Mammalian Target of Rapamycin Contributes to the Acquired Apoptotic Resistance of Human Mesothelioma Multicellular Spheroids*

When grown as three-dimensional structures, tumor cells can acquire an additional multicellular resistance to apoptosis that may mimic the chemoresistance found in solid tumors. We developed a multicellular spheroid model of malignant mesothelioma to investigate molecular mechanisms of acquired apoptotic resistance. We found that mesothelioma cell lines, when grown as multicellular spheroids, acquired resistance to a variety of apoptotic stimuli, including combinations of tumor necrosis factor-related apoptosis-inducing ligand (TRAIL), ribotoxic stressors, histone deacetylase, and proteasome inhibitors, that were highly effective against mesothelioma cells when grown as monolayers. Inhibitors of the phosphatidylinositol 3-kinase/Akt/mammalian target of rapamycin (mTOR) pathway, particularly rapamycin, blocked much of the acquired resistance of the spheroids, suggesting a key role for mTOR. Knockdown by small interference RNA of S6K, a major downstream target of mTOR, reproduced the effect of rapamycin, thereby confirming the role of mTOR and of S6K in the acquired resistance of threedimensional spheroids. Rapamycin or S6K knockdown increased TRAIL-induced caspase-8 cleavage in spheroids, suggesting initially that mTOR inhibited apoptosis by actions at the death receptor pathway; however, isolation of the apoptotic pathways by means of Bid knockdown ablated this effect showing that mTOR actually controls a step distal to Bid, probably at the level of the mitochondria. In sum, mTOR and S6K contribute to the apoptotic resistance of mesothelioma cells in three-dimensional, not in two-dimensional, cultures. The three-dimensional model may reflect a more clinically relevant in vitro setting in which mTOR exhibits anti-apoptotic properties.

standing the molecular foundations of apoptotic resistance could lead to improvements in the utility of current therapies or provide new therapies altogether. In vitro, apoptotic resistance of tumor cells is usually studied using two-dimensional monolayers. However, cells can attain increased resistance when grown in three-dimensional structures, a phenomenon referred as the acquisition of multicellular resistance (3). This acquired property of tumor cells may help explain why promising findings from in vitro studies have not been easily translated into therapy (4). Consequently, in the last decade, multicellular spheroids have become a valuable tool in the study of solid tumors by representing a threedimensional system of intermediate cellular complexity between monolayer cell cultures and tumors in vivo (5,6).
Human malignant pleural mesothelioma, an aggressive thoracic cancer that arises from the pleural mesothelium, is characterized by a profound resistance to standard anti-neoplastic therapies. At present, no curative therapy is available (7). To investigate the apoptotic resistance of mesothelioma, we are increasingly utilizing in vitro three-dimensional models. Indeed, the use of a three-dimensional model appears highly appropriate to mesothelioma, a tumor that arises from a twodimensional mesothelial layer into a solid, dense three-dimensional structure and that naturally forms aggregates in vivo, a feature once considered diagnostic (8,9). In previous studies using three-dimensional tumor fragment spheroids grown from human mesothelioma tumor, we showed that mesothelioma cells demonstrated a high degree of resistance compared with mesothelioma cells in monolayer culture (10). Using these tumor fragment spheroids, we previously found that blockade of the Akt/mTOR 2 pathway sensitized mesothelioma cells to treatment (10), suggesting that some of the resistance of tumor cells in their tissue environment derived from the PI3K/Akt/ mTOR pathway. Limited by the complexity of tumor fragments, we decided to develop a more tractable in vitro model of multicellular spheroids to investigate the role of this pathway further.
The PI3K/Akt/mTOR pathway plays a pivotal role in tumor cell survival and resistance to chemotherapy (11) and is acti-vated in most tumors (12), including mesothelioma (13,14). The roles of Akt and its myriad downstream targets are complex and interrelated (15), opening possibilities for specific targeting of certain functions. For example, in studies using animal models, the key tumorigenic and anti-apoptotic actions of Akt were shown to be mediated by its downstream kinase, mTOR (16,17). From these and others studies, mTOR has been found to mediate survival, through either of its two downstream targets, S6K or 4E-BP1/eIF4E (18,19), suggesting that blockade of mTOR would be highly beneficial. Blockade of mTOR, however, can also lead to a rebound upstream activation of Akt that can perhaps negate or limit the effects of mTOR inhibition, especially if Akt has survival activity separate from its effects via mTOR (20). Nonetheless, finding an important role for Akt or for mTOR in mesothelioma cell survival would open the door to the use of promising highly specific and multitargeted inhibitors (21).
In this study, based on our prior work in human tumor fragments, we have developed an in vitro three-dimensional model using mesothelioma cell lines to determine whether mesothelioma cells acquire multicellular apoptotic resistance. Then, using pharmacologic inhibition and RNA interference, we evaluated the contribution of the PI3K/Akt/mTOR pathway to this resistance and discovered an important role for the kinase, mTOR. Finally, we investigated the step(s) in the apoptotic pathways where mTOR exerts its anti-apoptotic effects.

