Tumor-Specific Efficacy of Transforming Growth Factor-βRI Inhibition in Eker Rats

The cytokine transforming growth factor β1 (TGF-β1) is the prototypic member of the TGF-β superfamily, which is composed of related growth factors such as activins and bone morphogenic proteins (1–3). TGF-βs are secreted as latent molecules (large latent complex) requiring local activation for receptor binding, which releases a 25 kDa active dimer composed of identical polypeptide chains (4, 5). TGF-βs exert a diverse range of biological functions regulating cell proliferation (6, 7), apoptosis (8, 9), angiogenesis (10), immune function (11, 12), and synthesis of extracellular matrix components (13–15). The functional significance of TGF-β is underscored by the existence of three mammalian isoforms (TGF-β1, TGF-β2, and TGF-β3) that have both overlapping and isoform-specific functions. Whereas TGF-β isoforms are highly conserved at the amino acid level (70-80% homology) and bind identical receptors, their expression and function are highly regulated, as their promoters and mechanisms for activation of their respective latent forms are different (5, 16).

TGF-βs signal through a heterotetrameric receptor complex that consists of dimers of type I and type II receptors, both of which are required for signal transduction (17–20). The TGF-β type II receptor binds ligand and the TGF-β type I receptor, also named activin-like kinase 5 (ALK5), is a serine/threonine kinase that phosphorylates intracellular secondary messengers Smad2 and Smad3. The phosphorylated Smad proteins bind Smad4, and the complex translocates to the nucleus to act as transcriptional regulators for responsive genes mediating a wide range of TGF-β functions.

TGF-β signaling participates in opposing ways to tumorigenesis, serving both inhibitory and promoting functions (21–23). It is now well recognized that whereas TGF-βs act as tumor suppressors early in the pathogenesis of epithelial lesions, in later stages of this disease, they may promote progression, epithelial to mesenchymal transition, and mediate metastasis. In addition, TGF-βs affect the tumor microenvironment by being immunosuppressive and angiogenic. Escape from TGF-β–mediated growth repression occurs in a significant proportion of epithelial tumors (4, 23, 24) and has been shown to be due to both down-regulation of TGF-β receptors and/or mutations in components of the signaling pathway (4, 25, 26). For example, in renal cell carcinoma (RCC), which arises from the epithelial cells of the renal nephron, loss of both type I and type II TGF-βRs occurs with a high frequency and is associated with tumor progression (27, 28).

There is a distinct difference between the role of TGF-β in the pathogenesis of mesenchymal and epithelial lesions. Whereas TGF-β inhibits the growth of epithelial cells, it is mitogenic for mesenchymal cells and has been implicated in the pathogenesis of mesenchymal diseases such as fibrosis and in the development of mesenchymal tumors such as uterine leiomyoma (29). Uterine leiomyoma are benign myometrial neoplasms that are the most common gynecologic tumor of women. There is strong evidence that TGF-β plays a central role in the pathogenesis of these tumors by contributing to tumor growth through stimulation of both myometrial cell proliferation and production of the abundant extracellular matrix characteristic of this disease (30, 31).

Eker rats carry a germ line defect in the tuberous sclerosis complex 2 (Tsc2) tumor suppressor gene (32, 33). The protein product of the Tsc2 gene, tuberin, inhibits mTOR activation, functioning as a negative regulator of AKT signaling (34–37). Eker rats develop spontaneous mesenchymal (uterine leiomyoma) and epithelial (RCC) lesions with a high frequency (38–40). Previous data have established that Eker rat leiomyomas share many phenotypic and molecular characteristics with the cognate human disease (29). Loss of function of the Tsc2 tumor suppressor gene in Eker rats results in the development of spontaneous uterine leiomyoma (41), and loss of function of this tumor suppressor gene also occurs in a significant proportion of human leiomyomas. Using tissue microarrays, it has been estimated that ∼50% of human leiomyomas exhibit absent or reduced expression of the Tsc2 gene product, tuberin (42), showing the relevance of this tumor suppressor gene for both the human and murine disease. Tumor-derived cell lines have also been established from Eker rat tumors, facilitating in vitro mechanistic studies (43, 44). As a result, this in vivo/in vitro model has been extensively used as a preclinical model to elucidate mechanisms of tumorigenesis and evaluate the efficacy of chemotherapeutic agents (39, 45).

Eker rats heterozygous for the Tsc2 mutation also develop multifocal, bilateral RCC with 100% incidence by 12 months of age (40, 46). Tumors develop from early preneoplastic lesions and progress through adenoma to carcinoma (47). Rat RCC are solid, chromophilic lesions, and although these tumors differ from the clear cell type most often observed in humans, they share many similarities with their human counterpart (46). Several genes are involved in human RCC, including von Hippel-Lindau (VHL), tuberous sclerosis complex 2 (TSC2), fumarate hydratase, and Birt-Hogg-Dube (48). RCC that result from loss of VHL are the most common, and inactivation of VHL leads to stabilization of hypoxia-inducible factor (HIF) 1α and 2α and overexpression of genes that promote tumorigenesis and angiogenesis (49–52). Recent evidence suggests that the involvement of von Hippel-Lindau and Tsc-2 in the development of RCC may affect similar molecular pathways. Renal tumors that arise in patients with both tuberous sclerosis and von Hippel-Lindau show a high degree of vascularity as compared with unaffected kidneys (53). Tsc2-null rat RCC also exhibit constitutively high expression of HIF2α (54), making dysregulation of HIF2α expression a common theme in both human and rodent RCC (54, 55). Thus, the Eker rat model for RCC is an excellent surrogate for the human disease, and this model is currently being used in preclinical studies for therapeutic agents of RCC (56, 57).

