Report

TGF-ß Signaling in Fibroblasts Modulates the Oncogenic Potential of Adjacent Epithelia

See allHide authors and affiliations

Science  06 Feb 2004:
Vol. 303, Issue 5659, pp. 848-851
DOI: 10.1126/science.1090922

Abstract

Stromal cells can have a significant impact on the carcinogenic process in adjacent epithelia. The role of transforming growth factor–β (TGF-β) signaling in such epithelial-mesenchymal interactions was determined by conditional inactivation of the TGF-β type II receptor gene in mouse fibroblasts (Tgfbr2fspKO). The loss of TGF-β responsiveness in fibroblasts resulted in intraepithelial neoplasia in prostate and invasive squamous cell carcinoma of the forestomach, both associated with an increased abundance of stromal cells. Activation of paracrine hepatocyte growth factor (HGF) signaling was identified as one possible mechanism for stimulation of epithelial proliferation. Thus, TGF-β signaling in fibroblasts modulates the growth and oncogenic potential of adjacent epithelia in selected tissues.

Epithelial-mesenchymal interactions play a critical role in development and cancer progression. In addition to the mesenchymal cell regulation of germ layers during organogenesis, tissue fibroblasts regulate the proliferation and differentiation of epithelial tissues (1, 2). Transformed stroma can induce malignancy in lung and mammary epithelia (3, 4), and conversely, normal fibroblasts have been reported to convert malignant epithelia in the prostate and skin to morphologically benign lesions (5, 6). Known mediators of epithelial-mesenchymal interaction include members of the TGF-β family (79). This family is responsible for context-dependent inhibition or stimulation of cell proliferation and neoplastic transformation (1012). TGF-β exerts its effect by binding the TGF-β type II receptor (TβRII) and subsequently recruiting the type I receptor (TβRI) for downstream cytoplasmic signaling through multiple parallel signaling pathways, including the Smad proteins (13).

To study the role of TGF-β signaling in stromal fibroblasts in vivo, we generated mice conditionally inactive for Tgfbr2, the gene encoding TβRII. Tgfbr2floxE2/floxE2 mice, having loxP sites at introns 1 and 2 of Tgfbr2 (14), were crossed with mice expressing Cre recombinase under the control of the FSP1 (fibroblast specific protein-1; S100A4) promoter [(15); fig S1]. FSP1 is selectively expressed in fibroblasts (16, 17) during development after embryonic day 8.5 (18). In mice expressing green fluorescent protein (GFP) under the control of the FSP1 promoter, FSP.GFP (16), FSP1-expressing fibroblasts were seen in the interstitial stroma of mature tissues throughout the mouse, including the prostate, forestomach, and skin (fig. S1). Selective Cre-mediated recombination within the prostate and forestomach stromal fibroblasts were observed in FSP. Cre-Rosa26r mice (15, 19) (fig. S1). We did not observe any phenotypic differences between the Tgfbr2floxE2/floxE2 and Tgfbr2fspKO (homozygous knockout) tissues examined, including the skin, esophagus, lung, kidney, and liver (fig. S2). However, the male Tgfbr2fspKO mice were sterile, and both males and females died at about 8 weeks of age. Thus, the bigenic line was propagated by backcrossing Tgfbr2wt/fspKO (heterozygous knockout) mice to C57BL/6 Tgfbr2floxE2/floxE2 mice, and the experiments described below were done with F6 and F7 generation mice in the C57BL/6 strain.

We determined whether FSP1.Cre-mediated recombination of Tgfbr2 produced a functional loss of TβRII in skin and prostate by 125I-labeled TGF-β–affinity cross-linking experiments (fig. S3). Primary cultures of skin keratinocytes and prostate epithelia exhibited TβRI- and TβRII-associated binding; however, the lack of TβRII binding in the Tgfbr2fspKO fibroblasts resulted in a subsequent loss of TβRI binding. Binding specificity was verified by putting 125I-labeled TGF-β1 ligand in competition with a fivefold excess of unlabeled TGF-β1. Dermal and prostate fibroblasts from Cre (Tgfbr2floxE2/floxE2) and Tgfbr2wt/fspKO mice showed equivalent levels of TGF-β1–bound TβRII. Of note, 125I-labeled TGF-β1 binding to TβRIII in the Tgfbr2fspKO fibroblasts was less than that in the Tgfbr2floxE2/floxE2 and Tgfbr2wt/fspKO mice, possibly the result of a loss of positive-feedback stimulation. The lack of TGF-β binding to TβRIII in cultured keratinocytes has been reported previously (20).