Reagents and Inhibitors
Inhibitors and apoptotic agents were purchased and prepared in DMSO as follows, with the stock concentration shown: rapamycin (5 mM), LY294002 (10 mM), MG-132 (20 mM), trichostatin A (10 mM), anisomycin (5 mg/ml) (all from Sigma-Aldrich) and PI-103 (1 mM) (a kind gift of Dr. Kevan Shokat, San Francisco, CA). Sodium butyrate (1 M stock solution) (Sigma-Aldrich) was prepared in water. Final concentrations of inhibitors were chosen based on prior work (10,22). Concentrations of apoptotic agents were derived from the literature (23)(24)(25). Anisomycin was used at a low, subtoxic concentration (25 ng/ml) that we have previously shown does not inhibit the proliferation of our cells (23). In this study, we also show that anisomycin at this concentration does not affect the rate of protein synthesis (supplemental Fig. S5). DMSO was used as a vehicle control as appropriate. Recombinant human TRAIL was from R&D Systems (#375-TEC, Minneapolis, MN).

Cell Cultures
The human mesothelioma cell lines M28 (from Dr. Brenda Gerwin, NCI, National Institutes of Health, Bethesda, MD), REN (from Dr. Roy Smythe, University of Texas M.D. Anderson Cancer Center, Houston, TX), and VAMT (from Dr. B. Gerwin, NCI, NIH) were all cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum in a 37°C humidified incubator with 5% CO 2 . Normal human mesothelial cells were cultured from ascites fluid from patients without infection or malignancy according to a protocol approved by the UCSF Committee on Human Research. All cells were confirmed to be negative for mycoplasma every 2 months by PCR analysis as previously described (26).

Spheroid Generation
Spheroids of a consistent cell number and size were generated in non-adhesive round bottomed 96-well plates. The 96-well plates were rendered non-adhesive by coating them with a 1:24 dilution of polyHEMA (120 mg/ml, Sigma-Aldrich) in 95% ethanol and drying them at 37°C for 48 h (27). Plates were sterilized by UV light for 30 min prior to use. To generate spheroids and monolayers, cells were plated 24 h before each experiment; cells were added into each well of the 96-well plates to form spheroids or into each well of cell culture-treated 6-well plates to form monolayers. Then, before each experiment, 15-20 spheroids were transferred to each well of a 24-well poly-HEMA-coated plate. The number of spheroids per well was chosen to match the numbers of cells plated per well in the monolayers (150,000 -200,000). Apoptotic agents with or without inhibitors (and the appropriate DMSO vehicle control) were added to spheroids and to monolayers at the same time.
Microspheroids generated from an average of 300 cells per spheroid were produced as described (28). In brief, 3% agarose gels (Ultrapure agarose, Invitrogen) were cast using micromolds (Dr. Jeffrey Morgan, Brown University). After setting, gels were equilibrated overnight with culture medium. Cells were trypsinized, counted, and resuspended to the desired cell density and then overlaid onto the gel, where cells sedimented into calibrated recesses. Microspheroids recovered from micromolds by centrifugation were then transferred to 24-well polyHEMA-coated plates for experiments, as above.

RNA Interference
M28, REN, or VAMT cells (5 ϫ 10 6 ) were pelleted and resuspended in 100 l of nucleofection buffer (solution V, Amaxa Biosystems, Cologne, Germany) with 3 g of the appropriate siRNA duplex. This suspension was transferred to a sterile cuvette and nucleofected using program T-20 on a Nucleofector II device (Amaxa Biosystems). After recovery for 30 min in complete Dulbecco's modified Eagle's medium, the cells were plated and allowed to grow for 24 h. Cells were then trypsinized, counted, and plated as monolayers and spheroids for 24 h and exposed to apoptotic stimuli (48 h after transfection). The siRNA sequences (Ambion, TX) were: S6K #1, CUG UUA GUU UCA CAU GAC CdTdT; S6K #2 (to a non-overlapping sequence), AAA CAC UCC UGC CAU GUC CdTdT; Bid, UAU UCC GGA UGA UGU CUU CdTdT; and scramble, ACG UGA CAC GUU CGG AGA AdTdT.

