Mesenchymal Tumorigenesis Driven by TSC2 Haploinsufficiency Requires HMGA2 and Is Independent of mTOR Pathway Activation

TSC is regarded as a tumor suppressor gene syndrome that affects an estimated 1 in 6,000 people (1, 2). It is associated with the widespread development of mesenchymal tumors including renal angiomyolipomas (AML; ref. 3) and pulmonary lymphangiomyomatosis (LAM). Other clinical manifestations of TSC include epilepsy, neurocognitive dysfunction, and autism (1). Approximately 26% to 39% of women with TSC will develop the lung manifestation of TSC, LAM, which is a progressive lung disease that occurs in women of childbearing age and is characterized by the abnormal proliferation of smooth muscle-like cells in the airways, vasculature, and lymphatics of the lung (4, 5).

Mutations in TSC genes, which are regarded to be classical tumor suppressor genes, were identified from the familial form of the disease in both the TSC1 gene on chromosome 9q34 (6) or the TSC2 gene on chromosome 16p13 (7). The proteins of each TSC gene (TSC1, hamartin and TSC2, tuberin) bind to each other to form a functional heterodimer that acts as the tumor suppressor (8). TSC occurs as both a familial and sporadic condition but the majority of cases arise sporadically with a 3- to 4-fold greater prevalence of TSC2 mutations occurring compared with TSC1 (9). Biallelic inactivation (the “two-hit” hypothesis) of either one of the TSC genes activates its downstream target Rheb, which stimulates the phosphorylation and activation of the mTOR complex (mTORC1; refs. 10, 11). mTORC1 is known to regulate cell growth, proliferation, translation, and autophagy (12, 13) and its activation is said to be central to the tumorigenic pathway in tuberous sclerosis.

A major molecular player in mesenchymal tumorigenesis is the architectural transcription factor High Mobility Group A2 (HMGA2; ref. 14). HMGA2 is primarily expressed in the undifferentiated mesenchyme (15, 16) and is absent in normal adult tissues (17). HMGA2 is central to the processes of mesenchymal differentiation and proliferation and initiates mesenchymal tumorigenesis (18, 19) when misexpressed in a number of human mesenchymal tumors of a differentiated phenotype (18). The causality of HMGA2 expression in tumorigenesis was demonstrated in vivo in the development of mesenchymal tumors in transgenic mice (20). As a transcription factor, downstream targets of HMGA2 have been identified in mesenchymal tumors (21), with IGF2BP2 (IGFII-mRNA binding protein) being the most well-documented target. Importantly for this study, HMGA2 expression has been observed in LAM (22).

The TSC2 disease mouse model exhibits a phenotype similar to the human disease, with the Tsc2^+/− mice displaying a wide spectrum of mesenchymal tissue tumor phenotypes by 15 months of age (including liver, lungs, and extremities; refs. 23–25). The tumor phenotype is revealed in the heterozygous Tsc2^+/− mice since the Tsc2-null embryos die prenatally, (23, 24). Significantly, Tsc2^+/− rodents have been used as preclinical models (26) and have led to newly developed human trials testing mTOR inhibitor therapies with the assumption that the mTORC1 pathway is activated (27) in TSC2-associated diseases. The combination of mesenchymal tumors in TSC, the critical role of HMGA2 in mesenchymal tumorigenesis, the presence of HMGA2 expression in LAM, and the availability of both Hmga2 (28) and Tsc2 (24) genetic mouse models, led to the following experiments examining the HMGA2 pathway in tumor pathogenesis in TSC.

LAM (n = 29) and AML (n = 12) samples imbedded in paraffin were used in this study. The use of these samples was approved by the Institutional Review Board of The University of Medicine and Dentistry of New Jersey (Newark, NJ).

Tsc2^+/− mice (C57Bl/129SvJ; ref. 24) and Hmga2^+/− mice were crossed to produce double-heterozygous mice (Tsc2^+/−, Hmga2^+/−), which were then intercrossed to produce double-homozygous HMGA2 mice (Tsc2^+/−, Hmga2^−/−). The study population was produced from the double heterozygote cross utilizing littermate controls. Mice were fully characterized pathologically with no differences observed from the published phenotype (24, 25). The following experimental groups were used throughout the study: Tsc2^+/− Hmga2^+/+, Tsc2^+/− Hmga2^+/− and Tsc2^+/− Hmga2^−/−. Mice were genotyped first for the Hmga2 locus and then for the Tsc2 locus. Tail snips were used to prepare DNA for PCR genotyping analysis, as previously described (24, 28). Both male and female mice were used for subsequent analysis. A group of mice from each cohort were randomly sacrificed at the end of the experimental period every four weeks up to 16 months.