The inhibitor, SB-525334, blocks the ATP-binding site of the TGF-β type I receptor, ALK5, and inhibits TGF-β–induced ALK5 serine/threonine kinase activity, thereby preventing phosphorylation of the Smad transcription factors and subsequent gene activation (58). Analogues of this compound have been shown to inhibit TGF-β1–induced up-regulation of collagen Iα1 and plasminogen activator inhibitor 1 (PAI-1) mRNA by TGF-β1 in renal epithelial A498 carcinoma cells (59) due to inhibition of Smad2/3 activation of these genes. These compounds are now being evaluated for use in chronic organ remodeling diseases in which proliferation, malignant transformation, and fibrosis are a major component. In addition, as blockade of TGF-β signaling has been proposed as a cancer therapeutic because of its ability to block metastases and the immunosuppressive and angiogenic functions of TGF-β, evaluation of this strategy in preclinical models is warranted (60).

We have now evaluated the efficacy of a TGF-β signaling blockade using SB-525334 in a series of preclinical experiments in the Eker rat model. Similar to human leiomyomas, leiomyomas that developed in female Eker rats expressed both type I and type II TGF-β receptors, express several isoforms of TGF-β, and exhibited elevated TGF-β signaling relative to normal myometrium. In response to treatment with SB-525334, TGF-β signaling in these cells was inhibited and the incidence and multiplicity of uterine leiomyomas was significantly reduced. However, SB-525334 increased mitoses and decreased apoptosis in renal epithelial cells and significantly exacerbated renal tumorigenesis, as evidenced by an increase in renal tumor multiplicity in treated animals.

In vivo study. Animals were maintained and handled according to NIH guidelines and in Accreditation of Laboratory Animal Care–accredited facilities. The protocols involving the use of these rats were approved by the M.D. Anderson Cancer Center Institutional Animal Care and Use Committee. Animals were maintained on a 12 h light/dark cycle, with food and water provided ad libitum. To determine the effects of a TGF-β receptor inhibitor on uterine leiomyoma, female Eker rats 12 or 14 months old were given SB-525334 (GlaxoSmithKline) at a dose of 200 mg/L drinking water (estimated dose of 10 mg/kg/d) or received normal drinking water for 2 and 4 months. At 16 months of age, animals were sacrificed by CO₂ asphyxiation and tissues were harvested and either snap-frozen in liquid nitrogen and stored at −80°C or fixed in 10% neutral buffered formalin and paraffin embedded. To further analyze the effects of SB-525334 on kidneys, 9-month-old male Eker rats were given plain drinking water or the compound in drinking water at 200 mg/L for 2 months. Rats were then sacrificed and tissues were harvested, fixed, and stored as described above. For histology, tissues were stained with H&E, and kidneys and multiple sections of female reproductive tract (uterus, vagina, and cervix) were examined microscopically by a pathologist blinded as to treatment group (see below). All tumors and proliferative lesions were identified and evaluated as previously described (61).

In vitro analyses.In vitro experiments were conducted to examine the effects of SB-525334 on cells from the Eker rat leiomyoma-derived cell line, ELT-3 (43). Cells were maintained in DF8 medium (43) for 24 h, then starved in DMEM/F12 medium (Invitrogen) + 1% fetal bovine serum (Hyclone) for 24 h. To determine dose-response of ELT-3 cells to SB-525334, cells were treated for 1 h with vehicle (DMSO), TGF-β3 (R&D Systems), and SB-525334 at 0.5, 1, and 2 μmol/L, respectively, or TGF-β3 + SB-525334 at 0.5, 1, or 2 μmol/L, then harvested for Western analysis for quantitation of SMAD phosphorylation. Treatment with 2 μmol/L of SB-525334 resulted in maximal inhibition of phosphorylation and the 2 μmol/L dose was used in subsequent experiments.

Western analysis. Purified rabbit IgG antipeptide antibodies to human TGF-β1, TGF-β2, and TGF-β3 were non–cross-reacting and have been previously described (62). Rat leiomyoma and myometrial tissue lysates (20 μg) were subjected to SDS-PAGE (acrylamide gradient, 5-10%) and transferred to polyvinylidene difluoride membranes. The membranes were incubated in 3% nonfat dry milk blocking buffer overnight at 4°C and separately incubated with each anti–TGF-β isoform antibody (TGF-β1 and TGF-β2 at 2.0 μg/mL and TGF-β3 at 1.0 μg/mL) in blocking buffer for 3 h, followed by streptavidin horseradish peroxidase–conjugated goat anti-rabbit secondary antibody for 1 h at room temperature, and finally, the Super Signal West Dura Kit (Pierce) was used for detection on X-ray film (BioMax, Eastman Kodak). The protein bands were quantified by densitometry using an EDAS 290 (Eastman Kodak) and the Kodak 1D3.6 image analysis software. The blots were stripped and reprobed with an antibody to γ-tubulin (1:10,000; Sigma). The net intensity for each band was obtained by comparison with tubulin for each sample and blot. Other antibodies used for Western analysis according to the instructions of the manufacturer were: TGF-β receptor type I and type II (Santa Cruz Biotechnology), SMAD2/3 (1:1,000, Upstate), and phospho-SMAD2 (1:500; Cell Signaling Technologies).

