Transforming growth factor α disrupts the normal program of cellular differentiation in the gastric mucosa of transgenic mice

The glands of the gastric mucosa are simple tubular structures consisting of a complex mixture of diverse epithelial cell types that undergo continuous renewal, a developmental process that continues throughout life. Generation of the appropriate number of each cell type in a particular region of the stomach at a particular time requires the execution of delicately balanced mechanisms that regulate cellular proliferation, differentiation and senescence. Daughters from progenitor cells, located within a central proliferative zone of the gastric tubules called the isthmus, follow specific programs of differentiation as they migrate either upward toward the pit, forming surface mucous cells predominantly, or downward toward the base, forming chief cells, parietal cells and enteroendocrine cells (Karam and Leblond, 1992, 1993). Surface mucous cells secrete neutral mucins that protect the mucosa from acid, chief cells secrete a variety of pepsinogens, parietal cells produce gastric acid and enteroendocrine cells help regulate mucosal growth and secretory activity. It is widely accepted that extracellular factors capable of receptor-mediated signal transduction play a prominent role in the complex regulation of stomach function, including cellular growth and differentiation (reviewed by Lipkin, 1987). These include true hormones such as gastrin, cholecystokinin and secretin, and local regulatory peptides such as somatostatin and transforming growth factor-α (TGFα).

TGFα, a member of the epidermal growth factor (EGF) family of peptides, induces a wide variety of biological responses (reviewed by Salomon et al., 1990). TGFα is best known as a highly potent mitogen, and has been associated with malignant transformation (Derynck et al., 1987; Salomon et al., 1990). Its restricted expression in some normal and embryonic tissues, and its effects on specific developmentally responsive primary and established cultured cells raise the possibility that TGFα participates in the regulation of differentiation as well as proliferation (Twardzik, 1985; Lee et al., 1985; Wilcox and Derynck, 1988; Butterwith et al., 1992; Nakafuku and Kaziro, 1993). TGFα activity is mediated by the EGF receptor proto-oncoprotein, a powerful tyrosine kinase capable of phosphorylating both itself and additional substrates upon ligand-specific activation/dimerization (reviewed by Carpenter, 1987).

TGFα and EGF have been shown, both in vivo and in vitro, to inhibit acid secretion in the stomach (Gregory, 1975; Bower et al., 1975; Finke et al., 1985; Rhodes et al., 1986; Guglietta et al., 1994), stimulate gastric mucosal growth (Johnson and Guthrie, 1980; Dembinski et al., 1982; Chen et al., 1991), and stimulate the production and secretion of mucins (Yoshida et al., 1987; Kelly and Hunter, 1990). TGFα and EGF receptor are both produced in the adult and neonatal mammalian stomach. TGFα has been localized by in situ hybridization and by immunohistochemical visualization to the mature, non-proliferative compartment of the gastric mucosa, suggestive of a role in cellular differentiation (Beauchamp et al., 1989; Thomas et al., 1992; Nasim et al., 1992; Yasui et al., 1992).

To elucidate the role of TGFα in a genetically defined in vivo model system, we generated transgenic mice overexpressing a human TGFα cDNA in a wide variety of tissues by using a mouse metallothionein (MT) I promoter (Jhappan et al., 1990). These transgenic mice developed multiple phenotypic abnormalities, including cancer of the liver and mammary gland, and fibrosis and ductular metaplasia of the pancreas. In the latter, TGFα overexpression stimulated redifferentiation of acinar cells to simple ductal cells, mucinproducing cells and, less frequently, insulin-producing cells, further supporting a role for TGFα in cellular differentiation (Bockman and Merlino, 1992; Wang et al., 1993). We have reported that TGFα overexpression induced structural and functional alterations in the stomach of transgenic mice, particularly line MT100, resulting in a phenotype reminiscent of Ménétrier’s disease in humans (Takagi et al., 1992). TGFα transgenic mice demonstrated severe hypertrophic gastropathy and achlorhydria, the failure to secrete detectable gastric acid (Takagi et al., 1992). Similar results have been found in transgenic mice overexpressing rat TGFα generated by Sandgren and coworkers (Sandgren et al., 1990; Dempsey et al., 1992).

