|Home | About | Journals | Submit | Contact Us | Français|
In the present study, we demonstrated that human skin cancers frequently overexpress TGF-β1 but exhibit decreased expression of the TGF-β type II receptor (TGF-βRII). To understand how this combination affects cancer prognosis, we generated a transgenic mouse model that allowed inducible expression of TGF-β1 in keratinocytes expressing a dominant negative TGF-βRII (ΔβRII) in the epidermis. Without ΔβRII expression, TGF-β1 transgene induction in late-stage, chemically induced papillomas failed to inhibit tumor growth but increased metastasis and epithelial-to-mesenchymal transition (EMT), i.e., formation of spindle cell carcinomas. Interestingly, ΔβRII expression abrogated TGF-β1–mediated EMT and was accompanied by restoration of membrane-associated E-cadherin/catenin complex in TGF-β1/ΔβRII compound tumors. Furthermore, expression of molecules thought to mediate TGF-β1–induced EMT was attenuated in TGF-β1/ΔβRII–transgenic tumors. However, TGF-β1/ΔβRII–transgenic tumors progressed to metastasis without losing expression of the membrane-associated E-cadherin/catenin complex and at a rate higher than those observed in nontransgenic, TGF-β1–transgenic, or ΔβRII-transgenic mice. Abrogation of Smad activation by ΔβRII correlated with the blockade of EMT. However, ΔβRII did not alter TGF-β1–mediated expression of RhoA/Rac and MAPK, which contributed to increased metastasis. Our study provides evidence that TGF-β1 induces EMT and invasion via distinct mechanisms. TGF-β1–mediated EMT requires functional TGF-βRII, whereas TGF-β1–mediated tumor invasion cooperates with reduced TGF-βRII signaling in tumor epithelia.
TGF-β1 is a potent growth inhibitor for epithelial cells, and this function contributes greatly to its role in tumor suppression (1). Paradoxically, TGF-β1 is overexpressed in many malignant human tumors including squamous cell carcinomas (SCCs) (2–4) and in various cancers in experimental animals, including skin tumors (for review, see ref. 5). Studies have shown that TGF-β1 overexpression at early stages of carcinogenesis provides tumor-suppressive effects primarily via growth inhibition (1), whereas TGF-β1 overexpression at late stages promotes tumor progression, metastasis, and epithelial-to-mesenchymal transition (EMT), potentially via loss of adhesion molecules, angiogenesis, proteinase activation, and immune suppression (1, 6).
TGF-β1 exerts its effects primarily via a receptor complex comprising a type I and a type II receptor (TGF-βRI and TGF-βRII). To date, only 1 TGF-βRII has been identified that is essential for TGF-β binding and for assembly of the TGF-βRI/TGF-βRII complex (7). When TGF-β binds to a TGF-βRI/TGF-βRII complex, TGF-βRI phosphorylates Smad2 and Smad3. Phosphorylated Smad2 (pSmad2) and pSmad3 form heteromeric complexes with Smad4 and translocate into the nucleus to regulate TGF-β–responsive genes (7). Several Smad-independent TGF-β signaling pathways have also been identified, including the RhoA and MAPK pathways (7). Whereas TGF-β1 is overexpressed in cancer cells, TGF-βRII is often lost in cancer cells (1). It is commonly accepted that the tumor-suppressive effect of TGF-β1 requires functional TGF-βRII; therefore, inactivation of TGF-βRII is an important mechanism by which tumor cells escape TGF-β1–mediated growth inhibition and progress to malignancy. However, with respect to the role of TGF-βRII in TGF-β1–mediated tumor promotion, the data are controversial. In vitro studies have shown that TGF-βRII is required for TGF-β1–mediated tumor invasion (8, 9). Decreased breast cancer metastasis was seen in a model expressing a dominant negative TGF-βRII (ΔβRII) transgene in human breast-derived cell lines (10). In contrast, increased skin cancer metastasis and prostate cancer metastasis were observed in the respective ΔβRII-transgenic models (11, 12). Clinical studies also reveal controversial results regarding patient outcome when the prognosis is correlated with the loss of TGF-βRII in cancer cells. For instance, loss of TGF-βRII expression correlates with poor prognosis in esophageal cancer (13) and renal carcinoma (14) but correlates with a better survival rate in colon cancer (15) and gastric cancer (16). These observations highlight the complex nature of the functions of TGF-βRII in carcinogenesis, which could be stage and/or tissue specific. Although TGF-β1 overexpression and loss of TGF-βRII have been observed in human HaCaT keratinocytes with oncogenic potential (17, 18), whether these changes occur in primary SCCs in the skin has not been reported.