Cell fractionation. To examine phospho-SMAD2, SMAD2/3, and TGF-β receptor type I and II localization, ELT-3 cells were treated for 1 h with vehicle (DMSO), TGF-β3, SB-525334 (2 μmol/L), or TGF-β3 + SB-525334, and harvested for fractionation. For whole cell extracts, cells (80% confluent in 150 mm plates) were washed twice with ice-cold PBS, scraped into 200 μL of cold 1× lysis buffer [from Cell Signaling Technologies: 20 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA, 1% Triton X-100, 2.5 mmol/L sodium PPi, 1 mmol/L β-glycerophosphate, 1 mmol/L Na₃VO₄, 1× Roche complete protease inhibitor], homogenized by sonication and pelleted by centrifugation at 14,000 rpm at 4°C for 10 min. The supernatant was collected and stored at −80°C for further analysis. To prepare nuclear and cytosolic fractions, cells (80% confluent in 150 mm plates) were washed twice with ice-cold PBS and scraped into 75 μL of ice-cold buffer A [10 mmol/L HEPES (pH 7.9), 10 mmol/L KCl, 0.1 mmol/L EDTA, 0.4% NP40, 1 mol/L DTT, 200 mmol/L phenylmethylsulfonyl fluoride, 4 mol/L aprotinin, and 1 μL leupeptin], incubated at room temperature for 5 min and centrifuged at 14,000 rpm at 4°C for 10 min. The resulting cytosolic supernatant was transferred to a new microcentrifuge tube and stored at −80°C for further analysis. The remaining pellet was washed with 350 μL of buffer A, and centrifuged at 14,000 rpm at 4°C for 5 min. The supernatant was discarded and the pellet was resuspended in buffer B [20 mmol/L HEPES (pH 7.9), 0.4 mol/L NaCl, 1 mmol/L EDTA, 10% glycerol, 1 mol/L DTT, 200 mmol/L phenylmethylsulfonyl fluoride, 4 mol/L aprotinin, and 1 μL leupeptin] at a volume approximately equal to that of the pellet. Samples were placed on a rotator at 4°C for 2 h, then centrifuged at 14,000 rpm at 4°C for 10 min. The supernatant (nuclear fraction) was collected and stored at −80°C for further analysis.

Immunohistochemistry. Paraffin sections were deparaffinized, rehydrated, and subjected to heat-induced antigen retrieval using 1× citrate buffer in a pressure cooker. Sections were treated with 3% hydrogen peroxide for 5 min and blocked for endogenous biotin using an avidin/biotin blocking system (DakoCytomation Corporation). For phospho-SMAD2 labeling, nonspecific antibody binding was blocked by incubating slides with 10% goat serum in PBS for 30 min. Slides were drained and incubated at 4°C overnight with polyclonal phospho-SMAD2 (1:50; Calbiochem). Following the primary antibody, slides were incubated with EnVision Plus–labeled polymer, anti-rabbit horseradish peroxidase (DakoCytomation Corporation) at room temperature for 30 min. Staining development was monitored as sections incubated in 3,3-diaminobenzidine (BioGenex). Slides were counterstained, dehydrated, cleared, and coverslipped.

Several antibodies were used to assess tissue proliferation rates and apoptotic indices. For female reproductive tract tissues, following a 15-min protein block (DakoCytomation Corporation), bromodeoxyuridine monoclonal antibody (1:500; BD Biosciences) was applied to uterine and leiomyoma sections and incubated at room temperature for 1.5 h. Following primary antibody, biotinylated rabbit anti-mouse F(ab)′ (Accurate Chemicals) was added and incubated at room temperature for 15 min. Kidney sections were treated with a monoclonal anti-human topoisomerase IIα clone SWT3D1 (1:50; DakoCytomation Corporation) or a monoclonal anti-rat Ki-67 clone MIB-5 (1 μg/mL; DakoCytomation Corporation) which was applied for 30 min. Omission of primary antibody and an isotype-matched mouse IgG (Southern Biotechnology) were used as controls. For topoisomerase IIα labeling, sections were incubated in mouse EnVision horseradish peroxidase–labeled polymer for 30 min. To enhance staining for Ki-67, the Catalyzed Signal Amplification system (DakoCytomation Corporation) was used.

Tissue sections were read by board-certified veterinary pathologists (blinded with respect to treatment group) who had extensive experience with rodent tissues and Eker rat proliferative lesions. The entire reproductive tract was evaluated for proliferative changes on H&E-stained sagittal sections of the vaginal and cervical regions as well as multiple cross-sections of the uterine horns. Additionally, terminal nucleotidyl transferase–mediated nick end labeling (TUNEL), topoisomerase II, and Ki-67 immunostaining for each rat were scored separately by region: renal cortex (including glomeruli, tubular epithelium, and vasculature), distal medullary collecting ducts, outer stripe of the outer medulla (OSOM), inner stripe of the outer medulla (ISOM), as well as the TUNEL, topoisomerase II, and Ki-67 score for renal tumors (adenomas or adenocarcinomas). Tumors were not included in the scores for any region in which they resided (for example, cortical TUNEL scores did not include staining involving a renal cortical neoplasm, but did include hyperplastic foci). Scoring was done by counting the actual number of obviously positive cells (borderline positive staining was not counted) in a 100× microscopic field. Ten fields were examined and averaged for the cortex, three for the distal medulla, five each for the OSOM and ISOM, and two fields for the renal tumors. For TUNEL staining, the following specific criteria were used to distinguish real staining from artifacts: necrotic areas were common in tumors; however, these universally stained positive and were disregarded, as were all positive cells that were free-floating within the tubular lumina. Other disregarded, positively staining cells included any positive cells along the edges of these necrotic foci, or along cut tissue edges anywhere in the kidney. Inflammatory cells, including a number of positively staining intravascular lymphocytes, were not included in the counts. Hyaline cast staining was also disregarded.

RNA isolation and quantitative real-time PCR. Total RNA was isolated from uterine tumor samples and ELT-3 cells with commercially available kits (Ambion, Invitrogen). Residual DNA was removed using DNase I (Ambion) for 30 min at 37°C followed by inactivation by incubation for 2 min at 20°C with a DNase inactivation reagent (Ambion). For cDNA synthesis, 1 μg of total RNA, random hexamers, and SuperScript II RT (Invitrogen) were combined and one cycle was done for 10 min at 25°C, 50 min at 42°C, and 15 min at 70°C. To finalize cDNA synthesis, RNase H was added followed by incubation at 37°C for 20 min to digest the remaining RNA. cDNA was diluted 10-fold prior to PCR amplification.

