Using a dominant‐negative mutant receptor (DNR) approach in transgenic mice, we have functionally inactivated transforming growth factor‐β (TGF‐β) signaling in select epithelial cells. The dominant‐negative mutant type II TGF‐β receptor blocked signaling by all three TGF‐β isoforms in primary hepatocyte and pancreatic acinar cell cultures generated from transgenic mice, as demonstrated by the loss of growth inhibitory and gene induction responses. However, it had no effect on signaling by activin, the closest TGF‐β family member. DNR transgenic mice showed increased proliferation of pancreatic acinar cells and severely perturbed acinar differentiation. These results indicate that TGF‐β negatively controls growth of acinar cells and is essential for the maintenance of a differentiated acinar phenotype in the exocrine pancreas in vivo. In contrast, such abnormalities were not observed in the liver. Additional abnormalities in the pancreas included fibrosis, neoangiogenesis and mild macrophage infiltration, and these were associated with a marked up‐regulation of TGF‐β expression in transgenic acinar cells. This transgenic model of targeted functional inactivation of TGF‐β signaling provides insights into mechanisms whereby loss of TGF‐β responsiveness might promote the carcinogenic process, both through direct effects on cell proliferation, and indirectly through up‐regulation of TGF‐βs with associated paracrine effects on stromal compartments.
The transfoming growth factors‐βs (TGF‐β) are multifunctional cytokines which regulate cell growth, differentiation and function (Roberts and Sporn, 1990), and recent evidence suggests that they constitute part of an important tumor suppressor pathway (Markowitz and Roberts, 1996). One or more of the three mammalian TGF‐βs is expressed in nearly every tissue in the body, implicating this as a widely used regulatory system throughout development and adulthood (Flanders et al., 1989; Millan et al., 1991). However, despite a wealth of data on the multitude of biological activities of TGF‐βs in vitro, in most cases the precise roles played by the TGF‐βs in a particular in vivo setting are not known. This is in part because TGF‐β action is strongly contextual. Thus, the specific effect of TGF‐β on a particular cell appears to be an integrated function of the target cell type, its differentiated state and its environmental context, particularly regarding the nature of the extracellular matrix and the activities of other cytokines (Nathan and Sporn, 1991). For example, while TGF‐β is a potent inhibitor of the growth of keratinocytes in vitro, transgenic overexpression of TGF‐β in the skin can lead to an unexpected stimulation of keratinocyte proliferation in the basal state, but results in growth inhibition in a hyperplastic setting, following treatment with the phorbol ester TPA (Cui et al., 1995). This demonstrates that the physiological roles of TGF‐βs in a particular tissue in vivo may not be readily predictable from its in vitro effects on cells derived from that tissue.
As one approach to the molecular dissection of the in vivo biology of TGF‐β, we wished to generate animal models in which TGF‐β function is experimentally compromised in select tissues. The type I and type II TGF‐β receptors (TβRI and TβRII) are activated by ligand‐dependent formation of hetero‐oligomeric complexes, in which TβRII transphosphorylates and activates TβRI, thereby initiating the signal transduction cascade (Wrana et al., 1994). In order to eliminate the TGF‐β response in target tissues, we have used transgenic overexpression of a dominant‐negative mutant form of the TβRII (Brand et al., 1993; Chen et al., 1993), resulting in tissue‐restricted functional inactivation of the TGF‐β receptor complex. This approach of local functional inactivation of genes has attractive advantages when compared with germline null mutations generated by gene targeting. First, loss of function can be targeted to specific cells in selected organs with appropriate transcriptional control elements, and potential embryonic lethality can be circumvented. Because of the important roles of TGF‐βs during embryonic development, null mutations in the TβRII gene are expected to result in embryonic lethality. Second, the problem of isoform redundancy, a potential issue with the three mammalian TGF‐β isoforms, can be eliminated. Third, confounding systemic effects such as widespread inflammation observed in multiple organs of the TGF‐β1 null mouse (Kulkarni et al., 1993) can be avoided.
