Rapamycin-Insensitive Up-Regulation of Adipocyte Phospholipase A2 in Tuberous Sclerosis and Lymphangioleiomyomatosis

Tuberous sclerosis syndrome (TSC) is an autosomal dominant tumor suppressor gene syndrome affecting multiple organs, including renal angiomyolipomas and pulmonary lymphangioleiomyomatosis (LAM). LAM is a female-predominant interstitial lung disease characterized by the progressive cyst formation and respiratory failure, which is also seen in sporadic patients without TSC. Mutations in TSC1 or TSC2 cause TSC, result in hyperactivation of mammalian target of rapamycin (mTOR), and are also seen in LAM cells in sporadic LAM. We recently reported that prostaglandin biosynthesis and cyclooxygenase-2 were deregulated in TSC and LAM. Phospholipase A2 (PLA2) is the rate-limiting enzyme that catalyzes the conversion of plasma membrane phospholipids into prostaglandins. In this study, we identified upregulation of adipocyte AdPLA2 (PLA2G16) in LAM nodule cells using publicly available expression data. We showed that the levels of AdPLA2 transcript and protein were higher in LAM lungs compared with control lungs. We then showed that TSC2 negatively regulates the expression of AdPLA2, and loss of TSC2 is associated with elevated production of prostaglandin E2 (PGE2) and prostacyclin (PGI2) in cell culture models. Mouse model studies also showed increased expression of AdPLA2 in xenograft tumors, estrogen-induced lung metastatic lesions of Tsc2 null leiomyoma-derived cells, and spontaneous renal cystadenomas from Tsc2+/− mice. Importantly, rapamycin treatment did not affect the expression of AdPLA2 and the production of PGE2 by TSC2-deficient mouse embryonic fibroblast (Tsc2 −/−MEFs), rat uterine leiomyoma-derived ELT3 cells, and LAM patient-associated renal angiomyolipoma-derived “mesenchymal” cells. Furthermore, methyl arachidonyl fluorophosphate (MAFP), a potent irreversible PLA2 inhibitor, selectively suppressed the growth and induced apoptosis of TSC2-deficient LAM patient-derived cells relative to TSC2-addback cells. Our findings suggest that AdPLA2 plays an important role in promoting tumorigenesis and disease progression by modulating the production of prostaglandins and may serve as a potential therapeutic target in TSC and LAM.

Although most biochemical and signaling effects in cells lacking TSC1 or TSC2 are thought to occur through activation of mTORC1, there is evidence that several abnormalities in TSC2null cells are independent of mTORC1 [16]. For example, B-Raf kinase activity is reduced in TSC2-null cells due to high Rheb-GTP levels, but is independent of mTORC1 [17,18]; and Tsc1 or Tsc2-null MEFs have a higher percentage of cilium-containing cells compared to controls, and rapamycin treatment has no effect on this observation [19]. Notch activation may also be regulated by TSC2 in an mTORC1-independent manner [5], although differing results have also been reported [20]. Recently we reported that TSC2 negatively regulates COX-2 expression, prostaglandin production and tumorigenesis in an mTORC1insensitive, but mTORC2-dependent manner [10].
Despite our previous findings that COX-2 expression and prostaglandin production were controlled by TSC2 [20], the relationship between prostaglandin biosynthesis and TSC/LAM pathogenesis has not been extensively studied. Here, we show upregulation of adipocyte-specific AdPLA2 expression in LAM nodule cells relative to non-LAM lungs. We confirmed this finding by showing that the levels of AdPLA2 transcript and protein were higher in LAM lungs compared with control lungs. AdPLA2 accumulation is evident in LAM lung nodules and renal angiomyolipomas. Moreover, we found that TSC2 acts as a negative regulator of the expression of AdPLA2 and the production of prostaglandins in vitro and in vivo. Rapamycin treatment did not affect the expression of AdPLA2 and the production of PGE 2 by TSC2-deficient cells. Finally, we show that pharmacologic inhibition of PLA2 selectively decreases the growth and promotes apoptosis of TSC2-deficient patient-derived cells relative to TSC2-addback cells.