Real-time PCR was done using the ABI 7700 Detection System (Applied Biosystems) according to the instructions of the manufacturer. Reactions were conducted in a 25 μL volume reaction mixture containing 10 mmol/L of primers and a 10 mmol/L of FAM-labeled probe. TaqMan universal PCR master mix was used, which contained nucleotides, Taq DNA polymerase, and buffers. The PCR reaction conditions were as follows: 10 min denaturation step, followed by 40 cycles at 95°C for 15 s and 60°C for 1 min. To confirm the specificity of PCR products, each primer pair was subjected to a melting curve analysis and agarose gel electrophoresis.

Statistics and data analyses. Only those comparisons with statements of nonsignificant differences or with estimates of significance (P values) were compared with formal statistical tests. Other statements of differences were based on visual or observational comparisons.

The statistical comparisons of the uterine sample data from the SB-525334–treated animals compared with the uterine sample data from the age-matched animals used several different statistical tests. The comparisons of the bromodeoxyuridine proliferative index, the TUNEL apoptosis index, and the different TGF-β Western blot expression levels used the nonparametric Wilcoxon-Mann-Whitney test. For these same animals, the comparison of the protein samples with and without the lower molecular weight form of TGF-β3 used the two-sided Fisher exact test. No multiple comparison adjustment calculations were made for these comparisons. The leiomyoma incidence, multiplicity, and size comparisons used the Cochran-Mantel-Haenzel test to account for the separate 2-month and 4-month treatment groups. The adjustment for the multiple comparisons across the seven incidence, multiplicity, and size tests used the step-down Bonferroni method (63).

The statistical comparisons of the proliferative and apoptosis indices of the renal samples from the SB-525334–treated animals and the data from the age-matched animals all used the one-sided Wilcoxon-Mann-Whitney test with the critical side predetermined by the expected tumor increase. The adjustment for the multiple comparisons across the four regions of the kidney used the step-down Bonferroni method within the separate staining methods.

The analyses for all of the real-time PCR mRNA measurements were based on the log of the gene expression measurement value. The log-scaled experimental replicate values were calculated as the difference between the average of the triplicate log expressions values for the target gene (TGF-β or PAI) and glyceraldehyde-3-phosphate dehydrogenase from the same tissue and experimental replicate (three replicates of the triplicate measurements for all except the PAI expression from the uterine tissue, which had either one or two replicates of the triplicates). Previous gene expression studies have shown that the log-scaled mRNA levels have an approximate normal distribution. Based on this historical evidence, the tissue and treatment comparisons used t test and ANOVA test methods.

The comparisons of the TGF-β gene expression between the leiomyoma and normal tissue used separate two-sample t tests for each isoform. These t tests used the mean of the three experimental replicates for the separate tissue sources. This gave sample sizes of three for the leiomyoma tumor tissue and one for the normal uterine tissue. The adjustment for the multiple comparisons across the three TGF-β genes used the step-down Bonferroni method. The additional comparison of the PAI gene expression between the leiomyoma and normal tissue used weighted ANOVA methods to account for the single experimental replicate of the normal tissue. This comparison used the log-scaled value of the limit of detection level as the normal tissue expression value and the mean of the two experimental replicates from the four-tumor sample sources (the use of the scaled limit of detection level provided a conservative estimate of the statistical significance of the tumor difference). No multiple comparison adjustment was required.

The comparisons of the PAI gene expression for the different in vitro treatments of the ELT-3 cell line used simple ANOVA of the log-scaled expression levels. The adjustment for the multiple comparisons across the six pair-wise treatment comparisons used the step-down Bonferroni method.

TGF-β signaling in Eker rat uterine leiomyomas. A series of in vitro/in vivo studies were conducted to investigate TGF-β expression and signaling in uterine leiomyoma in the Eker rat model, using primary tumors, normal myometrium, and a leiomyoma-derived cell line, ELT-3 (43). Both normal myometrium and leiomyomas expressed abundant type I and type II TGF-βRs, as did the leiomyoma-derived ELT-3 cell line (Fig. 1A and B).

TGF-β expression was more complex, exhibiting both tissue-specific and isoform-specific patterns of expression. Relative to normal myometrium, and similar to what has been shown in human leiomyomas, Eker rat leiomyomas and ELT-3 cells expressed TGF-β as determined by real-time PCR and Western analysis (Figs. 1–3). Only TGF-β3 mRNA expression was determined to be significantly elevated in tumors versus normal myometrium (multiplicity adjusted P = 0.013 and unadjusted P = 0.0043; Fig. 2). There was no significant difference between TGF-β1 or TGF-β2 expression in tumors versus normal myometrium. At the protein level, leiomyomas variably expressed the bioactive dimer (∼M_r 25,000) of all three TGF-β isoforms and protein expression was generally concordant with mRNA levels (Fig. 3). Although TGF-β1 and TGF-β3 mRNA expression was higher in tumors, at the protein level, there was no significant difference in TGF-β1 and TGF-β3 expression in tumor versus normal tissue (Fig. 3). However, the TGF-β3 isoform was expressed as two prominent bands (∼27 and 17 kDa; added together for quantification). The lower molecular weight variant of TGF-β3 was observed in 12 of 12 tumors and as a very faint band in one of five normal tissues. A minor band of ∼18.5 kDa, which might have been a minor proteolytic fragment of the dimer, was seen in five of five normal tissues but not in tumors. Interestingly, the TGF-β2 isoform also exhibited a tumor-specific expression pattern, with leiomyomas having readily detectable levels of TGF-β2, whereas expression of this isoform was barely detectable or absent in all normal myometrial samples examined (Fig. 3). Therefore, although all tumors expressed TGF-β receptors and one or more TGF-β isoforms, it was not clear from examination of these components of the TGF-β signaling pathway alone that tumors exhibited differential activation of TGF-β signaling relative to normal myometrium.

