Utility of a PI3K/mTOR Inhibitor (NVP-BEZ235) for Thyroid Cancer Therapy

Background We assessed the utility of the dual PI3K/mTOR inhibitor NVP-BEZ235 (BEZ235) as single agent therapy and in combination with conventional chemotherapy for thyroid cancer. Methodology/Principal Findings Eight cell lines from four types of thyroid cancer (papillary, follicular, anaplastic, medullary) were studied. The cytotoxicity of BEZ235 and five conventional chemotherapeutic agents alone and in combination was measured using LDH assay. Quantitative western blot assessed expression of proteins associated with cell cycle, apoptosis and signaling pathways. Cell cycle distribution and apoptosis were measured by flow cytometry. Murine flank anaplastic thyroid cancers (ATC) were treated with oral BEZ235 daily. We found that BEZ235 effectively inhibited cell proliferation of all cancer lines, with ATC exhibiting the greatest sensitivity. BEZ235 consistently inactivated signaling downstream of mTORC1. BEZ235 generally induced cell cycle arrest at G0/G1 phase, and also caused apoptosis in the most sensitive cell lines. Baseline levels of p-S6 ribosomal protein (Ser235/236) and p27 correlated with BEZ235 sensitivity. Growth of 8505C ATC xenograft tumors was inhibited with BEZ235, without any observed toxicity. Combination therapy of BEZ235 and paclitaxel consistently demonstrated synergistic effects against ATC in vitro. Conclusions BEZ235 as a single therapeutic agent inhibits thyroid cancer proliferation and has synergistic effects in combination with paclitaxel in treating ATC. These findings encourage future clinical trials using BEZ235 for patients with this fatal disease.


Introduction
Thyroid cancer is the most common endocrine malignancy, originating from thyroid follicular cells (papillary, follicular, poorly differentiated and anaplastic cancer) or parafollicular C cells (medullary cancer). The incidence of thyroid cancer has increased over the past 3 decades, primarily from an increase in the detection of papillary cancer. In contrast, the incidence of follicular and poorly differentiated thyroid cancer remains unchanged [1,2]. Most patients with well-differentiated cancer, including papillary (PTC) and follicular (FTC) thyroid cancer have a favorable prognosis. Nevertheless, about 5% patients develop radioactive iodine refractory tumors and usually cause death within 5 years [3]. Anaplastic thyroid cancer (ATC) is a rare, highly aggressive, and often fatal disease, with a median survival of just 6 months. Medullary thyroid cancer (MTC) accounts for 3-5% of thyroid malignancy. The frequency of regional and distant metastatic disease in MTC diminishes survival rates [4,5]. Novel therapies for refractory and aggressive thyroid cancer are needed to improve currently poor outcomes for these patients.
The PI3K/mTOR pathway is important for cell metabolism, survival and proliferation. Class I A PI3Ks are heterodimers containing a p85 regulatory and a p110 catalytic subunits which phosphorylate phosphatidylinositol-4,5-biphosphate (PIP 2 ), yielding phosphatidylinositol-3,4,5-triphosphate (PIP 3 ). PIP 3 combines with phosphoinositide-dependent protein kinase 1 (PDK1) to phosphorylate AKT at Thr308. In addition, the phosphorylation of AKT at Ser 473 by mTORC2 is required for full activity of AKT. Activation of AKT phosphorylates mTORC1, which subsequently phosphorylates S6 kinase1 and 4E-BP1, leading to G1/S cell cycle progression and inhibition of apoptosis. PTEN is a tumor suppressor that dephosphorylates PIP 3 and inactivates this pathway [6,7]. Alterations of this signaling pathway frequently occur in malignancies and are potential targets for cancer therapy [6][7][8].
In thyroid cancer, genetic alterations affecting the PI3K/ mTOR pathway have been identified. PIK3CA (encoding p110a of class I A PI3K) copy number gain correlates with increased PIK3CA protein expression. PIK3CA copy number gain occurs more frequently than genetic mutations of PIK3CA or PTEN in thyroid cancer [9][10][11][12][13][14]. More PIK3CA/AKT1 mutations and PIK3CA copy gain are identified in ATC as compared to well differentiated cancer, suggesting that PI3K/mTOR pathway activity is involved in the process of cancer de-differentiation [3,11,12]. For MTC, RET proto-oncogene mutations occur in almost all familiar cases (25% of MTC) and about half of sporadic MTC. This gain-offunction rearrangement enhances PI3K/mTOR signaling transduction [15,16]. In sporadic MTC without RET mutations, over 50% of tumor samples show activation of AKT or mTOR by immunohistochemistry [16].