Gene expression analysis
Re-analysis of previously published expression array data (GEO accession number GSE10072 [25,26]; GSE19804 [25,26] and GSE12027 [27]) were performed using the online tool GEO2R. Expression levels of PLA2 family members were compared among LAM cells collected by laser-capture microdissection from LAM nodules (LAM) and Non-LAM lungs (NL) including lung cancer/ tumor.

Ethics statement
The study protocol was reviewed and approved by the Partners Human Research Committee (PHRC) of the Brigham and Women's Hospital and The Peking Union Medical College in China. After explanation of the description of the study, the risks and benefits of the participation, all participants signed a written consent form.

Human samples
Lung tissue from LAM and control subjects was obtained from the National Disease Research Interchange (NDRI) and the Brigham and Women's Hospital-Pulmonary Division Lung Tissue Biorepository. Informed consent was obtained for all lung tissues under Partners approved IRB protocols. Sera from 11 clinical LAM patients (10 sporadic LAM, one TSC-LAM) were collected from Peking Union Medical College in China, with informed consent obtained for research use.
Quantitative RT-PCR RNA from cultured cells and lung tissues was isolated using RNeasy Mini Kit (Qiagen). Gene expression was quantified using One-Step qRT-PCR Kits (Invitrogen) in the Applied Biosystems Real-Time PCR System and normalized to beta-actin.

Cell viability assay
Cells were seeded at a density of 5610 4 /ml in 96-well plates for 24 hr, and then treated with inhibitors or vehicle for 24 hr. Cell viability was determined by MTT assay (Sigma). Values are expressed as mean 6 SEM; n = 8/group. *P,0.05; Student's ttest.

Animal studies
The Brigham and Women's Hospital-Children's Hospital of Boston Standing Committee on Animals approved all procedures described according to standards as set forth in The Guide for the Care and Use of Laboratory Animals. Tsc2 +/2 mice (kindly provided by Dr. Kwiatkowski) develop renal cystadenomas at high frequency by 15 months of age [33]. In the current study, 15 months-old Tsc2 +/2 C57Bl6 mice were treated with celecoxib (0.1% in mouse chow) or vehicle (n = 3 mice) for one month, as described in [10]. Mice were sacrificed and kidneys were harvested. Xenograft tumor model (n = 8 mice/group) was established as previously described [11]. Animal health was monitored three days/week during the entire tumor experiments. The endpoint of the xenograft tumor study was the onset of the clinical signs of pain/distress including 1) animals are in constant pain (hunched posture, sluggish movement, vocalization when handled); 2) bilateral tumors (two subcutaneous tumors/mouse) have caused inactivity, became ulcerated and/or larger than 15% of the animal's body weight (tumor volume ,1,000 mm 3 ); 3) animals have lost more than 20% of their body weight. All mice were euthanized by carbon dioxide (CO 2 ) inhalation via compressed gas in response to the onset of the above distress.

Immunohistochemistry
Immunohistochemistry was performed on paraffin-embedded 10 mm-sections using antibodies against AdPLA2. Slides were deparaffinized and antigen retrieval was performed using sodium citrate retrieval solution pH 6. Sections were stained by the immunoperoxidase technique using DAB substrate (Invitrogen) and counterstaining with hematoxylin.

Immunoblotting analysis
Protein samples were analyzed by SDS-PAGE using 4-12% NuPAGE Gel (Invitrogen), and transferred to a nitrocellulose membrane. Immunoblotting was performed by standard methods using HRP-conjugated secondary antibodies, and chemiluminescence using Supersignal West Pico Chemiluminescent substrate  Quantification of prostaglandin metabolites and VEGF-D PGE 2 and 6-keto-PGF 1a were measured using Enzyme Linked Immuno-Sorbent Assay (ELISA) kits (Cayman Chemical). Levels of secreted prostaglandins were normalized to protein concentrations and expressed as pg/mg protein. Serum levels of PGE 2 and VEGF-D were quantified using ELISA kits.

Statistical analyses
Statistical analyses were performed using Student's t-test when comparing two groups. Results are presented as means 6 SEM.

