Blocking Transforming Growth Factor β Signaling in Transgenic Epidermis Accelerates Chemical Carcinogenesis: A Mechanism Associated with Increased Angiogenesis¹

The TGF-β³ family has multiple functions in regulating cell proliferation and differentiation, tissue remodeling and repair, and immunomodulation (for reviews, see Refs. 1 and 2). In epithelial tissues, TGF-β is a potent growth inhibitor, playing an important role in the maintenance of tissue homeostasis (3, 4). Because malignant epithelial tumors develop mechanisms of escaping from TGF-β-induced growth inhibition (5, 6, 7), it is believed that TGF-β plays an important role in preventing cancer cell growth at earlier stages. In agreement with this, Glick et al. (8) reported that skin tumors devoid of TGF-β1 are associated with a high risk for malignant conversion. They also reported that grafts of v-ras^Ha-initiated TGF-β1-null keratinocytes progressed much more rapidly to SCCs than did wild-type TGF-β1 keratinocytes initiated with v-ras^Ha (9). Consistent with these observations, transgenic mice coexpressing both TGF-β1 and TGF-α in mammary glands exhibited a reduced rate of spontaneous or chemically induced carcinomas, as compared with transgenic littermates, which only expressed TGF-α (10). Tang et al. (11) recently reported that mice hemizygous for the TGF-β1 gene showed accelerated chemical carcinogenesis in the liver and lung, indicating that TGF-β1 is a tumor suppressor with haploid-insufficiency. Conversely, many reports documented high levels of TGF-β in malignant tumors and metastases both in clinical specimens as well as those induced experimentally in animal models (5, 6, 7, 12, 13, 14), thus suggesting a direct promoting role for TGF-β in late-stage carcinogenesis.

If the role of TGF-β is switched from a tumor suppressor to a tumor promoter, this must occur after cells have lost their responsiveness to TGF-β-induced growth inhibition. Because TGF-β ligand is present at both early and late stages of carcinogenesis, it is possible that this role reversal occurs because of altered functions of TGF-β receptors. Although the type I and II TGF-β receptors (TGF-βRI and TGF-βRII) form tetra-heteromers for TGF-β signaling, TGF-βRII is considered to be the primary receptor (15). Upon ligand binding, the intrinsic serine/threonine kinase of TGF-βRII phosphorylates TGF-βRI to activate its serine/threonine kinase activity (16, 17), and this activated receptor complex transduces TGF-β signals (18). Deletion of the serine/threonine kinase domain of TGF-βRII produces a dominant negative form (ΔβRII), which is able to block the growth inhibitory function of TGF-β in vitro (19, 20) and in transgenic animals (21, 22, 23), which suggests that TGF-βRII is required to mediate the growth inhibitory effect of TGF-β. Mutations in TGF-βRII were initially detected in TGF-β-resistant cancer cell lines (24, 25, 26). Similar mutations have then been reported in primary cancers of the colon (25, 27, 28, 29, 30), head and neck (31), ampulla (32), and pancreas (33). Recently, germ-line mutations of TGF-βRII have been identified in familial colorectal cancer (29). These data suggest that inactivation of TGF-βRII may be one mechanism by which epithelial tumor cells escape from TGF-β-induced growth inhibition and progress to malignancy. To test this hypothesis in the skin, we have generated transgenic mice expressing ΔβRII in the epidermis, using a truncated mouse loricrin expression vector (ML.ΔβRII). ML.ΔβRII mice exhibit marked hyperplasia/hyperkeratosis at birth, which suggests that the ΔβRII can block TGF-β-mediated growth inhibition in the epidermis (21). This is further supported by the fact that primary keratinocytes isolated from the epidermis of ML.ΔβRII transgenic mice were resistant to exogenous TGF-β1-induced growth inhibition (21). However, only a few ML.ΔβRII mice have developed spontaneous papillomas, which suggests that other genetic/epigenetic events are required for the development of overt lesions.

In the present study, we have subjected ML.ΔβRII mice to a chemical carcinogenesis protocol and observed an increased susceptibility compared with nontransgenic mice, with greatly accelerated benign papilloma formation, malignant conversion, and metastasis. We also provide evidence that the accelerated chemical carcinogenesis observed in mice expressing the ΔβRII in the epidermis is associated with increased angiogenesis.