By 3 weeks of age, Tgfbr2fspKO male mice displayed an abundance of stromal fibroblasts lacking TβRII and nuclear phosphorylated Smad2 in the anterior, dorsolateral, and ventral prostate lobes. In contrast, almost all of the sparse stromal fibroblasts from wild-type and Tgfbr2floxE2/floxE2 animals demonstrated TβRII expression and phosphorylated Smad2 in the nucleus (Fig. 1, A and B; fig. S3). The Tgfbr2fspKO stromal phenotype was accompanied by epithelial hyperplasia and foci of hyperchromatic nuclei with atypia, suggestive of intraepithelial neoplasia (100% of 32 Tgfbr2fspKO mice examined). This developed by 5 to 7 weeks of age and occurred predominantly in the anterior and dorsolateral prostate lobes (Fig. 1, A and B). DNA amplification from primary cultures of Tgfbr2floxE2/floxE2 and Tgfbr2fspKO prostate fibroblasts and epithelia demonstrated specific Cre-mediated recombination of Tgfbr2 only in Tgfbr2fspKO fibroblasts (fig. S3). Immunostaining for TβRII and phosphorylated Smad2 showed no signal in the Tgfbr2fspKO prostate stroma (fig. S4; Fig. 1, C and D). A high percentage of the abundant stromal cells lacked nuclear phosphorylated Smad2 staining, which provided additional evidence for abrogation of TGF-β signaling due to recombination of Tgfbr2. Expression of the cell proliferation marker Ki-67 was increased sixfold in Tgfbr2fspKO prostatic fibroblast and epithelial compartments in comparison with Tgfbr2floxE2/floxE2 littermates (n = 12; Fig. 1, E and F). Thus, as a result of the loss of TGF-β signaling in fibroblasts, fibroblast and luminal epithelial proliferation was increased and was associated with the development of preneoplastic lesions in the prostate epithelia.

Fig. 1.

Loss of TβRII expression in fibroblasts results in prostate intraepithelial neoplasia (PIN). Dorsolateral prostate glands of Tgfbr2floxE2/floxE2 mice at 7 weeks of age (A, C, and E) were compared with those of matched Tgfbr2fspKO (B, D, and F) mice. (A and B) Hematoxylin- and eosin-stained sections of dorsolateral prostate exhibit histology in Tgfbr2fspKO mice analogous to low-grade PIN lesions. A single layer of epithelia (arrowhead) with a minor stromal component (asterisk) is found in the Tgfbr2floxE2/floxE2 prostate. In contrast, the Tgfbr2fspKO prostate epithelia is hyperplastic with nuclear atypia (arrowhead), and the accompanying abundant stroma (asterisk) appear activated based on the significant cytoplasmic to nuclear ratio. (C and D) Nuclear immunostaining of activated phosphorylated Smad2 was observed in Tgfbr2fspKO and Tgfbr2floxE2/floxE2 epithelia (arrowheads). Although Tgfbr2floxE2/floxE2 prostate fibroblasts stained for phosphorylated Smad2, most Tgfbr2fspKO fibroblasts lacked such staining (asterisk). (E and F) Elevated nuclear Ki-67 expression was observed in the epithelia (arrowheads) and fibroblasts (asterisks) of Tgfbr2fspKO mice compared with the infrequent staining in Tgfbr2floxE2/floxE2 mice. Scale bar represents 500 μm and 250 μm for the insets.

Seven-week-old Tgfbr2fspKO mice consistently exhibited invasive squamous cell carcinoma of the forestomach that resulted in a significant decrease in stomach volume (100% of 12 Tgfbr2fspKO mice examined; fig. S5). The forestomach of Tgfbr2fspKO mice demonstrated increased abundance of fibroblasts in the submucosa underlying areas of carcinoma (Fig. 2, A and B). Amplification of DNA from Tgfbr2fspKO forestomach stroma, obtained by laser-capture microdissection, showed specific Cre-mediated recombination of Tgfbr2 (fig. S2). No recombination was observed in the Tgfbr2fspKO epithelia or either compartment of the Tgfbr2floxE2/floxE2 forestomach. Immunostaining for TβRII and phosphorylated Smad2 corroborated these findings; a high percentage of the abundant Tgfbr2fspKO stromal cells lacked TβRII expression and nuclear phosphorylated Smad2 staining, as evidence for abrogation of TGF-β signaling due to recombination of Tgfbr2 (Fig. 2, C and D; fig. S4).