Identification of adipocyte-specific AdPLA2 upregulation in pulmonary LAM nodule cells
Our previous studies had identified an estradiol-enhanced prostaglandin biosynthesis signature in Tsc2-deficient (TSC22) cells [20]. Since prostaglandin biosynthesis is initiated by PLA2 acting on membrane phospholipids, we examined its expression using public available expression array data sets from pulmonary LAM nodule cells collected by laser-capture microdissection (GEO data set GSE12027) [27], lung cancer/tumor (GEO data set GSE10072 and GSE19804) [25,34], and control lungs (GEO data set GSE10072 and GSE19804) [25,34] ( Figure 1A). The transcript levels of 18 PLA2 family members in LAM nodule cells were compared to non-LAM lungs (NL), respectively ( Figure 1B). Of note, two PLA2 transcripts, PLA2G16 (AdPLA2) and PLA2G4C, were significantly higher by two-fold in LAM nodule cells compared to control non-LAM lungs (p,0.05, Figure 1C).

Expression of AdPLA2 is upregulated in pulmonary LAM nodules and angiomyolipomas
To validate the expression array data, the transcript levels of AdPLA2 were examined in six LAM lungs and six control lungs using real-time RT-PCR. LAM lungs (LAM) exhibited a twofold increase of AdPLA2 transcript relative to control lungs (NL) (p, 0.01, Figure 2A), confirming the overexpression of AdPLA2 identified using LAM expression array analyses ( Figure 1). Moreover, immunoblotting analysis showed that LAM lungs positive for Phospho-S6 (S235/236) accumulated higher levels of AdPLA2 protein relative to control normal lungs (NL) (Figure 2B). Furthermore, immunohistochemical staining of two LAM specimens showed abundant accumulation of AdPLA2 in smooth muscle-like cells ( Figure 2C). In addition, confocal microscopy showed that smooth muscle actin-positive cells were stained with AdPLA2, whereas pulmonary artery cells were negative with AdPLA2 staining ( Figure 2C).
Serum levels of VEGF-D have been used a biomarker for LAM [35,36]. We recently reported that serum levels of PGE 2 were higher in women with LAM relative to healthy women [20]. To determine the correlation of the levels of VEGF-D and PGE 2 in LAM, we collected sera from 11 LAM patients ( Table 1) and measured the levels of PGE 2 and VEGF-D using ELISA. Serum levels of PGE 2 segregated into two groups: 5-10 pg/mL (high-PGE 2 ) and 10-20 pg/mL (low PGE 2 , Figure 2D). Interestingly, serum levels of VEGF-D were lower in LAM subjects (VEGF-D 19376520 pg/mL) with lesser PGE 2 compared to LAM subjects (VEGF-D 729862529 pg/mL) with higher PGE 2 , indicative of a strong correlation between PGE 2 and VEGF-D (p,0.05, Figure 2E). These data suggest that serum PGE 2 could be useful as a diagnostic marker and for assessment of disease severity.

TSC2 negatively regulates AdPLA2 expression in a rapamycin-insensitive manner in vitro
To define the molecular mechanisms responsible for AdPLA2 upregulation in LAM, we first tested whether TSC2 plays a role in regulating AdPLA2 expression. AdPLA2 expression was higher by 3.5-fold in TSC2-deficient LAM patient-derived cells (TSC22) compared with TSC2-addback cells (TSC2+) (p,0.001, Figure 3A). Because PLA2 catalyzes the conversion of plasma membrane phospholipids to prostaglandins, we next measured the production of PGE 2 in cells lacking TSC2. TSC2-deficient LAM patient-derived cells secreted ,75% higher levels of PGE 2 compared with TSC2-addback cells (p,0.01, Figure 3B), indicative of active AdPLA2.

