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

Background Human gastrointestinal mucosa regenerates vigorously throughout life, but the factors controlling cell fate in mature mucosa are poorly understood. GATA transcription factors direct cell proliferation and differentiation in many organs, and are implicated in tumorigenesis. GATA-4 and GATA-6 are considered crucial for the formation of murine gastrointestinal mucosa, but their role in human gastrointestinal tract remains unexplored. We studied in detail the expression patterns of these two GATA factors and a GATA-6 down-stream target, Indian hedgehog (Ihh), in normal human gastrointestinal mucosa. Since these factors are considered important for proliferation and differentiation, we also explored the possible alterations in their expression in gastrointestinal neoplasias. The expression of the carcinogenesis-related protein Indian hedgehog was also investigated in comparison to GATA factors. Methods Samples of normal and neoplastic gastrointestinal tract from children and adults were subjected to RNA in situ hybridization with 33P labelled probes and immunohistochemistry, using an avidin-biotin immunoperoxidase system. The pathological tissues examined included samples of chronic and atrophic gastritis as well as adenomas and adenocarcinomas of the colon and rectum. Results GATA-4 was abundant in the differentiated epithelial cells of the proximal parts of the gastrointestinal tract but was absent from the distal parts. In contrast, GATA-6 was expressed throughout the gastrointestinal epithelium, and in the distal gut its expression was most intense at the bottom of the crypts, i.e. cells with proliferative capacity. Both factors were also present in Barrett's esophagus and metaplasia of the stomach. GATA-6 expression was reduced in colon carcinoma. Ihh expression overlapped with that of GATA-6 especially in benign gastrointestinal neoplasias. Conclusion The results suggest differential but overlapping functions for GATA-4 and GATA-6 in the normal gastrointestinal mucosa. Furthermore, GATA-4, GATA-6 and Ihh expression is altered in premalignant dysplastic lesions and reduced in overt cancer.


Background
The structure of human gastrointestinal mucosa is highly complex. The mucosal morphology and enzymatic function are reached mainly during mid-gestation, but the mucosa undergoes regeneration throughout life. The mature mucosa comprises multipotent stem cells, their immediate progeny, and various differentiated cell types that serve digestive, absorptive, or immunologic functions. The maintenance of mucosal homeostasis requires a balance between rapid regeneration and cell turnover. Little is known, however, about the factors determining tissue renewal and cell fate in the gastrointestinal tract.
GATA proteins are zinc finger transcription factors that regulate cellular development and differentiation. The GATA family contains six members; GATA-1, GATA-2, and GATA-3 function mainly in hematopoietic cell lineages [1], whereas GATA-4, GATA-5, and GATA-6 are found in organs of endodermal origin [2]. In the gastrointestinal tract, GATA-4 and GATA-6 are present from the onset of formation of the primitive gut tube through the organization of normal postnatal epithelium [3][4][5][6]. GATA-4 is considered to be one of the first transcription factors binding to chromatin during early endodermal differentiation [7]. This binding has been proposed to initiate the opening of the chromatin and enable the binding of other transcription factors to DNA [8]. Mice with targeted deletion of Gata4 or Gata6 die during early embryonic development: GATA-6 is crucial in the development of visceral endoderm, whereas GATA-4 is needed for the formation of the heart tube [4,9].
GATA-4 and GATA-6 are both believed to facilitate the terminal differentiation of the epithelium of the intestinal villi [10]. In mice chimeric for Gata4 -/-and wild type cells, the gastric epithelial lineages were found to differentiate poorly during embryonic development [6]. In mice harbouring a tissue-specific deletion of Gata4 in the gut, jejunal gene expression was noted to change to the ilealtype, underscoring the role of GATA-4 in the development of the proximal small intestine [11]. In the gastrointestinal tract, GATA proteins regulate a variety of target genes involved in digestive functions and tissue regeneration. For example, two genes expressed in enterocytes of the small intestine, lactase phlorizin hydrolase (LPH) and fatty acid binding protein (FABP), are regulated by GATA-4, GATA-5, and GATA-6 [12][13][14][15]. Moreover, GATA factors are related to gastroprotective Trefoil Factor Family (TFF) peptides and mucin protein genes expressed in the stomach and colon mucosa [16,17].
In addition to normal growth, GATA proteins have been proposed to be associated with tumorigenesis in the adrenals, lungs and ovaries [18][19][20]. GATA-6 upregulates expression of Indian hedgehog (Ihh) [21]; the hedgehog signalling is known to be critical for proper gut morphogenesis [22]. Ihh is present in mature human gastric and colon epithelium [23,24], and its expression is altered during carcinogenesis of the gastrointestinal tract [23][24][25]. The role of Ihh in tumor biology, however, remains controversial. Ihh is believed to stimulate proliferation and induce carcinogenesis in the gastrointestinal tract [25,26], but it may also act as an antagonist of the Wnt-pathway to restrict tumor progression [23].
In this study we compare the expression of GATA-4, GATA-6, and Ihh in normal and pathological human gastrointestinal tissue samples. We show that GATA-4, GATA-6 and Ihh have particular expression patterns in the gastrointestinal mucosa, depending on the segment in the longitudinal or crypt-villus axis, as well as the specific cell type. In addition, the expression patterns of these factors in gastrointestinal neoplasms differ from those in the normal gastrointestinal tract. We suggest that GATA-4 and GATA-6 may maintain normal tissue renewal and differentiation in mature gastrointestinal mucosa and that alterations in their expression accompany premalignant and malignant changes in the gut mucosa.