Fig. 2.

Squamous cell carcinoma develops in the forestomachs of Tgfbr2fspKO mice. Seven-week-old matched Tgfbr2floxE2/floxE2 (A, C, E) and Tgfbr2fspKO (B, D, F) were compared. (A and B) Hematoxylin and eosin staining of tissue sections show normal stratified squamous epithelia (arrowhead) with sparse fibroblast in the submucosa (asterisk) in Tgfbr2floxE2/floxE2 mice. Tgfbr2fspKO mice have invasive squamous cell carcinoma (arrowhead) with abundant stromal cells (asterisk) and hyperkeratinization (k). (C and D) Nuclear immunostaining of phosphorylated Smad2 was observed in Tgfbr2fspKO and Tgfbr2floxE2/floxE2 epithelia. The stromal fibroblasts of Tgfbr2floxE2/floxE2 forestomach showed activated, phosphorylated Smad2 expression (asterisk), however many Tgfbr2fspKO fibroblasts in comparison lacked similar nuclear staining (asterisk). (E and F) Nuclear Ki-67 expression was observed in the basal epithelia of Tgfbr2floxE2/floxE2 mice, but Tgfbr2fspKO mice had additional expression in the proliferating carcinoma-associated epithelia and fibroblasts. Scale bar represents 500 μm and 250 μm for the insets.

All Tgfbr2fspKO forestomach squamous cell carcinomas invaded into the fundus and stomach body by 7 weeks of age. There was a general loss of parietal cells in the glandular stomach in Tgfbr2fspKO mice, associated with decreased acidity of the stomach contents (pH 5.5-7.0, compared with pH 3.0 in Tgfbr2floxE2/floxE2 mice). Examination of Tgfbr2fspKO forestomach at 2 and 10 days postnatally revealed no abnormalities, which suggested that forestomach development does not require stromal TGF-β signaling and indicated that the carcinoma evolved rapidly in the mice between 2 and 7 weeks of age. The minimal monocytic infiltration in the stomach submucosa observed in Tgfbr2fspKO mice might be due to loss in the integrity of the epithelial lining (21). The stomach tumors, coupled with disruption of the stomach lining, were most likely the cause of death in the Tgfbr2fspKO mice by 8 weeks of age. As deduced from Ki-67 staining, the proliferation in the forestomach was limited to the basal cells in Tgfbr2floxE2/floxE2 mice. However, areas associated with spontaneous squamous cell carcinoma in 7-week-old Tgfbr2fspKO mice exhibited a fourfold increase in the proliferation of fibroblasts and squamous carcinoma cells in the forestomach that invade into the stomach body (n = 8, Fig. 2, E and F). Together, these data indicate that TGF-β signaling in fibroblasts not only limits fibroblast proliferation in vivo, but can also inhibit the proliferation of adjacent epithelial cells in a tissue-specific manner.

To investigate possible mechanisms underlying the development of neoplasia in the Tgfbr2fspKO mice, we performed Western blot analyses of prostate and forestomach tissue extracts. The expression level of TGF-β1 was comparable in the Tgfbr2floxE2/floxE2 and Tgfbr2fspKO mice (Fig. 3A, table S1). By contrast, the expression level of the transcription factor often elevated in transformed cells, c-Myc (22), was higher in Tgfbr2fspKO forestomach and prostate tissues than in Tgfbr2floxE2/floxE2 mice (Fig. 3A, table S1). Because the physiology and cellular composition of mouse forestomach is comparable to human esophagus, we evaluated the expression of prognostic markers and paracrine factors implicated in esophageal carcinoma. The Tgfbr2fspKO forestomach samples exhibited diminished expression of the cdk inhibitors, p21 and p27, compared with Tgfbr2floxE2/floxE2 controls. Similar comparisons in the prostate showed decreased expression of p27 in the Tgfbr2fspKO mice, but similar p21 expression (Fig. 3A, table S1). We further studied the signaling by two mitogenic growth factors known to be regulated by TGF-β, hepatocyte growth factor (HGF), and epidermal growth factor (EGF/TGF-α). Western blots for activated EGF receptor-1 did not exhibit any consistent pattern in the Tgfbr2fspKO and Tgfbr2floxE2/floxE2 forestomach. However, Western blotting for the phosphorylated c-Met (activated HGF receptor), showed a more than twofold increased abundance in Tgfbr2fspKO forestomachs examined compared with Tgfbr2floxE2/floxE2 littermates (Fig. 3A, table S1). Total levels of c-Met were not significantly different in Tgfbr2floxE2/floxE2 and Tgfbr2fspKO forestomach tissues. Western and enzyme-linked immunosorbent assay (ELISA) analysis was used to further examine HGF expression. The inactive pro-HGF expression in the forestomach tissue extracts in Tgfbr2fspKO mice were almost 2 times that in Tgfbr2floxE2/floxE2 mice, with no significant differences in the prostate tissues (Fig. 3B, table S1). However, the medium of conditioned primary fibroblast cultures demonstrated that Tgfbr2fspKO fibroblasts from both forestomach and prostate secreted at least 3 times as much active-HGF as Tgfbr2floxE2/floxE2 fibroblasts (Fig. 3C), which indicated a paracrine mechanism for activation of c-Met.