TSC2 negatively regulates AdPLA2 expression in vivo
To determine whether TSC2 regulates AdPLA2 expression in vivo, we first used xenograft tumors from mice inoculated with TSC2-deficient ELT3-V3 (TSC22) cells and TSC2-reexpressing ELT3-T3 (TSC2+) cells. AdPLA2 protein levels were markedly higher by 3.5-fold in TSC2-xenograft tumors with increased immunofluorescent staining of smooth muscle actin (a-actin), phospho-S6 (P-S6) (S235/236) and AdPLA2 in pulmonary LAM nodules from two LAM subjects (LAM-1 and LAM-2). (D) Serum levels of VEGF-D and PGE 2 were measured using ELISA from 11 clinical LAM patients. (E) The correlation of serum levels of PGE 2 and VEGF-D was analyzed. *P,0.05, ***P,0.005, Student's t-test. doi:10.1371/journal.pone.0104809.g002 AdPLA2 as a Therapeutic Target in TSC and LAM PLOS ONE | www.plosone.org phospho-S6 compared to that of TSC2+ tumors ( Figure 4A). Mice bearing xenograft tumors of ELT3-V3 (TSC22) produced higher urinary levels of PGE 2 compared with mice bearing ELT3-T3 (TSC2+) xenograft tumors (p,0.05, Figure 4B). The fact that LAM is a female predominant disease made us hypothesize that estradiol plays an important role in disease progression. To assess the impact of estradiol on AdPLA2 expression in vivo, ovariectomized female scid mice inoculated with Tsc2-deficient ELT3 cells were treated with estradiol or placebo. Xenograft tumors were developed, as expected. The accumulation of AdPLA2 in xenograft tumor was markedly elevated in mice receiving estradiol compared with placebo treatment ( Figure 4C). Importantly, AdPLA2 accumulation was evident in lung metastatic lesions of ELT3 cells in mice treated with estradiol compared with placebo treatment ( Figure 4C). These data suggest that estradiolenhanced AdPLA2 expression was correlated with xenograft tumor growth and lung metastasis of Tsc2-deficient cells.
We next examined spontaneously arising renal cystadenomas in Tsc2 +/2 mice [33]. Renal Tumors exhibited increased expression of AdPLA2 relative to stromal cells, although some normal kidney tubule cells were moderately positive for AdPLA2, while glomeruli were negative for AdPLA2 ( Figure 4D). We recently reported that TSC2 negative regulates cyclooxygenase-2 (COX-2) and prostaglandin biosynthesis [10]. To examine the impact of COX-2 inhibition on the upstream regulator, AdPLA2, Tsc2 +/2 mice were subjected to short-term treatment with COX-2 specific inhibitor-Celecoxib or vehicle control before harvesting renal tumors for immunoblotting. The levels of AdPLA2 expression from a Tsc2 +/2 mouse treated with Celecoxib remained high in renal tumors compared to adjacent normal kidney tissues ( Figure 4D), suggesting that COX-2 targeted therapy does not have a feedback effect on AdPLA2 expression.

Pharmacological inhibition of PLA2 selectively suppresses the growth of TSC2-deficient LAM patientderived cells
To determine whether elevated PGE 2 production has any biologic consequence, we examined the cell growth in response to PGE 2 and PGI 2 stimulation. PGE 2 or PGI 2 stimulation for 72 hr led to a 60% or 55% increase of the growth of TSC2-deficient cells compared with vehicle control, respectively (p,0.01, Figure 5A), although the growth of TSC2-addback cells was not sensitive to PGE 2 or PGI 2 stimulation ( Figure 5A). Because mTORC1 inhibition had no effect on AdPLA2 expression, we examined whether the PLA2 inhibitor, methyl arachidonyl fluorophosphonate (MAFP), a selective, active-site directed, irreversible inhibitor of PLA2 [41], had an effect on the growth of TSC2deficient cells. Treatment with 3 mM MAFP significantly reduced the number of TSC2-deficient cells compared to vehicle control (p,0.05, Figure 5B). Importantly, TSC2-addback cells were resistant to 3 mM MAFP treatment, although 4.5 mM MAFP reduced cell numbers by 80% relative to vehicle control ( Figure 5B). To further define the optimal dose of MAFP in promoting cell death, we treated TSC2-deficient cells with 1, 2, 3,  Figure 5C). Our data indicate that MAFP selectively suppresses the growth of TSC2-deficient cells relative to TSC2-addback cells. Moreover, 2 mM MAFP treatments led to an increased caspase 3 cleavage, and 5 mM MAFP caused higher levels of cleaved caspase 3 and PARP compared with control treatment ( Figure 5D). Together, Table 1. Clinical profile of LAM subjects.   these data indicate that inhibition of PLA2 selectively suppresses the survival of TSC2-deficient LAM patient-derived cells.