Ten-week-old ML.ΔβRII heterozygous mice and their nontransgenic littermates (ICR) were divided into groups for treatments with DMBA/TPA, DMBA, and TPA, respectively. The back skin of each mouse was carefully shaved two days before the treatment. Mice in the TPA group received a topical application of 5 μg of TPA (dissolved in 50 μl of acetone) once a week for 60 weeks. Mice in the DMBA group received one treatment of DMBA with either 50 μg or 250 μg (dissolved in 100 μl acetone). Mice in the DMBA/TPA group were treated with a subcarcinogenic dose of DMBA (50 μg), followed by TPA promotion beginning one week after DMBA initiation, i.e., 5 μg of TPA was applied to the initiated area of the skin once a week for 20 weeks.

Biopsied tumors were fixed in 10% neutral-buffered formalin at 4°C overnight, embedded in paraffin, sectioned to 6-μm thickness, and stained with H&E. Immunofluorescence analysis for keratin 13 (K13) was performed on frozen sections as described previously (34). Vascularization was visualized by immunofluorescence analysis using a rat antimouse CD31 (PECAM-1) antibody (PharMingen), and a rabbit antimouse keratin 14 (K14) antibody. Frozen sections were washed with PBS twice for 10 min, fixed in cold acetone for 10 min, and rinsed with water. The CD31 antibody (5 μg/ml) and the K14 antibody (1:500), diluted in 12% BSA/PBS (w/v), were applied to the sections at room temperature for overnight. Sections were then washed as above and a secondary antibody, biotinylated goat antirat IgG (1:100, PharMingen) was applied for 1 h at room temperature. Tissue sections were washed again and a 1:400 dilution Streptavidin-Texas Red (Life Technologies, Inc.) and a 1:40 dilution of FITC-labeled, antirabbit IgG (Dakopatts) antibody was applied for 30 min at room temperature. Sections were then washed, air dried, coverslipped and photographed.

Microvessel counting was performed as described by Bolontrade et al. (35). Briefly, each stained section was screened with a ×10 eyepiece and ×10 objective magnification to identify the areas of highest vascularization. The number of vessels in each tumor were counted and averaged in five areas of highest vascular density with a ×10 eyepiece and ×20 objective magnification (defined as the area unit). The percentage stromal area covered by vessels was determined in the same fields screened for vessel counts by computer analysis. Five to seven tumors were counted in each group. The number of vessels/area unit, as well as the percentage of stromal area covered by vessels is expressed as mean ± SD.

Total RNA was isolated from chemically induced tumors with RNAzol B (Tel-Test, Inc.) as described previously (34). RPAs were performed using the RPA II kit (Ambion Inc., Austin, TX) and ³²P-riboprobes. ML.ΔβRII transgene expression was detected using a riboprobe specific for the ΔβRII (21). The riboprobe template of murine VEGF was prepared from the VEGF cDNA clone (provided by Dr. Georg Breier, Max-Plank Institute, Bad Nauheim, Germany), and linearized at the ClaI site within the vector, which is upstream of the start codon. The TSP-1 cDNA clone (provided by Dr. Peter J. Polverini, University of Michigan, Ann Arbor, MI) containing bp 491–822 of murine TSP-1 (in pGem7, Promega) was linearized at the 5′ ClaI site and used to generate the TSP-1 riboprobe. To normalize each RNA sample for differences in loading, a [³²P]GAPDH riboprobe was included in each analysis. The intensity of protected bands was determined by densitometeric scanning of X-ray films.

An analysis for mutations in the endogenous c-ras^Ha and p53 genes was performed under conditions described previously (34). To screen for mutations in the endogenous TGF-βRII gene, tumor RNA was reverse-transcribed. The resultant cDNA was amplified by PCR using the T7- or SP6-linked primers encompassing:

(a) the extracellular domain and the transmembrane domain (forward: 5′-T7-GGGGGCTCGGTCTATGACGA-3′; reverse: 5′-SP6-CCTTCCGGTGGAACGCCGTG-3′); and

(b) the cytosolic domain (forward: 5′-T7-TCATCCTGGAGGACGACCGC-3′; reverse: 5′-SP6-TTTGGTAGTGTTCAGAGAG-3′).