The mouse skin chemical carcinogenesis model is a useful tool for studying different stages of carcinogenesis. This model mimics the multistage nature of cancer development in humans, i.e., initiation, promotion, and malignant conversion (19). Initiation is induced by topical application of a subcarcinogenic dose of a particular carcinogen, e.g., dimethylbenzanthracene (DMBA). Following initiation, a tumor promoter, e.g., 12-O-tetradecanoylphorbol-13-acetate (TPA), is applied to elicit benign papillomas. After promotion, papillomas can persist, regress, or convert to malignancy (19). In this experimental model, expression of endogenous TGF-β1 is induced by TPA at a very early stage and persists in malignant carcinomas (5), and loss of endogenous TGF-βRII protein occurs at the stage of carcinoma formation (6). When this 2-stage carcinogenesis protocol was applied to transgenic mice overexpressing TGF-β1 in the epidermis, TGF-β1 transgene expression inhibited benign tumor formation yet enhanced progression to carcinomas (6, 20). Since TGF-β1 transgene expression began at a relatively early stage in both studies, it remains to be determined whether the tumor promotion effect is the result of long-term TGF-β1 overexpression, which creates a selective advantage for TGF-β1–resistant tumors, or whether late-stage tumors by their own nature can escape TGF-β1–mediated growth inhibition. In either case, it is not clear whether functional TGF-βRII is required for TGF-β1–mediated tumor promotion. In contrast to in vitro studies showing that TGF-β receptors are required for TGF-β1–mediated tumor invasion (8, 9, 21, 22), our previous results have shown that TGF-β1–transgenic tumors exhibited earlier loss of TGF-βRII and signaling Smads (6). These results suggest that downregulation of specific TGF-β signaling components in tumor epithelia selectively abolishes TGF-β1–mediated growth inhibition but may not affect all of the tumor-promoting activities of TGF-β1.
In the present study, we found that human skin cancers often overexpress TGF-β1 but exhibit decreased expression of TGF-βRII. To further investigate how this combination affects cancer progression, we bred ΔβRII mice with gene-switch–TGF-β1 mice in which TGF-β1 transgene expression can be induced in ΔβRII tumor epithelia at specific stages of skin carcinogenesis (23). We provide in vivo evidence that loss of functional TGF-βRII in tumor epithelia selectively blocks the molecular and pathological alterations required for TGF-β1–mediated EMT but cooperates with TGF-β1 for tumor invasion.
To determine whether expression of TGF-β signaling components is altered in human skin cancer, we performed immunostaining for TGF-β1, TGF-βRII, and nuclear pSmad2, a marker for activation of TGF-β signaling (7). Among the tissue samples examined were 8 normal skin, 14 actinic keratoses (AKs), 9 carcinomas in situ (CISs), and 34 SCCs. In normal skin, TGF-β1 exhibited weak staining in the epidermis and dermis (Figure (Figure1).1). TGF-βRII exhibited staining at the highest intensity in normal epidermis compared with the other tissue samples examined (Figure (Figure1).1). As shown in Table Table1,1, in the progression from AK to CIS to SCC, there was a clear overall trend toward increased TGF-β1 and decreased TGF-βRII expression. The combination of both increased TGF-β1 and reduced TGF-βRII was detected at a rate of 7% in AK, 22% in CIS, and 38% in SCCs, with expression in normal epidermis serving as a baseline (Table (Table1).1). The increase in TGF-β1 staining appeared patchy in AKs and gradually became uniform when progressing to CISs and SCCs (Figure (Figure1).1). The decrease in TGF-βRII staining appeared uniform or patchy in AKs and CISs (Figure (Figure1),1), but the reduction was generally uniform in SCCs (Figure (Figure1).1). Nuclear staining of pSmad2 was not detected in normal skin (Figure (Figure1),1), which suggests that the baseline level of TGF-β signaling via the Smad pathway is very low. Nuclear staining for pSmad2 was detectable in 3 AK samples and 4 CIS samples, all of which exhibited both increased TGF-β1 staining intensity and the presence of TGF-βRII (Figure (Figure1).1). Staining of pSmad2 was not detected in 32 of 34 SCC samples (94%; Table Table1)1) regardless of the levels of TGF-β1 expression (Figure (Figure11).
Gene-switch–TGF-β1 mice (designated hereafter as TGF-β1–transgenic mice), in which epidermal TGF-β1 expression can be specifically regulated, and mice expressing ΔβRII in the epidermis (ΔβRII-transgenic mice) were generated in the ICR strain as described previously (23, 24). As we have previously reported, ΔβRII-transgenic neonates exhibit hyperproliferative epidermis due to blocked TGF-β1–induced growth inhibition but normalizes in adulthood (24), whereas gene-switch–TGF-β1 skin is normal prior to TGF-β1 transgene induction by RU486 but undergoes epidermal growth inhibition upon TGF-β1 induction (23). These mice were crossbred to generate TGF-β1/ΔβRII–transgenic mice, in which TGF-β1 transgene expression can be induced in tumor epithelia with constitutive ΔβRII expression. As shown in Table Table2,2, littermates were divided into different groups based on genotypes and exposed to the 2-stage skin chemical carcinogenesis protocol, which is outlined in Figure Figure2A.2A. TGF-β1 transgene expression was induced by topical application of a progesterone antagonist, RU486 (20 μg/mouse, 3 times/week) to bigenic mouse skin, beginning at 20 weeks after DMBA initiation (Figure (Figure2A).2A). As previously documented (25), the dose of RU486 used in this study is at least 1,000-fold lower than that required to exert any antagonistic effect on endogenous steroid receptors and does not affect the tumor kinetics or tumor types when this skin chemical carcinogenesis protocol is used on male or female mice. Nevertheless, we included RU486-treated nontransgenic and monogenic mice as controls (Table (Table2).2). Once lesions developed that appeared to be SCCs (ulcerated lesions), tumor-bearing mice were euthanized. All tumors on each euthanized mouse were excised and preserved for analyses. After all visible skin tumors were excised, necropsy was performed on each mouse to identify potential metastatic lesions. Tumor types (primary and metastatic lesions) were confirmed by histological analysis. The experiment was terminated by 40 weeks because the majority of transgenic mice had been euthanized due to development of aggressive skin tumors.