Analysis of Apoptosis
Annexin V Binding-Monolayers and spheroids were exposed to apoptotic stimuli, trypsinized under identical conditions and then resuspended in binding buffer (10 mM Hepes/ NaOH, pH 7.4, 140 mM NaCl, 2.5 mM CaCl 2 ). Cells were stained with annexin V and propidium iodide and analyzed, using standard techniques as described previously (29).
Hoechst Staining-After the indicated apoptotic stimuli, monolayers and spheroids were disaggregated to single cells with trypsin under identical conditions, taking ϳ5 min (Invitrogen). Trypsin was inactivated with serum-containing medium, and cells were pelleted at 2000 revolutions/minute for 10 min at 4°C, fixed with 2.5% glutaraldehyde (Sigma-Aldrich), stained with 8 g/ml Hoechst 33342 (Molecular Probes, Invitrogen), and placed onto slides. For each condition, at least 100 cells were counted in triplicate by investigators blinded to the experimental conditions. Cells with distinctive signs of nuclear condensation were considered apoptotic. After direct comparison with the annexin V assay, we found Hoechst staining to be a more consistent and reliable quantitative method for assessing apoptosis in both spheroids and monolayers and used it for the analysis of most experiments.

Immunohistochemistry
Spheroids were collected, fixed in 10% formalin in PBS overnight at 4°C, and set in 5% agar pellets that were embedded in paraffin. From the paraffin blocks, 8 m sections were cut and mounted on glass slides, deparaffinized, rehydrated, and boiled in sodium citrate solution (pH 6.0) with 0.1% Tween 20 in a pressure cooker for 10 min for antigen retrieval. After cooling for 30 min at room temperature, sections were blocked with hydrogen peroxide solution for 20 min to remove endogenous peroxidase. Primary antibodies to P-Akt or P-S6K were applied overnight at 4°C. The secondary antibody conjugated to a horseradish peroxidase-labeled polymer was applied for 30 min and detected by the 3,3-diaminobenzidine tetrahydrochloride method (EnVision, Doublestain, DAKO, Carpinteria, CA).

Immunoblotting
After exposure to apoptotic stimuli, monolayers or spheroids were lysed in boiling lysis buffer (2.5% SDS, 250 mM Tris-HCl, pH 7.4). Total protein concentration of samples was evaluated with a colorimetric assay (DC Protein Assay, Bio-Rad). 30 -50 g of total cell lysates was loaded in reducing conditions (0.2 M Tris, pH 6.8, 5% SDS, 3% glycerol, 0.01% bromphenol blue, and 200 mM dithiothreitol). After separation in SDS-PAGE (5-15% acrylamide) and transfer to polyvinylidene difluoride (Immobilon, Millipore, Billerica, MA), membranes were blocked with a protein-free Tris-buffered saline blocking buffer (Pierce) and probed with antibodies diluted in 5% milk or 5% bovine serum albumin, as appropriate, at 4°C overnight. Secondary antibodies were from Amersham Biosciences. The signal was detected by the enhanced SuperSignal West Pico Chemiluminescent Substrate (Pierce). For each experiment, a representative blot is shown, and tubulin staining of the same blot is used for confirmation of equal protein loading.

Statistical Analysis
Data from each experiment are expressed as mean Ϯ one S.D. of at least three different experiments. One-or two-way analysis of variance was used to evaluate statistical significance, and the Tukey test was performed to detect where the differences lay (Prism version 4.0, GraphPad Software, Inc.). A p value of Ͻ0.05 was considered significant; the significance reached is specified in the text.

Mesothelioma Cells Readily Formed Multicellular Spheroids within 24 h-
To determine the characteristics and kinetics of spheroid formation by mesothelioma cells, we used two epithe-  MAY 9, 2008 • VOLUME 283 • NUMBER 19 lial (M28 and REN) lines and one sarcomatous (VAMT) line as representative of the histological subtypes of this tumor. Spheroids can be generated by various methods (30); after evaluating several techniques, we found that plating cells in polyHEMAcoated non-adhesive 96-well plates generated the most reproducible and consistent spheroids (27). When seeded in these non-adhesive conditions, cells from all tested mesothelioma cell lines aggregated and formed multicellular spheroids within 24 h. Spheroids of a range of cell numbers were tested for the presence of baseline apoptosis. Spheroids formed from 25,000 cells or greater showed baseline cleavage of PARP in M28 and REN cell lines as evidence of apoptosis (Fig. 1A). In addition, because spheroids formed from 5,000 or fewer cells by this technique were loosely formed, we chose spheroids formed from 10,000 cells each for further experiments (Fig. 1B). Spheroids assumed a discoid shape with a width of 500 -1000 m and an average thickness of 100 m, a shape we considered an advantage for minimizing diffusion distances. To confirm that acquired resistance was not dependent on a particular size or shape, we also generated microspheroids of ϳ300 cells per spheroid by means of micromolded non-adhesive hydrogels (28) and utilized them in some of the apoptosis studies (supplemental Fig. S1A). These microspheroids were spherical in shape, with most in the range of 50 -100 m in diameter and a few with a maximal diameter of 150 m. Thus, in both threedimensional structures, the mean maximal diffusion distance would be 25-75 m, a distance in which cells can be sustained by diffusion alone (31).