Mice were monitored for tumors over a period of 16 months. Randomly selected cohorts of six mice from each genotype were sacrificed every four weeks, starting at 3 months age, to macroscopically examine appearance of tumors. No tumors were observed in any genotype prior to animals turning 9 months old. Tumor tissue was analyzed from mice sacrificed at 16 months of age and step sections were performed and stained with hematoxylin and eosin (Fisher Scientific) for histochemical examination for the presence of tumors.

Total RNA was isolated from tissue using the RNeasy Mini Kit (Qiagen) and reversed transcribed using the TaqMan Reverse Transcriptase Kit (Applied Biosystems). Detection and expression of genes were accessed by real-time PCR using SYBR green fluorescence and the following primers: Hmga2 5′-CAGCAGCAAGAGCCAACCTG-3′ (forward) and 5′-TGTTGTGGCCATTTCCTAGGT-3′ (reverse); IGF2BP2 5′-ACGTGGAGCAAGTCAACACAG-3′ (forward) and 5′-ACTGATGCCCACTGAGCTTCTC-3′ (reverse); glyceraldehyde-3-phosphate dehydrogenase (G3PDH) 5′-AAGCCCATCACCATCTTCCAG-3′ (forward) and 5′-CCTTCTCCATGGTGGTGAAGAC-3′ (reverse). The 7800HT real-time sequence detection system (Applied Biosystems) was used along with the GeneAmp 7900 SDS software and converted into threshold cycle (ΔΔC_t) values.

Immunoreactivity assays were performed on paraffin-embedded samples (6 μm), as previously described (22). Sections were stained with polyclonal anti-HMGA2 (16), anti-IGF2BP2 (ProteinTech Group; 11601-1-AP), anti-tuberin (Santa Cruz Biotechnology; sc-893), anti-pho-mTOR (Ser2448; Cell Signaling Technology; #2971) and anti-pho-p70 S6 kinase (T389; Abcam; ab32359) antibodies all raised in rabbits. Normal tissue or isotype control rabbit IgG were used as negative controls in each assay. In addition, Western blot analysis was performed to verify antibody specification with only the expected protein size observed with the tuberin and pho-p70 S6 kinase antibodies, both for wild-type mouse lung homogenate and for ELT3 cells (uterine leiomyoma cells derived from the Eker rat; Supplementary Fig. S2B and S2D; ref. 29). The tuberin antibody is raised against a peptide mapping the C-terminus region of human tuberin. The specificity of the tuberin peptide antibody interaction was verified by peptide-competitive assays, resulting in no detectable signaling (Fig. 5). In addition, to further verify the specificity of the C-20 tuberin antibody used throughout this study, IHC was performed on LAM samples with two additional antibodies [Abcam anti-tuberin (EP1107Y; ab52936) and Abcam anti-tuberin (Y320; ab32554); a positive tuberin staining for the Y320 antibody is shown in Supplementary Fig. S2C]. These antibodies recognize residues of human tuberin or C-terminus of human tuberin. Immunoblots were also performed on mouse tissue to confirm specificity of the C-20 tuberin antibody.

Microdissected tissue was homogenized in protein lysis buffer (50 mmol/L HEPES, pH 7.5, 150 mmol/L NaCl, 1% Triton X-100, 1% glycerol, 1 mmol/L EDTA, 10mmol/L NaF, 2 μg/mL leupeptin, 1 μg/mL pepstatin A, 10 mmol/L Na₃VO₄, and 1 mmol/L phenylmethylsulfonyl fluoride) and 14.5 μg of protein was separated on 5%, 10%, and 15% SDS-polyacrylamide gels and transferred to nitrocellulose membranes. Rabbit antibodies against tuberin (Santa Cruz Biotechnology), anti-pho-p70 S6 kinase (T389; Abcam), anti-S6K (E175; Abcam), HMGA2 (16), tubulin and Actin (Santa Cruz Biotechnology; sc-1616), were detected with enhanced chemiluminescence reagents (Pierce) as previously described (30).

