Transcription factors GATA-4 and GATA-6 in normal and neoplastic human gastrointestinal mucosa

Normal gastrointestinal biopsy specimens were obtained during esophagogastroduodenoscopy and colonoscopy performed for diagnostic purposes. The specimens (Table 1) included tissue from gastric, duodenal, ileal, colonic and rectal mucosa. Most of the samples (37 out of 55) were collected from children. All biopsy specimens were diagnosed as morphologically normal by a pathologist.

The pathological samples from distal esophagus and stomach were collected during routine esophagogastroduodenoscopy from adult patients with chronic and atrophic gastritis (Table 2). Esophageal biopsies contained intestinal metaplasia (also called Barrett's esophagus), a precancerous state of the esophageal epithelium. Neuroendocrine cell hyperplasia was found in three stomach samples. Colon and rectum samples were obtained in resections for adenomas and primary adenocarcinomas. Some of them contained dysplasia (n = 5) and invasive carcinoma (n = 2).

Radioactive in situ hybridization was performed on normal gastrointestinal samples (Table 1). Paraffin embedded tissue sections (n = 22) were deparaffinized, and frozen sections (n = 9) were fixed in 40 g/L paraformaldehyde in phosphate-buffered saline (PBS). We applied the procedure described earlier [27]. Human GATA-4 and GATA-6 cDNAs were prepared as previously described [28]. Tissue sections were incubated with antisense or sense riboprobes labeled with ³³P (1.2 × 10⁶ cpm, 1000–3000 Ci/mmol, Amersham Pharmacia Biotech, Arlington Heights, IL) in a total volume of 80 μl. The slides were assessed by three researchers independently and blinded under a dark field light microscope (Leica DMRXA microscope, Leica, Switzerland).

Deparaffinized sections of normal and pathological tissue (Tables 1 and 2) were dehydrated, and antigen retrieval was improved by microwave cooking for 20 min in 10 mM citric acid, pH 6. The endogenous peroxidase reaction was inhibited by treatment with 3 g/L of H₂O₂. The sections were then subjected to immunohistochemistry, using polyclonal goat anti-human GATA-4 (sc-1237, dilution 1:200) or Ihh (sc-1196, dilution 1:50) antibodies, or polyclonal rabbit anti-human GATA-6 antibody (sc-9055, dilution 1:50). The antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA). The specificity of the staining was assessed using preimmune serum or PBS instead of the primary antibody during the staining protocol. In addition, our earlier studies with consecutive samples for RNA in situ hybridization and immunohistochemistry on adrenal samples have revealed the reliability of the used anti-GATA-4 and anti-GATA-6 antibodies in human tissues [28]. In some instances, GATA-6 gives a cytoplasmic staining as previously reported [15]. Although cytoplasmic staining for GATA proteins usually has not been considered specific, there is also some evidence that GATA proteins are not always transported to the nucleus immediately after translation [29]. The avidin-biotin immunoperoxidase system (Vectastain Elite ABC Kit, Vector laboratories, Burlingame, CA) and 3,3'-diaminobenzidine (Sigma-Aldrich, St. Louis, MO) were used to visualize antibody binding. The tissues were counterstained with Harris hematoxylin. The Alcian blue/periodic acid-Schiff staining method and hematoxylin-eosin staining were used to detect goblet and Paneth cells, respectively. These results were compared with GATA immunohistochemistry. All the slides were analysed under a light microscope (Leica DMRXA microscope) and reassessed by a pathologist.

The double-staining method was performed on normal paraffin-embedded sections. The sections were prepared as described above, with the following modifications: After staining for GATA-4 (sc-1237) or GATA-6 (sc-9055) with 3,3'-diaminobenzidine, the slides were incubated with the second primary antibody for 1 h at +37°C. The second primary antibody was either a polyclonal goat anti-human chromogranin A antibody (sc-1488, dilution 1:50, Santa Cruz Biotechnology) used to detect neuroendocrine cells in stomach, duodenum, and large intestine, or monoclonal anti-human hydrogen/potassium adenosine triphosphatase (H+/K+-ATPase) antibody against α (119102, dilution 1:50, Calbiochem^®, EMD Chemicals Inc., Darmstadt, Germany) or β subunit (MA3-923, dilution 1:1000, Affinity BioReagents Inc., Golden, CO) to identify the parietal cells in the stomach. The immunoreactivity was visualised by Vector SG^® (Vector Laboratories).