We report here the effects of functional inactivation of TGF‐β in target tissues in mice with expression of a dominant‐negative mutant TβRII (DNR) under control of a metallothionein 1 (MT1) promoter (Palmiter et al., 1993). The results show that TGF‐β negatively controls growth of pancreatic acinar cells and is essential for the maintenance of a differentiated acinar phenotype in the exocrine pancreas in vivo. Furthermore, loss of TGF‐β responsiveness in DNR‐positive acinar cells results in an unexpected increase in expression of TGF‐β1 and TGF‐β3 in vivo in the exocrine pancreas. This is associated with classic correlates of TGF‐β overexpression such as fibrosis, angiogenesis and macrophage infiltration. These results have implications for understanding the complex roles of the TGF‐β system in tumorigenesis.
Generation of DNR transgenic mice and analysis of expression
The DNR transgene encodes the extracellular and transmembrane domains of TβRII and was expressed under control of a mouse MT1 promoter and MT locus control regions (LCRs) (Figure 1A) (Palmiter et al., 1993). We generated two lines of transgenic mice, line AM3 with three copies, and line BF1 with one copy per haploid genome, respectively, in the strain FVB/N. Northern blotting revealed DNR RNA expression in both lines. In line AM3, the highest zinc‐induced levels of DNR RNA were observed in the pancreas, liver, colon and small intestine (Figure 1B). Kidney, stomach and salivary gland had substantially lower levels of expression (Figure 1B). DNR expression was consistently lower in tissues from line BF1 mice when compared with line AM3 tissues (Figure 1C). In liver and pancreas, substantial levels of DNR were expressed without zinc induction (Figure 1C). To show the cellular localization of DNR protein in situ, we stained sections of pancreas and liver with anti‐human TβRII(1–28) antibody. Cell surface‐associated expression of DNR was observed in acinar cells in the pancreas and hepatocytes in the liver (Figure 2A and C). DNR expression was heterogeneous in both organs and was highest at 2 months of age.
Dominant‐negative mutant function of DNR in hepatocytes and acinar cells
Ligand binding and receptor complex formation. Primary hepatocytes and pancreatic acinar cells were prepared from either line AM3 or from non‐transgenic FVB/N mice. Ligand binding by DNR was assessed by receptor affinity labeling with [125I]TGF‐β1. Ligand–receptor complexes were undetectable in lysates of affinity‐labeled non‐transgenic FVB/N hepatocytes, indicating low levels of endogenous receptors (Figure 3A). In hepatocyte lysates from line AM3, a prominent complex of ∼40 kDa represented DNR bound to [125I]TGF‐β1, as confirmed by immunoprecipitation of affinity‐labeled lysates with anti‐human TβRII(1–28) (Figure 3A). Analogous experiments showed similar results using primary acinar cell preparations (Figure 3B).
Loss of TGF‐β responsiveness. Primary pancreatic acinar cell cultures from control FVB/N mice showed a marked inhibition of [3H]thymidine incorporation when treated with either TGF‐β1, TGF‐β2, TGF‐β3 or activin, a member of the TGF‐β superfamily structurally closely related to TGF‐β (Figure 3C). In contrast, TGF‐β isoforms had no effect on [3H]thymidine incorporation in acinar cells from line AM3 (Figure 3C). However, activin inhibited [3H]thymidine incorporation in transgenic acinar cells (Figure 3C), indicating that DNR expression only inactivates TGF‐β signaling, but not signaling by other members of the TGF‐β superfamily. Primary hepatocytes from control FVB/N mice were growth inhibited by TGF‐β isoforms (Figure 3D), but not by activin (data not shown). None of the three TGF‐β isoforms inhibited growth of hepatocytes from line AM3 (Figure 3D). In addition, induction of fibronectin protein secretion by TGF‐β1, as seen in FVB/N hepatocytes, was absent in line AM3 hepatocytes (Figure 3E). These results indicate that DNR expression in cells from transgenic mice completely abrogates both the growth inhibition mediated by the three TGF‐β isoforms, and the TGF‐β mediated induction of matrix‐associated genes.
Phenotypic characterization of transgenic DNR mice
On macroscopic inspection, the pancreas of line AM3 mice of 5 months and older had a lighter color when compared with the pancreas from non‐transgenic littermates. Relative pancreatic weight expressed as a percentage of total body weight was significantly decreased in line AM3 mice when compared with FVB/N mice at 8 months of age (0.82 ± 0.10% versus 1.08 ± 0.03%, P <0.05). All other organs including the liver, colon and small intestine were macroscopically normal.