SP6 sequence is: 5′-TCATTTAGGTGACACTATA-3′. T7 sequence is: 5′-GATAATACGACTCACTATA-3′. The PCR products were then purified through a PCR purification kit (Qiagen), and analyzed using the Mismatch II kit (Ambion, Inc., Austin, TX), as well as automated sequencing using internal primers.

Tumors were fixed in 10% formalin-PBS (pH 7.4), embedded in paraffin, and sectioned on polylysine slides. The sections were deparaffinized and hydrated. The endogenous peroxidase was quenched by incubating sections in 0.6% H₂O₂ in methanol at room temperature for 30 min. Slides were rinsed four times with water and washed three times with Tris-buffered saline (5 min each), and the sections were treated with bovine testicular hyaluronidase [1 mg/ml in 0.1 m sodium acetate (pH 5.5)] at 37°C for 30 min. The sections were then washed as above and blocked with 5% normal goat serum at room temperature for 30 min. The TGF-β1 antibody (LC, a gift from Dr. Kathleen C. Flanders, NIH, Bethesda, MD), which recognizes intracellular TGF-β1 precursor, was applied to each section (5 μg/ml in Tris-buffered saline with 1% BSA) at room temperature overnight. The immune complex was detected by the avidin-biotin-peroxidase complex using Vectastain kits (Vector Lab, Burlingame, CA). Immune complexes were visualized with diaminobenzidine and counterstained with hematoxylin.

Although ML.ΔβRII mice exhibited a hyperplastic skin phenotype (21), they rarely developed spontaneous tumors. We, therefore, assessed their susceptibility to chemical carcinogenesis. DMBA was used as an initiator followed by repeated promotion with TPA. In total, 30 transgenic mice and 30 nontransgenic littermates in each line of ML.ΔβRII (lines B9223 and B9273; Ref. 21) were treated. The results from both transgenic lines were virtually identical. ML.ΔβRII mice exhibited accelerated papilloma formation, with the average appearance of papillomas by 6 weeks after DMBA initiation compared with 9 weeks in control siblings (Fig. 1,A). By 10 weeks after the DMBA initiation, all of the ML.ΔβRII mice developed papillomas, whereas only 70% of the nontransgenic mice developed papillomas at the termination of this study (50 weeks; Fig. 1,A). Additionally, the number of the tumors in ML.ΔβRII mice averaged nine papillomas/mouse at the end of the promotion stage (20 weeks, Fig. 1,B), 2-fold greater than that of control siblings (four papillomas/mouse). Furthermore, papillomas in control mice were prone to regression after the termination of TPA treatment (20 weeks, Fig. 1,B), and only 10% of control mice developed carcinomas by 40 weeks after DMBA initiation (Fig. 1,C). However, ML.ΔβRII papillomas grew autonomously and converted to SCCs as early as 25 weeks after DMBA initiation, and 90% of mice developed SCC by 40 weeks (Fig. 1,C). The SCCs in ML.ΔβRII mice further metastasized to inguinal lymph nodes as early as 30 weeks after DMBA initiation, and 50% of ML.ΔβRII mice had already developed metastases by 40 weeks. In contrast, control siblings did not develop any metastases by 50 weeks after DMBA initiation. Previous reports using this chemical protocol revealed that metastasis generally occurs at late stages and exhibits a poorly differentiated histotype. However, metastases in ML.ΔβRII mice occurred at earlier stages (Fig. 2,B) when the primary ML.ΔβRII tumors were still classified as carcinomas in situ or well-differentiated SCC (Fig. 2,A). Consistent with the differentiation stage of the primary lesions, both primary and metastatic lesions showed expression of keratin 13 (Fig. 2, C and D), an early marker for progression of skin tumors that is usually lost in late-stage SCC (36). Only a few undifferentiated spindle cell carcinomas developed in ML.ΔβRII mice by 30 to 40 weeks after DMBA initiation, which also metastasized to inguinal lymph nodes (data not shown).