Among control mice with different genotypes (Table (Table2),2), we did not observe a difference in tumor kinetics. Therefore, tumor kinetics data among these mice were pooled. For the same reason, tumor kinetics data among different genotypes representing ΔβRII transgenics (Table (Table2)2) were also pooled. Prior to RU486 application (i.e., prior to induction of TGF-β1), tumor kinetics in TGF-β1–transgenic mice was essentially the same as that in control mice (Figure (Figure2B)2B) and was also similar to that in nontransgenic mice without RU486 treatment as we reported previously (6, 11). In contrast, TGF-β1/ΔβRII–transgenic mice exhibited earlier tumor formation and a 2- to 3-fold increase in tumor number compared with control mice (P < 0.01; Figure Figure2B).2B). The tumor kinetics of TGF-β1/ΔβRII–transgenic mice was essentially the same as that of ΔβRII-transgenic mice (Figure (Figure2B).2B). After withdrawal of TPA promotion, the tumor kinetics in each group did not significantly change, despite the fact that regression appeared to be more noticed in control and ΔβRII-transgenic mice than in TGF-β1–transgenic or TGF-β1/ΔβRII–transgenic mice (P > 0.05). However, malignant conversion rates varied significantly among these groups. By 25 weeks after DMBA initiation, i.e., 5 weeks with TGF-β1 transgene induction, 73% of the TGF-β1/ΔβRII–transgenic mice developed SCCs (Table (Table3).3). Similar to the findings in our previous reports (6, 11), at 25 weeks, none of the control mice had developed SCCs, and 20% of the TGF-β1–transgenic mice and 11% of the ΔβRII-transgenic mice had developed SCCs (Table (Table3).3). At other time points throughout the entire course of the carcinogenesis experiment, TGF-β1/ΔβRII–transgenic mice had the highest malignant conversion rate, followed by TGF-β1–transgenic mice, then ΔβRII-transgenic mice (Table (Table3).3). In contrast, papillomas in control mice progressed to SCCs very slowly, which is typically observed in experiments using this carcinogenesis protocol (6, 11), having the lowest rate at all time points compared with any transgenic group in this study (Table (Table3).3). Almost all of the TGF-β1/ΔβRII–transgenic tumors were classified as SCCs, ranging from well to poorly differentiated (Figure (Figure2C).2C). Approximately 30% of the TGF-β1–transgenic mice developed spindle cell carcinoma (SPCC; Figure Figure2D),2D), a typical EMT tumor type. SPCC cells generally constituted the entire lesion (Figure (Figure2D),2D), but in some cases, they were adjacent to SCC cells (data not shown). SPCC formation rarely occurred in ΔβRII-transgenic, TGF-β1/ΔβRII–transgenic, or control mice. Furthermore, TGF-β1– or ΔβRII-transgenic tumors metastasized to lymph nodes beginning 30 weeks after DMBA initiation, and by 40 weeks, both groups of mice developed metastatic lesions at a similar rate (~25%; Table Table3).3). TGF-β1/ΔβRII–transgenic mice also began to develop metastatic lesions at around 30 weeks. By 40 weeks, 60% of the TGF-β1/ΔβRII–transgenic mice developed metastatic lesions in lymph nodes and/or lungs (Figure (Figure2,2, E and F). All of the metastatic lesions in TGF-β1/ΔβRII–transgenic mice were SCCs but not SPCCs. In contrast, only 1 control mouse developed metastasis by 40 weeks after DMBA initiation (Table (Table33).
We have previously shown that both TGF-β1– and ΔβRII-transgenic mice exhibited increased malignant conversion and metastasis (6, 11); the latter correlated with increased endogenous TGF-β1 expression (11). To determine whether the kinetics of malignant conversion and metastasis observed in this study correlated with TGF-β1 levels, we examined TGF-β1 expression levels in SCCs 25 weeks after DMBA initiation. The porcine TGF-β1–transgenic transcripts were detected only in TGF-β1– or TGF-β1/ΔβRII–transgenic SCCs (Figure (Figure3A).3A). Analysis of total levels of TGF-β1 protein in these tumor samples revealed that chemically induced SCCs in control mice exhibited an approximately 3-fold increase in endogenous TGF-β1 protein compared with normal skin (169.6 ± 17.7 pg/mg protein vs. 53.6 ± 13.3 pg/mg protein; n = 5; P < 0.01; Figure Figure3B).3B). Similar to what we have previously observed, ΔβRII-transgenic SCCs exhibited further elevation in endogenous TGF-β1 protein level (233.3 ± 42.9 pg/mg protein; n = 5; P < 0.05 compared with control SCCs). TGF-β1 protein levels were the highest in TGF-β1–transgenic SCCs (374.7 ± 83.5 pg/mg protein; n = 5; P < 0.01 compared with ΔβRII-transgenic tumors), and TGF-β1/ΔβRII–transgenic tumors exhibited TGF-β1 protein levels similar to those in TGF-β1–transgenic tumors (341.5 ± 109.4 pg/mg protein; n = 5). Notably, the variations in TGF-β1 protein levels among these groups were very similar to those in human head and neck SCCs, as we have previously reported (4).