Histological analysis of the pancreas showed severe abnormalities, including ductular transformation, neoangiogenesis, inter‐ and intralobular fibrosis and adipose replacement of acini in the exocrine pancreas of both transgenic lines AM3 and BF1 (see Table I; Figures 4 and 5). The islets of Langerhans appeared normal, consistent with the absence of DNR expression in endocrine cells. The abnormalities in the exocrine pancreas were generally more severe and occurred at a younger age in line AM3 when compared with line BF1, consistent with increased levels of DNR expression in the AM3 pancreas (Figure 1C). We focused therefore on line AM3 for a detailed analysis of the phenotype in the pancreas.
The earliest feature of the phenotype was the appearance of aberrant non‐acinar cells in the exocrine pancreas from 3 weeks of age. These cells increased progressively in number and formed primitive ductular structures (Table I; Figure 4B). Strong staining with AE1/AE3 antibody, a marker of ductal epithelium, was observed in most of the aberrant non‐acinar cells, confirming their ductal cell phenotype (Figure 4D). In addition, acinar cells in line AM3 mice were progressively replaced by adipose cells (Table I; Figure 4F). Tubular complexes, reflecting acinoductular metaplasia (Bockman, 1981), were found in some younger line AM3 mice and appeared frequently in older transgenic mice, often transforming entire lobules (Table I; Figure 4G). Within the pancreas of transgenic lines AM3 and BF1, the severity of the described abnormalities correlated well with the level of DNR expression in affected lobules. Older animals (>14 months) were also examined for the presence of eosinophilic foci of acinar cells, considered a marker of focal hyperplasia in rats (Eustis et al., 1990). Eosinophilic foci were not observed in non‐transgenic FVB/N mice (n = 12), but were found in 21% (7/33) (P = 0.001) of line AM3 pancreas (Figure 4H).
To better define the abnormalities in the transgenic pancreas, we applied specific staining techniques. Inter‐ and intralobular fibrosis developed by 5 months of age in line AM3 mice, and increased progressively (Table I). Reticulin staining of pancreas sections for collagen demonstrated significant expansion of the extracellular matrix in 8‐month‐old line AM3 mice when compared with FVB/N mice (18.6 ± 5.8% versus 3.5 ± 0.5% reticulin‐stained section surface, respectively; P <0.05) (Figure 5A and B). Consistent with this, RNA expression levels of collagen 1, fibronectin and tissue inhibitor of metalloproteinase 1 (TIMP‐1) were increased in line AM3 pancreas when compared with non‐transgenic pancreas at 1, 3, 6 and 10 months of age (data not shown). Increased angiogenesis (see also Figure 4B) and macrophage infiltration were revealed by immunostaining for von Willebrand factor VIII (Figure 5D) and Mac‐2 antigens (Figure 5F), respectively.
TGF‐β regulates growth and affects differentiation of acinar cells in the exocrine pancreas in vivo
Proliferation and apoptosis. To address possible mechanisms underlying the observed pancreatic phenotype, we analyzed proliferative and apoptotic rates in the pancreas. Immunohistochemistry for proliferating cell nuclear antigen (PCNA) demonstrated a large increase in the proportion of acinar cells in cell cycle in line AM3 pancreas when compared with control FVB/N pancreas at 6 weeks (Table II; Figure 6A and B) and 8 months of age (Table II). This was accompanied by large numbers of apoptotic cells in line AM3 pancreas, but not in FVB/N pancreas (Table II; Figure 6C and D). The aberrant ductular epithelial cells present in line AM3 pancreas showed no evidence of increased proliferative activity, as demonstrated by the low frequency of PCNA staining in these cells (1.2 ± 0.4% at 6 weeks of age), when compared with acinar cells (35.0 ± 5.1% at 6 weeks of age) (Figure 6A). These results indicated that inactivation of TGF‐β signaling in acinar cells causes an increase in proliferation and apoptosis throughout adult life in transgenic mice. In contrast, rates of proliferation and apoptosis in the liver were not significantly different between line AM3 mice and FVB/N littermates (data not shown).
