Transgenic expression of dominant negative tuberin through a strong constitutive promoter results in a tissue-specific tuberous sclerosis phenotype in the skin and brain.

Tuberous sclerosis (TS) is a common autosomal dominant disorder caused by loss or malfunction of hamartin (tsc1) or tuberin (tsc2). Many lesions in TS do not demonstrate loss of heterozygosity for these genes, implying that dominant negative forms of these genes may account for some hamartomas and neoplasms in TS. To test this hypothesis, we expressed a dominant negative allele of tuberin (DeltaRG) behind the cytomegalovirus promoter in NIH3T3 cells and transgenic mice. This allele binds hamartin but has a deletion in the C terminus of tuberin, leading to constitutive activation of rap1 and rab5/rabaptin. Expression of DeltaRG in NIH3T3 cells led to a strong induction of reactive oxygen species, induction of vascular endothelial growth factor, and malignant transformation in vivo. Expression of DeltaRG driven by the constitutive cytomegalovirus promoter led to high level expression in all murine tissues examined, including skin, kidney, liver, and brain. Surprisingly, mice expressing the DeltaRG transgene developed a fibrovascular collagenoma in the dermis, which closely resembles the Shagreen patch observed in human patients with TS. In addition, numerous small subpial collections of external granule cells in the cerebellum were observed, which may be the murine equivalent of subependymal giant cell astrocytomas or tubers commonly seen in TS patients. Thus, expression of a dominant negative tuberin in multiple tissues can lead to a tissue-specific phenotype resembling some of the findings in human TS. Our data are the first to demonstrate that specific signaling abnormalities underlie specific hamartomas in a model of a human genetic disorder.


Tuberous sclerosis (TS) is a common autosomal dominant disorder caused by loss or malfunction of hamartin (tsc1) or tuberin (tsc2). Many lesions in TS do not
demonstrate loss of heterozygosity for these genes, implying that dominant negative forms of these genes may account for some hamartomas and neoplasms in TS. To test this hypothesis, we expressed a dominant negative allele of tuberin (⌬RG) behind the cytomegalovirus promoter in NIH3T3 cells and transgenic mice. This allele binds hamartin but has a deletion in the C terminus of tuberin, leading to constitutive activation of rap1 and rab5/rabaptin. Expression of ⌬RG in NIH3T3 cells led to a strong induction of reactive oxygen species, induction of vascular endothelial growth factor, and malignant transformation in vivo. Expression of ⌬RG driven by the constitutive cytomegalovirus promoter led to high level expression in all murine tissues examined, including skin, kidney, liver, and brain. Surprisingly, mice expressing the ⌬RG transgene developed a fibrovascular collagenoma in the dermis, which closely resembles the Shagreen patch observed in human patients with TS. In addition, numerous small subpial collections of external granule cells in the cerebellum were observed, which may be the murine equivalent of subependymal giant cell astrocytomas or tubers commonly seen in TS patients. Thus, expression of a dominant negative tuberin in multiple tissues can lead to a tissue-specific phenotype resembling some of the findings in human TS. Our data are the first to demonstrate that specific signaling abnormalities underlie specific hamartomas in a model of a human genetic disorder.
Tuberous sclerosis is an autosomal dominant disorder that causes morbidity and mortality because of seizures, mental retardation, hamartomas, and benign and malignant neoplasms (1)(2)(3). The hamartomas found in TS 1 include collageno-mas, which can infiltrate skin and muscle, as well as abnormal giant neurons, which calcify and cause seizures (tubers) (4 -6). Various types of hamartomas including cortical tubers and brain neoplasms can cause seizures or compress vital brain structures. Neoplasms of the kidney can cause renal failure, pain, and life-threatening hemorrhage, and renal neoplasms remain the major cause of morbidity and mortality in adults with TS. No medical therapy exists to slow the progression of these hamartomas and tumors or to shrink them once they become problematic. A common feature of all TS associated hamartomas and benign and malignant tumors is that these tumors are highly angiogenic. Indeed, our laboratory has demonstrated that these lesions produce the angiogenic stimulator vascular endothelial growth factor (VEGF) (7).