These chemical carcinogenesis experiments suggest roles of ΔβRII in carcinogenesis at both the early and late stages. To determine at which stage ΔβRII influences skin carcinogenesis in vivo, DMBA and TPA were applied separately to ML.ΔβRII mice. If epidermal cells only contain an initiation event, such as c-ras^Ha mutations, typically induced by a subcarcinogenic dose of DMBA (37), they will not develop tumors unless a constitutive promoting stimulus, e.g., TPA-treatment, is provided (37, 38). ML.ΔβRII mice failed to develop tumors with a subcarcinogenic DMBA application, suggesting that ΔβRII did not serve as a constitutive tumor promoter. However, when the DMBA treatment was increased to 250 μg, 5 times higher than the initiating dose, 54% of ML.ΔβRII mice developed papillomas within 12 months, the earliest was 1 month after the treatment (Fig. 3A). In addition, 50% of these papillomas converted to SCCs or spindle cell carcinomas. The nontransgenic siblings did not develop tumors with the same treatment, indicating that this dose of DMBA is still below a full carcinogenic dose for normal mice of this strain. Therefore, inactivation of TGF-βRII seems to be sufficient, in combination with additional genetic insults elicited by the higher dose of DMBA, to serve as a full-strength tumor promoter.

Consistent with previous reports (37, 38), topical application of TPA alone did not induce tumor formation in nontransgenic mice, indicating that an initiation event is required for overt lesions to occur. However, when TPA was applied to ML.ΔβRII mice, they developed papillomas as early as 4 months of weekly TPA promotion (Fig. 3,B), and 50% of TPA-treated ML.ΔβRII mice developed papillomas within 12 months (Fig. 3 B). This result suggests that ΔβRII can act as an initiation event. Although these papillomas persisted, they did not convert to SCC by the end of this study (over 18 months).

Expression of ΔβRII at both RNA and protein levels in ML.ΔβRII transgenic epidermis has been documented previously (21). To determine whether the ML.ΔβRII transgene was causally involved in accelerating chemically induced tumor formation and progression, RPA was used to detect transgene expression in ML.ΔβRII tumors. As shown in Fig. 4, ML.ΔβRII transgene was strongly expressed in both papillomas and carcinomas as well as in metastatic lymph nodes.

Consistent with previous reports that greater than 90% of DMBA-initiated skin tumors contain c-ras^Ha mutations (37), most ML.ΔβRII tumors elicited by the two-stage carcinogenesis protocol exhibited c-ras^Ha mutations. In nine tumors that we have analyzed, seven had an A→T transversion at codon 61 (Gln→Leu), and one possessed a C→T transition at codon 11 (Ala→Val) of the c-ras^Ha gene (Table 1). In four DMBA-elicited tumors, one possessed a G→A transition at codon 12 (Gly→Glu), and two had an A→T transversion at codon 61 (Gln→Leu) of the c-ras^Ha gene (Table 1). In addition, metastatic lesions exhibited the same c-ras^Ha mutations as found in the primary tumors. In contrast, in seven papillomas analyzed from TPA-treated ML.ΔβRII mice, only one papilloma possessed a G→T transition at codon 12 (Gly→Glu); six tumors possessed the wild-type sequence of the entire c-ras^Ha gene (Table 1), further supporting the conclusion that the ΔβRII may serve as an alternative initiation event.

Previous chemical carcinogenesis experiments have associated p53 loss with malignant conversion in skin carcinogenesis (39, 40); therefore, ML.ΔβRII tumors were analyzed for mutations in the endogenous p53 gene. In total, eight papillomas, eight carcinomas, and six metastases were analyzed, and all of them were devoid of spontaneous mutations in the coding region of the p53 gene (data not shown). This result suggests that the ability of ΔβRII to accelerate malignant conversion is independent of p53 mutations.

Because mutations in the endogenous TGF-βRII have been found in various malignant tumor cells, we also analyzed ML.ΔβRII tumors for mutations in the endogenous TGF-βRII gene. All of the 12 ML.ΔβRII carcinomas and 3 metastases analyzed possessed wild-type TGF-βRII coding sequence (data not shown), which suggests that the ΔβRII can accelerate skin carcinogenesis via its dominant negative effect on wild-type TGF-βRII.