Differentiation and dedifferentiation. The accumulation of largely non‐proliferating ductular cells and duct‐like structures with the loss of acini suggested that normal differentiation and/or maintenance of a differentiated phenotype were disturbed in the transgenic pancreas. Toluidine blue staining of semi‐thin (0.5 μm) sections of line AM3 pancreas showed that acini were replaced by ductular structures harboring transitional cells, characterized by ductal morphology, but containing zymogen granules (Figure 6E). Within the same structure, cells with remnant acinar morphology were present (Figure 6E). Dedifferentiation of acini into ductal structures as so‐called ‘tubular complexes’ has been described in association with neoplastic and inflammatory conditions in the pancreas (Bockman, 1981). Tubular complexes were present in severely affected pancreatic lobules in line AM3 mice with increasing age (Table I; Figure 4G). These findings suggest that inactivation of TGF‐β signaling in acinar cells in vivo is associated with dedifferentiation of acini into duct‐like structures and tubular complexes. In addition, ductal cells may accumulate in association with high rates of acinar cell turnover (see Table II) in the transgenic pancreas.
Inactivation of TGF‐β signaling results in a selective increase of expression of TGF‐β isoforms
Because increased expression of matrix‐associated genes and angiogenesis in vivo have been associated with increased TGF‐β activity (Roberts et al., 1986), we examined levels of expression of TGF‐β isoforms. By Northern blot analysis, expression of TGF‐β1 and TGF‐β3 RNA were markedly increased in line AM3 pancreas, and TGF‐β2 RNA levels were slightly increased (Figure 7A). Similar results were obtained at 1, 2, 6 and 10 months of age. When compared with primary acinar cells from FVB/N control littermates, primary acinar cells from line AM3 mice had increased levels of TGF‐β1 RNA (Figure 7B). Immunostaining with TGF‐β isoform‐specific antibodies showed little staining for TGF‐β1, TGF‐β2 and TGF‐β3 in the exocrine pancreas of FVB/N mice (Figure 8). However, TGF‐β1 protein was markedly increased in acinar cells and aberrant non‐acinar cells, and TGF‐β3 protein was markedly increased exclusively in acinar cells in the pancreas of line AM3 (Figure 8). TGF‐β2 staining was not changed (Figure 8). These results suggest that inactivation of TGF‐β signaling in acinar cells in vivo leads to increased expression of TGF‐β1 and TGF‐β3 isoforms.
Using a dominant‐negative mutant receptor approach in transgenic mice, we have demonstrated for the first time the feasibility of inactivating TGF‐β in select tissues in the whole animal. Specifically, we have shown that TGF‐βs play an essential role in maintaining epithelial homeostasis and the differentiated phenotype in the exocrine pancreas, as demonstrated by severely perturbed cellular proliferation parameters and development of characteristic histologic abnormalities in the exocrine pancreas of two transgenic lines. While hepatocytes and acinar cells derived from the transgenic mice showed no growth inhibition or gene induction in response to any of the three TGF‐β isoforms, growth inhibition induced by activin, the TGF‐β superfamily member most closely related to TGF‐β, was not affected. This confirms that the dominant‐negative effect of this construct is confined to TGF‐β1, 2 and 3.
Mechanisms underlying the disruption of homeostasis
Our data clearly indicate that TGF‐β is an essential negative regulator of pancreatic acinar cell growth in vivo, consistent with in vitro data (Logsdon et al., 1992). We observed no changes in hepatocyte proliferation, apoptosis or collagen deposition in vivo in the liver of DNR transgenic mice when compared with normal mice, despite total loss of responsiveness to TGF‐β in DNR transgenic hepatocytes in vitro. This suggests that TGF‐β may not be a direct inhibitor of hepatocyte proliferation in the normal liver, and that the relative contribution of endogenous TGF‐βs to epithelial homeostasis may vary from organ to organ. Since it has been shown that transgenic overexpression of active TGF‐β1 in the liver can suppress the early proliferative response after partial hepatectomy (Böttinger et al., 1996), it is possible that TGF‐β may play a role in growth control in the liver when proliferative homeostasis is perturbed. We are currently investigating this possibility. However, it is also possible that, despite the in vitro data, in vivo expression of the DNR construct in hepatocytes may not be adequate to inactivate the function of TβRII.