BEZ235 is a dual PI3K/mTOR inhibitor that reduces PI3K and mTOR kinase activity by competitive binding to the ATPbinding cleft of these enzymes [17]. BEZ235 may treat cancers through induction of G0/G1 cell cycle arrest and apoptosis, and has recently entered phase II clinical trials [17][18][19][20][21][22]. This study was conducted to evaluate the efficacy of BEZ235 in treating thyroid cancer from four major pathological types, including PTC, FTC, ATC and MTC. We also explored combination effects of BEZ235 and currently employed chemotherapeutics against four ATC cell lines.
Effects of BEZ235 on cell cycle and apoptosis KAT4C, 8505C, TT, BHP7-13 and WRO82-1 were exposed to BEZ235 for 24 or 72 hours ( Figure 3A). Compared with untreated control cells, BEZ235 at 6.25, 25 and 100 nmol/L induced increasing cell fractions at G0/G1phase in all cell lines. A representative cell line BHP7-13 demonstrated the effect of BEZ235 on cell cycle distribution ( Figure 3B).
The effects of BEZ235 (100 nmol/L) on the expression of G0/ G1 related cell cycle proteins was examined ( Figure 3C, Figure  S2B and S3). The transition of G0/G1 to S phase is a complex process involving cyclins, cyclin-dependent kinases (CDKs) and associated proteins, including p53, p21, p27 and cyclin D1 [31]. p53 is a tumor suppressor that may activate p21, and both p21 and p27 are inhibitors of G1 cyclin-dependent kinases.
BEZ235 increased p27 by 4 to 24 hours in all cell lines, with maximal increases of 3.5 to 723.2-fold in KAT4C, TT and BHP7-13. Such a ratio of increase was not measurable in 8505C due to undetectable signal in untreated cells. Statistical analysis was performed for p27 expression in TT, and achieved significance at 8 hours as compared with basal expression (P = 0.019, t-test) ( Figure S3C). This data is consistent with our finding of G0/G1 cell cycle arrest. BEZ235 had varied effects on the expression of p53, p21 and cyclin D1 in 4 cell lines, showing that there may be varied contributions of these proteins on affecting cell cycle progression. A consistent increase of p27 in BEZ235 treated cells likely exerted inhibitory effects on cell cycle progression at G0/G1. It is possible that G0/G1 cell cycle arrest is the result of a combined effect of more than one G0/G1 related protein affected by BEZ235. Alternatively, the induction of p27 alone may be enough to arrest cell cycle progression. In a p53 negative PC3M cell line, BEZ235 induced p27 expression and complete cell cycle arrest at G1 phase, without inducing p21 [17]. This study demonstrated G1 cell cycle inhibition induced by BEZ235 does not necessary require association with p53 and p21.
The inhibition of the PI3K/mTOR pathway may also lead to apoptosis [6,8]. The ability of BEZ235 to cause apoptotic cell death in thyroid cancer cells was explored ( Figure 3D). Compared with control, BEZ235 at 25 and 100 nmol/L significantly induced apoptosis as measured by the proportion of sub-G1 cells at 96 hours in KAT4C. Similar findings were observed in KAT18, with BEZ235 at 6.25, 25 and 100 nmol/L driving an increasing proportion of apoptotic cells. However, BEZ235 failed to show any increase of sub-G1 cells in 8505C, TT, BHP7-13, and WRO82-1.
To validate the induction of apoptosis in KAT4C and KAT18 by BEZ235, caspase-3 was assessed by immunoblot after 72 hours of treatment ( Figure 3E, Figure S4). In general, higher doses of BEZ235 induced more degradation of apoptotic executioner caspase-3, with less than 12% of caspase-3 detected at 100 nmol/ L in both cell lines. This data suggests that apoptotic mechanisms account for the cytotoxicity of BEZ235 in KAT4C and KAT18. These findings are consistent with previous reports of BEZ235 causing apoptosis in some, but not all, cell lines [17][18][19][20][21][22].The underlying mechanisms of the varied abilities of BEZ235 to induce apoptosis at different doses and in different cell lines remain unclear [6][7][8].