In this communication, we describe a major age-dependent disorganization of the gastric mucosa in MT-TGFα transgenic mice, resulting in a pathological accumulation of surface mucous cells and a concomitant depletion in differentiated, functional parietal and chief cells. Analysis of this in vivo transgenic model system indicates that chronic TGFα overexpression disrupts the dynamic balance between proliferation and terminal differentiation responsible for appropriate epithelial cell renewal in the stomach.

Transgenic mouse line MT100 bearing the MT-TGFα fusion gene was created as described previously (Jhappan et al., 1990). Mice used in this study were rederived and free of adventitious agents. Animals were housed up to 5 per cage in polycarbonate cages with hardwood chip bedding. Mice were given Ziegler Brothers mouse chow and water ad libitum. Animals were cared for and maintained in accordance with guidelines established by the National Institutes of Health.

Analysis of gastric secretions was as previously described (Takagi et al, 1992). Animals were fasted for 24 hours prior to experiment. For stimulation of acid secretion 500 μg/kg body weight of pentagastrin (Sigma Chemical Company) in 0.9% NaCl was administered into the peritoneal cavity at time of gastric ligation; the pylorus was ligated under methoxyflurane (Metofane, Pitman-Moore) anesthesia. Mice, recovered without food or water for between 2 and 2.5 hours, were killed and the acid content of their stomachs determined by titration with NaOH.

Routinely, resected stomachs were ligated at the esophagus and duodenum, inflated by injection of fixative and immersed in fixative. After 5 minutes, the stomach was opened along the greater curvature, flattened and immersed in the same fixative. Typically, 4 sections from the fundus and 2 sections from the pyloris were cut perpendicularly and embedded in paraffin. Thin sections were stained with either hematoxylin and eosin (H&E), or diastase-resistant periodic acid Schiff (PAS).

Rates of DNA synthesis were determined using immunohistochemical detection of BrdU incorporation. Briefly, 0.2 mg BrdU/gm body weight was administered IP to 2- to 6-week-old male mice 1 hour prior to killing. Stomach tissue was fixed in 70% ethanol, paraffin-embedded and sectioned at 5 μm. Stomach sections were treated with 0.01% trypsin for 3 minutes at 37°C and 0.3% hydrogen peroxide/methanol for 30 minutes to block endogenous peroxidase activity. After exposure to BrdU-specific primary antibody and biotinylated secondary antibody, sections were stained with the avidin/biotin complex and 3,3′-diaminobenzidine (Vectastain Kit, Vector Laboratories).

Apoptotic and necrotic cell death were assessed using the ApopTag in situ apoptosis detection kit, according to the manufacturer (Oncor), and light and electron microscopic analysis. Statistical analysis was performed using 2-tailed P values based on the Student’s t-test.

Total tissue RNA was isolated and analyzed using northern blot hybridization as described (Takagi et al., 1992). Unless otherwise indicated, 15 μg per lane of total RNA from the whole stomach were loaded on agarose gels and hybridized with one of a variety of ³²P-radiolabeled antisense riboprobes. Pepsinogen C and H⁺,K⁺-ATPase probes were prepared by reverse transcriptase-PCR from rat stomach RNA and subcloned into pGEM7. The rat pepsinogen C probe corresponded to nucleotides 515-715 (Ichihara et al., 1986) and the rat ATPase probe to amino acid residues 285-435 (Shull and Lingrel, 1986). A rat spasmolytic polypeptide probe was prepared as described by Jeffrey et al. (1994), and rat somatostatin and gastrin probes were prepared as described by Brand and Stone (1988). TGFα transgene activity was determined by hybridization to a 925 bp human TGFα probe (obtained from Graeme Bell), labeled either as an RNA riboprobe, or by random priming of the BglII-BglII cDNA insert.

For localization of transcripts encoding H⁺,K⁺-ATPase, spasmolytic polypeptide and somatostatin, stomachs were excised, immersion-fixed in neutral buffered formaldehyde and paraffin-embedded. 4 μm sections were permeabilized with proteinase K, postfixed in 4% paraformaldehyde in PBS, acetylated with acetic anhydride in 0.1 M triethanolamine and dehydrated. Hybridization was at 55°C in a 25 μl solution containing 0.02% Denhardt’s, 10% dextran sulfate, 50% formamide, 0.3 mg/ml bovine rRNA, 10 mM dithiothreitol, 300 Mm NaCl, 10 mM Na2HPO4, 10 mM Tris-HCl, 5 mM EDTA and 1×10⁶ cts/minute of [α-³⁵S]UTP-labeled anti-sense riboprobes (see above). The following morning tissue was washed with increasing stringency to 0.5× SSC, 65°C. Sections were dipped in NBT-2 nuclear emulsion (Eastman Kodak) and exposed for 6 to 10 days before development. Sections were counterstained with H&E.