In general, histopathology of ML.ΔβRII tumors exhibited increased vascularization in comparison with nontransgenic tumors. To easily visualize the vascular density, tumors were subjected to immunofluorescence staining using an antibody against CD31, a marker of endothelial intercellular junctions as well as platelets and leukocyte subsets (41). Increased neovascularization was shown in ML.ΔβRII papillomas and carcinomas, in comparison with tumors in nontransgenic mice (Fig. 5). Microvascular density (number of vessels/unit area) in ML.ΔβRII papillomas increased to 15 ± 2.8, compared with 6.3 ± 1.9 in nontransgenic papillomas (P < 0.005). The percentage of the stromal area covered by vessels in ML.ΔβRII papillomas also increased to 53 ± 24%, compared with 29 ± 12% in nontransgenic papillomas (P < 0.05). Angiogenesis was also obvious in chemically induced SCCs, with ML.ΔβRII SCCs exhibiting a larger number of vessels and increased area of vascularization compared with SCCs in nontransgenic control. Microvascular density in ML.ΔβRII SCCs increased to 25 ± 5.3 versus 11 ± 2.5 in nontransgenic SCCs (P < 0.005); and the percentage of the stromal area covered by vessels was 64 ± 18% in ML.ΔβRII SCCs versus 36 ± 15% in nontransgenic SCCs (P < 0.05).

We then examined changes in VEGFs (a critical factor for angiogenesis and tumor cell growth; Ref. 42) and TSP-1 (an angiogenesis inhibitor; Ref. 43) by RPA. Although there were variations in VEGF and TSP-1 expression levels in nontransgenic tumors, ML.ΔβRII tumors showed more consistent expression levels of VEGF and TSP-1. ML.ΔβRII papillomas and SCCs exhibited a 2 to 3-fold increase in VEGF expression and a 2 to 4-fold decrease in TSP-1 expression, in comparison with nontransgenic papillomas and SCCs, respectively (these values were normalized with the intensity of the corresponding GAPDH signal, Fig. 6).

Previous chemical carcinogenesis studies revealed that carcinogens or tumor promoters induce TGF-β1 expression in skin tumors (44, 45). It is not clear whether TGF-β1 induction is a negative feedback system for cells against a tumor promotion event. In the present study, we examined endogenous TGF-β1 expression by immunohistochemistry using an antibody specific for intracellular TGF-β1. TGF-β1 was not detectable in normal epidermis (not shown) but was overexpressed in suprabasal cells of papillomas in nontransgenic mice (Fig. 7A). Interestingly, TGF-β1 staining in ML.ΔβRII papillomas was not only stronger in suprabasal cells but was also observed in proliferative cells (Fig. 7,B). Carcinomas in nontransgenic mice expressed a higher level of TGF-β1 than papillomas (Fig. 7,C versus 7,A), with TGF-β1 positive cells primarily located in the terminally differentiated layers (Fig. 7,C). In contrast, more cell layers in ML.ΔβRII carcinomas expressed TGF-β1, including proliferative cells (Fig. 7 D). The elevation of endogenous TGF-β1 in ML.ΔβRII tumors may represent a compensatory effect due to the expression of ΔβRII throughout the entire epidermis. Because ΔβRII is able to block TGF-β1-induced growth inhibition in keratinocytes (21), higher levels of TGF-β1 in ML.ΔβRII tumors may facilitate its paracrine effect on angiogenesis and thereby accelerate tumor growth and invasion.

The detection of mutations in TGF-βRII in human cancer suggests that inhibiting TGF-β signaling may be an important mechanism in the malignant progression of epithelial tumors. To test this in an in vivo model, ML.ΔβRII transgenic mice were exposed to a chemical carcinogenesis protocol. Consistent with the previous report on transgenic mice expressing ΔβRII in the basal cells of the epidermis [K5.ΔβRII, (45)], ML.ΔβRII mice exhibited a rapid progression of papillomas to malignant carcinomas. Although we used a lower initiation dose of DMBA and a reduced frequency of TPA promotion in ML.ΔβRII mice than used in the study of K5.ΔβRII mice (45), ML.ΔβRII mice exhibited a more rapid benign papilloma appearance and a more rapid metastasis relative to nontransgenic controls than K5.ΔβRII mice. This higher susceptibility of ML.ΔβRII mice to chemical carcinogenesis than K5.ΔβRII mice may reflect differences in the genetic backgrounds of these transgenic models (ICR for ML.ΔβRII mice versus FVB/N for K5.ΔβRII), although the FVB/N strain is generally more susceptible to skin chemical carcinogenesis (46). A more likely explanation for the differences in susceptibility of these two models to chemical carcinogenesis may lie in expression patterns of the ΔβRII transgene. Unlike K5.ΔβRII mice, which express ΔβRII in the basal compartment, our ML.ΔβRII mice express ΔβRII throughout the entire epidermis, including both basal (proliferative) and suprabasal (differentiated) keratinocytes (21). The difference in the site of synthesis of ΔβRII may also account for the fact that ML.ΔβRII mice show significant epidermal hyperplasia (21), whereas mice expressing ΔβRII exclusively in basal keratinocytes exhibited a normal epidermis (45). Thus, our present chemical carcinogenesis experiments demonstrate that a constitutive block in TGF-β signaling throughout the entire epidermis accelerates skin carcinogenesis at both early and late stages.