BEZ235 sensitivity correlates with baseline expression of p-S6 ribosomal protein (Ser235/236) and p27 The Dm of BEZ235 spans a broad spectrum across 8 different cell lines, with an 11-fold difference between KAT4C and WRO82-1.To explore potential biomarkers that correlate with sensitivity of BEZ235, the baseline expression of proteins involved in the PI3K/mTOR and RAS/RAF/ERK pathways, and cell cycle-associated proteins were evaluated in 6 cell lines ( Figure 4A). The sensitivity of the six cell lines to BEZ235 was ordered according to the Dm value.

BEZ235 therapy of murine flank tumors
Athymic nude mice with flank xenografts of 8505C were used to study the therapeutic effects and safety of BEZ235 in vivo. Animals with established flank tumors of similar starting volumes were treated with oral BEZ235 (50 mg/kg) or vehicle daily for 25 days and followed until day 31 ( Figure 5A). BEZ235 retarded tumor growth and the difference of tumor volumes between BEZ235 and control mice was statistically significant at day 21 (94.1626.4 mm 3 and 335.1678.5 mm 3 , P = 0.009, t-test) and day 24 (109.9629.6 mm 3 and 403.16112.8 mm 3 , P = 0.019, t-test). The difference of tumor volumes lost significance at day 28, 4 days after treatment had been discontinued (402.76123.6 mm 3 and 750.36198.3 mm 3 , P = 0.15, t-test). Representative mice were photographed on the last day of treatment ( Figure 5B). BEZ235 did not significantly reduce body weight between the two groups during the study period ( Figure 5C). Oral BEZ235 reduced phosphorylation of AKT and S6 ribosomal protein by 2 hours and achieved its greatest inhibitory effects at 4 hours with just 2% and 10% of pAKT and pS6 remaining, respectively. This inhibitory effect persisted for 24 hours in xenograft tumors ( Figure 5D, Figure S5). PCNA was slightly and transiently repressed from 2 to 4 hours. However, significant degradation of caspase-3 occurred by 4 hours, with only 4% of caspase-3 remaining by 24 hours. Immunohistochemistry of p-S6 ribosomal protein (Ser235/236) in tumors treated with BEZ235 similarly demonstrated a reduction in staining, most markedly from 2 to 6 hours ( Figure 5E).

Interaction of BEZ235 and chemotherapy in ATC cells
We studied the combination of chemotherapeutic agents and BEZ235 against ATC. Five chemotherapeutic agents (paclitaxel, irinotecan, etoposide, 5-FU and doxorubicin) demonstrated cytotoxic effects in a dose-and time-dependent manner in four ATC cell lines ( Figure S6). The data were used to calculate the median effect dose (Dm) on day 4 using CompuSyn software ( Figure 6A).
These Dm of paclitaxel, irinotecan and etoposide are achievable in patients' serum, and are clinically relevant [32][33][34]. Although doxorubicin is used clinically for anaplastic thyroid cancer, the relatively high Dm in these cell lines implied that clinically relevant dosing is not achievable in patients serum and therefore was excluded from these studies. Interactions between BEZ235 and paclitaxel, irinotecan, and etoposide were evaluated ( Figure 6B). The combination of BEZ235 and paclitaxel significantly improved cytotoxicity over single agent therapy in four cell lines. BEZ235 combined with irinotecan also enhanced therapeutic efficacy, particularly when more cells were affected. BEZ235 plus etoposide only slightly increased cytotoxicity.