To determine if TGF-β signaling differed between normal and tumor tissues, we next examined SMAD phosphorylation, localization, and expression of PAI, a highly sensitive TGF-β–regulated gene, in tumors versus normal myometrium. Relative to normal myometrium, tumors and ELT-3 cells had abundant nuclear phosphorylated Smad, which correlated with levels of PAI expression. As shown in Fig. 4, leiomyomas exhibited abundant nuclear immunoreactivity to a phospho-SMAD antibody, in contrast with normal myometrium in which immunoreactivity was scattered or only barely detectable. Concordant with this observation, leiomyoma-derived ELT-3 cells exhibited nuclear phospho-SMAD as determined by cell fractionation (Fig. 5A). Leiomyomas also expressed high levels of PAI transcripts, as detected by real-time PCR, whereas PAI transcripts were undetectable in the normal myometrium (Fig. 2D). Therefore, TGF-β signaling was activated in Eker rat leiomyomas, similar to what is thought to be the case for human leiomyomas, in which this signaling pathway is believed to play an important role in tumor pathogenesis (29).

The ALK5/type I TGF-βR inhibitor SB-525334 blocks TGF-β signaling in uterine leiomyoma cells. The presence of an active TGF-β signaling pathway in Eker rat leiomyomas suggested that these rats could be used as a preclinical model to examine the efficacy of inhibition of TGF-β signaling for uterine leiomyoma. To show proof-of-principle that the TGF-βR inhibitor SB-525334 could inhibit TGF-β signaling in leiomyomas, in vitro studies were first conducted using ELT-3 cells. As shown in Fig. 5B, ELT-3 cells exhibited a dose-dependent inhibition of signaling in response to TGF-β following treatment with SB-525334. Decreased SMAD phosphorylation in response to doses of SB-252334 ranging from 0.5 to 2 μmol/L were observed, and inhibition of signaling was confirmed by cell fractionation experiments that showed decreased phospho-SMAD in the nucleus of treated cells (Fig. 5A). In response to TGF-β, levels of nuclear phospho-SMAD increased in ELT-3 cells, and nuclear translocation was effectively inhibited by SB-525334 (Fig. 5B). In addition, as determined by real-time PCR, TGF-β induction of PAI transcription was also significantly inhibited by SB-525334 (multiplicity adjusted P = 0.00001 and unadjusted P < 0.00001; Fig. 2E) in contrast with basal PAI expression, which was not decreased in the presence of the inhibitor (Fig. 2E). Therefore, because SB-525334 was efficacious at inhibiting TGF-β signaling in leiomyoma cells in vitro, additional in vivo experiments were done to examine the effect of SB-525334 on leiomyomas in Eker rats.

SB-525334 treatment is efficacious for uterine leiomyoma. Female Eker rats were given SB-525334 (n = 25) or vehicle (n = 27) in drinking water for 2 to 4 months and sacrificed at 16 months of age. As shown in Fig. 6A, the incidence rate estimate for uterine leiomyomas was lower for animals treated with SB-525334 for either 2 (14-16 months of age) or 4 (12-16 months of age) months duration. Similarly, the multiplicity of uterine leiomyomas was also reduced in both 2- and 4-month treatment groups (Fig. 6B). The stratified analyses of the combined data from the 2- and 4-month treatment groups revealed that SB-525334 treatment was associated with statistically significant reductions in uterine leiomyoma incidence and multiplicity. As shown in Table 1, tumor incidence in vehicle-treated controls was 78%, comparable with the historical tumor incidence in this model (65%). In SB-525334–treated animals, the incidence of leiomyomas was significantly reduced, with only 40% of the animals having gross and/or microscopic uterine lesions (multiplicity adjusted P = 0.043 and unadjusted P = 0.0065). Leiomyoma multiplicity was also reduced significantly, decreasing from 1.26 lesions per animal in the control group to 0.56 lesions per animal in the treated group (multiplicity adjusted P = 0.043 and unadjusted P = 0.0061). Although the pooled average size of individual tumors was reduced from 4.67 cm³ in control animals to 0.88 cm³ in the treated animals, the size distributions of grossly observable tumors were not significantly different between the groups (Fig. 6C).

Tumors present in SB-525334–treated animals were further characterized in terms of histology and mitotic and apoptotic indices. Tumor phenotype in treated and control animals was similar, with tumors from both groups exhibiting the same characteristic typical, epithelioid or mixed histology previously described in this model. Quantitation of bromodeoxyuridine incorporation in the leiomyomas of treated versus control animals revealed no significant difference in the proliferative index of the two groups (6.2 versus 7.3 positive cells/field, respectively; P = 0.49). This was also the case for the apoptotic index of leiomyomas in treated versus control animals (0.35 versus 0.48 TUNEL-positive cells/field; P = 0.24), which were not significantly different from each other. Therefore, leiomyomas present in the treated animals at the end of the study exhibited no decrease in proliferation, or any increase in apoptosis in the presence of SB-525334, suggesting that they were resistant to inhibition of TGF-β signaling (and thus growth inhibition) by SB-525334. As tumors that persisted in treated animals continued to express TGF-β receptors (Fig. 1B), resistance may have been due to decreased dependence on TGF-β signaling for growth, rather than loss of expression of the SB-525334 target ALK5 receptor. The fact that a 4-month duration of treatment had no advantage over a 2-month treatment was also consistent with the presence of a subpopulation of tumors refractory to blockade of TGF-β signaling by inhibition of the ALK5/type I receptor.