The tumor kinetics of the TGF-β1–transgenic tumors suggests that after a full course of TPA promotion, papillomas developed the ability to escape TGF-β1–mediated tumor suppression. To determine whether tumors at this stage have lost TGF-β1–mediated growth inhibition or whether tumor cells still respond to TGF-β1–mediated growth inhibition but only clonal TGF-β1–resistant cells further progress to malignancy, we performed in vivo BrdU labeling on expressing and control tumors at 21 weeks (i.e., after the first 3 RU486 treatments). As expected, TGF-β1–transgenic epidermis adjacent to tumors (Figure (Figure4B)4B) exhibited a lesser degree of hyperplasia compared with control epidermis adjacent to tumors (Figure (Figure4A).4A). The BrdU labeling index was 50 ± 3 cells/mm epidermis in TGF-β1–transgenic epidermis adjacent to tumors (n = 5; Figure Figure4B),4B), a nearly 3-fold reduction compared with that in control epidermis adjacent to tumors (143 ± 5 cells/mm epidermis; n = 5; P < 0.01; Figure Figure4A).4A). The BrdU-labeling index in TGF-β1–transgenic papillomas (176 ± 18.2 nuclei/mm2 tumor epithelia; n = 5; Figure Figure4D)4D) was similar to that of control papillomas (157 ± 16.1 nuclei/mm2 tumor epithelia; n = 5; P > 0.05; Figure Figure4C).4C). These results suggest that chemically induced papillomas at this stage no longer respond to TGF-β1–mediated growth inhibition.
To determine whether the resistance to TGF-β1–mediated growth inhibition in late-stage TGF-β1–transgenic papillomas was due to the loss of TGF-β signaling at this stage, we performed immunohistochemical staining for pSmad2. In control papillomas, approximately 30% of the papilloma cells exhibited nuclear pSmad2 staining (Figure (Figure5).5). This is likely due to elevated endogenous TGF-β signaling during skin carcinogenesis (11), as nuclear staining of pSmad2 was almost completely absent in ΔβRII-transgenic papillomas at the same stage (Figure (Figure5).5). However, after 3 RU486 treatments, TGF-β1–transgenic papillomas exhibited a further increase in pSmad2 staining, i.e., approximately 90% of the cells exhibited nuclear pSmad2 staining (Figure (Figure5).5). This suggests that the ability of these tumor cells to escape from TGF-β1–mediated growth arrest is not due to abrogation of TGF-β signaling. Nuclear staining of pSmad2 was absent in TGF-β1/ΔβRII–transgenic papillomas (Figure (Figure5),5), which indicates that ΔβRII expression blocked TGF-β1–mediated Smad activation. Staining of pSmad2 in SCCs of TGF-β1–transgenic mice after 5 weeks of TGF-β1 transgene expression was negative in the nucleus and very weak in the cytoplasm (Figure (Figure5).5). The SCCs from control, ΔβRII-, and TFGβ1/ΔβRII–transgenic mice failed to exhibit any staining for pSmad2 (Figure (Figure5),5), which is consistent with our previous report that Smad proteins are lost in SCCs during skin chemical carcinogenesis (26).
Previously, we have observed that TGF-β1 transgene induction resulted in earlier loss of membrane-associated E-cadherin/catenin complexes (6). To determine whether loss of E-cadherin/catenin complexes is associated with EMT and/or metastasis, we examined expression of E-cadherin, β-catenin, and γ-catenin in ΔβRII-, TGF-β1–, and TGF-β1/ΔβRII–transgenic tumors. Similar to nonexpressing tumors, ΔβRII-transgenic SCCs at 25 weeks after DMBA initiation retained membrane-associated E-cadherin and β- and γ-catenins (Figure (Figure6).6). In contrast, TGF-β1–transgenic SCCs at the same stage had lost membrane-associated E-cadherin and β- and γ-catenins, which were detected in the cytoplasm (Figure (Figure6).6). Western blot analysis revealed no significant differences in the levels of these proteins between control and TGF-β1–transgenic tumors (data not shown), which suggests that loss of membrane-associated forms of these molecules was due to redistribution to the cytoplasm. Interestingly, localization of E-cadherin and β- and γ-catenins in TGF-β1/ΔβRII–transgenic tumors was similar to that in ΔβRII-transgenic tumors (Figure (Figure6),6), which indicates that expression of ΔβRII was able to abrogate TGF-β1–mediated redistribution of the E-cadherin/catenin complex. Membrane-associated E-cadherin/catenins were eventually lost in some of the late-stage ΔβRII-transgenic SCCs or TGF-β1/ΔβRII–transgenic SCCs at approximately the same stage as when these changes occurred in control SCCs (data not shown). Similar to cells from primary tumors of each genotype, metastatic tumor cells that developed from TGF-β1–transgenic tumors exhibited cytoplasmic staining of the above-mentioned adhesion molecules (data not shown). In contrast, metastatic cells that originated from ΔβRII-transgenic SCCs (data not shown) or TGF-β1/ΔβRII–transgenic SCCs (Figure (Figure6)6) still exhibited membrane-associated cell adhesion molecules, which suggests that metastasis in these tumors occurred without or prior to the loss of membrane-associated E-cadherin/catenins.