These results demonstrate that BEZ235 and paclitaxel had the best combined effects in treating ATC. The calculated DRI of the chemotherapy agent means the fold of the drug dose that could be reduced in the presence of BEZ235 ( Figure 6D)

Discussion
BEZ235 effectively inhibited cell proliferation in eight thyroid cancer lines originating from four major histologic types. ATC were the most sensitive, followed by follicular undifferentiated, medullary and well-differentiated thyroid cancer cell lines. Cancer cells harboring a PI3K gain-of-function mutation or a PTEN deletion demonstrate higher PI3K/mTOR pathway activity and greater sensitivity to BEZ235 [22]. Our data suggest that ATC relies on PI3K/mTOR activity, and interruption of this pathway with BEZ235 impairs ATC growth more significantly than other thyroid cancer histologies. The relatively low median effect doses of BEZ235 in all of the thyroid cancer lines (,44 nmol/L) suggest that BEZ235 may have utility for treating a spectrum of thyroid malignancy. Refractory cancers that develop activation of PI3K/ Figure 2. mTORC1downstream signaling is inhibited and p-ERK1/2 is activated by BEZ235 at 100 nmol/L. p-AKT (Thr308) and p-AKT (Ser473) was increased in 8505C for more than 24 hours. p-AKT (Thr308) and p-AKT (Ser473) was decreased transiently in TT and BHP7-13. p-4E-BP1 (Thr70), p-4E-BP1 (Thr37/46) and p-S6 ribosomal protein (Ser235/236) were consistently reduced from 2 hours through over 24 hours in three cell lines. BEZ235 caused a rapid increase of p-ERK1/2 (Thr202/Tyr204) by 2 to 4 hours in three cell lines. doi:10.1371/journal.pone.0046726.g002 Thyroid Cancer and NVP-BEZ235 PLOS ONE | www.plosone.org mTOR signaling in the process of tumor de-differentiation may be particularly attractive targets for therapy.
The failure of BEZ235 to repress p-AKT (Ser473) in 8505C and KAT4C may be related to a negative feedback inhibition. It has been previously shown that inhibition of mTORC1 leads to inactivation of S6 kinase 1, which may subsequently overwhelms the inhibitory effect of BEZ235 on mTORC2, activates mTORC2, and increases p-AKT (Ser473) in 8505C [35]. Prior reports also showed that lower doses of BEZ235 fail to inhibit p-AKT (Ser473) in some cell lines, and higher doses of BEZ235 may overcome the negative feedback of mTORC1/S6 kinase 1 feedback loop [18,20]. Although p-AKT (Ser473) was activated in 8505C and KAT4C, BEZ235 had better inhibitory effects in these cell lines as compared to TT and BHP7-13, suggesting that other molecules affected by BEZ235 play a more important role in determining therapeutic outcome.
We found that the expression of p-S6 ribosomal protein (Ser235/236) and p27 correlate with the sensitivity of BEZ235 in thyroid cancer. S6 ribosomal protein is a downstream of S6 kinase 1, which is activated by mTORC1 [6][7][8]. Phosphorylation of S6 ribosomal protein increases translational control of protein synthesis and enhances cell growth. Cells with higher levels of p-S6 ribosomal protein (Ser235/236) are more susceptible to BEZ235, suggesting that the inhibition of mTORC1 and S6 ribosomal protein is the major therapeutic effect of BEZ235.
p-S6 ribosomal protein (Ser235/236) has also recently been recognized as a marker to predict therapeutic effect of an mTOR inhibitor in sarcoma [36]. In this study, thyroid cancer lines with higher expressions of p-S6 ribosomal protein (Ser235/236) also showed lower levels of p27. This finding suggests S6 ribosomal protein is a suppressor of p27 in thyroid cancer, and may explain why both p-S6 ribosomal protein (Ser235/236) and p27 were predictors of sensitivity to BEZ235. Interestingly, p27 was previously noted to have an inverse association with the activity of the PI3K/mTOR pathway in thyroid cancer cells, and repression of this pathway increases p27 [37]. 4E-BP1 is another protein downstream of mTORC1. Inhibition of mTORC1 leads to dephosphorylation (activation) of 4E-BP1, enhancing the binding of 4E-BP1to eIF4E, and blocking protein translation and cell proliferation [7,8]. Although BEZ235 affects both S6 ribosomal protein and 4E-BP1 efficiently, only p-S6 ribosomal protein (Ser235/236) expression predicts for sensitivity to BEZ235 in this study. In ovarian cancer, biomarkers predicting susceptibility of BEZ235 were reported [22], and the expression of p-4E-BP1 (Thr37/46) did correlate with the sensitivity of BEZ235.
In addition to inhibiting cell cycle progression, BEZ235 caused apoptosis in two of six cell lines. The inhibition of cell cycle progression is a known effect of BEZ235, even at lower doses (#100 nmol/L). However, apoptosis appears in only some cancer cell lines and is more apparent at higher doses (100-1000 nmol/L) of BEZ235 [17][18][19][20][21][22]. BEZ235 efficiently inhibited mTORC1, a molecule controlling both cell cycle and apoptosis that might lead to cell cycle arrest and apoptosis in KAT4C and KAT18 [7]. Understanding the mechanisms through which BEZ235 contributes to apoptotic cell death will require further study.