Mutations in TGF-βRII have been detected in malignant cell lines and cancers (24, 25, 26, 27, 28, 29, 31, 31, 32, 33), suggesting that inactivation of TGF-βRII is an important mechanism for malignant progression at late stages of carcinogenesis. Previous reports on transgenic mice expressing ΔβRII in the mammary gland and lung (47) or basal epidermal keratinocytes (45) have consistently demonstrated that loss of TGF-β responsiveness plays a late role in chemical carcinogenesis. Interestingly, we observed an accelerated benign papilloma formation in ML.ΔβRII mice at a very early stage (Fig. 1, A and B). More surprisingly, we observed that TPA application alone, without DMBA initiation, induced papilloma formation in ML.ΔβRII mice (Fig. 3,B). These results suggest that blocking TGF-βRII signaling can serve as an initiating event in skin carcinogenesis. Supporting this, most of the TPA-induced papillomas in ML.ΔβRII mice did not exhibit c-ras^Ha mutations (Table 1), a common initiating event in chemical carcinogenesis (37). Such alternative initiating events have also been described in transgenic mice expressing TGF-α in the epidermis, in which the lack of c-ras^Ha mutations has been documented in TPA-induced papillomas (48). Because TPA-induced papilloma formation in ML.ΔβRII mice is much slower than that elicited by the DMBA/TPA two-stage carcinogenesis protocol (Fig. 3 versus Fig. 1) as well as that of the TPA-treated ras^Ha transgenic mice (49), the initiating role of ΔβRII seems to be weaker than c-ras^Ha activation. Nevertheless, this novel observation suggests that blocking TGF-βRII signaling may serve as a predisposing event in skin carcinogenesis.

In addition to a role in initiation, if ΔβRII also serves as a tumor promoter, ML.ΔβRII mice would develop spontaneous papillomas. To date, only a few ML.ΔβRII mice have developed spontaneous papillomas, which argues against ΔβRII playing a role in promotion. Amendt et al. (45) reported that K5.ΔβRII mice developed carcinomas by repeated DMBA treatments. The strain of their transgenic mice is FVB/N, which develops an unusually high incidence of SCC when exposed to chemical carcinogenesis protocols (46), and the dosage of DMBA used in their study was 4-fold higher than that used in the present study. To clarify whether ΔβRII by itself is sufficient to serve as a constitutive tumor promoter, our present study used two different doses of DMBA. Under a subcarcinogenic, initiation dosage (50 μg), ML.ΔβRII mice failed to develop papillomas without further TPA promotion, which suggests that ΔβRII by itself is insufficient to serve as a constitutive tumor promoter. However, when ML.ΔβRII mice were treated with 250 μg of DMBA, five times higher than the initiating dose, they developed papillomas (Fig. 3) that further converted to carcinomas and metastases. In this case, the c-ras^Ha mutation elicited by DMBA (Table 1) was likely to provide the initiation event, and the cooperation of ΔβRII with mutated c-ras^Ha and/or other genetic insults provided a promoting event to achieve tumor formation and progression. Because tumor formation in response to DMBA alone was only observed in ML.ΔβRII mice and not in nontransgenic mice, inactivation of TGF-βRII seems to be sufficient—in combination with additional genetic insults elicited by the higher dose of DMBA—for a full carcinogenic effect.