Inhibition of TGF-β signaling by SB-525334 promotes the development of RCC. In addition to uterine leiomyomas, Eker rats are genetically predisposed to develop multiple, bilateral RCC. Susceptibility to renal lesions (adenomas, carcinomas, and cysts) is 100% penetrant in these animals, which made it possible to also assess the effect of SB-525334 treatment on these epithelial tumors. In contrast to its efficacy for uterine leiomyoma, SB-525334 had an adverse effect on the development of renal lesions in treated animals. The gross appearance of the kidneys of 16-month-old female rats treated with SB-525334 were remarkable for both the size and number of tumors present in this organ. As shown in Fig. 7, examination of the kidneys of Eker rats treated with the TGF-β inhibitor revealed that, in general, neoplastic lesions in the kidneys of treated animals were more pronounced than in kidneys from control animals. The macroscopic and microscopic features of the renal tumors present in treated animals were identical to those previously described in Eker rats (64) and included renal adenomas, adenocarcinomas, and atypical hyperplasias of both tubular and cystic types. Although both treated and control rats had a high multiplicity of large solid and cystic masses within the renal cortices bilaterally, in treated animals, large, often coalescing lesions were so numerous that frequently they greatly distorted the normal kidney architecture (Fig. 7A and B). Quantitation of grossly observable tumors confirmed that animals treated with SB-525334 had significantly more tumors than vehicle-treated controls (16.86 versus 7.93 lesions/animal, P < 0.001).

In addition to a genetic predisposition conferred by the Tsc2 gene defect, the development of renal tumors in aged rats is promoted by a characteristic renal nephropathy that occurs in older animals (61). This chronic progressive nephropathy occurs spontaneously in many strains of rats as a result of renal tubule degeneration, and results in a compensatory proliferation of tubular epithelial cells and an increased production of extracellular matrix in chronic progressive nephropathy lesions. To determine if inhibition of TGF-β signaling was directly affecting epithelial progenitor cells that give rise to RCC (as opposed to exacerbation of chronic progressive nephropathy in aging animals, which could indirectly promote tumor development), we examined the effect of SB-525334 on young animals prior to the development of chronic progressive nephropathy. For this study, young male rats (<9 months of age) were exposed to SB-525334 in the drinking water for 2 months. Kidneys from exposed and control animals were then evaluated for changes in proliferative (Ki-67 or topoisomerase II) and apoptotic (TUNEL) indices of tubular epithelial cells and nascent tumors.

In the kidney, proliferative indices varied throughout the renal nephron and appeared to be segment-specific (Table 2), with proliferation generally increasing from the papilla to the cortex. In vehicle-treated controls, the tip of the papilla at the distal medulla generally had a very low proliferative index, evidenced by an extremely small number of cells staining positively for either Ki-67 or topoisomerase II. Proliferative indices increased more proximally in the inner medulla approaching the area of the inner stripe, and continued to increase from the medulla towards the cortex, with progressively higher numbers of proliferative cells in the inner and outer stripe of the medulla. The highest proliferative indices were observed in cells of the cortex. Large numbers of positively staining cells were also associated with hyperplastic proximal convoluted tubules and cortical tumor cells. Scattered glomerular mesangial cells and rare interstitial fibroblasts were also positive.

SB-525334 treatment caused a roughly 2-fold increase in epithelial cell proliferation in all regions of the kidney as assessed by Ki-67 staining, which was concordant with topoisomerase II staining (Table 2). As shown by Ki-67 staining, cell proliferation in response to TGF-βRI inhibition was significantly increased in all four regions of the kidney (multiplicity adjusted and unadjusted P < 0.05 for the cortex, OSOM, ISOM, and distal medulla). In addition, incidental adenocarcinomas were present in some kidney sections of these young animals. Lesions in the SB-525334–exposed animals had a higher proliferative index than lesions present in vehicle-exposed animals, as assessed by both Ki-67 (198 versus 91.08 positive cells/field) and topoisomerase II (314 versus 236 positive cells/field) staining. However, the limited number of tumors present in these young animals (eight in the controls and five in the SB-525334–treated animals) precluded any assessment of statistical significance between the proliferative index of SB-525334–treated and vehicle-exposed tumors.

Apoptosis in the kidney exhibited a more complex pattern (Table 3). In vehicle-treated controls, TUNEL positivity was most often associated with tubular or duct epithelial cells and interstitial myofibroblasts. Glomerular mesangial cells, podocytes, vascular smooth muscle cells, and endothelial cells were only rarely positive. The overall staining pattern was often very focal, with a concentration of positively staining epithelium within a section of tubule or set of tubules. TUNEL staining also tended to be much more common in tubules that were hyperplastic or undergoing atypical dysplasia than in normal tubules, characteristic of the increased cell turnover occurring in these lesions. The junctional area of the distal medullary collecting ducts and the ISOM tended to be more positive than any other region, including the OSOM and particularly the cortex. Therefore, apoptosis seemed to be much higher in regions of the kidney in which tumors did not develop (ISOM and distal medulla) than in the regions of the kidney in which the tumors were likely to arise (cortex). Tumors present in control kidneys in general had only scattered positive cells, except in foci of coagulative necrosis and along the edges of necrotic areas.