mTOR Supports Acquired Survival of Mesothelioma Spheroids
Mesothelioma Cells in Spheroids Acquire Resistance to Diverse Apoptotic Stimuli-Our group and others have found that mesothelioma cells in monolayer culture undergo apoptosis in response to combinatorial strategies using TRAIL plus either DNA or other damaging agents (23,32) or histone deacetylase or proteasome inhibitors (33,34). To assess whether mesothelioma three-dimensional cell structures acquired multicellular resistance, we tested these combinations of apoptotic agents on cells grown as monolayers and as spheroids. Compared with monolayers, spheroids from all cell lines showed a significant resistance to these TRAIL-containing regimens ( Fig. 2A).
To evaluate whether the apoptotic agents penetrated into the spheroid (35,36), we performed immunohistochemistry on spheroids exposed for 6 h to TRAIL or to no treatment. TRAIL was detected evenly throughout the spheroid in TRAIL-exposed but not in unexposed spheroids (Fig. 2B). We also assessed the response of microspheroids, which had a lower mean diffusion distance. The microspheroids also demonstrated apoptotic resistance, suggesting that limited diffusion was not the explanation for the resistance (supplemental Fig. S1B).
To determine whether apoptotic resistance was limited to TRAIL and its engagement of the extrinsic apoptotic pathway, we sought non-TRAIL agents that would induce apoptosis in FIGURE 2. Multicellular spheroids acquire resistance to apoptosis. A, mesothelioma spheroids display resistance to various apoptotic stimuli. M28, REN, and VAMT monolayers and spheroids exposed to TRAIL (1 ng/ml) plus the proteasome inhibitor MG-132 (2.5 M) or the histone deacetylase inhibitor trichostatin A (250 nM) for 24 h were studied for apoptosis using annexin V staining and confirmed by examination of condensed nuclei following Hoechst staining. Spheroids consistently showed resistance to the different apoptotic stimuli in all cell lines tested (n ϭ 3, *, p Յ 0.05 monolayer versus spheroids, mean Ϯ S.D.). B, TRAIL diffuses uniformly within spheroids. After M28 spheroids were exposed to TRAIL (1 ng/ml) for 6 h, immunohistochemistry was performed to detect TRAIL within spheroids. TRAIL diffusion into the spheroids was complete within 6 h as demonstrated by the uniform staining of spheroids exposed to TRAIL, with no staining seen in unexposed spheroids (control) (representative images of four separate experiments). C, spheroids also display resistance to non-TRAIL combinations at 24 h. Combinations of non-TRAIL apoptotic agents that were found to induce apoptosis in mesothelioma cells in monolayers were then tested on monolayers and spheroids of M28, REN, and VAMT cells. Monolayer cells underwent apoptosis following MG-132 (2.5 M) together with either sodium butyrate (10 mM) or trichostatin A (250 nM), whereas the same cells grown as spheroids were relatively resistant. Each agent, MG-132, trichostatin A, and sodium butyrate, had little effect alone, with apoptosis being Ͻ10% (not shown). Apoptosis was measured by quantification of nuclear condensation of Hoechst-stained cells, in this and remaining apoptosis studies (n ϭ 3, *, p Յ 0.01 monolayer versus spheroids, mean Ϯ S.D.). D, spheroids are also resistant to TRAIL plus the sensitizing agent, anisomycin, at 6 h. After exposure to TRAIL (1 ng/ml) plus anisomycin (25 ng/ml) for 6 h, all three mesothelioma cell lines showed apoptosis, whereas spheroids displayed resistance. The same resistance was seen at 12 and 24 h (not shown) (n ϭ 3, *, p Յ 0.001 monolayer versus spheroids, mean Ϯ S.D.).