Since TGF-β1/ΔβRII–transgenic tumors exhibited the highest rate of metastasis, we suspected that ΔβRII expression may not block the effects of TGF-β1 on pathological alterations related to invasion. Therefore, we examined angiogenesis. Immunofluorescence staining using a CD31 antibody on SCCs 25 weeks after DMBA initiation revealed that the percentage of stromal area covered by vessels was 36.5% ± 4.9% in control SCCs, 56.9% ± 9.1% in TGF-β1–transgenic SCCs, and 59% ± 6.9% in ΔβRII-transgenic SCCs (Figure (Figure7,7, A and B). TGF-β1/ΔβRII–transgenic SCCs exhibited the greatest amount of vessels covering the stromal area (70.1% ± 9.3%; Figure Figure7,7, A and B). We also examined these tumors for expression levels of VEGF and its receptor VEGFR1. VEGF was expressed at comparable levels among TGF-β1–, ΔβRII-, and TGF-β1/ΔβRII–transgenic SCCs (about 5-fold higher than in control SCCs; Figure Figure7C).7C). In control tumors, VEGFR1 expression was too low to be detected by RNase protection assay (RPA) (Figure (Figure7C).7C). In contrast, ΔβRII-transgenic SCCs exhibited a detectable level of VEGFR1 (Figure (Figure7C).7C). VEGFR1 expression levels in TGF-β1– and TGF-β1/ΔβRII–transgenic SCCs ranged from 2- to 5-fold higher than those in ΔβRII-transgenic SCCs (Figure (Figure7C).7C). We then examined SCCs at 25 weeks after DMBA initiation for expression levels of MMP-2 and MMP-9, both of which degrade the basement membrane (27) and also induce angiogenesis (28). Low levels of MMP-2 and MMP-9 expression were detected in control SCCs (Figure (Figure7D).7D). Both TGF-β1– and ΔβRII-transgenic SCCs exhibited a 2- to 4-fold increase in expression of MMP-2 and MMP-9 compared with control SCCs (Figure (Figure7D).7D). The highest level of MMP-2 and MMP-9 expression was observed in TGF-β1/ΔβRII–transgenic SCCs, which exhibited a 5-fold increase in MMP-2 expression and a 10-fold increase in MMP-9 expression compared with control SCCs (Figure (Figure77D).
To understand how ΔβRII selectively blocked TGF-β1–mediated EMT but not metastasis in skin carcinogenesis in vivo, we examined expression levels of signaling components of the Notch, Rho/Rac, and MAPK pathways that have been suggested to mediate TGF-β1–mediated tumor invasion and EMT (29, 30). As shown in Figure Figure8A,8A, expression of Jagged 1 (Jag1), a Notch ligand, and Hey1, a downstream target, was elevated 7- to 10-fold in the TGF-β1–transgenic SCCs compared with control SCCs. However, expression of Jag1 and Hey1 in TGF-β1/ΔβRII–transgenic SCCs was similar to that in control SCCs. With respect to expression of Rho/Rac signaling molecules, elevated expression of RhoA and Rac1 was observed in SCCs from ΔβRII-, TGF-β1–, and TGF-β1/ΔβRII–transgenic mice, ranging from 7- to 10-fold compared with their expression levels in control SCCs (Figure (Figure8A).8A). RhoC expression levels were similar among control and the various transgenic SCCs (data not shown). Among MAPK signaling components, mRNA levels of Erk1, Erk2, and JNK1 in SCCs from all 3 transgenic groups were increased approximately 5- to 20-fold compared with levels in control SCCs (Figure (Figure8B).8B). Expression levels of JNK2 were 2- to 3-fold higher in transgenic SCCs than in control SCCs (Figure (Figure8B).8B). Since activation of these molecules is independently regulated at the protein level by phosphorylation, we performed Western blot analysis for the total and phosphorylated forms of each of these molecules. Higher levels of total and phosphorylated Erk1/2 and JNK1/2 proteins were observed in the various transgenic SCCs compared with control SCCs (Figure (Figure8C).8C). Expression levels of another MAPK component, p38 kinase, were similar among transgenic and control SCCs at both the mRNA and protein levels (data not shown).
In this study, we observed that TGF-β1 was overexpressed and TGF-βRII was reduced in more than 50% of the human skin SCC samples we analyzed, and nearly 40% of the SCCs exhibited the combination of TGF-β1 overexpression and TGF-βRII reduction. In our transgenic mouse models, we found that late-stage TGF-β1 overexpression in chemically induced skin papillomas did not exert a tumor-suppressive effect. ΔβRII expression selectively blocked TGF-β1–mediated EMT but cooperated with TGF-β1 for tumor invasion.