Importantly, as shown in Table 3, in SB-525334–exposed animals, apoptosis was significantly decreased in the region of the kidney in which tumors arise, specifically the cortex. TUNEL-positive cells per field for treated versus control kidneys, respectively, was significantly reduced in the cortex (18 versus 40 with a multiplicity adjusted P = 0.01 and unadjusted P = 0.003, respectively). Apoptosis was also reduced in the OSOM (11 versus 17) and distal medulla (185 versus 200), although the reduction in apoptosis in these regions of the kidney were not statistically significant. Interestingly, in the ISOM, the apoptotic fraction increased in exposed animals (148.14 versus 72.0 TUNEL-positive cells), the same region of the nephron that showed the highest level of cell proliferation in response to SB-525334 (Table 2), although again, this change was not statistically significant. Consistent with the increase in tumor multiplicity observed in SB-525334–treated animals, the number of TUNEL-positive cells in the microscopic lesions of treated animals was lower than that of lesions from control animals (66 versus 85 positive cells/field); however, the number of tumors present was too small to draw statistical inferences. Taken together, the increased epithelial cell proliferation in SB-525334–exposed animals, combined with decreased apoptosis in the region of the kidney that is the primary site for tumor development in this model argues that the TGF-β blockade induced by this inhibitor had directly promoted the epithelial tumor development in animals genetically predisposed (initiated) to develop these tumors.

TGF-β signaling has been implicated in the pathogenesis of uterine leiomyoma and RCC via opposite mechanisms: increased TGF-β signaling promotes the development of uterine leiomyoma whereas escape from growth inhibition by TGF-β occurs with a high frequency in RCC (28, 29). Using Eker rats that are genetically predisposed to develop uterine leiomyoma and RCC with a high frequency, we found that the ALK5/type I TGF-βR inhibitor, SB-525334, was able to block TGF-β signaling in uterine leiomyoma cells. Similar to their human counterpart, we found that primary tumors and ELT-3 cells expressed type I and type II TGF-βRs, expressed TGF-β, and had elevated levels of nuclear phospho-SMAD. SB-525334 efficiently inhibited TGF-β–mediated signaling in these cells as shown by inhibition of SMAD phosphorylation, translocation to the nucleus, and induction of PAI expression. In female Eker rats treated with SB-525334 for 2 to 4 months, TGF-βRI blockade with this inhibitor significantly decreased the incidence and multiplicity of uterine leiomyomas. However, in the kidney, treatment with this inhibitor was mitogenic, reduced apoptosis in cortical epithelial cells, and greatly exacerbated the development/progression of RCC. These data show that although pharmacologic inhibition of TGF-β signaling can be efficacious for a mesenchymal tumor such as leiomyoma, systemic blockade of this important growth-inhibitory signaling pathway has the adverse effect of promoting the development of epithelial lesions.

Many cytokines and growth factors are produced by uterine leiomyomas, which might contribute to tumor growth through paracrine and/or autocrine mechanisms. These include TGF-β, insulin-like growth factors 1 and 2, basic fibroblast growth factor, platelet-derived growth factor, and epidermal growth factor (65). TGF-β has been of particular interest, and previous studies on human leiomyomas have found that these tumors express TGF-β receptors and SMADs and overexpress TGF-β1 and TGF-β3 compared with normal myometrium (66, 67). Consequently, the downstream targets of TGF-β signaling, such as tissue inhibitor of matrix metalloproteases, collagen, fibronectin, and PAI (66), which promote extracellular matrix production, are also overexpressed in these tumors (68). Recently, transcriptional profiling identified additional TGF-β–responsive genes overexpressed in leiomyoma cells, including interleukin 11, which plays a major role in other fibrotic disorders (69, 70).

One of the hallmarks of uterine leiomyoma, which distinguishes these benign tumors from malignant uterine leiomyosarcoma, is their low mitotic index. Although these tumors become quite large, often reaching baseball or grapefruit size, by definition, uterine leiomyoma have fewer than five mitoses per high-powered field. Given the low mitotic index of uterine leiomyoma, it is likely that growth factors contribute to tumor growth by stimulating both cell proliferation and the production of the abundant extracellular matrix that is the hallmark of these tumors. TGF-β3 has been shown to stimulate cell growth, collagen synthesis, and fibronectin expression in cell cultures derived from human leiomyomas (66). Responsiveness to TGF-β may be isoform- and tumor-specific, as previous studies found that whereas TGF-β1 and TGF-β3 both inhibited the growth of normal myometrial smooth muscle cells in vitro, in leiomyomas, TGF-β3 stimulated growth and TGF-β1 had no effect on the growth of these cells in culture (71). To some extent, the different effects of TGF-βs on cell growth in different studies is likely related to cell density and dose, as has been shown for other cell types in culture (10, 72–74). Nonetheless, taken together, it is clear that increased expression and/or responsiveness to TGF-β, particularly the TGF-β3 isoform, contributes to increased growth and production of the abundant extracellular matrix deposition characteristic of leiomyomas.

In contrast to the abundant data on TGF-β signaling in human leiomyoma, this is the first study to examine TGF-β expression and responsiveness in the Eker rat leiomyoma model. As shown in human leiomyomas, we observed an intact TGF-β signaling pathway in Eker rat uterine leiomyomas, however, some differences between the rat and human disease were evident. Whereas TGF-β1 and TGF-β3 were overexpressed at the RNA level in the rat leiomyomas, TGF-β1 and TGF-β3 isoform protein levels were not significantly elevated in leiomyomas compared with normal age-matched myometrium. In contrast, the expression of TGF-β2 in rat leiomyomas seemed to be tumor-specific and a low molecular weight variant of TGF-β3 was observed in all the tumors. There was a slight expression of this variant in one normal myometrium, which possibly may be predictive of tumor formation. Importantly, the presence of TGF-β and its cognate receptors does not necessarily indicate that it is functionally active because TGF-β exists as a latent molecule requiring activation for ligand-receptor interaction and downstream signaling. The fact that SMAD2 was activated and that PAI mRNA was highly expressed in leiomyomas compared with normal myometrium indicates that despite equal protein levels of TGF-β, in contrast with normal myometrium, the tumors show evidence of remarkably high activated TGF-β, which is consistent with the observed fibrogenic response in these tumors. These data on TGF-β signaling in Eker rat leiomyomas add to our body of knowledge regarding the extent of similarity of tumors that develop in this widely used preclinical model relative to the cognate human disease, and furthermore, suggests that the Eker rat may be a valuable preclinical model for testing the inhibition of this pathway as a therapy for this disease.