ATC is by far the most aggressive of the four major histologic types of thyroid cancer. Chemotherapy has been applied to treat patients with ATC with response rates around 20 to 50%. Novel strategies to improve outcomes are needed. Among three combination therapy regimens, BEZ235 combined with paclitaxel had the best synergistic effect in four ATC cell lines. Cancer cells with activation of PI3K/mTOR signaling are more resistant to paclitaxel, and the co-administration of PI3K/mTOR inhibitors with paclitaxel improves therapeutic effects [38][39][40]. This finding is of clinical relevance since paclitaxel, a microtubule stabilizer, has shown to achieve a 53% response rate in patients with ATC in a phase II clinical trial [41]. The combination of BEZ235 with a microtubule depolymerizing drug vincristine also revealed promising effect in the treatment of sarcoma [20]. Our data showed that the combinational effects of BEZ235 with inhibitors of DNA topoisomerase type I (irinotecan) or type II (etoposide) in treating ATC were largely antagonistic. Similar antagonistic effects of inhibition both topoisomerase activity and PI3K/AKT pathway have been observed in ovarian cancer cells [42]. Cells in S phase are more susceptible to topoisomearse inhibitors. BEZ235-induced accumulation of cells at G0/G1 phase may therefore reduce the therapeutic advantage of concurrent therapy with topoisomerase inhibitors, explaining the unfavorable combination effects of BEZ235 with irinotecan and etoposide.
Daily treatment of BEZ235 significant retarded 8505C xenograft tumor growth during the therapeutic period. The inhibitory effect was less prominent after the discontinuation of therapy, suggesting that prolonged treatment may be necessary to maintain therapeutic efficacy. BEZ235 significantly degraded caspase-3 in 8505C xenograft tumors, indicating this compound may induce apoptosis in vivo. After discontinuation of BEZ235, the volume of 8505C xenograft tumors subsequently increased. This enlargement of tumor following cessation of therapy might be due to normalization of cell size and resumption of cell proliferation, as mTOR and its downstream proteins S6 Kinase 1 and 4E-BP1 play pivotal roles in this respect [43]. No significant weight loss or illness was observed during study period, suggesting that this therapy may have a promising safety profile.
A prior report demonstrated that the presence of genetic alterations of PTEN, PIK3CA and AKT1 correlated well with sensitivity to an AKT inhibitor, but had weaker correlations with an mTOR inhibitor [44]. This discrepency suggests that mTOR activity does not depend solely on PI3K/AKT activity. The reported data of genetic alterations in 8 thyroid cancer cell lines was summarized (Appendix S2). Limited genetic changes on RAS/RAF/ERK and PI3K/mTOR pathways were identified. The available data do not show any correlation between genetic aberrations of PI3K/mTOR pathways and sensitivity of BEZ235, as no mutation of this pathway in these cell lines reported. Mutations of p53 and BRAF also seem to not be correlated with Figure 5. BEZ235 decreases p-AKT(Thr308), p-S6 ribosomal protein (Ser235/236), caspase-3, and retards the growth of ATC flank xenograft tumors without significant toxicity in nude mice. A, daily oral gavage of BEZ235 (50 mg/kg) represses 8505C tumor growth. The differences of tumor volume between BEZ235 and control group achieved statistical significance on days 21 and 24. Significance was lost on day 28, 4 days after treatment had been discontinued. B, representative mice with 8505C xenograft tumors (arrows) were photographed on last day of treatment (day 24). C, BEZ235 did not result in any weight loss, in treated and control mice (P.0.05 for both comparison). D, BEZ235 represses p-AKT (Thr308) and p-S6 ribosomal protein (Ser235/236) by 2 hours and caspase-3 by 4 hours in vivo. PCNA was slightly reduced at early period. E, expression of p-S6 ribosomal protein in tumor cells is detected by immunostaining at the indicated periods (magnification, 6200). These results were in consistent with Western blot results. doi:10.1371/journal.pone.0046726.g005 Figure 6. Combination therapy of BEZ235 with paclitaxel has synergistic effects against ATC. A, Dm was calculated using cytotoxicity data using CompuSyn software for each drug on each cell line at day 4. B, the cytotoxic effects of BEZ235 and chemotherapeutic agents (paclitaxel, irinotecan and etoposide) alone or in combination after a 4-day treatment in ATC were evaluated using LDH assays. The combination of BEZ235 and sensitivity of BEZ235, because both sensitive (8305C) and less sensitive (WRO82-1) cell lines harbor the mutations. Exploration of genetic changes of mTOR, the major target of BEZ235, might someday provide information that may predict therapeutic efficacy. Inhibition of mTORC1 may activate the MAPK pathway through a PI3K-dependent feedback loop [45], which explains why BEZ235 was seen here to activate ERK1/2 in thyroid cancer cells. Combining BEZ235 with an inhibitor targeting MAPK pathway may be a potential approach to enhance therapeutic efficacy. Recently, the combination of a pan-RAF inhibitor and BEZ235 was shown to induce cell cycle arrest at G0/G1 phase, and had beneficial combination effects in treating thyroid cancer [46]. Similarly, in combination of mTOR and MEK inhibitors also demonstrated therapeutic advantage in treating thyroid cancer, providing further evidence that targeting both PI3K/ mTOR and MAPK pathways is a potential therapeutic strategy for this disease [47,48].