Our previous study demonstrated that TGF-β1 transgene expression switches its function from a tumor suppressor to a tumor promoter after mid-stage skin carcinogenesis (6). Since TGF-β1 expression was induced at a relatively early stage in that study, we could not rule out the possibility that long-term TGF-β1 overexpression selects for TGF-β1–resistant tumors that have a high malignant potential. To address this issue in the present study, we induced TGF-β1 transgene expression after a full course of TPA promotion and performed BrdU labeling shortly after TGF-β1 transgene induction in late-stage papillomas. We found that these papillomas were resistant to TGF-β1–mediated growth inhibition. Loss of TGF-β1–mediated growth arrest is apparently not due to spontaneous loss of TGF-β signaling, since increased nuclear pSmad2 protein was observed in TGF-β1–transgenic papillomas. This contrasts with our previous observation that TGF-β1 transgene expression accelerates the loss of its signaling components during skin chemical carcinogenesis (6). Thus, a relatively long term of TGF-β1 overexpression is required for downregulation of TGF-β1 signaling components, and the resistance to TGF-β1–mediated growth inhibition in late-stage papillomas observed in this study might be attributable to the accumulation of more genetic insults than in the earlier-stage papillomas, which can override TGF-β1–mediated growth arrest.
In vitro experiments using ΔβRII cell lines (8, 31), or xenografted cells transfected with ΔβRII (32) demonstrated that ΔβRII prevents TGF-β1–mediated EMT. Consistent with these reports, our results showed that TGF-β1–mediated SPCC formation requires functional TGF-βRII in tumor epithelia during skin carcinogenesis, which indicates that the role of TGF-βRII in TGF-β1–mediated EMT cannot be compensated for in vivo by other mechanisms. It has been suggested that TGF-β1–mediated downregulation of E-cadherin, either at the transcriptional level (33, 34) or via disassembly of membrane-associated E-cadherin (35, 36), plays an important role in EMT. In cultured keratinocytes, TGF-β1 induces the disassembly of E-cadherin via complex signal transduction pathways, including the Notch, MAPK, and Rho/Rac pathways (29, 37). In the present study, we found that TGF-β1–transgenic SCCs exhibited either increased expression levels or activation of components of the above-mentioned signaling pathways. However, the blockade of EMT by ΔβRII was only correlated with reversed expression levels of Jag1 and Hey1 in TGF-β1/ΔβRII–transgenic tumors, which suggests that changes in the other molecules may be functionally redundant for TGF-β1–mediated EMT. A recent study has shown that expression of Hey1 is not only activated by Notch signaling, but can also be activated directly by TGF-β1 in a Smad-dependent manner (37). Our current study shows that TGF-β1–transgenic SCCs exhibited a very low level of pSmad2. This result is consistent with our previous observation that TGF-β1 induces loss of membrane-associated E-cadherin when tumors begin to lose Smads (6). However, one report shows that Smad2 is required for TGF-β1–mediated EMT in cultured SCC cells (22). It is possible that a very low level of pSmad2 in SCCs is sufficient for inducing expression of Jag1 and Hey1. Alternatively, since TGF-β1 transgene expression was able to induce Jag1 and Hey1 expression before pSmad2 was lost (data not shown), it is possible that once Jag1 and Hey1 levels were elevated by Smads, they were maintained at a high level via their own positive feedback loop (37, 38) and thus no longer required Smads. Since blockade of TGF-β1–induced Jag1 and Hey1 correlated with abrogation of TGF-β1–mediated loss of membrane-associated E-cadherin/catenin and SPCC formation, our data suggest that these 2 molecules play an indispensable role in TGF-β1–mediated EMT during skin carcinogenesis in vivo.
Although ΔβRII expression blocked TGF-β1–mediated EMT, malignant conversion and metastasis still occurred at the highest rate in TGF-β1/ΔβRII–transgenic tumors. Furthermore, many metastatic cancer cells from TGF-β1/ΔβRII– or ΔβRII-transgenic SCCs still retain membrane-associated E-cadherin/catenin. These results suggest that even though loss of membrane-associated E-cadherin is required for EMT, this event is dispensable for tumor metastasis. In addition to being involved in EMT, the Rho/Rac pathway is documented as a metastasis-associated pathway (39), and elevated levels of Erk and JNK have been observed in invasive skin SCCs (40, 41). Specifically, both the RhoA/Rac and the MAPK pathways have been documented to significantly affect cell motility (42, 43). Once cancer cells have the ability to move out of the primary lesion, local proteinases and angiogenesis play crucial roles in distant metastasis formation. To this end, both the RhoA/Rac and the MAPK pathways have been documented to induce MMP expression and angiogenesis (42–44). Our data show that ΔβRII expression also upregulated the levels of RhoA, Erk, and JNK. This result may reflect 1 of the following 2 possibilities. First, the current level of ΔβRII expression preferentially blocked Smad signaling but could not completely abrogate upregulation of RhoA, Rac1, Erk, and JNK. Therefore, increased endogenous TGF-β1 in ΔβRII-transgenic tumors or TGF-β1 transgene expression in TGF-β1/ΔβRII–transgenic tumors could upregulate these molecules in tumor cells via residual TGF-βRII signaling. Alternatively, levels of RhoA, Rac1, Erk, and JNK in ΔβRII- or TGF-β1/ΔβRII–transgenic tumors may be elevated as a result of a secondary effect of ΔβRII expression, as tumors with ΔβRII transgene expression developed earlier than did TGF-β1–transgenic tumors and thus presumably accumulated more genetic insults than TGF-β1–transgenic tumors at the same time point. It is hopeful that future studies using an approach that completely ablates TGF-βRII in keratinocytes will further clarify this issue. If the first possibility proves to be true, ablation of TGF-βRII in keratinocytes will block both EMT and metastasis. Therefore, tumors with partial and complete loss of TGF-βRII (e.g., via mutation or promoter methylation vs. gene deletion) will have a different outcome and prognosis. However, if the second possibility proves to be true, ablation of TGF-βRII in keratinocytes will still uncouple TGF-β1–mediated EMT and metastasis.