Localization of TGFα transcripts in the gastric mucosa by in situ hybridization was performed according to Fox and Cottler-Fox (1993). Both antisense and sense RNA probes were used; endogenous mouse TGFα transcripts were barely detectable in the nontransgenic stomach under the conditions used (not shown).

To quantify changes in the thickness and regional composition of the gastric mucosa, sections from 3 month old control (n=2) and trans-genic (n=5) mouse stomachs were stained with toluidine blue and viewed through a microscope equipped with a camera lucida. Three lines were traced for each section: (1) the muscularis mucosae to establish the base, (2) the luminal surface and (3) a line approximating the junction between the region containing surface mucous cells above and the remaining glandular base below. Distances between lumen and muscularis mucosae were determined for each section, as were distances from lumen to the junctional line (the latter determining the thickness of the pit region) by averaging 2 to 3 measurements for each section. The thickness of the remaining glandular base region was calculated from the first 2 thicknesses. The ratio of mucosa occupied by surface mucous cells was calculated by dividing pit region thickness by mucosal thickness. Statistical comparison was by Student’s t test.

To determine changes in cellular volume fractions, toluidine bluestained sections of stomachs from 3-month-old mice were photographed at a magnification of 80×, which was sufficient to encompass the thickness of the mucosa in the controls, and also to encompass all parietal cells in the transgenics while including the basal extent of the glands. Between two and four photographs were prepared for each of 7 animals (2 control and 5 transgenic) at a magnification of 400×. A transparent grid was placed over each photograph and the cell type falling under each intersection was tallied. The groups enumerated were parietal cells, zymogenic cells, undifferentiated cells, mucous neck cells and intersections falling on anything else, which included spaces, lumina, connective tissue, blood vessels and surface mucous cells. Volume fractions were calculated for each cell type based on the fraction of intersections falling on each type, compared with the total number of intersections tallied. Transgenics were compared with controls for each cell type using the Student’s t-test.

Gastrin was measured in 4- to 5.5-month-old male mice by a radioim-munoassay using rabbit antiserum 2604, which was raised against synthetic human heptadecapeptide gastrin (gastrin 17) (Rehfeld et al., 1972). Tyrosine monoiodinated human gastrin 17 tracer was used in all assays and synthetic human gastrin 17 was used as a standard (Stadil and Rehfeld, 1972). Radioimmunoassays using antiserum 2604 react with component I, gastrin 34 and gastrin 17, sulfated and nonsulfated, with equal potency.

Stomachs of MT-TGFα transgenic mice (line MT100) demonstrated a dramatic age-dependent weight increase which, although not readily apparent in juveniles, became obvious between 2 to 3 months of age (Fig. 1). Histological analysis indicated that the increase in gastric mass was due almost exclusively to an accumulation of surface mucous cells in the pit (Table 1). The vast majority of these demonstrated positivity for diastase-resistant PAS staining, indicating that they were neutral mucin-producing cells (not shown). Further information on the nature of these cells was obtained by in situ hybridization using a radiolabeled rat spasmolytic polypeptide cRNA probe, a marker for mucin-producing cells (Poulsom and Wright, 1993). In the nontransgenic gastric mucosa, spasmolytic polypeptide was localized to a discrete band in the pit where mucous cells are typically found (Fig. 2A,B). In contrast, grains representing spasmolytic polypeptide were found throughout the expanding transgenic mucosa and were particularly high at the base near the submucosa (Fig. 2C-F).

Northern blot hybridization was used to monitor the expression of the human TGFα transgene during postnatal stomach development. Fig. 3 shows that TGFα transgene expression was not detectable in newborn mice, and was variable and generally low at 1 week of age. However, at 3 weeks, transgene expression had increased to about 70% of the adult level, which was much higher than endogenous TGFα gene expression in nontransgenic mice (Fig. 3). In situ hybridization was used to determine the spatial pattern of TGFα expression. There appeared to be a general correlation between the level of TGFα expression and disease progression. TGFα transcripts were poorly represented near the antrum, which was spared severe hypertrophic changes. In contrast, TGFα transcripts were highly expressed in the mature surface mucous cell population of the severely affected transgenic fundus (Fig. 4C,D). TGFα transcripts were detected in surface mucous cell precursors as well. Grains were heaviest in luminal surface mucous cells exposed to the contents of the stomach and at the glandular base (Fig. 4E,F). In the glands, grains were heaviest over more undifferentiated cell types and less dense over parietal cells (Fig. 4G). Typically, cells with mitotic figures tended to have fewer grains.