GATA-4 and GATA-6 mRNA expression was first analyzed in the normal gastrointestinal mucosa by in situ hybridization. Samples representing stomach, duodenum, terminal ileum, colon, and rectum showed differences in the expression of GATA-4 and GATA-6 along the longitudinal axis and the crypt-villus axis. GATA-4 mRNA was abundant in the superficial two thirds of the mucosa of the proximal gastrointestinal tract (Fig. 1A-A' and 1C-C') but was absent from the distal gut (Fig. 1E-E' and 1G-G'). In contrast, GATA-6 mRNA was highly expressed in the basal regions of the mucosa along the whole gastrointestinal tract, diminishing somewhat towards the distal end (Fig. 1B-B', 1D-D', 1F-F', and 1H-H', data on rectum not shown).

Immunostaining was used to delineate the gastrointestinal cell types expressing GATA-4 and GATA-6. GATA-6 immunoreactivity was widespread in the mucosa. In the corpus, GATA-6 expression was abundant in the glands, consisting of chief cells and parietal cells (Fig. 2B with inset). In the antrum and pylorus, the staining for GATA-4 was strongest in the neck and surface epithelium, diminishing towards the base, where GATA-6 exhibited moderate immunoreactivity. In the duodenum, there was no significant GATA-4 immunoreactivity in the Brunner glands, which were positive for GATA-6 (data not shown), but enterocytes lining the villi were highly GATA-4 positive (Fig. 2D). GATA-6 predominated in the crypts of Lieberkühn (Fig. 2E). The mucosa of the terminal ileum (Fig. 2G), colon (Fig. 2J) and rectum (data not shown) generally lacked GATA-4 protein. In 3 out of 17 ileum and colon samples, however, a faint staining for GATA-4 was detected in the epithelium adjacent to a lymph node. Distinct positivity for GATA-6 was detected especially in the bottom of the crypts of the distal gut (Fig. 2K). It is noteworthy that neither GATA-4 nor GATA-6 was found in neuroendocrine cells along the gastrointestinal tract (Fig.3A, 3A', 3C, 3C', 3F, and 3F'). Thus, GATA-4 and GATA-6 presented a cell-specific expression pattern throughout the gastrointestinal tract (Table 3). There were no age-dependent differences in the expression patterns.

In the stomach, Ihh expression resembled that of GATA-6; the glandular region, especially in the corpus, showed strong immunoreactivity for Ihh (Fig. 2C). Apart from intense Ihh expression in the neuroendocrine cells throughout the GI tract (Fig. 2F and 2I, arrowheads), the gut mucosa was mostly Ihh-negative. Staining along the brush border of the villus was considered non-specific. Only occasional villus and crypt enterocytes of the small intestine (Fig. 2F) as well as colonocytes and lamina propria of the superficial colon mucosa showed immunoreactivity for Ihh (Fig. 2L, arrowhead).

GATA-4 and Ihh were not found in normal squamous epithelium of the esophagus, while GATA-6 was expressed in the epithelium near the gastroesophageal junction (data not shown). Barrett's esophagus (data not shown) as well as intestinal metaplasia of the stomach were nevertheless highly positive for GATA-4, GATA-6 and Ihh (Fig. 4A, 4A', and 4A").

Neuroendocrine cell hyperplasia was present in 3 atrophic gastric samples. As described above, normal neuroendocrine cells expressed neither GATA-4 nor GATA-6 but were positive for Ihh. On the other hand, hyperplastic neuroendocrine cells strongly expressed both GATA-4 and GATA-6 (Fig. 4B and 4B'). Interestingly, Ihh expression was weak or absent in these hyperplastic clusters of neuroendocrine cells (Fig. 4B").

No expression of GATA-4 was detected in dysplastic lesions or adenomas of the colon (Fig. 4C). In contrast, GATA-6 was positive in colon adenomas and low grade dysplasias; moreover, GATA-6 expression was intense in high grade dysplasias (Fig. 4C' with inset). The expression pattern of Ihh in tubulovillous adenomas and epithelial dysplasias resembled that of GATA-6 (Fig. 4C" with inset). Ihh was not detected in serrated adenoma (data not shown).

In colon carcinoma, only low GATA-6 expression was detected (Fig. 4D'), and there was no correlation to histological grade of the tumor (Table 4). GATA-6 expression seemed to be diminished in carcinoma cells as compared to the adjacent normal colon mucosa. On the contrary, our results suggest that GATA-6 expression is stronger in the parts of the carcinoma nearest the surrounding stroma, i.e., in areas representing the invasive edge of the tumor (data not shown). GATA-4 and Ihh were not present in any of the colon carcinoma samples studied (Fig. 4D and 4D").