Concomitant with increased proliferation, we observed increased apoptosis in the transgenic acini throughout adult life. Induction of apoptosis has been shown to accompany enhanced cell proliferation resulting from many forms of perturbation of the cell cycle, including oncogene expression or inactivation of negative regulators of the cell cycle (Evan et al., 1992; Jacks et al., 1992; Howes et al., 1994; Naik et al., 1996; White, 1996). Since TGF‐β actually induces apoptosis in many epithelial cell types (Oberhammer et al., 1992; Hsing et al., 1996), the increased apoptosis observed in DNR‐positive acini most likely represents an appropriate response to the abnormal proliferation and differentiation, as opposed to a direct effect of loss of TGF‐β function in these cells.
The increased numbers of primitive ductules in the transgenic pancreas at all ages, and the presence of tubular complexes, suggest that loss of negative growth regulation by TGF‐β in the acinar cells may be incompatible with the acquisition or maintenance of the differentiated phenotype. It is well documented that acinar cells, which derive from ductal cells during normal development (Githens, 1993), can retrodifferentiate into cells with the ductal phenotype, forming ductules and tubular complexes in experimental models and human diseases associated with abnormal growth in the pancreas (Logsdon, 1995). Furthermore, when acinar cells are stimulated to grow in vitro, they dedifferentiate into a more fetal‐like phenotype with ductal characteristics, while on cessation of growth, differentiated acinar markers are re‐acquired (De Lisle and Logsdon, 1990). Thus acinar dedifferentiation can be a direct consequence of increased proliferation in this normally quiescent compartment.
Several lines of evidence support the conclusion in our transgenic model that the abnormally proliferating acinar cells may dedifferentiate into cells with a ductal phenotype. First, we have observed the presence of intermediate cell types with features of both acinar and ductal morphology (Figure 6). Similar intermediate cell types consistent with acinoductular retrodifferentiation have been seen using ultrastructural analysis in the pancreas of MT‐TGF‐α transgenic mice (Bockman and Merlino, 1992). Second, the affected lobules in the transgenic animals are characterized by a net loss of acinar cells, despite greatly increased proliferation in the acinar compartment. In contrast, a net gain of ductular cells occurs in these lobules without increased proliferation in the ductal compartment. These observations are most consistent with acinoductular metaplasia. It should be emphasized here that the ductular cells do not express the DNR transgene. Dedifferentiated ductular cells in tubular complexes of pancreas of MT‐TGF‐α transgenic mice also fail to express transgenic TGF‐α, suggesting that the MT1 promoter is not expressed in ducts and ductules (Jhappan et al., 1990). Hence, these cells may re‐acquire responsiveness to TGF‐β‐mediated growth inhibition. Finally, the absence of inflammation or necrosis in most of the affected lobules suggests that increased acinar proliferation and dedifferentiation result from loss of TGF‐β function in the DNR transgenic pancreas, rather than as a secondary consequence of other pathologic changes.
Replacement of pancreatic acini within lobules by adipose tissue is often observed in humans and animals accompanying atrophy of acini from any cause including aging (Seifert, 1984; Eustis et al., 1990). In the DNR transgenic pancreas, where acinar homeostasis is severely disturbed by abnormal proliferation, differentiation and apoptosis, the sometimes dramatic replacement of acini by intralobular adipose tissue is probably a non‐specific consequence of the resulting acinar atrophy, and not directly related to loss of TGF‐β function.
TGF‐β autoregulation and carcinogenesis
Profound changes were observed in the stromal compartment of the DNR transgenic pancreas, including neoangiogenesis, fibrosis and low‐level infiltration by macrophages. All these are similar to changes seen in other systems when TGF‐β activity is increased (Roberts et al., 1986; Pierce et al., 1989), and consistent with the fibrosis and cellular infiltrates observed in the exocrine pancreas of mice transgenic for TGF‐β1 expressed under the control of an insulin promoter (Lee et al., 1995). We therefore examined the pancreas for expression of the three isoforms of TGF‐β and found strikingly increased levels of TGF‐β1 and TGF‐β3 in transgenic acinar cells, indicating that inactivation of TGF‐β signaling results in increased expression of TGF‐β1 and TGF‐β3 in these cells.