Mouse Tsc2 LOH analysis was performed on tissue from Tsc2 heterozygote mice, as described by Onda and colleagues (24), after microdissection of paraffin-embedded sections under 100× magnification. When the wild-type allele was present at less than 50% of the mutant allele, this was judged as evidence of LOH. Human TSC2 LOH analysis was undertaken by combining methodologies from Smolarek and colleagues (31) and Lesma and colleagues (32). Microdissected areas of tissue were verified for tuberin immunoreactivity from serial sections. Microsatellite markers, Kg8, D16S283, D16S291, and D16S525 that are within 600 kb of TSC2 locus on chromosome 16p13.3 were examined (33, 34). PCR was performed with 6-FAM–labeled sense primers (Invitrogen). PCR was performed as described by Lesma and colleagues (32). The PCR products were purified using SigmaSpin Post-Reaction Clean-up Columns (Sigma-Aldrich) and underwent fragment analysis (Genewiz). Samples were examined using the program peak scanner (ABI) to view amplicant peaks. All analysis was repeated at least twice for confirmation.

Data evaluation was carried out using the ABI Prism sequence detection system and Microsoft Excel. All data in the text and figures are expressed as mean ± SEM. and were analyzed using a two-tailed t test. A P value of less than 0.05 was considered significant. The effect of the Hmga2 deficiency on the tumor incidence in the Tsc2^+/− mice was analyzed by the Mann–Whitney U nonparametric statistical tests.

Initially, Hmga2 expression was assessed in the various tumors of Tsc2^+/− mice and misexpression of Hmga2 observed in 100% of the tumors (n = 227) isolated from Tsc2^+/− mice (n = 45) at 16 months of age (Fig. 1A and B). Hmga2 misexpression was also observed in the kidney of Tsc2^+/−Hmga2^+/+ mice as early as 12 months of age, with levels of expression increasing up to 16 months (Fig. 1B), correlating with an age dependent increase in tumor severity (15, 25, 35). Furthermore, HMGA2 immunoreactivity was detected and restricted to the nucleus of kidney tumor cells (Supplementary Fig. S1A). Similarly, the mesenchymal tumor cells of the lung and extremity hemangiosarcoma (ExH) from the Tsc2^+/−Hmga2^+/+ mice also expressed HMGA2 (Fig. 1A and Supplementary Fig. S1B). As expected, HMGA2 was not detected in normal mouse adult tissues (Supplementary Fig. S1A; ref. 17). Consistent with the mRNA expression, the HMGA2 protein was detected in kidney, liver, and lung tumors of Tsc2^+/− mice (Fig. 1C). Specific expression for Igf2bp2 was observed in the tumors of the Tsc2^+/−Hmga2^+/+ mice with a total absence of statistically significant expression of Igf2bp1 and Igf2bp3 (the other two members of the IGF2BP family) in all of the tumor cell types as compared with their wild-type counterpart tissue (Fig. 1D). As was seen for HMGA2, Igf2bp2 expression increased in renal tumors of increasing age and severity (Fig. 1E). Importantly, Igf2bp2 expression was directly proportional to Hmga2 expression in renal tumors and was therefore not observed in the renal tumors of Tsc2^+/−Hmga2^−/− by qPCR (Fig. 1F) and immunohistochemistry (Supplementary Fig. S1A). Immunoreactivity for IGF2BP2 was observed in the cytoplasm of HMGA2-positive tumor cells from Tsc2^+/−Hmga2^+/+ mice (Fig. 1G). In conclusion, HMGA2 and IGF2BP2 are expressed in the tumors that develop in the Tsc2 heterozygous mouse.

To investigate the human relevance of the mouse studies, the expression of HMGA2 and IGF2BP2 was analyzed in the human mesenchymal LAM and AML tumor samples (n = 29 and n = 12 for each tumor group, respectively). HMGA2 (as seen in a previously described separate cohort of 21 LAM samples; ref. 22) and IGF2BP2 expression was observed in all of the LAM samples analyzed (Fig. 2; Table 1). Similarly, all 12 AML samples were found to express HMGA2 and IGF2BP2 (Fig. 2; Table 1). Therefore, consistent with the mouse studies, all of the human tumor samples also expressed HMGA2.

