hood and Cheresh 2002); αVβ5 integrin activates tumor angiogenesis (Hood and Cheresh 2002), and αVβ6 integrin activates TGF-β1 at the tumor cell/stroma interface and enhances migration in a fibronectin-rich ECM (Breuss et al. 1995; Kawashima et al. 2003; Hazelbag et al. 2007). Accordingly, significant upregulation of αVβ3, αVβ5, and αVβ6 integrins was observed in LN metastasis (Figure 4).

However, in comparison to the surrounding NN mucosa, the primary tumors en masse did not display significant upregulation of the expression of αV family integrins (except αVβ3, which also reflects integrin EMT). This may be because the remodeling processes are also enhanced also in NN mucosa of cancer-bearing stomachs. Considerable expression of αV family integrins in the NN mucosa, as well as in gastric carcinomas, has also been observed by Kawashima et al. (2003) at the mRNA and protein levels.

Staining with the APAAP-linked polymer detection system appeared to be more sensitive and specific in staining of integrins than IF staining; it revealed smaller amounts of mesenchymal and αV integrins than IF, giving less background and a clearer picture than IF or HRP detection systems (unpublished data). One more advantage over IF staining was much better preservation of the UGC general morphology and cancer cell nuclei, which provides the opportunity to pursue the dynamic changes of integrin at different tumor depths during invasion through various tissues.

In conclusion, our results have demonstrated that the novel method of double APAAP staining with subsequent quantitative analysis of digital images is applicable to analyses of integrins and other peptide expression in UGC, in which the recognition of cancer cells is not easy. In comparison with routine IHC/IF staining, this method has allowed us to overcome the problem of crossreactivity in double staining, to produce more-quantitative results, and to introduce integrated indexes (such as EMi). Thereby we could demonstrate EMT and integrin repertoire alterations in tumor progression.