Fig. 3.

Paracrine signaling contributes to epithelial proliferation and transformation. Protein extracts from (A) forestomach and prostate (anterior, dorsolateral lobes) whole tissues of matched 6-weekold Tgfbr2floxE2/floxE2 (+/+) and Tgfbr2fspKO (ko) mice were analyzed for the expression of the indicated proteins by Western blotting. Phospho-specific antibodies for EGF receptor and c-Met were used to evaluate the activated forms of these receptor proteins. The Western blots were quantified from extracts of at least three independent mice and normalized to α-tubulin, loading control (table S1). (B) HGF (90 kD, pro-HGF) expression from tissue lysates was blotted with actin as the loading control (table S1; n = 4). (C) Conditioned media from primary cultures of forestomach and prostate fibroblasts were analyzed for the expression of active HGF by ELISA assay (n = 7; error bars, SD; P value determined from unpaired two-tailed Student's t test).

To localize phosphorylated c-Met and c-Myc overexpression, we immunostained adjacent sections of prostate and forestomach (fig. S6). Proliferating basal cells of the prostate and forestomach from Tgfbr2floxE2/floxE2 mice exhibited c-Myc staining independent of phosphorylated c-Met expression. However, PIN-associated prostate and tumor-associated forestomach tissues from Tgfbr2fspKO mice displayed parallel epithelial overexpression of c-Myc and phosphorylated c-Met (fig. S6). In light of evidence suggesting that TGF-β inhibits HGF expression (23), the loss of TGF-β signaling in stromal fibroblasts may in fact result in HGF-mediated cell-cycle regulation by suppressing cyclin-dependent kinase inhibitors, p27 and p21, as well as induction of c-Myc (24, 25).

In conclusion, we provide evidence that TGF-β signaling in fibroblasts differentially affects epithelial growth and oncogenesis. Despite the striking phenotypic changes in the prostate and forestomach, many other tissues appeared histologically normal (fig. S2). Because FSP1.Cre-mediated TβRII recombination in fibroblasts occurs after embryonic day 8.5 (18), when commitment to organogenesis is complete in many tissues, our mouse model implicates a role for stromal TGF-β signaling in tissue maturation. The conservation of structure and function within the TGF-β family of proteins, including bone morphogenetic proteins and activin, which mediate their actions through receptors other than TβRII (13), likely resulted in functional compensation for TGF-β loss in most tissues and may have masked other roles of stromal signaling.

Previous studies have indicated that stromal cells influence the carcinogenesis process in adjacent epithelia (27), but the specific paracrine factors and signaling pathways involved have not been identified. The Tgfbr2fspKO mouse model illustrates that a signaling pathway known to suppress cell-cycle progression when activated in epithelial cells can also have an indirect inhibitory effect on epithelial proliferation when activated in adjacent stromal fibroblasts in vivo. Loss of this inhibitory effect can result in increased epithelial proliferation and may even progress to invasive carcinoma in some tissues

Supporting Online Material

www.sciencemag.org/cgi/content/full/303/5659/848/DC1

Materials and Methods

Figs. S1 to S6

Table S1

References and Notes

References and Notes

View Abstract

Navigate This Article