With respect to the role of loss of TGF-βRII signaling in malignant progression, ML.ΔβRII tumors showed a correlation between DMBA-elicited c-ras^Ha mutations and malignant conversion. Both the DMBA/TPA two-stage protocol and the higher dose of DMBA treatment induced high rates of papilloma conversion to carcinomas with similar rates of c-ras^Ha mutations (Table 1). However, none of the TPA-elicited papillomas progressed to carcinomas, and the rate of c-ras^Ha mutation was also very low in this group (one of seven, Table 1). These data suggest that, although ΔβRII plays important roles in both the earlier and later stages of carcinogenesis, c-ras^Ha activation seems to be necessary to cooperate with the loss of TGF-β signaling for malignant conversion. Supporting this hypothesis, Glick et al. (9) reported that v-ras^Ha transduction of TGF-β1 null keratinocytes accelerated malignant conversion. However, because a subcarcinogenic dose of DMBA (despite the induction of c-ras^Ha mutations at this dosage) did not induce tumor formation in ML.ΔβRII mice, other genetic insults (induced by the higher dose of DMBA) or tumor promotion events (induced by TPA) are still required for tumorigenesis in ML.ΔβRII mice.

Metastasis is usually associated with late-stage, poorly differentiated SCC (37). However, a unique feature of ML.ΔβRII tumors is that metastatic lesions developed from well-differentiated SCCs, as identified by both histopathology and keratin staining (Fig. 2). This rapid progression to metastasis was correlated with increased vascularization in ML.ΔβRII tumors (Fig. 5), which suggests that angiogenesis may play a pivotal role in ML.ΔβRII tumor progression and invasion. Because TGF-β is believed to induce angiogenesis (2, 50) and has been reported to induce expression of VEGFs in cultured keratinocytes (51), epithelial tumor cells, and their stromal fibroblasts (52), we were somewhat surprised to find that ML.ΔβRII tumors exhibited increased angiogenesis (Fig. 5) that was correlated with increased VEGF expression and decreased TSP-1 expression (Fig. 6). This may have resulted from increased expression of endogenous TGF-β1 in ML.ΔβRII tumors (Fig. 7). It is possible that ΔβRII mainly blocks TGF-β-induced growth inhibition in the epidermis, whereas other TGF-β functions, such as TGF-β-induced VEGF expression, are not blocked by ΔβRII in this transgenic model. Although it has been debated whether TGF-βRI and TGF-βRII can signal independently (19, 20), recent studies suggest differential suppression of TGF-β functions may depend on the expression level of the ΔβRII (18). In agreement with this postulation, Sankar et al. (53) reported that the truncated TGF-βRII blocked only TGF-β1-induced growth inhibition without affecting angiogenesis, whereas the truncated TGF-βRI blocked TGF-β1-induced angiogenesis. Alternatively, if the intact type II receptor is still required for mediating TGF-β-induced angiogenesis, increased angiogenesis in ML.ΔβRII tumors may be a result of a paracrine effect of the increased TGF-β on stromal cells that do not express ΔβRII and are fully capable of TGF-β signaling. This potential paracrine effect could be further enhanced in ML.ΔβRII tumors inasmuch as TGF-β1 was expressed at higher levels in ML.ΔβRII tumors than nontransgenic tumors, particularly in proliferative cells (Fig. 7) that are closely associated with the tumor stroma.

Because TSP-1 has been reported to be transcriptionally activated by the tumor suppressor p53 (43) and VEGF expression can also be elevated by mutant p53 (54), changes in TSP-1 and VEGF expression in ML.ΔβRII tumors may occur via the inactivation of p53. However, we did not detect p53 mutations in ΔβRII tumors. Furthermore, we have shown that chemically induced tumors in either p53 knockout mice (p53^−/−) or transgenic mice expressing a dominant negative p53 mutant (p53^m) in the epidermis did not exhibit altered expression levels of either TSP-1 or VEGF (55). Therefore, increased angiogenesis resulting from ΔβRII expression seems to be independent of p53 status. Our present study suggests that tumors that possess both elevated TGF-β and loss of functional TGF-βRII may have a poor prognosis. The mutant TGF-βRII not only allows tumor cells to escape from TGF-β-induced growth arrest, but the subsequent elevation in TGF-β expression levels increases angiogenesis, which facilitates tumor growth and invasion.

We thank Dr. Dennis R. Roop for his support of this work. We also thank Dr. Peter Polverini for providing the TSP-1 cDNA clone, Dr. Georg Breier for providing the VEGF cDNA clone, Dr. Kathleen C. Flanders for providing the TGF-β1 antibody, and Dr. Dennis R. Roop and Kristin Liefer for critical comments on the article. Mattie Brooks provided excellent technical assistance.