mTOR Supports Acquired Survival of Mesothelioma Spheroids
monolayers and tested them in spheroids. As in our prior experience, no single agent (MG-132, gemcitabine, trichostatin A, sodium butyrate, or etoposide) was found to induce apoptosis in mesothelioma cell monolayers (data not shown). Combinations of every two agents were then tried; MG-132 plus trichostatin A or MG-132 plus sodium butyrate were the most effective at inducing apoptosis of cells in monolayers. When these apoptotic stimuli were tested in spheroids, all cell lines consistently showed resistance (Fig. 2C), indicating that acquired resistance was a general feature and was not limited to TRAILcontaining therapies.
For further studies of acquired resistance, we chose to use the combination of TRAIL with anisomycin at a low concentration (25 ng/ml) that we have shown induces JNK activation without toxicity (23) (see also supplemental Fig. S5). The combination of TRAIL plus anisomycin induces apoptosis rapidly and consistently in our mesothelioma cells (23). In response to TRAIL plus anisomycin, monolayers demonstrated extensive apoptosis after only 6 h, whereas spheroids, whether treated for 6 h or up to 24 h (data not shown), failed to undergo apoptosis (Fig. 2D). TRAIL plus anisomycin was chosen for further experiments.
Inhibitors of PI3K/Akt/mTOR Reduce the Apoptotic Resistance of Multicellular Spheroids-Our laboratory has previously reported that inhibitors of the PI3K/Akt/mTOR pathway are able to sensitize spheroids obtained from actual human mesothelioma tumor to TRAILcontaining apoptotic combinations (10). Thus, we aimed to investigate the effectiveness of these inhibitors in the multicellular spheroids. Inhibitors included rapamycin, a highly specific inhibitor of one complex of mTOR (TORC1), PI-103, a novel dual inhibitor of PI3K and both mTOR complexes (TORC1 and TORC2) (22), and LY294002, a broader and less specific inhibitor of PI3K (37). At baseline, although all three cell lines showed high levels of S6K phosphorylation (P-S6K), only two of the three showed Akt phosphorylation (P-Akt) (Fig. 3A). Rapamycin completely blocked P-S6K, as evidence of effective mTOR inhibition and consistently resulted in increased P-Akt (Ser-473), a positive feedback described in other systems (38,39). PI-103 completely blocked both P-S6K and P-Akt. LY294002 inhibited P-Akt and reduced but did not completely block P-S6K, suggesting other inputs to S6K activation other than Akt (40). Similar findings were observed in spheroids (data not shown).
To learn whether inhibition of the PI3K/Akt/mTOR pathway reduced apoptotic resistance of spheroids, we then tested the inhibitors in monolayers and spheroids of the three cell lines exposed to TRAIL plus anisomycin. The inhibitors had little effect on the apoptotic response of monolayers but consistently reduced the resistance of spheroids to TRAIL plus ani- A, PI3K/Akt/mTOR pathway is inhibited differently by rapamycin, PI-103, and LY294002. M28, REN, and VAMT cells grown in monolayers were exposed to nothing, rapamycin (5 nM), PI-103 (1 M), or LY294002 (5 M) for 4 h and then studied by immunoblot for expression of P-Akt (Ser-473) and P-S6K (Thr-389) as markers of Akt and mTOR activity, respectively. Rapamycin strongly inhibited mTOR activity in all tested cells but also produced a feedback Akt activation, PI-103 potently inhibited both Akt and mTOR, and LY294002 inhibited Akt and partially inhibited mTOR. B, PI3K/Akt/mTOR inhibitors reduce apoptotic resistance of spheroids, as measured by nuclear morphology. M28, REN, and VAMT cells grown as monolayers or spheroids were treated with TRAIL (1 ng/ml) plus anisomycin (25 ng/ml, TϩA) with either no inhibitor, rapamycin (5 nM), PI-103 (1 M), or LY294002 (5 M) for 6 h. Apoptosis was measured by determination of characteristic nuclear morphology of Hoeschtstained cells. Inhibitors increased apoptosis in spheroids without a comparable effect in monolayers (n ϭ 3, *, p Ͻ 0.05 compared with no inhibitor, mean Ϯ S.D.). C, PI3K/Akt/mTOR inhibitors reduce apoptotic resistance of spheroids, as measured by PARP cleavage. For M28 cells studied under the same conditions as in Fig. 3B, apoptosis was assessed by PARP cleavage using immunoblot analysis. Inhibitors increased PARP cleavage induced by TRAIL plus anisomycin in spheroids while monolayers were not significantly affected. The pancaspase inhibitor z-VAD-fmk (50 M) inhibited PARP cleavage confirming a caspase-dependent apoptosis. D, rapamycin reduces apoptotic resistance of spheroids to two different apoptotic regimens similarly. M28, REN, and VAMT spheroids were treated with TRAIL (1 ng/ml) plus anisomycin (25 ng/ml) for 6 h or with trichostatin A (250 nM) plus MG-132 (2.5 M) for 24 h in the presence of rapamycin (5 nM), PI-103 (1 M), or DMSO control. The experiments using TRAIL plus anisomycin are those shown in Fig. 3B, whereas the experiments using trichostatin plus MG-132 are shown here for the first time. Apoptotic resistance is defined here as the difference between the apoptosis of the monolayer and spheroid divided by that of the monolayer. Rapamycin reduced apoptotic resistance of spheroids to the two apoptotic combinations from 73 Ϯ 14% to 44 Ϯ 17%, or by ϳ40%. PI-103, although targeting a broader range of kinases, was no more effective than rapamycin (n ϭ 3). MAY 9, 2008 • VOLUME 283 • NUMBER 19 somycin (Fig. 3B). Apoptosis was also confirmed with analysis of PARP cleavage by immunoblot (Fig. 3C). As expected, rapamycin also reduced apoptotic resistance of M28 cells when grown as microspheroids (supplemental Fig. S1B).