Cytotoxicity assays
Cells were plated at 2610 4 cells per well in 24-well plates in 1 mL media. After overnight incubation, the agents or vehicle were added at the indicated concentration. Six serial 1:1 dilutions were tested starting at the following doses: BEZ235 at 100 nmol/ L, paclitaxel at 68 nmol/L (8305C, 8505C and KAT4C) and 20 nmol/L (KAT18), irinotecan at 8 mmol/L, etoposide at 10 mmol/L, 5-FU at 20 mmol/L and doxorubicin at 1 mmol/L over a 4-day treatment course. Cytotoxicity was determined daily for BEZ235 group and on day 3 and 4 for the chemotherapy groups. Cells were washed with PBS and lysed with Triton X-100 (1.35%, Sigma) to release intracellular lactate dehydrogenase (LDH), which was quantified with a Cytotox 96 kit (Progmega) at 490 nM by spectrophotometry (BT-MQX200R, Bio-Tek Instruments). Each experiment was performed in triplicate and results are shown as the percentage of surviving cells determined by comparing the LDH of each sample relative to control samples which are considered 100% viable. Median effect doses (Dm) on day 4 were calculated for each cell line and drug using CompuSyn software [27].
For combination therapy experiments, ATC cells were treated with BEZ235 and a chemotherapeutic drug (paclitaxel, irinotecan or etoposide) at a fixed dose ratio. Cells were incubated with vehicle, BEZ235, chemotherapeutic agent, or both simultaneously for a 4-day course and cytotoxicity was measured. Five serial 1:1 dilutions were examined at the following starting doses for 8305C, 8505C, KAT4C and KAT18: BEZ235 at 26.8, 37.2, 15.6 and 26.4 nmol/L, paclitaxel at 52. 8, 22.4, 28 and 17.6 nmol/L, irinotecan at 13.6, 14, 13.6 and 11.6 mmol/L, etoposide at 11.2, 2.56, 6.8 and 5.2 mmol/L, respectively. The doses chosen were based on the Dm determined previously.

Western blots
Cells were plated at 8610 5 cells in 100-mm Petri dishes in 8 mL media overnight and treated with BEZ235 or vehicle for indicated periods. Cells were plated overnight for baseline expression of untreated cells. Cell pellets were dissolved in radio-immunoprecipitation assay buffer and protease inhibitor cocktail, vortexed and clarified by centrifugation. Total protein (20-50 mg) was electrophoresed on 10-12% Tris-HCl gels, transferred to polyvinylidene difluoride membranes, blocked, and exposed to primary antibodies followed by a secondary antibody conjugated to horseradish peroxidase. Signals were developed using an enhanced chemiluminescence kit (Bionovas Biotechnology).