Normal samples
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.

Pathological samples
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).

Tissue preparation and ethical considerations
Tissue specimens, originally collected for diagnostic purposes, were fixed in formalin and embedded in paraffin for further investigations, then cut into 6-μm sections, coded, and evaluated; the evaluator had no knowledge of the specimen. The specimens used for in situ hybdization experiments were frozen in liquid nitrogen and stored at -70°C until used. The use of the samples in this study was approved by the Ethics Committee of the Hospital for Children and Adolescents and the Ethics Committee of the Department of Internal Medicine, Helsinki University Central Hospital, and the Finnish National Authority of Medicolegal Affairs.

In situ hybridization
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 33 P (1.2 × 10 6 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).

Immunohistochemistry
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 H 2 O 2 . 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.

GATA-4 and GATA-6 protein expression show cell type specificity along the normal human gastrointestinal tract
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 moder-ate 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.

Ihh is expressed in normal gastrointestinal mucosa
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, GATA-6, and Ihh expression is induced in Barrett's esophagus and intestinal metaplasia of the stomach
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").

GATA-4 and GATA-6, but not Ihh, are strongly expressed in hyperplastic neuroendocrine cells of atrophic stomach mucosa
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").

GATA-6 and Ihh are expressed in dysplastic lesions and adenomas of the colon
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).
GATA-4 and GATA-6 mRNA expression in normal gastrointestinal mucosa

GATA-6 expression is low in colon carcinoma
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").