Previous in vitro studies in several TGF‐β‐responsive epithelial and fibroblast cell lines have indicated complex patterns of auto‐ and cross‐regulation among TGF‐β isoforms (Bascom et al., 1989; Kim et al., 1990). Here we show for the first time in vivo that lack of response to TGF‐β and overexpression of TGF‐β may be mechanistically related, and this may have important implications for understanding the roles of the TGF‐β system at different stages in the carcinogenic process. While TGF‐β is a growth inhibitor for most epithelia (Roberts and Sporn, 1990), and transgenic overexpression of TGF‐β protects the mammary gland from tumorigenesis induced by chemical carcinogens or oncogenes (Pierce et al., 1995; G.Merlino, J.Jakubczak and G.Smith, unpublished data), expression of TGF‐β isoforms is actually increased in many advanced human cancers (Gorsch et al., 1992; Friess et al., 1993; Gold and Korc, 1994; Gold et al., 1994). The elevated TGF‐β is proposed to promote tumor progression, primarily through paracrine effects on stromal elements, such as increased angiogenesis and decreased immune surveillance. These observations are reconciled by the hypothesis that the TGF‐β system has tumor suppressor activity early in tumorigenesis when the TGF‐β response is intact, but has oncogenic effects later if the TGF‐β response in the tumor cell is lost.
Recently, it has been shown that TβRII is inactivated or deleted in several different types of human tumor (Markowitz and Roberts, 1996). Our data suggest this may have two deleterious consequences, namely (i) loss of response of the tumor cell to the growth inhibitory effects of TGF‐β, and (ii) consequent up‐regulation of the expression of TGF‐β protein, which can then further promote tumorigenesis through unwanted paracrine effects on adjacent, TGF‐β‐responsive stromal tissue. Clearly, it will be important to determine whether the up‐regulation of TGF‐β that we see in the pancreatic acinar cells when TGF‐β responsiveness is abolished, is also observed in other cell types.
Implications for pancreatic cancer
Our observations of perturbed proliferation and differentiation in the transgenic pancreas may have important implications for our understanding of pancreatic carcinogenesis in humans. Pancreatic cancers in humans exhibit ductal morphology in most cases, but it is controversial whether the duct cell is the origin of pancreatic cancer. Tumors with ductal morphology in several rodent models of pancreatic carcinogenesis may have their origin in hyperproliferative and dysplastic acinar cells (Flaks, 1984; Scarpelli et al., 1984). Recent studies using transgenic mice with overexpression of mitogenic oncogenes such as TGF‐α, c‐myc and SV40 T antigen in acinar cells directly support the potential of acinar cells to form tumors with ductal characteristics (Ornitz et al., 1987; Jhappan et al., 1990; Sandgren et al., 1990, 1991). Furthermore, cells of mixed acinar–ductular morphology have been observed in human pancreatic carcinoma specimens (Parsa et al., 1985). These data are consistent with the hypothesis that pancreatic cancer could arise from dedifferentiated acinar cells (Logsdon, 1995). Thus loss of TGF‐β function and associated acinar dedifferentiation may represent a first step on the pathway to pancreatic neoplasia. Interestingly, recent genetic studies have shown that >50% of pancreatic cancers in humans have defects in the gene DPC‐4 (Hahn et al., 1996), which encodes an Mad‐related protein involved in TGF‐β signaling (Lagna et al., 1996; Zhang et al., 1996). Furthermore, expression of TβRI was attenuated in some pancreatic cancers with resistance to TGF‐β‐mediated growth inhibition (Baldwin et al., 1996). This is further evidence for the critical role which TGF‐β plays in pancreatic homeostasis.