Discussion
Recently we reported that COX-2 and prostaglandin biosynthesis pathway is aberrantly activated in TSC2-deficient cells [10]. PGE 2 and PGI 2 , in particular, two major prostaglandin metabolites of the eicosanoid end-products from the COX-2 mediated branch of the arachidonate pathway, were significantly elevated in TSC2-deficient cells, in preclinical models of TSC/LAM, and in sera from women with LAM [10]. Importantly, pharmacological inhibition of COX-2 led to apoptosis of TSC2-deficient cells, and blocked the progression of Tsc2-deficient xenograft tumors and kidney cystadenomas in preclinical models [10]. The findings of elevated COX-2 activity and prostaglandin production in our recently published study [10] led us hypothesize that phospholipase A2 (PLA2), the first rate-limiting enzyme responsible for the conversion of plasma membrane phospholipids to arachidonate [42], is activated in cells lacking TSC2 and/or with mTORC1 hyperactivation.
Phospholipase A2s (PLA2s) belongs to a superfamily of 15 distinct members comprising four major clusters, including the secreted sPLA2, cytosolic cPLA2, calcium-independent iPLA2, and platelet activating factor (PAF) acetyl hydrolase/oxidized lipid lipoprotein associated (LP) PLA2 [43][44][45]. Each type of PLA2 To determine which PLA2(s) plays a critical role in TSC/LAM pathogenesis, we analyzed the publicly available expression array data sets from laser capture microdissected LAM nodule cells [27] and non-LAM lung data sets [25,26]. Surprisingly, we found that the adipocyte-specific PLA2 (AdPLA2, also called PLA2G16) was upregulated in LAM nodule cells relative to non-LAM lungs (Figure 1). AdPLA2 is a major PLA2 enzyme in adipose tissue and regulates lipolysis through PGE 2 [46,47]. This finding is of particular interest because of the ''mesenchymal'' features of LAM, including the expression of smooth muscle actin [48] and melanocytic markers HMB45 [49] and/or gp100 [50], the metastatic potential [51], and epithelial-mesenchymal transition [26]. Pathologically, renal angiomyolipomas are composed of immature smooth muscle cells, aberrant blood vessels, and fat cells [52]. Fat cells are also evident in a primary cultures derived from a LAM-associated angiomyolipoma [29]. Thus, the identification of AdPLA2 is in agreement with the characteristics of LAM patientderived cells.
Our current study demonstrated that the expression of AdPLA2 is negatively regulated by TSC2. However, the elevated AdPAL2 expression was not affected by pathway blockade, including mTORC1 inhibitor rapamycin, mTORC1/2 inhibitor Torin 1, Akt inhibitor PI-103, or MEK1/2 inhibitor PD98059, which represents uncovered regulatory mechanisms. Several precedent studies have documented possible mTORC1-independent cellular outcomes in cells lacking TSC1 or TSC2 (reviewed in Neuman et al. [53]). B-Raf kinase activity is reduced in TSC2-deficient cells due to Rheb-GTP, but independent of mTORC1 [17,18]. Akt activation in Tsc2 2/2 p53 2/2 MEFs was reduced due to impaired mTORC2 activity [54]. Tsc1 2/2 or Tsc2 2/2 p53 2/2 MEFs had more abundant cilia relative to the counterpart controls, and rapamycin treatment had no effect on the cilia formation [19]. We had previously found that the overexpression of matrix metalloproteinase (MMP) in TSC2-deficient LAM patient-derived cells is insensitive to rapamycin [30]. Recently, we discovered that TSC2 negatively regulates COX-2 expression and prostaglandin production in a rapamycin-insensitive but mTORC2-dependent manner [10]. The current study adds to the accumulating evidence that mTORC1-independent regulation of signaling pathways may contribute to the pathogenesis and progression of TSC/LAM.
The Multicenter International LAM Efficacy of Sirolimus Trial (The MILES trial) demonstrated that the mTORC1 inhibitor Sirolimus stabilizes lung function and improves quality of life in LAM patients. However, upon drug discontinuation, lung function decline resumed [15], indicating that mTORC1 inhibitor has a cytostatic but not cytotoxic effect on LAM cells. There is an unmet need for novel strategies to promote cell death in TSC/LAM. In summary, we report that AdPLA2 expression is elevated in TSC2deficient patient-derived cells compared to TSC2-addback cells, and that the PLA2 inhibitor selectively suppresses the proliferation of TSC2-deficient cells. We anticipate that PLA2 inhibitors may provide a novel therapeutic strategy for TSC and LAM.