As mentioned above, Western analysis also identified a low molecular weight TGF-β3 isoform that was consistently found in the leiomyoma samples. There are precedents for a switch in TGF-β isoform expression in a variety of fibrotic diseases (75, 76) and cancer (4). Because the promoter regions for the TGF-β isoform genes are very different, the functional redundancy of TGF-β is ensured by the presence of a variety of gene transactivators that respond to changes in the intracellular milieu. Interestingly, a 5′ truncated version of TGF-β3 mRNA with higher translational efficiency, driven by methylation-specific regulation of alternative promoters for TGF-β3, was previously observed in a number of human breast cancer cell lines (77, 78). Therefore, whereas it is reasonable to consider that in this model the lower molecular weight form may represent a tumor-associated form of TGF-β3, confirmation of this as well as a determination of function will require further study.

Currently, the only medicinal therapy for leiomyomas is gonadotropin-releasing hormone agonists, which work by shutting down the entire reproductive axis. These agonists are efficacious at abrogating both bleeding and size-related symptoms, but the hypoestrogenic hormonal milieu induced by these drugs produces such significant side effects that therapy cannot be extended beyond 6 months. Gonadotropin-releasing hormone agonists also inhibit TGF-β expression, and the reduced expression of this cytokine may contribute to tumor shrinkage via reduction of the extracellular matrix component. However, due to the negative health effect of gonadotropin-releasing hormone therapy, particularly drug-induced menopause due to disruption of the hypothalamic-pituitary axis, there is still a need for the development of new medicinal therapies for this disease. Thus, direct inhibition of TGF-β signaling, without disruption of the hypothalamic-pituitary axis seems to be an optimal candidate approach. Indeed in vitro experiments using other ALK5 inhibitors have shown potent antitumor effects. SB-431542 inhibited the tumor-promoting effects of TGF-β in cancer cell lines including TGF-β–induced epithelial to mesenchymal transition, migration, invasion as well as vascular endothelial growth factor production (79). The preclinical data presented here, that ALK5 inhibition was very effective in reducing the incidence and multiplicity of uterine tumors, indicate that TGF-β signaling is a rational target for this disease.

Our data also indicate that caution must be used when considering TGF-β inhibition as a systemic therapy. In contrast to previous in vivo studies using cell lines that are refractory to the growth-inhibitory effects of TGF-β (27, 80–84), we found that systemic blockade of TGF-βR signaling exacerbated the growth of de novo epithelial tumors in the kidney. It is now understood that TGF-β signaling can switch from growth-inhibitory to oncogenic during the progression of epithelial tumors via paracrine effects on stromal cells, stimulation of angiogenesis, and immune suppression by this cytokine (4, 21, 85–87). In the case of cells that have acquired resistance to TGF-β growth inhibition, systemic blockade of TGF-β signaling can inhibit tumor growth and metastasis (27, 80–84). Clearly, the present results show an increase in renal epithelial proliferation, decreased apoptosis, and enhanced development of RCC when the TGF-β pathway is inhibited with SB-525334 in Eker rats that are predisposed to develop these tumors. This suggests that ALK5 inhibition may also carry the risk of promoting the early growth of epithelial lesions which have not acquired resistance to the growth-inhibitory effects of this cytokine. In addition, it is important to consider that whereas many studies using antibodies directed against TGF-β are encouraging with respect to blocking both fibrosis and cancer metastasis in other disease models, inhibition of ALK5 directly might have different effects. For example, an antibody against TGF-β would not silence the basal kinase activity of ALK5 that may occur in the absence of ligand, and minimal cellular activity of SMADs may still be possible. In contrast, an ALK5 kinase inhibitor would block basal activity and has the potential, at high doses, to more effectively shut down TGF-β signaling. Therefore, ALK5 inhibitors such as SB-525334 might more effectively block the antiproliferative effects of TGF-β on epithelial cells and thereby allow epithelial neoplasms to escape growth inhibition.

The results obtained with SB-525334 in the Eker rat model indicate that blocking ALK5 activity will promote primary tumor formation when cells are initiated by genetic predisposition in animals with a high propensity to develop TGF-β–sensitive epithelial lesions. Indeed, it has been shown that decreased responsiveness to TGF-β is unable to initiate tumorigenesis without a prevailing oncogenic lesion (81). It should also be noted that this is the first ALK5 kinase inhibitor to be evaluated in this tumor model. Thus, there could be a compound-specific off-target renal effect unrelated to the ALK5 kinase activity that is interacting with the proliferative aspects of TGF-β blockade. In the future, these findings should be confirmed in other models with additional compounds to determine whether and how TGF-β signaling blockade increases the risk of epithelial neoplasms.

In summary, inhibitors of ALK5 have the potential to be efficacious, but may well carry an epithelial cancer liability. Our data suggest that the beneficial versus deleterious effects of inhibition of TGF-β signaling may be tumor-specific and support the general concept that TGF-β stimulates the growth of mesenchymal cells while suppressing the growth of epithelial lesions. In the example reported here, the ALK5 inhibitor SB-525334 was efficacious for mesenchymal uterine leiomyoma, but promoted the development of epithelial tumors in the kidney. In the future, the challenge in using therapeutics that block TGF-β activity through ALK5 will lie not only in designing specific inhibitors, but also in striking a balance between beneficial and deleterious cancer outcomes.

We thank T. Berry, R. Mirabile, and J. Blando for excellent technical assistance and S. Henninger for manuscript preparation.