Since angiogenesis and expression levels of MMP-2 and MMP-9 were the greatest in TGF-β1/ΔβRII–transgenic SCCs, ΔβRII apparently cooperated with TGF-β1 overexpression in these processes. Obviously, ΔβRII expression in tumor epithelia cannot block the paracrine effect of TGF-β1 on tumor stroma, an important factor in inducing MMPs and angiogenesis. However, this cooperation cannot be simply explained by TGF-β1 levels in TGF-β1/ΔβRII–transgenic tumors, since they were similar to those in TGF-β1–transgenic tumors. It is likely that ΔβRII expression allows the accumulation of other oncogenic events, which synergistically interact with the paracrine effect of TGF-β1 overexpression on the upregulation of MMPs and VEGF. These oncogenic events may have occurred at stages prior to the spontaneous loss of TGF-β1–induced growth inhibition in wild-type tumors. Alternatively, the oncogenic events elicited by ΔβRII expression may also result from the blocking of other TGF-β1–mediated tumor-suppressive effects that are independent of TGF-β1–mediated growth arrest (1). Although in either case, the tumor suppressive effects of TGF-β1 should counteract its paracrine effect on tumor invasion, these effects appeared to be abrogated in TGF-β1/ΔβRII–transgenic tumors.
In summary, our study revealed that TGF-β1–mediated tumor invasion and EMT can be uncoupled. Figure Figure99 depicts the potential underlying mechanisms of these 2 distinct functions of TGF-β1. TGF-β1–mediated EMT relies on functional TGF-βRII to deregulate epithelial adhesion molecules, depending, at least in part, on the activation of Jag1 and Hey1. Jag1 and Hey1 are upregulated via a Smad-dependent pathway at early stages, and their expression levels are maintained via their own positive feedback loop once expression of Smads is lost. In contrast, TGF-β1–mediated metastasis largely relies on proteinase activation and angiogenesis, which are mainly mediated by increased levels of RhoA/Rac, Erk, and JNK. ΔβRII expression either cannot completely block TGF-βRII–mediated activation of these pathways and therefore activates these pathways via increased endogenous TGF-β1, or it leads to activation of these pathways via other oncogenic events accumulating in the absence of TGF-β1–mediated growth arrest. In addition, the accumulated additional oncogenic events in TGF-β1/ΔβRII–transgenic tumors due to blocking TGF-β1–mediated tumor-suppressive effects in tumor epithelia cooperate with the paracrine effect of TGF-β1 on tumor stroma to achieve tumor invasion and metastasis. Our results warrant future studies using both clinical samples and experimental models to further determine whether skin SCCs that have both increased TGF-β1 and reduced TGF-βRII will have a poor prognosis.
We collected tissue samples between the years 2001 and 2004 from consenting patients under a protocol approved by the Institutional Review Board at the Oregon Health & Science University. Normal human skin from healthy volunteers was obtained via punch biopsy or cosmetic surgeries that removed the excess skin. Skin tumors were surgically removed. H&E-stained sections were prepared from all tissue samples and were reviewed by a dermatopathologist for diagnosis.
Immunohistochemistry was performed on formalin-fixed tumor sections as previously described (11). Staining for TGF-β1 and TGF-βRII was performed as previously described (45) using TGF-β1 (R&D Systems) or TGF-βRII antibodies (Santa Cruz Biotechnology Inc.). Blocking peptides specific for each antibody were included as negative controls. Cell adhesion molecules in mouse tumor samples were stained as previously described (6), using antibodies specific for E-cadherin, β-catenin, and γ-catenin (BD Biosciences — Pharmingen). Staining of pSmad2 was performed using a pSmad2 antibody (Cell Signaling Technology). The immune complexes were detected by the avidin-biotin-peroxidase complex using Vectastain kits (Vector Laboratories) and visualized with diaminobenzidine. Sections were counterstained with hematoxylin. Protein levels detected by immunohistochemistry were visually evaluated. A double-blind evaluation was performed by 2 investigators using the methods described by Yang et al. (46) with modifications. Briefly, signal intensity was scored as 0 (no intensity), 1 (weak intensity), 2 (moderate), or 3 (strong). For each tissue section, the percentage of positive cells within each of the 4 intensity categories was estimated and multiplied by the corresponding intensity score. The 4 partial scores obtained in each section were added and expressed as a final score. The staining intensity in normal epidermis was determined to be 1 for TGF-β1, 3 for TGF-βRII, and 0 for pSmad2. Therefore, a score of 1.5 or higher for TGF-β1 was considered as increased staining; less than 2.5 for TGF-βRII as decreased staining; and above 0 for Smad2 as positive staining.