mTOR Supports Acquired Survival of Mesothelioma Spheroids
We concluded from these results that the PI3K/Akt/mTOR pathway contributes to acquired apoptotic resistance and that, because rapamycin was as effective as other inhibitors with a broader range of targets, the resistance due to PI3K/Akt/mTOR was mostly mediated by mTOR. A major role for mTOR was further supported by considering that mTOR inhibition with rapamycin reduced acquired apoptotic resistance despite an associated increase of P-Akt (see Fig. 3A).
The apoptotic resistance was calculated as the difference between the percentage of apoptosis for monolayer and spheroid divided by that of the monolayer and shown for TRAIL plus anisomycin (the experiments presented in Fig. 3B) and for additional experiments using trichostatin A plus MG-132. Overall, rapamycin reduced the apoptotic resistance of spheroids by ϳ40%, from 73 Ϯ 14% to 44 Ϯ 17% (Fig. 3D). Because PI-103, despite its broader inhibitory effect, was not significantly more effective than rapamycin, we decided to focus on mTOR by using rapamycin for further studies.

PI3K/Akt/mTOR Activity Is Down-regulated in Spheroids Compared with Monolayers but Remains Higher Than in Normal Mesothelial Cells-Because inhibitors had a greater effect
in spheroids than in monolayers, we evaluated the activity of the PI3K/Akt/mTOR pathway in spheroids. Somewhat surprisingly, the baseline phosphorylation of PI3K/Akt/mTOR pathway members in all spheroids was reduced compared with that of the monolayers (Fig. 4A). Nonetheless, the levels of P-Akt, detected by phosphorylation at either Ser-473 or Thr-308, and P-S6K in spheroids were clearly elevated compared with that of normal mesothelial cells (Fig. 4B).
To determine whether spheroids displayed regional variation in activity of this pathway, we stained M28 spheroids with phospho-specific antibodies to localize P-Akt and P-S6K. P-Akt and P-S6K were found to be uniform throughout the spheroid (Fig. 4C).
S6K Mediates Apoptotic Resistance in Spheroids-To identify the molecular pathway underlying the resistance to apoptosis displayed by spheroids and to confirm a role for mTOR, we interrupted mTOR signaling more specifically by ablating the activity of one of its downstream effectors. We selected S6K as the target for ablation, because it has been recognized as a major kinase involved in many diseases, including cancer (41), and is known to be the target of mTOR that is fully inhibited by rapamycin (42). The other major target of mTOR, 4E-BP1, has rapamycin-insensitive phosphorylation sites and thus may not be fully inhibited by rapamycin (43). Furthermore, interference with 4E-BP1/eIF4E results in phosphorylation events downstream of mTOR by unknown mechanisms (19,44).
We utilized an RNA interference strategy to knock down S6K for the time needed for spheroid formation and for exposure to FIGURE 4. In spheroids, the baseline activity of Akt/mTOR pathway is down-regulated but remains more activated than in non-malignant mesothelial cells. A, Akt/mTOR activity in spheroids is reduced compared with monolayers. Akt/mTOR pathway activity of M28, REN, and VAMT mesothelioma cells was reduced in the spheroids (s) as demonstrated by a lower expression of P-Akt (Ser-473), P-S6K (Thr-389), and P-4E-BP1 (Thr-37/46) compared with that in the monolayers (m). The total levels of Akt, S6K, and 4E-BP1 show no consistent change, and the tubulin levels confirm equal loading. B, Akt/mTOR activity in spheroids is greater than in normal mesothelial cells. Normal mesothelial cells (NMC) monolayers and both monolayers and spheroids from M28, REN, and VAMT cells were analyzed by immunoblot for P-Akt (Ser-473 or Thr-308) and P-S6K. In this experiment, cells were grown in reduced serum (2% fetal bovine serum) to reduce growth factor stimulation of the Akt/mTOR pathway. Although phosphorylation of Akt/mTOR pathway members in spheroids is lower than in the monolayer, spheroids nonetheless demonstrate a greater activation of the pathway than in their non-malignant counterparts. Below, the P-S6K panel has been overexposed to show more clearly the higher level of P-S6K in spheroids than in normal mesothelial cells. C, Akt and S6K phosphorylation is uniform throughout the spheroid. Staining of M28 spheroids with antibodies against P-Akt (Ser-473) and P-S6K (Thr-389) shows a uniform pattern of Akt and S6K phosphorylation throughout the spheroid. For the P-Akt, specificity is shown by a lack of staining with inclusion of a P-Akt Ser-473 blocking peptide (negative control). For the P-S6K staining, no primary antibody was used as a control and resulted in no staining (not shown). Two magnifications (20ϫ and 40ϫ) are shown.