To determine if TGFα overexpression induced the accumulation of surface mucous cells through enhanced proliferation, rates of DNA synthesis were determined using BrdU incorporation and immunohistochemical staining. In the normal mucosa, the vast majority of DNA synthesis occurs throughout the life of the animal in the isthmus between the pit and the base of each tubule (Fig. 5A). However, in the 2-week-old MT100 mouse stomach, a distortion in this pattern was already evident; proliferative cells were frequently found outside the isthmus toward the glandular base and near the submucosa (Fig. 5B). Significantly, little proliferative activity was detected in the accumulating mature surface mucous cells of young adult transgenic mice (Fig. 5C). In older adults, elevated DNA synthesis was associated with more atypical, non-mucoid cells found in pathological regions near the luminal surface (Fig. 5D).

Quantification of BrdU-labeled nuclei revealed a persistent increase in the rate of DNA synthesis in gastric glands of juvenile transgenic mice relative to nontransgenic siblings (Table 2). The rates of DNA synthesis at 2.3 and 5.7 weeks of age were 57% and 78% higher, respectively, in the transgenic stomach.

Cell death in the gastric mucosa of juvenile mice was assessed through direct immunoperoxidase detection of digoxigenin incorporation onto free 3′-OH ends of fragmented genomic DNA. The number of gastric cells undergoing apoptosis or necrosis was determined in transgenic and control mice 3 and 4.7 weeks of age. Table 3 shows that the rate of cell death in the glandular region of the juvenile MT100 gastric mucosa was not overtly different from nontransgenic siblings. However, an increase in cell death was detected in the transgenic pit region (Table 3).

The pH of gastric secretions was essentially unchanged in juvenile MT100 mice up to 6 weeks of age, but was neutral in adults where mucosal thickening was exaggerated (Takagi et al., 1992). Histological analysis of the gastric mucosa demonstrated that this change in acid production was caused by a decrease in parietal cells. Large, mature parietal cells were numerous in the stomachs of both control and transgenic mice at 2 weeks of age. By 1 month of age, control gastric glands were composed predominantly of parietal cells (Fig. 6B,C). In contrast, the non-mucoid portion of the transgenic gastric glands was composed predominantly of undifferentiated cells (Table 4), with a few smaller parietal cells (Fig. 6D).

The functional status of the parietal cell population was determined using a parietal cell-specific H⁺,K⁺-ATPase marker. In nontransgenic mice, in situ hybridization demonstrated H⁺,K⁺-ATPase expression throughout the fundus, from the glandular base to all but the most luminal surface mucous cells (Fig. 7A-D). In 3-week-old MT100 transgenic mice, H⁺,K⁺-ATPase staining was still quite strong and widespread, but more disorganized in appearance (Fig. 7E,F). In contrast, expression of H⁺,K⁺-ATPase in adult MT100 mice was restricted to the less severely affected oxyntic mucosa near the antrum, at the base adjacent to the submucosa (Fig. 7G,H).

To quantify and confirm changes in H⁺,K⁺-ATPase seen by in situ hybridization, RNA from stomachs of adult transgenic and control mice were analyzed by northern blot hybridization. Fig. 8 shows that H⁺,K⁺-ATPase RNA was dramatically reduced in the stomach of an adult MT100 mouse relative to its nontransgenic counterpart. In comparison, the same pair of samples exhibited no difference in the level of transcripts encoding spasmolytic polypeptide, the mucous cell marker (Fig. 8). Table 5, which presents results from an expanded analysis of 4 transgenic and 5 control mice, shows that, in adult MT100 mice, H⁺,K⁺-ATPase RNA was lowered by a factor of 15. When weight-adjusted to account for mucous cell accumulation, this parietal cell RNA marker was determined to be reduced 6.7-fold.