Band density was imaged and quantified using Molecular Imager VersaDoc MP 4000 system (Bio-Rad). The ratios of p-AKT, p-4E-BP1, p-S6 ribosomal protein and p-ERK1/2 to the correlated total protein, and the ratios of p53, p21, p27 and cyclin D1 to a-tubulin in each cell line were calculated. Relative paclitaxel showed synergistic effects in all cell lines. The combination of BEZ235 and irinotecan had enhanced cytotoxic effects at higher doses, but revealed limited or no beneficial therapeutic effects at lower doses. The combination of BEZ235 plus etoposide displayed little additional cytotoxicity. C, the CI of BEZ235 and each chemotherapeutic agent was calculated using CompuSyn software. BEZ235 plus paclitaxel exhibited synergistic effects in all cell lines, with CI,0.8 at most conditions. BEZ235 plus irinotecan had synergistic to antagonistic effects in 8305C, 8505C and KAT18, and antagonistic effects in KAT4C. BEZ235 plus etoposide demonstrated mostly antagonistic effects in all cell lines. The horizontal dash lines at CI = 1 were drawn for discrimination of synergism (CI,1) and antagonism (CI.1). D, the typical ranges of DRI values for chemotherapeutics in combination with BEZ235. With the presence of BEZ235, paclitaxel mostly had the greatest DRIs in 8505C, KAT4C and KAT18, and also demonstrated favorable DRIs in 8305C. doi:10.1371/journal.pone.0046726.g006 expression was calculated using untreated cells and KAT4C baseline values as a reference, or as indicated otherwise.

Cell cycle and apoptosis assessment
Cells were plated at 1610 5 cells per well in 6-well plates in 2 mL media overnight. BEZ235 or vehicle were added at indicated doses for 24 hours (KAT4C, 8505C, BHP7-13 and WRO82-1) or 72 hours (TT), adherent cells were trypsinized, washed with PBS, fixed with cold 70% ethanol and incubated with RNase A (100 mg/mL; Sigma) and propidium iodide (5 mg/mL; Sigma) at 37uC for 15 min. Cell cycle distribution was assessed by DNA content detected by flow cytometry (BD FACScalibur Flow Cytometer, BD Biosciences).
Cells were plated at 2610 4 cells per well in 6-well plates in 2 mL media overnight and treated with BEZ235 for 96 hours. Floating cells and trypsinized adherent cells were collected and samples were prepared as described above. Apoptotic sub-G1 cells were detected by DNA content by flow cytometry. Each condition was performed in triplicate.

Flank xenograft tumor therapy
Eight-week-old athymic female nude mice (National Laboratory Animal Center, Taiwan) were anesthetized with intraperitoneal injection of ketamine hydrochloride (90 mg/kg; Nang Kuang Pharmaceutical Co.) and xylazine hydrochloride (9 mg/kg; Bayer) before implantation of thyroid cancer cells. 8505C flank tumors were established by injecting 1610 6 cells in 100 mL PBS into the subcutaneous flanks of nude mice. When tumors reached 6 mm in mean diameter, mice (n = 7-8 per group) were treated with BEZ235 (50 mg/kg) or placebo by oral gavage daily for 25 days. Tumor dimensions were serially measured with electronic calipers twice a week, and the volumes were calculated by the formula a 2 6b60.4, where a represents the smallest diameter and b is the perpendicular diameter. The body weight of each animal was followed as a marker of toxicity.
Tumor levels of p-AKT, p-S6 ribosomal protein, PCNA and caspase-3 were evaluated in mice treated with a single dose of BEZ235 (50 mg/kg). At indicated periods, carbon dioxide was used for euthanasia, tumors were then harvested, mixed with protein extraction buffer (GE Healthcare), homogenized and sonicated on ice. After centrifugation, clarified supernatants were aliquoted and stored at 280uC until western blot was performed. Tumors were also fixed in 10% formalin and paraffin embedded. Sections (5 mm) were incubated with rabbit p-S6ribosomal protein antibody (1:100) at room temperature for 30 minutes, followed by poly-horseradish peroxidase anti-rabbit IgG reagent, and diaminobenzidine utilized to visualize the complex. Sections were counterstained with hematoxylin. This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of Chang Gung Memorial Hospital and the protocol was approved by the Committee of Laboratory Animal Center at Chang Gung Memorial Hospital, Linkou (Permit Number: 2008112401).

Quantitative analysis of drug interactions and statistical analyses
Interactions between BEZ235 and the chemotherapeutic drugs for each cell line were determined by calculating the combination index (CI) of Chou-Talalay equation where CI,1 is synergism, = 1 is addictive, and .1 is antagonism [28,29]. The dose-effect analysis was produced using the computer software, CompuSyn [27,30], following the dose and effect data entries. The dose reduction index (DRI) was also determined. DRI represents the fold dose-reduction permitted by the combination, for a given effect level, when compared with each drug alone [30].