TRAIL plus anisomycin. Knockdown of S6K significantly increased the apoptotic response of spheroids to TRAIL plus anisomycin when compared with control spheroids (Fig. 5). Moreover, the S6K knockdown reproduced the effects that rapamycin had on control spheroids, and the addition of rapamycin to S6K-kd spheroids did not further increase their apoptotic rate, suggesting that the effects of S6K-kd and rapamycin were identical. The same results were obtained using an siRNA duplex targeting a separate, non-overlapping region of S6K mRNA (data not shown). Data are shown for M28 cells; similar results were seen in REN and VAMT cells (not shown).
Bcl-2 and FLIPs Are Up-regulated in Some Spheroids but Are Not Modulated by Rapamycin or S6K Knockdown-To find possible mediators of the mTOR-dependent resistance, we measured the levels of a panel of pro-/anti-apoptotic proteins known to be important in mesothelioma resistance to apoptosis (45). Two anti-apoptotic proteins, Bcl-2 and FLIPs, which could account for acquired resistance, were up-regulated in two of the mesothelioma cell lines upon spheroid formation (Fig. 6A). However, neither rapamycin nor S6K knockdown reduced the level of Bcl-2 or FLIPs in spheroids, suggesting that these antiapoptotic proteins did not account for the resistance mediated by mTOR (Fig. 6B).
mTOR Inhibits Apoptosis at the Level of the Mitochondria-In mesothelioma cells, the synergistic combination of TRAIL with mitochondria-sensitizing agents has been shown to engage both the death receptor and the mitochondrial pathways of apoptosis (46). To determine where mTOR might affect the apoptotic machinery, we measured caspase-8 cleavage following TRAIL plus anisomycin in monolayers and spheroids in the presence or absence of rapamycin. Interestingly, when treated with TRAIL plus anisomycin, spheroids demonstrated a reduced caspase-8 cleavage compared with monolayers; mTOR inhibition restored caspase-8 cleavage and apoptosis (Fig. 7A). We considered that the increase in TRAIL-induced caspase-8 cleavage after mTOR inhibition could indicate either an action of mTOR at the level of the death receptor or alternatively at the level of the mitochondria, where amplification of apoptotic signals leads to activation of multiple caspases leading to a feedback activation and cleavage of caspase-8 (47) (Fig. 8).
We then ablated the pro-apoptotic protein Bid (Bid-kd) using RNA interference in order to block the mitochondrial amplification step downstream of TRAIL and to assess whether mTOR was acting proximal to Bid, at the level of death receptors, or distal to Bid, at the mitochondria (see Fig. 8). In control cells, as expected, cells in monolayer culture were sensitive to TRAIL plus anisomycin-induced apoptosis, whereas cells in spheroids were resistant but rendered more sensitive after rapamycin. In cells without Bid, however, monolayers and spheroids failed to undergo apoptosis after TRAIL plus anisomycin, and rapamycin did not increase the apoptotic rate in the spheroids (Fig. 7B). We then evaluated caspase-8 cleavage in Bid-kd spheroids following TRAIL plus anisomycin with or without rapamycin. Importantly, the ability of rapamycin to increase caspase-8 cleavage in the control spheroids following FIGURE 5. S6K ablation inhibits the acquired apoptotic resistance of spheroids and reproduces the effect of rapamycin. Upper panel, the knockdown of S6K (S6K-kd) reduced the resistance of M28 spheroids to TRAIL plus anisomycin (see the asterisk). The effect of S6K-kd was the same as the effect of rapamycin on control cells. Indeed, when combined, rapamycin did not add to the effect of S6K-kd, suggesting that they acted identically. Conversely, monolayers were not significantly affected either by S6K siRNA or by rapamycin (n ϭ 3, *, p Յ 0.001 scramble versus S6K duplexes, mean Ϯ S.D.). Lower panel, S6K was effectively ablated by siRNA, as measured 48 h after transfection, the time when agents are added. TRAIL plus anisomycin was lost in the Bid-kd spheroids (Fig. 7C). The fact that Bid was absolutely required in order for rapamycin to enhance apoptosis and caspase-8 cleavage indicates that mTOR inhibits an apoptotic step distal to Bid, presumably at the level of the mitochondria, not at the death receptor.