Changes in H⁺,K⁺-ATPase gene expression in the transgenic stomach were next examined as a function of time. Fig. 9 shows that H⁺,K⁺-ATPase RNA levels were nearly normal in MT100 mice less than 1 month of age; however, by 2 months, H⁺,K⁺-ATPase RNA levels had fallen markedly and thereafter remained low. In contrast, H⁺,K⁺-ATPase RNA levels were essentially unchanged in adult FVB/N mice (not shown).

To determine if the transgenic gastric mucosa could be induced to secrete acid, adult MT100 mice were stimulated with pentagastrin. Fig. 10 shows that the gastric mucosa of the transgenic mice failed to respond to pentagastrin. In contrast, gastric acid production was clearly stimulated in control mice exposed to pentagastrin.

Histological analyses demonstrated that TGFα induced a massive depletion of mature chief cells in MT100 mice. In control mice at 2 weeks of age, tiny zymogen granules were present in cells throughout the gastric mucosa. By 1 month, zymogen granule-containing cells were concentrated in the basal third of the mucosa and larger zymogen granules were present. By 2 months, the adult situation was reached with mature chief cells containing many large granules distributed along the basal third of the mucosa (Fig. 6B,C). In striking contrast, mature chief cells never developed in the MT100 mucosa. Small zymogen granules were present at first in the transgenic mucosa, but they remained tiny, with only a few in precursor chief cells. By 1 month, clear differences were evident in transgenic zymogen granule number and size compared with controls. By 2 months even chief cells containing tiny granules were extremely rare. Significantly, mucous neck cells, the precursors of chief cells, were more numerous in the gastric mucosa of transgenic mice greater than 1 month of age relative to controls (Table 4). Mucous neck cells formed the bases of many transgenic gastric glands, occupying the position taken by chief cells in the control mucosa (Fig. 6A,D).

To evaluate further the functional status of the chief cell population, northern blot hybridization was performed on the same RNA samples from adult transgenic and control stomachs using a chief cell-specific rat pepsinogen C radiolabeled probe. Fig. 8 shows that, like H⁺,K⁺-ATPase transcripts, pepsinogen C RNA was dramatically reduced in the transgenic gastric mucosa; the decrease was determined to be 6.3-fold after adjustment for increased stomach weight (Table 5).

Specific representative endocrine cells were also examined by northern blot hybridization. Fig. 8 shows that somatostatin and gastrin transcripts, specific for D cells and G cells, respectively, were not as severely reduced in the transgenic stomach as parietal and chief cell markers. Table 5 indicates that, after adjustment for changes in stomach weight, somatostatin RNA levels were virtually unchanged, while gastrin transcripts were reduced 2.8-fold.

To determine if the distribution of endocrine cells was affected in the MT100 stomach, a somatostatin riboprobe was chosen for in situ hybridization. Fig. 2 shows that, in the transgenic stomach (I, J), much like the control (G, H), somatostatin-producing D cells were well represented and scattered throughout the mucosa. In the transgenic mucosa many somatostatin-producing cells were closer to the luminal surface relative to the spasmolytic polypeptide-producing surface mucous cells (compare Fig. 2J and D); this was not seen in the control gastric mucosa (compare Fig. 2H and B).

Serum from 5 transgenic and 5 control male mice 4 to 5.5 months of age was examined for changes in circulating gastrin. Serum gastrin levels in transgenic mice (111±31 pg/μl; mean±s.d.) were comparable to those found in controls (129±25 pg/μl; mean±s.d.), indicating that G cells were present and functional in the antrum of the MT100 stomach.

Gastric epithelial renewal, like embryonic development, requires balanced regulation of cellular proliferation, migration, differentiation and senescence. When mechanisms responsible for regulating any of these processes are compromised, pathological conditions develop; in the human stomach, these include hyperplastic and hypertrophic gastropathies, such as Ménétrier’s disease (Ménétrier, 1888). Here we report that overexpression of a TGFα transgene disturbs the dynamic balance responsible for appropriate gastric postnatal development and renewal.

Every MT100 mouse stomach demonstrated an age-dependent increase in mass caused by an accumulation of surface mucous cells, identified as positive for neutral mucins by diastase-resistant PAS staining and for spasmolytic polypeptide by in situ hybridization. Analysis of BrdU uptake showed that this accumulation was not due to proliferation of mature surface mucous cells in the pit; however, a significant increase in DNA synthesis was detected in the glandular portion and the isthmus, known to contain surface mucous progenitors (Karam and Leblond, 1992). Normally, surface mucous cells live about 3 days, during which time they migrate from their point of origin in the glandular isthmus toward the luminal surface where they undergo apoptosis or necrosis (Kataoka et al., 1985; Lee, 1985; Karam and Leblond, 1993). Quantification of apoptotic/necrotic cells indicated that accumulation of surface mucous cells in the transgenic pit was not caused by enhanced cell survival.