Next, the potential genetic interaction between the Tsc2 pathway and Hmga2 expression in tumor pathogenesis was examined in vivo by the generation of mice of various Hmga2 genotypes on a Tsc2 heterozygous background. In the TSC2 heterozygous background and the HMGA2 wild-type or heterozygous genotype, these cohorts revealed a tumor incidence at a comparable frequency as in the original published studies (23, 24) in various organs (Fig. 3A and B). As expected, mice wild type at both loci did not develop tumors during the course of this study. However, when the Tsc2^+/− mice were crossed into the HMGA2-null background, activation of HMGA2 was required in the majority of renal tumors. Despite the incidence of renal tumors (>80%) in the two mouse groups being similar, the number of tumors per mouse was substantially reduced in the Tsc2^+/−Hmga2^−/− mice (1.26 tumors/mouse) compared with Tsc2^+/−Hmga2^+/+ mice (4.63 tumors/mouse; P < 0.001; Fig. 3B). Consequently, there appears to be an HMGA2-independent pathway for a minority of kidney tumors in the Tsc2^+/− mice. Most dramatically, there was complete absence of the mesenchymal tumors (in the extra-renal sites) in the Tsc2^+/− mice on the Hmga2-null background (Fig. 3A). Hence, tumor pathogenesis in the extra-renal tumors has an absolute requirement for the expression of HMGA2 in the Tsc2^+/− mice (P = 0.0038).

The interaction between mTORC1 activity on HMGA2 expression was examined by treating tuberin deficient rat and human cells with rapamycin. HMGA2 was expressed in rat (ELT3 V3) and human cells deficient in tuberin (621-101), respectively (Fig. 3C and D). Rapamycin treatment, as expected, blocked phosphorylation of p70 S6K (Fig. 3E) but did not affect the level of expression of HMGA2 (Fig. 3C and D). These findings demonstrate that HMGA2, while expressed in the TSC-null cells, is independent of mTOR activity.

To further investigate the potential interaction between HMGA2 and the TSC2 pathway in tumor pathogenesis in TSC, the tuberin and mTOR profiles in the mouse mesenchymal tumors of the liver, lung, and ExH tumors in the Tsc2^+/− mice were determined. Surprisingly, all mouse mesenchymal tumors expressed tuberin as demonstrated by antibody reactivity towards the carboxy terminus of tuberin (Fig. 4A; Supplementary Table S1) and qPCR (Supplementary Fig. S2A). Because of these unanticipated results, IHC studies were performed with two additional antibodies against tuberin and the findings were confirmed. Furthermore, Western blot analysis demonstrated the specificity of the antibodies by exhibiting a single molecular species at the correct molecular weight (Supplementary Fig. S2B). Importantly, the downstream mTOR signaling pathway was not activated in the extra-renal mesenchymal tumors as demonstrated by an absence of immunoreactivity for phosphorylation of p70 S6 kinase and further confirmed by the phosphorylation status of mTOR (Fig. 4A; Supplementary Table S1) confirming the presence of functional tuberin. Following this surprising result, experiments also established at the DNA level that LOH was not observed in the extra-renal tumors (Fig. 4C).

The status of tuberin expression was next investigated in the renal tumors observed in the Tsc2^+/− mouse on all three Hmga2 genetic backgrounds. The results were similar, independent of the status of Hmga2 expression and will therefore be considered collectively. Tuberin expression was investigated by IHC (Fig. 4B; Supplementary Table S1) and observed in over 80% of the mouse tumors with mTOR pathway activation being observed in only 31% of the renal tumors. Western blot analysis was performed on renal tumor and nontumor tissue from Tsc2^+/− mice and renal tissue from wild-type mice to further demonstrate the presence of tuberin protein as well as the variation of mTOR pathway activation in the Tsc2^+/− tumor samples (Fig. 4D). Most importantly, Tsc2 LOH was performed and identified (Fig. 4E) only in those tumors that exhibited loss of tuberin immunohistochemistry and activation of the mTOR pathway (Fig. 4E and F).

Equally captivating, tuberin expression was also observed in all (29/29) human LAM lesions (Fig. 5A; Table 1) and confirmed by the use of an additional tuberin antibody (Supplementary Fig. S2C). Upon investigation of the mTOR pathway in the LAM lesions, it was observed that 15 of 29 (51.7%) LAM samples stained for the phosphorylation of p70 S6 kinase and mTOR (n = 29; Fig. 5A; Table 1). Most importantly, approximately 50% of the tumors do not exhibit altered mTOR pathway activation and thus alternative pathways to the activation of the mTOR pathway must play a major role in the pathogenesis of LAM tumorigenesis. Parallel results were obtained for the human kidney AMLs, tuberin expression being observed in 8 of 12 samples (Fig. 5B; Table 1) and the remaining four AMLs that did not express tuberin were shown to have LOH at the TSC2 locus (Supplementary Table S2; Fig. 4E and F).