Although tuberin and hamartin deletions have been generated in mice, these models do not fully recapitulate human TS. Notably, whereas neoplasms occur commonly in these mice, hamartomas have not been described (8,9). These hamartomas include human skin lesions, including collagenomas and angiofibromas as well as certain brain lesions such as the abnormalities in neural migration that result in tubers and subependymal giant cell astrocytomas. Second, no phenotype/genotype correlation has been established between mutations in tuberin and hamartin and human phenotypes. However, loss of heterozygosity (LOH) is most common in neoplasms such as angiomyolipoma, lymphangiomyomatosis, and renal cell carcinoma but much less common in hamartoma tissue even when neoplasms and hamartomas coexist in the same patient (10,11). A potential reason for this lack of correlation is that LOH of these genes may result in a wide variety of signaling abnormalities in differing tissues, thus making it difficult to assign a phenotype to a specific signaling abnormality. In this study, we study the effect of a tuberin mutant (12) in a transgenic mouse when the transgene is expressed behind the cytomegalovirus promoter. Our transgenic mouse model directly addresses the contribution of a specific signaling abnormality to the TS phenotype in vivo. We demonstrate that dominant negative tuberin can induce signaling abnormalities in cultured cells and induce development of hamartomas in transgenic mice. Notably, these hamartomas occur in a tissue-specific manner even when the transgene is ubiquitously expressed in all tissues.

MATERIALS AND METHODS
Plasmids and Cell Lines-The plasmid pCMV⌬RG was obtained from Loren Fields (Indiana University, Indianapolis, IN). This plasmid was transfected into NIH3T3 cells using Lipofectin and selected with G418. Resistant colonies were pooled and labeled 3T3⌬RG. pCMV⌬RG was linearized with BgII and XbaI, gel-purified, and injected into mouse embryos.
Reactive Oxygen Generation-Confluent cells in 100-mm dishes were washed with 6 ml of Hanks' balanced salt solution and released by using 0.25% trypsin (w/v)/1 mM EDTA followed by the addition of 2 ml of Hanks' balanced salt solution. After pelleting, the cells were resuspended in 1 ml of Hanks' balanced salt solution. Dichlorofluorescein diacetate was added to a final concentration of 2 M and incubated for 1 h in the dark at room temperature. Dichlorofluorescein (DCF) fluorescence was determined by using a FACS from BD Biosciences (excitation wavelength, 488 nm; emission wavelength, 515-545 nm).
Reverse Trancription-PCR for VEGF-Total RNA was isolated using TRI reagent (Sigma). Reverse transcription-PCR was conducted with the Promega accession reverse transcription-PCR kit. Primers used were as follows: actin (728 bp), forward 5Ј-AAG ATG ACC CAG ATC ATG TTT GAG AC-3Ј and reverse 5Ј-CTG CTT GCT GAT CCA CAT CTG CTG G-3Ј; VEGF (445 bp), forward 5Ј-TCA TGC GGA TCA AAC CTC ACC AA-3Ј and reverse 5Ј-TCT CGC CCT CCG GAC CCA AAG T-3Ј. Reactions were performed in an Eppendorf master cycler. One PCR cycle at 45 C o for 45 min and 94 C o for 2 min followed by 40 PCR cycles under standard conditions with an annealing temperature of 60°C were performed. ␤-Actin mRNA was used as a reference message to normalize the content of total RNA. VEGF expression was calculated as the relative expression ratio to that of ␤-actin. All reactions were carried out in triplicate. Quantification was determined by triplicate repeats.
Generation of Transgenic Mice-Transgenic mice were produced by standard mechanisms. Briefly, the ⌬RG transgene (see Fig. 5) was purified using a Qiaex II extraction kit (Qiagen), resuspended in injection buffer at 2 ng/l, and microinjected into C57BL/6 fertilized embryos. Embryos were transferred into ICR recipient females, and resultant pups were screened for the presence of the transgene by both PCR (see Fig. 6) and Southern blot (data not shown). Fig. 6 shows the screening methodology and representative results for PCR screening of the transgenic mice. Primers TSC2F (GTCCAGGAGAGACTCAGGT-GCCAGT) and TSC2R (CTGTAAGGTCTGCAACTCCGGAGAA) span intron 8 and thus give a wild-type band of 568 bp and a transgene specific band of 281 bp) Three independent founder mice were identified (#922,928,945) and established as independent lines.