Comparisons were performed when appropriate using twosided Student's t test, and correlations using Pearson's coefficients (Excel, Microsoft). Results were expressed as the mean 6 SE.
In conclusion, BEZ235 effectively inhibits the proliferation of four different histologic types of thyroid cancer. The therapeutic effect and safety profiles are favorable in nude mice bearing 8505C xenograft tumors. Importantly, BEZ235 synergistically enhances the therapeutic effect of paclitaxel in treating ATC. These data support future clinical trials investigating the utility of BEZ235 as an agent to treat patients with refractory thyroid cancer.

Supporting Information
Appendix S1 Basal expression of p27 in thyroid cancer cell lines. (TIF) Appendix S2 Reported data of genetic alterations in thyroid cancer cell lines and Dm of BEZ235. (TIF) Figure S1 BEZ235 consistently inhibits mTORC1 signaling and activates p-ERK1/2. A, p-AKT (Thr308) was transient reduced by 2 hours in TT and BHP7-13 and increased by 4 hours in 8505C and KAT4C. B, p-AKT (Ser473) was reduced by 2 and 8 hours in TT and BHP7-13 and increased by 2 hours in 8505C and KAT4C. C, p-4E-BP1 (Thr70) was consistently reduced by 2 hours in 3 cell lines. D, p-4E-BP1 (Thr37/46) was reduced by 2 hours. The extent of repression is statistically significant from 2 to 24 hours in BHP7-13 compared with baseline level. E, p-S6 ribosomal protein (Ser235/236) was decreased by 2 hours. The inhibition was significant in 8505C and TT from 4 to 24 hours compared with basal level. F, p-ERK1/2 (Thr202/Tyr204) was activated by 2 to 4 hours in four cell lines. * denoted P,0.02. (TIF) Figure S2 BEZ235 inhibits p-S6 ribosomal protein (Ser235/ 236) and activates p-ERK1/2 and p27 in KAT4C. A, p-AKT (Thr308), p-AKT (Ser473) and p-ERK1/2 (Thr202/Tyr204) was increased by 4 hours and persisted for more than 24 hours. p-S6 ribosomal protein (Ser235/236) was reduced from 8 hours through over 24 hours. B, p27 was increased by 24 hours. p53, p21 and cyclin D1 were increased by 4 to 24 hours. (TIF) Figure S3 BEZ235 activates p27 expression in thyroid cancer cell lines. A, p53was decreased by 2 to 4 hours in 8505C and TT and increased in KAT4C by 8 hours. B, p21 was repressed in TT and BHP7-13 and increased in KAT4C by 8 hour. C, p27 was increased in all cell lines. The elevation of p27 achieved statistical significance at 8 hour in TT. There was no detectable signal in basal 8505C cells, therefore the reference data was at 4 h. D, Cyclin D1 was gradually decreased in 8505C and BHP7-13 and increased at certain time points in KAT4C and TT. There was no detectable signal in untreated KAT4C, therefore the reference data was 8 h. * denoted P = 0.019 (t-test). (TIF) Figure S4 BEZ235 degrades caspase-3 in vitro. Quantification of immunoblot showed caspase-3 was decreased in KAT4C and KAT18 at 72 hours at doses ranged from 6.25 to 100 nmol/L. (TIF) Figure S5 BEZ235 represses the expression of p-AKT, p-S6 ribosomal protein and capase-3 in vivo. The effects of BEZ235 on p-AKT, p-S6 ribosomal protein, PCNA and capase-3 in vivo. Quantification of immunoblot showed BEZ235 greatly repressed p-AKT (Thr308), p-S6 ribosomal protein (Ser235/236) and caspase-3 by 2 to 4 hours with durable effects. PCNA was slightly reduced at an early time point. (TIF) Figure S6 Five chemotherapeutic agents induce dose and time dependent cytotoxicity in 4 anaplastic thyroid cancer cell lines. Dose-response curves were obtained on day 3 and 4 from cells treated with serial dilutions of chemotherapeutic agents (paclitaxel, irinotecan, etoposide, 5-FU, doxorubicin) for a 4-day course on ATC cell lines using LDH assays. All drugs demonstrated dose and time dependent cytotoxicity in four cell lines. (TIF)