This study has identified a novel molecular pathway in the tumor pathogenesis of TSC. One hundred percent of the human and mouse tumors were found to express HMGA2. Furthermore, the downstream target of HMGA2, IGF2BP2, was also activated in all of the TSC tumors demonstrating that the HMGA2 pathway is activated in the tumor pathogenesis of TSC. This series of experiments was extended by the genetic demonstration that there was a complete absence of mesenchymal tumors on the mouse Hmga2-null background. Hence, tumor pathogenesis in TSC mesenchymal tumors has an absolute requirement for the activation of the HMGA2 pathway in the Tsc2^+/− mice. Interestingly, although the renal tumors were substantially reduced in the Tsc2^+/−Hmga2^−/− mice these animals still developed a small number of tumors. This finding is most likely due to the presence of an HMGA2-independent pathway in a minority of the kidney tumors in the Tsc2^+/− mice, which has been seen in other studies (36). Furthermore, differences in the kidney may occur due to the described epithelial origin of the tumor (37) as opposed to the pure mesenchymal phenotype of the extra-renal tumors.

The upstream pathway and mechanism by which HMGA2 mediates abnormal growth is an active area of research. Misexpression of HMGA2 in uterine leiomyoma (38) is the result of a loss of the regulatory miRNA let-7–binding sites in the 3′ untranslated region of HMGA2 mRNA (38). Interestingly, the MAPK extracellular signal–regulated kinase (Erk) pathway can regulate let-7 miRNA (39) and the HMGA2 pathway is known to be activated by Erk (40). Downstream of HMGA2 expression, the HMGA2 protein functions in a variety of cellular processes, such as cell growth, transcription regulation, neoplastic transformation, and progression (36). Multiple studies have demonstrated that misexpression of HMGA2 in differentiated mesenchymal cells leads to abnormal growth (18, 20, 22). Dependent upon cellular context, HMGA2 can activate a number of genes involved in the cell cycle including cyclin A (41) or E2F1 (42). In addition, HMGA2 activates the DNA repair gene ERCC1, which can potentially protect cancer cells from DNA damage–inducing reagents, such as methyl methanesulphonate (43). The current study shows that IGF2BP2, a downstream target gene of HMGA2 and known oncofetal protein (44) is also expressed in 100% of the LAM and TSC lesions. Future studies will concentrate on delineating which of these above molecular pathways are activated by the absolute requirement of HMGA2 for the pathogenesis of mesenchymal tumor formation in a TSC2 haploinsufficient background.

As TSC2 is a tumor suppressor gene (45), for the past decade, the primary genetic hypothesis for the TSC syndrome that has been followed states that the major mechanism for tumor pathogenesis observed in multiple organs is caused by an initial mutation at the TSC locus (often leading to a truncated protein), which is followed by a second-hit mutation said to involve large DNA deletions usually determined by LOH analysis (1) and subsequent activation of the mTOR pathway. Therefore, the prediction of this hypothesis is that the majority of tumors should have an absence of the TSC2 protein. However, absence of tuberin was rarely observed in this study and is similar to the low incidence of LOH that was previously observed in the brain phenotype of TSC as analyzed by IHC (46, 47) and deep sequencing analysis in the tubers of patients with TSC (48). Studies examining skin harmartomas (49) have shown that the tumor-proliferating cells also do not exhibit a loss of tuberin. In this case, the authors hypothesize that the surrounding fibroblasts secrete paracrine factors, which lead to the growth of the tumors. Specifically, for LAM, the lung manifestation of TSC, the current study (29/29) and others (50) have now demonstrated tuberin expression in 100% of LAM samples. In support of our novel findings, a recent report has demonstrated that mutations in TSC2/TSC1 are not present in all sporadic LAM patients (51). The described studies above suggest that the second-hit LOH hypothesis occurs in a tissue-specific manner, solely in the kidney, independent of whether the renal tumor is derived from the epithelium (mouse carcinoma) or mesenchyme (human angiomyolipoma). Furthermore, a consequence of tissue-specific LOH is that the hypothesized direct relationship between the kidney AML tumor cells and LAM lung cells shown by LOH analysis, the benign metastasis model (52), may be an infrequent or rare event.