Histology and Immunohistochemistry-Formalin fixed, paraffin embedded sections were prepared on poly-L-lysine coated slides. To block the endogenous peroxidase activity, the sections were treated with methanol containing 0.3% hydrogen peroxide for 15 min and then washed in PBS. They were stained using a standard avidin-biotin peroxidase kit (Nichirei Co., Tokyo, Japan) with antibodies against MCP-1 (Genzyme Techne, Minneapolis, MN) and anti-CCR-2B (see Fig.  8) (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) (both diluted in PBS, 1:250) (13). Subsequently, they were incubated with biotinylated secondary antibody followed by the streptavidin-peroxidase treatment. The sections were developed with 3,3Ј-diaminobenzidine solution as chromogen and then counterstained with hematoxylin and dehydrated, cleared, and mounted. Negative controls were prepared by omitting the primary antibodies and by their substitution with corresponding IgG at the dilution used for the specific antibodies in this study.
Primary Single Muscle Fiber Isolation and Culture-Mice were euthanized by CO 2 inhalation. The skin was prepared with 70% ethanol, and the legs were skinned and placed into room temperature Dulbecco's modified Eagle's medium. Muscles were dissected out and placed into collagenase type I (Worthington) 400 units in 10 ml of Dulbecco's modified Eagle's medium and digested at 37°C for 70 min. Individual live muscle fibers were collected under a dissecting microscope using polished glass pipettes and then cultured in Dulbecco's modified Eagle's medium ϩ 20% fetal bovine serum supplemented with antibiotics and glutamine.
Satellite HCl for 10 min, washed with PBS, then treated with 1% Triton-X in PBS for 4 min and washed again with PBS. Primary anti-bromodeoxyuridine antibody (rat monoclonal, Harlan Sera Labs) was applied in a 1:10 dilution at 4°C overnight. After PBS washes, the secondary antibody donkey anti-rat (Alexa Fluor 594, Molecular Probes) was incubated for 1 h at room temperature at 1:400 dilution. After PBS washes, the fibers were mounted using VectaShield with 4Ј,6-diamidino-2-phenylindole (Vector Labs, Burlingame CA). Fluorescent, stained satellite cells were counted as a function of unit length of fiber (one 200X field diameter) (14).
Satellite Cell Apoptosis-After 72 h in culture, fibers were fixed in 4% paraformaldehyde as above, washed in PBS, then permeabilized with 0.1 M sodium citrate ϩ 0.1% Triton-X on ice for 2 min. Terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling was then performed with antibody to the nick end DNA conjugated to fluorescein isothiocyanate (Roche Applied Science) for 1 h at 37°C. Fibers were mounted, and stained fluorescent satellite cells were counted as for proliferation studies.

RESULTS
The ⌬RG allele of tuberin was introduced into NIH3T3 fibroblasts to assess signaling abnormalities. Because various oncogenes have been shown to induce reactive oxygen, and reactive oxygen has been shown to contribute to cellular transformation (15)(16)(17), we assessed the effect of this mutant tuberin allele on generation of reactive oxygen. The ⌬RG allele of tuberin caused a significant increase in intracellular hydrogen peroxide as shown by DCF fluorescence (Fig. 1).
We have shown previously that PDGF-BB plays an important role in the pathogenesis of tuberous sclerosis lesions. Thus, we assessed the effect of the ⌬RG allele on response to exogenous PDGF-BB. Cells expressing ⌬RG or parental NIH3T3 were exposed to PDGF-BB. Cells expressing ⌬RG showed an increase in phosphorylation of a 185-kB protein consistent with PDGFR␤ ( Fig. 2A). In addition, cells expressing ⌬RG had increased levels of phosphorylation of akt at baseline and enhanced phosphorylation of akt in response to PDGF-BB (Fig. 2B).