Eosinophilic foci of acinar cells were observed in a significant number (21%) of DNR mice. In rats, eosinophilic foci indicate acinar cell hyperplasia and may represent early stages of the neoplastic process (Eustis et al., 1990). Hence, eosinophilic foci exclusively found in pancreas of DNR mice may represent early hyperplastic lesions with potential for malignant progression. The lack of such neoplastic progression in the pancreas of DNR transgenic mice, despite persistent hyperproliferation, focus formation and dedifferentiation in TGF‐β‐unresponsive acini, indicates that additional defects may be required for carcinogenesis to proceed. Alternatively, it may reflect the fact that transgene expression is lost when the cells dedifferentiate to a ductular morphology, and the growth inhibitory response is restored. However, mice bitransgenic for both the DNR and the hepatocyte growth factor (HGF) in the acinar cells do develop pancreatic cancer, which is not seen in mice expressing either transgene alone (J.L.Jakubczak and G.Merlino, unpublished data). These preliminary results indicate that loss of TGF‐β function in the exocrine pancreas may enhance susceptibility to tumorigenesis, and that the DNR transgenic mice may represent a useful model for studying the role of TGF‐β in pancreatic carcinogenesis in vivo.
Materials and methods
Construction of plasmids and transgenic mice
Human TβRII sequences between nucleotides −7 and +573, encoding extracellular and transmembrane domains followed by a stop codon at the 3′ end, were generated by PCR, ligated into the pCR vector (Invitrogen) to generate the construct pCRDNR and sequenced in their entirety. A 600 bp EcoRI fragment of pCRDNR containing DNR cDNA was blunt‐ended and subcloned into the NruI site of pMT‐LCR2999B4 (Palmiter et al., 1993) to generate pMT‐LCR‐DNR. Expression of the transgene is driven by a mouse MT1 promoter, and regulated by the presence of MT LCRs (Palmiter et al., 1993). A 19.1 kb SalI MT‐LCR‐DNR linear fragment (Figure 1A) was isolated and used for microinjections of zygote pronuclei. Transgenic mice were generated using inbred FVB/N zygotes as described previously (Jhappan et al., 1990). Of 19 mice born, two were positive for MT‐LCR‐DNR and were bred into lines AM3 and BF1. Both lines displayed autosomal inheritance of the transgene. Tail genomic DNA was prepared and tested either by Southern blotting (founder mice) or PCR analysis (subsequent generations) (Jhappan et al., 1990). Animals were cared for in accordance with NIH guidelines. When indicated, the activity of the MT1 promoter was induced by maintaining the animals on drinking water containing 25 mM ZnSO4.
RNA was isolated from mouse tissues by guanidium isothiocyanate extraction and column purification using the RNeasy kit (Qiagen) as described (Bonham and Danielpour, 1996). For Northern blot analysis, RNA was electrophoresed on 1% agarose gels and transferred to a filter. Filters were then hybridized with 32P‐labeled probes: human TβRII from pCRDNR, mouse TGF‐β1, mouse TGF‐β2 and mouse TGF‐β3.
Preparation of pancreatic acini and primary hepatocyte cultures
Pancreatic acini were isolated from adult non‐transgenic FVB/N and line AM3 mice as described (Logsdon and Williams, 1983). For proliferation studies, insulin (1 μM) and epidermal growth factor (1 nM) were added to the culture medium. Hepatocytes were isolated from 8‐week‐old male FVB/N or line AM3 mice by a two‐step collagenase perfusion of the liver followed by isodensity centrifugation in Percoll (viability >95%) as previously described (Böttinger et al., 1996).
Proliferation assays and fibronectin secretion assays
DNA synthesis in primary acinar preparations and primary hepatocytes was measured after incubation of cells with [3H]thymidine between 48 and 72 h after plating, followed by trichloroacetic acid (TCA) precipitation of cell extracts as described (Böttinger et al., 1996). [3H]Thymidine incorporation was measured in a liquid scintillation counter and normalized for DNA content per plate. Recombinant human TGF‐β1, porcine TGF‐β2 and recombinant human TGF‐β3 (all R&D Systems) and recombinant human activin (Genentech), were added between 24 and 72 h after cell plating as indicated.
Fibronectin production by hepatocytes was measured as described (Wrana et al., 1992). Fibronectin band intensities were quantified by densitometry and relative amounts expressed as x‐fold difference compared with fibronectin secreted by untreated cells.