The generation of transgenic mice and the procedures used were approved by the Institutional Animal Care and Use Committee of the Oregon Health & Science University. All of the transgenic lines were generated in the ICR strain, and transgene expression was achieved using a mouse loricrin (ML) promoter, which targets transgene expression to the interfollicular epidermis and through all stages of skin carcinogenesis, including metastatic lesions (11). We generated TGF-β1–transgenic mice by mating the ML.GLVPc transactivator line with the tk.TGF-β1 target line, in which a modified thymidine kinase (tk) promoter precedes the TGF-β1 transgene (23). ML.GLVPc/tk.TGF-β1 bigenic mice were also crossbred to ML.ΔβRII-transgenic mice (24) to generate ML.GLVPc/tk.TGF-β1/ML.ΔβRII–transgenic mice.
A chemical carcinogenesis protocol was applied to 8-week-old mice, as previously described (11). Briefly, 50 μg of DMBA was topically applied to the shaved back skin. One week after DMBA initiation, mice were topically treated with 5 μg TPA, once a week for 20 weeks. TGF-β1 transgene expression was induced by topical application of a progesterone antagonist, RU486, to skin tumors of bigenic TGF-β1– or TGF-β1/ΔβRII–transgenic mice (Figure (Figure2A).2A). RU486 was given at a dose of 20 μg/mouse (dissolved in 100 μl of 70% ethanol), 3 times per week, beginning at 20 weeks after DMBA initiation. Control mice and ΔβRII-transgenic mice were also treated with RU486 or with ethanol.
Tumors were fixed in 10% neutral buffered formalin at 4°C overnight, embedded in paraffin, sectioned to 6 μm, and stained with H&E. Tumor types were histologically determined following the criteria described by Aldaz et al. (47).
Tumor-bearing mice at 20 weeks after DMBA initiation were treated with RU486 (20 μg, 3 times/week). Eleven hours after the third RU486 treatment, BrdU (125 mg/kg in 0.9% NaCl) was injected (i.p.) into these mice. Papillomas and adjacent skin from TGF-β1–transgenic and control mice were dissected 2 hours after the injection, fixed, and processed as previously described (48). Sections were incubated with an FITC-conjugated monoclonal antibody to BrdU (BD Biosciences) and a guinea-pig antiserum to mouse keratin 14 (K14), which reacts with the epithelial component of papillomas. K14 was visualized using Alexa 594–conjugated anti-guinea-pig IgG (Invitrogen Corp.). Consecutive fields of BrdU-labeled cells were counted throughout the entire tissue section. Five papillomas and adjacent skins were analyzed in each group. The labeling index was expressed as the mean number of BrdU-positive cells/mm basement membrane ± SD. Immunofluorescence staining of CD31 and quantitation of blood vessels were performed as previously described (11).
Skin and tumor RNA was isolated with RNAzol B (Tel-Test Inc.) as previously described (24) and purified using a QIAGEN RNeasy kit (QIAGEN). Quantitative RT-PCR was performed as previously described (49). TaqMan Assays-on-Demand probes (Applied Biosystems) were used to detect transcripts of mouse Jag1, Hey1, RhoA, RhoC, Rac1, Erk1, Erk2, JNK1, JNK2, and p38. An 18S RNA probe was used as an internal control, and the data (cycle of threshold [CT] values) were analyzed using the ΔCT method. The results were averaged from 5 mice of each group, and 1 control sample with the lowest expression level of each individual molecule was assigned the value of 1 arbitrary unit. We used RT-PCR to detect TGF-β1 transgene expression in transgenic tumor samples, using primers specific for porcine TGF-β1 (23). We performed RPAs using the RPA II kit (Ambion) and 32P-labeled riboprobes. The MMP-2 and MMP-9 probes were made as previously described (6). The VEGF and VEGFR1 probes were from the mouse angiogenesis-1 multiprobe set from BD Biosciences — Pharmingen. A riboprobe for L32 or cyclophilin was used as a loading control. The intensity of bands representing protected transcripts was determined by densitometric scanning of x-ray films.
Skin and tumor proteins were extracted as previously described (49). The amount of TGF-β1 protein was measured by TGF-β1–specific ELISA as previously described (49). We performed Western blot analysis as previously described (49) using antibodies specific for total and phosphorylated Erk, JNK, and p38 (Cell Signaling Technology).
This work was supported by NIH grants CA79998 and CA87849 (to X.-J. Wang). We thank the Molecular Profiling Resource of the Departments of Dermatology and Otolaryngology, Oregon Health & Science University for providing skin SCC samples.
Nonstandard abbreviations used: ΔβRII, dominant-negative TGF-βRII; AK, actinic keratosis; CIS, carcinoma in situ; DMBA, dimethylbenzanthracene; EMT, epithelial-to-mesenchymal transition; Jag1, Jagged 1; K14, keratin 14; ML, mouse loricrin promoter; pSmad2, phosphorylated Smad2; RPA, RNase protection assay; SCC, squamous cell carcinoma; SPCC, spindle cell carcinoma; TGF-βRII, TGF-β type II receptor; tk, thymidine kinase promoter; TPA, 12-O-tetradecanoylphorbol-13-acetate.
Conflict of interest: The authors have declared that no conflict of interest exists.