Loss of tuberin has been associated with activation of protein synthesis, namely through the activation of the ribosomal proteins S 6 and S 6 kinase. Similarly, expression of the ⌬RG mutant tuberin led to activation of S 6 and S 6 kinase (Fig. 3). Expression of the ⌬RG allele also caused a strong induction of the angiogenic factor VEGF (Fig. 4).
Three lines of transgenic mice were established (922, 928, and 945). Integration of the mutant allele was confirmed by PCR using primers flanking exons 8 and 9 of tuberin. Genomic DNA gives a band of 567 bp, whereas transgene DNA (cDNA) gives a 281-bp band (Figs. 5 and 6). Among the seven 922 mice examined, five showed aggregates of granular cells at the surface of the cerebellum immediately underneath the pia. Aggregates ranged in size from 10 -25 cells in histologic cross sec-tions and took the form of linear arrangements along the surface of the molecular layer or as small nodules that extended into the subarachnoid space. Cells with aggregates had the morphology of mature granular cells with no evidence of cytologic atypia (Fig. 7, A-D). Solitary granular cells and small clusters (3-4 cells) were also noted at the surface of the cerebellum in similar frequency as wild type controls. In the seven 945 mice examined, three brains were noted to have similar aggregates of granular cells at the cerebellar surface, and the range of sizes was slightly smaller (10 -15 cells). Among the eight age-matched wild type mice, the cerebellum showed only rare individual cells or small clusters at the pial surface. In no cases did the aggregates extend outward into the subarachnoid space. No developmental or dysplastic changes resembling cortical tubers were noted in the cortex of cerebral hemispheres. No subependymal dysplastic growths were noted in along the ventricular system. No neoplastic growths were noted within the lateral ventricles of any brains.
The dermis and perimuscular area were remarkable for thickened muscle and a myxoid accumulation of cells in a perimuscular area, which was rich in mast cells (Fig. 7, E-G). These lesions histologically resemble collagenomas (Shagreen patches), which are present in a significant number of patients with tuberous sclerosis. These lesions are present in all of the transgenic mice but not in the wild type littermates. To assess whether the transgene resulted in increased proliferation or apoptosis of muscle satellite cells, we performed proliferation and apoptosis assays on wild type and transgenic mice. No differences in proliferation or apoptosis were observed, suggesting that these do not represent accumulations of muscle satellite cells (data not shown).
Monocyte chemotactic factor 1 (MCP-1) has been associated with various mesenchymal proliferations, including cirrhosis of the liver. Recently, MCP-1 has been shown to be up-regulated in human lesions of TS. 2 We examined the expression of MCP-1 and its receptor, CCR-2, and found that the expression of both of these genes is elevated in the dermal lesions of the transgenic mice but not in wild type mice (Fig. 8). DISCUSSION Tuberous sclerosis is a common autosomal dominant disorder characterized by an increased incidence of benign and malignant tumors as well as developmental hamartomas. In childhood, the major cause of morbidity and mortality are cerebral tubers, which histologically are giant cells with a tendency to calcify and become foci for seizures. Subependymal giant cell astrocytomas appear in a periventricular distribution and may cause ventricular obstruction through continued growth.
Lesions in TS can be divided into those that usually show LOH for tsc1/tsc2 and those that do not. Angiomyolipomas and lymphagiomyomatosis, clonal neoplasms with features of smooth muscle, fat, and melanocytes, commonly exhibit LOH (10,18). On the other hand, cerebral tubers usually do not exhibit LOH (19). We propose two potential explanations for the development of lesions in TS in the absence of LOH. First, we propose that certain mutations in tsc2 can act as dominant negatives and acquire novel functions through the activation of aberrant signaling pathways (12). A consequence of this is heightened sensitivity of certain cells to physiologic levels of growth factors. We and others have previously demonstrated that TS model cells are highly sensitive to PDGF (20,21). A second mechanism that we have observed is LOH of other tumor suppressor genes, and we have previously observed LOH of p16ink4a in a tumor arising in a mouse heterozygous for tuberin (22). Interestingly, although this tumor had lost p16ink4a, it retained a wild type tsc2 gene (22).