Affinity labeling and immunoprecipitation
Porcine TGF‐β1 was iodinated as described (Frolik et al., 1984) and used for the labeling of cell surface binding proteins on freshly prepared primary hepatocytes and acinar preparations as described (Geiser et al., 1992). Equal amounts of protein were incubated overnight at 4°C with 1.5 μg/ml of rabbit polyclonal antibodies in 500 μl of lysis buffer or rabbit IgG, respectively. Antigen–antibody complexes were precipitated with protein A–Sepharose and eluted by boiling in 2× SDS sample loading buffer (Novex) containing 5% β‐mercaptoethanol. Samples were subjected to SDS–PAGE, and complexes of [125I]TGF‐β1 ligand and bound receptors demonstrated by autoradiography.
Histology and immunohistochemistry
Tissues were fixed in 10% formaldehyde or 70% ethanol, as indicated, embedded in paraffin, sectioned and stained with hematoxylin and eosin (H&E), toluidine blue or reticulin by standard methods. The extent of fibrosis was quantified using the Leitz Quantimat 600 Image analysis system, in which the percentage area of posititive stain was determined by gray scale analysis. Immunohistochemical staining was performed using an indirect immunoperoxidase detection protocol (Vectastain Elite kit, Vector Laboratories). DNR was detected in ethanol‐fixed tissue sections after antigen retrieval with microwave treatment by polyclonal anti‐human TβRII (residues 1–28) (Upstate Biotechnology). All the following staining was performed on 10% formaldehyde‐fixed tissue sections. TGF‐β1 was detected with the rabbit polyclonal antibody LC(1‐30) (Flanders et al., 1989). TGF‐β2 and β3 were detected with rabbit polyclonal antibodies TGF‐β2(50‐75‐2) (Flanders et al., 1990) and TGF‐β3(50‐60‐3) (Flanders et al., 1991). PCNA was detected by mouse monoclonal antibody PC10 (Dako). Macrophage glycoprotein Mac‐2 was detected by mononclonal anti‐Mac‐2 antibody (ATCC). AE1/AE3 antibody (BioGenex Laboratories) was used to detect cytokeratins 1–8, 10, 14, 15, 16 and 19. von Willebrand factor was detected with a polyclonal anti‐factor VIII antibody (Dako). For cytokeratin and von Willebrand factor staining, antigen retrieval was performed by treating sections for 12 min with 0.1% trypsin and 0.1% calcium chloride in Tris‐buffered saline (TBS).
In vivo proliferation and apoptosis
PCNA‐positive acinar cells per high power field (HPF) (0.07 mm2) were counted. The mean of 10 fields was calculated and the number of PCNA‐positive acinar cells/mm2 computed. In some animals, the ratio of PCNA‐positive and total number of acinar cells was calculated. The percentage of proliferating ductular cells in line AM3 pancreas represents the average ratio of the number of PCNA‐positive and total ductular cells counted in 10 HPFs (0.07 mm2) per pancreas. Apoptotic cells were identified on H&E‐stained slides by a combination of pyknotic nucleus and condensed eosinophilic cytoplasm. A collection of apoptotic debris phagocytosed by another cell was counted as a single apoptotic cell. Milder changes of apoptosis such as cytoplasmic condensation and clumped hyperchromatic chromatin were not counted. Apoptotic cells per HPF (0.2 mm2) were counted in 20 fields. The mean of apoptotic cells in 20 HPFs was used to compute the number of apoptotic cells/mm2. In addition, we assessed apoptosis in the liver using a TUNEL labeling method (Trevigen).
We thank Drs Miriam Anver (NCI‐FCRDC, Frederick, MD), Dale Bockman (Medical College of Georgia, GA) and Nora Sarvetnick (Scripps Institute, La Jolla, CA) for useful discussions. We thank Dr Craig Logsdon (University of Michigan, Ann Arbor, MI) for advice in establishing pancreatic acinar cell cultures. We thank Drs Murray Korc (University of California, Irvine, CA) and Anita Roberts (NCI, Bethesda, MD) for critical reading of the manuscript. We are grateful to Genentech, Inc. (South San Francisco, CA) for the gift of recombinant human activin, to Dr Richard Palmiter (University of Washington, Seattle, WA) for the pMT‐LCR plasmids and to Heather Winbrow for technical assistance.
↵† E.P.Böttinger and J.L.Jakubczak contributed equally to this work
- Copyright © 1997 European Molecular Biology Organization