Discussion
Transcription factors play an important role in cell proliferation and differentiation, and consequently in tissue repair. We studied the expression of two GATA transcription factors in human gastrointestinal tract in detail. As our samples were derived from both children and adults, the wide age variation may have influenced the results. It is noteworthy, however, that in murine gastrointestinal tract, GATA-4 and GATA-6 expression patterns are rather stable from birth to adulthood [30].
In mature gastrointestinal mucosa, new cells are continuously generated from multipotent stem cells. Although the exact location of the stem cells of the gastrointestinal mucosa has not been determined, it has been suggested that in the stomach they lie in the isthmus, whereas in the intestine they may be located deeper in the crypts. The proliferating descendants of the stem cells enter the differentiation pathway and migrate into their specific sites in the mucosa. The cells eventually undergo apoptosis and are shed into the gut lumen [31]. It is of interest that the strongest GATA-6 expression occurs in the basal regions of the gut mucosa including cells with the highest proliferative capacity. Although GATA-6 expression has earlier been thought to decrease during enterocyte differentiation in vitro [10], we found GATA-6 in all mucosal layers, including the areas of highly differentiated cells. Ihh has been suggested to be regulated in part by GATA-6 [21], and we find that the expression patterns of these two factors partly overlap in the gut. In contrast to GATA-6, GATA-4 was localized to more differentiated cells (this study) [10]. In the murine small intestine, GATA-4 expression diminishes towards the villus tips the low expression thus associating with areas of apoptosis [15]. In cardiac myocytes, GATA-4 is related to anti-apoptotic factors [32], and its down-regulation is proposed to be essential for apoptosis [33]. We therefore suggest that the absence of GATA-4 from the villus tip enables enterocyte apoptosis and exfoliation of senescent cells.
In chronic gastrointestinal inflammation, such as esophagitis and gastritis, the renewal of normal tissue is disturbed, possibly leading to neoplastic tissue growth [34]. Interestingly, the expression of GATA-4, GATA-6 and Ihh appear to increase in two precancerous lesions such as Barrett's esophagus and intestinal metaplasias of the stomach. GATA-4 and GATA-6 are also expressed in hyperplastic neuroendocrine cells associated with atrophic gastritis. Some researchers have linked GATA-6 to normal murine neuroendocrine cells [35,36], but we The expression patterns are based on the number of positive cells detected and the intensity of staining: -no expression + individual positive cells or small areas with low expression ++ patches of moderate expression +++ large areas of moderate or strong expression detected neither GATA-4 nor GATA-6 in these cells (this study). The appearance of GATA factors in neuroendocrine cell hyperplasia raises the question of whether they contribute to the progression of neuroendocrine neoplasms. Interestingly, GATA factors have been shown to downregulate HDC [37] encoding a histamine synthesizing enzyme found in both normal and, in increasing amounts, in neoplastic neuroendocrine tissues [38].
In contrast to metaplasias of the proximal gastrointestinal tract, GATA-4 is not present in colon tumors, whereas GATA-6 and Ihh are moderately expressed in colon adenomas, but to a much lesser extent in carcinomas. Our results suggest that histological tumor grade does not significantly correlate with the level of expression of GATA-6 in cancer cells, although earlier in vitro studies have suggested that GATA-6 expression is strongest in the most undifferentiated colon carcinoma cells [10]. In our preliminary analyses, we found that intense GATA-6 immunoreactivity is characteristic of the border regions of malignant tissue and the invasive parts of the carcinoma. This may well be due to stromal signals that induce GATA-6 in the adjacent tumor regions.
The expression patterns of GATA-4 and GATA-6 in the longitudinal and crypt-villus axes are in accordance with the results of earlier studies on murine gastrointestinal tract [15]. In human fetuses, GATA-4 was found in the small intestine [39], and this expression is sustained in mature mucosa as well (this study). Also an earlier study based on RT-PCR analysis demonstrated the absence of GATA-4 from colon, and its presence in stomach [40]. A gene for hydrogen/potassium adenosine triphosphatase (H+/K+ ATPase) in the stomach, responsible for acid production in the parietal cells, is regulated by the gastrointestinal GATA factors [41]. In the parietal cells, GATA-6 positivity varied from one cell to another. This may reflect the fact that the structure and activity of the parietal cells depend on their developmental stage and location in the glands [31]. In our study, Ihh expression was intense in normal gastric glands, whereas Fukaya et al. [24] found Ihh only in gastric pits. The same Ihh antibodies were used in both studies, but the samples in Fukaya's study were neoplastic and their matched tissues. These differences are likely to explain the discrepancies between the two reports.
GATA-4, GATA-6, and Ihh protein expression in neoplastic gastrointestinal mucosa GATA proteins have been suggested to regulate genes encoding for TFF and mucin proteins [16,17] which protect the mucosa from harmful exogenous agents and are related to abnormal tissue growth and carcinogenesis [42][43][44]. It is of interest that both GATA-4 and GATA-6 proteins are present in TFF-expressing pit and surface epithelial cells. We detected GATA-6 also in human goblet cells, and others have found GATA-4 in murine goblet cells that express the mucin protein MUC2 [17]. Furthermore, TFF and mucin proteins have been shown to be present in esophageal and gastric metaplasias [45][46][47], similarly to GATA-4 and GATA-6 (this study). Particularly MUC2 expression in Barrett's metaplasia is considered to indicate a higher risk for carcinoma [46]. Collectively, these data support the interrelationship of GATA proteins, TFF, and mucins in the gastrointestinal endoderm.
A previous study suggested that inactivation of the Gata4 gene by methylation could be crucial during carcinogenesis [48]. It is tempting to speculate that GATA-4, in addition to promoting cellular differentiation in the human gastrointestinal tract, is also involved in the suppression of abnormal growth in the proximal gastrointestinal tract. When GATA-4 is inactivated by methylation, the fate of the cells may proceed towards malignant alterations. Likewise, the role of GATA-6 in neoplasias is controversial. In vascular smooth muscle cells, GATA-6 inhibits injury induced hyperplasia [49]. It is noteworthy that GATA-6 is present in adrenocortical adenomas, but diminishes in carcinomas, suggesting that also Gata6 may become methylated during tumorigenesis [50]. GATA-6 has been suggested to induce cell cycle arrest [51], to inhibit apoptosis and induce malignant cell growth [52]. Our preliminary results suggest high expression of GATA-6 in the invasive edges of the carcinomas. The enhanced GATA-6 action in these tumor areas may be linked to uncontrolled growth of the tumor cells. More detailed studies are, however, required to establish the role of GATA-6 in the gastrointestinal tumor growth.

Conclusion
The specific expression patterns of GATA-4 and GATA-6 and their abundance in the gastrointestinal mucosa (summarized in Figure 5) suggest that they have distinctive and central functions in the gut. Thus, GATA-4 may well be involved in digestive and absorptive functions in the proximal gut, whereas GATA-6 is likely to have a more ubiquitous task throughout the gastrointestinal tract. GATA-4 is expressed abundantly in the cell population programmed to differentiate, and GATA-6, on the contrary, is associated with the proliferative compartment of the mucosal cells. GATA-4 is abundant in intestinal metaplasias and GATA-6 in dysplastic lesions of the colon. Along with GATA-6, Ihh is also induced during abnormal growth. The significance of these factors in gastrointestinal neoplasia remains open, however. The present data offer a firm basis for further investigation of the role of GATA factors and Ihh in normal as well as pathological conditions of human gastrointestinal mucosa.