These data suggest that surface mucous cell accumulation is due, at least in part, to a mitogenic stimulation of progenitor cells by TGFα; this could occur through autocrine and/or paracrine mechanisms. Several lines of evidence support this hypothesis: TGFα has a mitogenic effect on the gastric mucosa (Johnson and Guthrie, 1980; Dembinski et al., 1982; Chen et al., 1991); surface mucous progenitors are replication competent (Karam and Leblond, 1993); a similar increase in the labeling index of epithelial cells (57%) is sufficient to induce hyperplasic changes in the gastric pits and glands of Zollinger-Ellison patients (Castrup et al., 1975); enhanced progenitor cell DNA synthesis is the mechanism responsible for gastrin-induced hyperplasia of parietal cells (Willems and Lehy, 1975). Expansion of the surface mucous cell population was not caused by hypergastrinemia, because serum gastrin levels were comparable in transgenic and control animals.

Perhaps the most significant feature of the adult MT-TGFα mouse is achlorhydria. TGFα can inhibit gastric acid secretion directly in normal parietal cells (Gregory, 1975; Bower et al., 1975; Finke et al., 1985; Rhodes et al., 1986; Guglietta et al., 1994). However, this mechanism alone cannot account for achlorhydria in the MT100 model; the pH of gastric secretions was normal in transgenic mice up to at least 6 weeks of age, long after TGFα transgene expression reached a maximal level (Takagi et al., 1992). In humans and other mammals, gastric acid producing capacity correlates with parietal cell number (Tongen, 1950; Card and Marks, 1960; Polacek and Ellison, 1966; Witzel et al., 1977). Here we demonstrate that achlorhydria in MT100 mice is due, in large part, to an age-dependent loss of fully differentiated, responsive parietal cells. The number and distribution of H⁺,K⁺-ATPase transcripts and identifiable parietal cells were relatively normal in transgenic mice less than 3 weeks of age (when transgenic TGFα levels were low; Fig. 11). However, in adults H⁺,K⁺-ATPase RNAs were reduced 6.7-fold and restricted to the distal portion of the stomach, which was spared severe hypertrophic changes and characterized by a relative reduction in TGFα transcripts. The time required for the loss of the majority of H⁺,K⁺-ATPase transcripts (>6 weeks) and by inference functional parietal cells, corresponded very well to the inhibition of acid production.

Significantly, parietal cell loss in the transgenic stomach was accompanied by a dramatic reduction in chief cells. This was demonstrated by the conspicuous absence of zymogen granules in the adult gastric mucosa and by the sharp decrease in pepsinogen C RNA in the whole stomach. The early depletion of chief cells coincided with activation of the TGFα transgene between 1 and 3 weeks of age (Fig. 11). Therefore, TGFα overexpression induced, either directly or indirectly, depletion of the two major non-mucoid exocrine cell types. The specificity of this effect was demonstrated through analysis of endocrine cells in the transgenic stomach; for example, examination of transgenic D cells revealed little difference in activity or distribution.

It is a formal possibility that the early disappearance of these specific glandular markers is due to an increase in apoptotic or necrotic cellular death associated with hypertrophic compression of the glandular compartment. However, no overt difference in cell death in the glandular portion was detected among transgenic and control juveniles, although a significant increase was detected in the transgenic pit. Moreover, levels of H⁺,K⁺-ATPase and pepsinogen C RNA had begun to fall before the stomach increased in weight (Fig. 11). It is therefore highly unlikely that this mechanism is responsible for the observed loss of parietal and chief cells in the juvenile transgenic stomach.