Recently, several other potential targets of tuberin have been discovered. Several groups have found that tuberin serves as an inhibitor of phosphoinositol-3 kinase/akt signaling, and overexpression of tuberin leads to decreased phosphorylation of 2 T. Darling, personal communication.  Fig. 6B shows MCP-1 staining in wild type skin. Fig. 8, B and C, show low power views of transgenic and wild type skin. Note the thickened muscle layers and additional cellularity in the transgenic mouse (C) compared with the wild type mouse (D). Fig. 6, E and F, represent CCR2 staining of transgenic (Fig. 8E) and wild type (F), respectively. The staining for CCR2 appears predominantly on mast cells (E).
akt (23)(24)(25). In addition, overexpression of tuberin has been shown to inhibit signaling through pathways that inhibit protein processing, such as target of rapamycin, ribosomal protein S 6, and S 6 kinase (Fig. 3) (25)(26)(27). These proteins are involved in the translation of proteins involved in cell growth, such as VEGF (Fig. 4) (28). However, systemic long term use of rapamycin in humans may not be an optimal therapy for TS, as rapamycin is a potent immunosuppressant and may lead to severe infections and neoplasms because of loss of immune surveillance (29). Other targets of tuberin have been proposed. In a study of Eker rat cells containing a tetracycline inducible tuberin, gene chip array has found several genes induced, including thiol-specific antioxidant and glutathione peroxidase, indicating that tuberin may protect cells against oxidative stress, and cells lacking tuberin or may have increased reactive oxygen species (30). We have previously shown that reactive oxygen species have been shown to contribute to mitogenesis and tumorigenesis (31).
In normal cerebellar development, external granular cells at the cerebellar surface are a proliferative population that eventually migrates inwardly to populate the internal granular cell layer. After this process early in life, the external granular layer is largely depleted with only scattered remnant granular cells. In the transgenic animals in this study, we found numerous microscopic aggregates of mature granular cells in the subpial region suggesting that their normal migration inward may have been disrupted. This particular phenotype is not encountered in the cerebellum of tuberous sclerosis patients. However, the cortical tubers and subependymal nodules that characterize this disease in the cerebral hemispheres are architecturally abnormal collections of neuronal and glial elements that are most likely caused by disordered migration and development. Subpial cerebellar aggregates of granular cells may be the murine manifestation of this disordered developmental program. No other central nervous system stigmata of tuberous sclerosis were noted in transgenic mice, and no neurologic symptoms could be ascribed to the cerebellar changes.
We have demonstrated that mutant tuberin can cause increased levels of intracellular reactive oxygen. Reactive oxygen has been implicated in the induction of fibrogenic cytokines, such as MCP-1, and increased expression of MCP-1 and its receptor CCR2 have been observed in fibrotic processes associated with reactive oxygen, such as bleomycin-induced scleroderma, idiopathic scleroderma, and hepatic fibrosis/cirrhosis (32)(33)(34)(35). Lesional skin from transgenic mice demonstrate elevated levels of MCP-1 compared with wild type littermate skin, suggesting that MCP-1 may play a pathogenic role in TS lesions. Interestingly, PDGF is a known stimulant of MCP-1, and the introduction of the ⌬RG allele of tuberin results in increased responsiveness to PDGF (35). This increased sensitivity could underlie in part the increased production of MCP-1 in our transgenic mice. Inhibition of MCP-1/CCR-2 interactions may be beneficial in the treatment of tuberous sclerosis, and mice deficient in either of these genes have a decreased fibrotic response to various experimental stimuli (32,36).
Prior studies have addressed signaling abnormalities in the absence of tsc2. However, none of these studies have addressed signaling abnormalities seen in the presence of a dominant negative tuberin. In this study, we show that a dominant negative tuberin can recapitulate many of the signaling abnormalities observed because of loss of tuberin. In vivo expression of mutant tuberin results in a unique phenotype not previously observed in nonhuman tuberous sclerosis models. Finally, we show that constitutive expression of a dominant negative gene leads to a tissue-specific phenotype resembling the human disease. Our model may be valuable in assessing contributions to tissue specificity and may prove useful in pharmacologic studies to prevent or treat tuberous sclerosis.