Despite the presence of tuberin in the human LAM and AML samples in the current study, the mTOR pathway was activated in 22/41 samples. In Fig. 6, a number of potential genetic mechanisms are outlined that could lead to an activated mTOR pathway. These include LOH of the TSC genes in the kidney AML, posttranslational modification of TSC proteins leading to increased mTORC1 activity, and ultimately uncontrolled cell proliferation (53), mTOR pathway activation downstream of TSC2 (54) or an inactivating mutation at the second locus as in the classical biallelic inactivation hypothesis of Knudson (55). Interestingly, TSC2 activity is regulated by phosphorylation from numerous upstream kinases (53, 56). Therefore multiple signaling pathways can potentially influence TSC2 activity and expression. Activation of mTOR can also occur by means independent of TSC (57). Akt and AMPK regulate the activity of the mTORC1 complex through two unrelated pathways (57).

These studies lead to the hypothesis outlined in the model described in Fig. 6, which now explains a number of outstanding questions in TSC. First, TSC2 haploinsufficiency is required only for tumor initiation and explains the increased incidence of tumors in patients with TSC (58, 59). A number of studies have noted the possible requirement of pathways other than mTORC1 activation for tumor development (60). Our studies identify the activation of the HMGA2 pathway as being at least one of these pathways after the initiating stimulus of loss of a functional TSC2 allele. In these studies in approximately 50% of the tumors the mTOR pathway is activated with tuberin still present. As described above, there are multiple mechanisms activating the mTOR pathway other than LOH (Fig. 6). We hypothesize that the activation of the mTOR pathway occurs later in tumor development after HMGA2 activation as all of the tumors express HMGA2. This hypothesis is significant when considering treatment options in patients given that mTOR inhibitors may be of limited utility in the tumors that have not activated the mTOR pathway (Fig. 6). This may also explain why a significant number of patients in a recent clinical trial utilizing mTOR inhibitors had no response or a decline in lung function when treated with sirolimus (rapamycin; ref. 27).

In conclusion, our results have important ramifications for the prevailing hypothesis in TSC and, by extension, to other tumor suppressor syndromes. These studies demonstrate that haploinsufficiency in a tumor suppressor gene is necessary for tumor initiation but biallelic inactivation is not required for tumor formation (58). In the current study, activation of the HMGA2 pathway was found to be essential for mesenchymal tumor formation. Also, our studies demonstrate that in those cancers, which give rise to tumors in multiple sites, tissue-specific genetic mechanisms exist for tumor pathogenesis even though the identical tumor-initiating stimulus is present in the various tissues. Interestingly, LOH occurred frequently in the renal tumors, which perhaps suggests that this tissue is more susceptible to a major deletion as a second-hit in the TSC2 allele as demonstrated by loss of tuberin in the mouse and human AMLs. Finally, many fundamental questions regarding TSC disease variability and heterogeneity, genotype–phenotype correlation, and tissue-specific pathology are still to be fully elucidated, which could potentially be linked to the differences in the varying pathways activated after the loss of the first TSC2 allele. Furthermore, the current work demonstrates the overall central importance of the HMGA2 pathway in the pathogenesis of TSC and mesenchymal tumorigenesis and may have greater general applicability to provide novel, potential drug targets for therapy in the broad spectrum of tumors in TSC.

No potential conflicts of interest were disclosed.

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conception and design: J. D'Armiento, K. Chada

Development of methodology: J. D'Armiento, D. Sankarasharma

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. D'Armiento, T. Shiomi, P. Geraghty, D. Sankarasharma

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. D'Armiento, T. Shiomi, S. Marks, P. Geraghty, D. Sankarasharma, K. Chada

Writing, review, and/or revision of the manuscript: J. D'Armiento, T. Shiomi, S. Marks, P. Geraghty, K. Chada

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. D'Armiento, D. Sankarasharma

Study supervision: J. D'Armiento, K. Chada

The authors thank Dr. David J. Kwiatkowski for the gift of the Tsc2^+/− mice, Tomoe Shiomi for assistance with histology, and Monica Goldklang for critical review of data and manuscript.

This study was supported by grants from the Columbia University LAM Center and the Department of Defense, W81XWH-05-1-0184 (K. Chada). J. D'Armiento is supported by NIHRO1-HL086936. P. Geraghty was supported by a T32 training grant from the NIH (HL007343).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.