We strongly favor the notion that TGFα can, upon reaching a threshold level of expression, interfere with and effectively block terminal differentiation of parietal and chief cells. The strongest evidence for this hypothesis is that TGFα expression results in a loss of parietal and chief cells without a concomitant loss of their less differentiated precursors. If this hypothesis is correct, it would then follow that only cells that were sufficiently differentiated when TGFα transgene expression reached the effective threshold level would be functional and these would be lost over time through normal attrition. The average life-span of mouse parietal cells is about 2 months (Karam and Leblond, 1993), which fits well with the observed age-dependent loss of H⁺,K⁺-ATPase transcripts following maximal expression of TGFα (Fig. 11). Although the more rapid decline in pepsinogen C transcripts is inconsistent with a natural loss of mature chief cells, whose average life span in the mouse is about 6 months (Karam and Leblond, 1993), this could be accounted for if mature chief cells were not fully formed before threshold TGFα levels were reached at about 3 weeks of age and thereafter failed to form altogether. In fact, only primitive chief cells with small, weakly labeled granules are found in the gastric mucosa of both control and TGFα transgenic mice less than 3 weeks of age (see Results; Kataoka et al., 1990).

Several alternative explanations are less likely. TGFα may be capable of redirecting differentiation of common stem cells, called undifferentiated granule free cells (Karam and Leblond, 1993), toward the formation of surface mucous cell progenitors in the pit at the expense of parietal and chief cell progenitors in the gland. This model is not compelling because no shortage of such undifferentiated precursors has been detected in the transgenic gastric mucosa. Alternatively, TGFα could reverse gastric exocrine cell maturation, which appears to be the cause of acinar cell conversion to ductular cells and mucin-producing cells in the MT-TGFα pancreas (Jhappan et al., 1990; Sandgren et al., 1990; Bockman and Merlino, 1992). However, in the transgenic stomach, there is no evidence for the existence of such transitional cell types, which are clearly evident in the transgenic pancreas.

Although apparently acting through different cellular mechanisms, TGFα transgene overexpression has broadly similar effects on the stomach and pancreas: selective reductions in non-mucoid exocrine cell content and gene expression. In both tissues, exocrine suppression by TGFα antagonizes the actions of gastrin-like peptide hormones. Gastrin, in addition to being a gastric acid secretagogue, stimulates growth of parietal cells and increases H⁺,K⁺-ATPase and pepsinogen gene expression in the stomach. In the pancreas, gastrin’s homolog, cholecystokinin, stimulates growth of the exocrine pancreas and expression of exocrine genes. We have shown that expression of gastrin in the pancreas partially reverses TGFα-induced exocrine suppression in gastrin/TGFα bitransgenic mice (Wang et al., 1993). Therefore, alterations in the balance between gastrin and TGFα stimulation may contribute to grossly altered mucosal differentiation in the MT100 stomach.

A remarkable finding was that gastric achlorhydria in MT100 mice did not result in hypergastrinemia or increased gastrin RNA levels. In most mammals, achlorhydria induces hypergastrinemia through reduced paracrine inhibition of antral gastrin cells by somatostatin. Thus, achlorhydria is usually accompanied by a decrease in somatostatin gene expression; in this respect, the response of the transgenic gastric mucosa to achlorhydria is also atypical. Further studies are required to determine how TGFα alters the responsiveness of antral D cells to changes in luminal acidity.

In conclusion, TGFα can disrupt cellular differentiation as well as proliferation when overexpressed in the stomach, suggesting that this EGF receptor-specific ligand can participate in normal developmental processes. These data broaden the functions attributable to TGFα and strengthen the contention that TGFα be defined as a multipurpose signaling molecule, and not solely as a growth factor. The MT-TGFα transgenic stomach appears to be a highly useful and relevant animal model for analysis of general mechanisms associated with the regulation of cellular differentiation and proliferation in tissues undergoing continuous renewal.

The authors thank Drs Jerrold Ward, Gilbert Smith, Thomas Sargent and Chamelli Jhappan for critical review of this manuscript and useful discussions. We acknowledge Dr Cecil Fox for advice and assistance with the TGFα in situ hybridization, Dr Robert Tarrone for statistical analysis, and Dr Mirian Anver, Shirley Hale and Barbara Kasprzak for help with tissue preparation and histopathological analysis. The authors thank Greg Oblak for technical assistance. S. T. is grateful to the following for support: Medical Research Council, Stockholm (14F-10044; 14R-10076); Kungliga Fysiografiska Sällskapet, Lund; Thelma Zoégas fond för medicinsk forskning, Helsingborg; Bokelunds medicinska fond, Lund; Sandozstiftelsen för medicinisk-biologisk forskning, Täby; Johan Petterssons stiftelse, Höör.