DISCUSSION
A compelling strategy for cancer therapy is to undermine cancer's extensive defenses against apoptosis, thereby either killing cancer cells outright or lowering their defenses against standard therapies (48). Understanding the molecular mechanisms of apoptotic resistance thus assumes an important role in counteracting cancer resistance to therapy. It has become apparent in recent years that the standard twodimensional monolayer in vitro model does not exhibit the full range of resistance to apoptosis seen in human cancers (49). Cancer cells have been found to acquire apoptotic resistance when grown as threedimensional structures that may model the apoptotic resistance of human solid tumors (3). In our study, by using a three-dimensional system, we have identified a role for mTOR in cell survival that was not apparent in the same cells grown as monolayers. Indeed, this strategy has allowed us to identify mTOR as a contributor to acquired multicellular resistance to apoptosis of mesothelioma thereby suggesting that mTOR could be a useful target in this highly resistant tumor. It is true that mTOR/S6K inhibition did not induce apoptosis by itself, but it enhanced the response to other apoptotic stimuli, presumably by releasing an inhibitory step at the level of the mitochondria. It is also worth pointing out that mTOR/S6K inhibition did not remove all acquired resistance, but it did account for a significant portion (roughly 40%) that would be amenable to currently available inhibitors. Our results provide a fresh insight into the functions of mTOR/S6K in mesothelioma and perhaps in other refractory solid tumors.
mTOR has been increasingly recognized as a crucial kinase in cancer and metabolic diseases because of its role in the integration of diverse mitogenic and metabolic inputs (41) and its recognition as a mediator of survival (16,17,19,50,51). In addition, with its own inputs and restraints, mTOR can act independently of the PI3K/Akt pathway (52,53). Several of our observations point to the importance of mTOR signaling, as apart from Akt signaling, in mesothelioma cells. For one, mTOR/S6K was activated, as judged by an elevated P-S6K, in all our cells, whereas P-Akt was elevated only in M28 and VAMT (see Fig. 4A). For another, despite the increase in P-Akt with FIGURE 7. Rapamycin increases caspase-8 cleavage in spheroids by actions distal to Bid, presumably at the mitochondria. A, rapamycin increases TRAIL-induced caspase-8 cleavage in spheroids. Spheroids but not monolayers grown from M28 cells displayed increased caspase-8 cleavage when rapamycin (5 nM) was added to TRAIL plus anisomycin, as shown by the appearance of p26 and p41/43 caspase-8 fragments in the immunoblot performed after 6 h. The increase in caspase-8 cleavage was associated with increases in apoptosis, as shown by PARP cleavage. B, rapamycin increases apoptosis in spheroids by actions requiring Bid. Bid was knocked down by siRNA to isolate the death receptor from the mitochondrial apoptotic pathways. Upper panel, in control M28 cells, as expected, monolayer cells were sensitive to TRAIL plus anisomycin-induced apoptosis, whereas spheroid cells were resistant but rendered more sensitive after rapamycin. In cells without Bid, however, monolayers and spheroids failed to undergo apoptosis after TRAIL plus anisomycin, and rapamycin did not increase the apoptotic rate in the spheroids (n ϭ 3, *, p Յ 0.001 no inhibitor versus rapamycin, mean Ϯ S.D.). Lower panel, Bid was effectively ablated by siRNA, as measured 48 h after transfection, the time when agents are added. C, rapamycin increases caspase-8 cleavage in spheroids by actions distal to Bid. Bid was knocked down by siRNA to permit identification of the site of action of rapamycin. Bid-kd prevented the increase in caspase-8 cleavage induced by rapamycin in control spheroids treated with TRAIL plus anisomycin. Because Bid is necessary for the rapamycin-induced increase in caspase-8 cleavage, mTOR is acting, not at the death receptor pathway, but distal to Bid. Caspase-8 cleavage paralleled apoptosis as shown by PARP cleavage. Addition of z-VAD-fmk (50 M) inhibited both caspase-8 cleavage and apoptosis. FIGURE 8. mTOR/S6K contributes to acquired multicellular apoptotic resistance at the level of the mitochondria. Scheme of the apoptotic signaling circuitry activated by TRAIL plus mitochondria sensitizers or stressors. Extrinsic (death receptor) and intrinsic (mitochondrial) apoptotic pathways, linked by the BH3-molecule Bid, can activate an effective apoptosis in mesothelioma cells (47). Truncated Bid (tBid), generated by caspase-8 activated at the death-inducing signaling complex (DISC) by TRAIL, signals to the mitochondria. When the mitochondrial threshold for apoptosis is lowered by concurrent DNA damaging or other sensitizers such as anisomycin, the mitochondria can respond to tBid and activate downstream caspases and activate caspase-8 by feedback. In our study, Bid was necessary for the ability of rapamycin to enhance caspase-8 cleavage and apoptosis indicating that mTOR/S6K acted distal to Bid. (Schematic diagram modified from Refs. 23 and 47).
rapamycin (see Fig. 3A), rapamycin was as effective as the other inhibitors in reducing apoptotic resistance. This suggests that, within the PI3K/Akt/mTOR pathway, mTOR is the major mediator of the acquired apoptotic resistance in mesothelioma spheroids. Finally, knockdown of S6K reproduced the effect of rapamycin, confirming the role of the mTOR/S6K arm in the survival of spheroids.
By silencing Bid expression and removing an essential amplification step between the death receptor and mitochondrial apoptotic pathways (54), we localized the anti-apoptotic activity of mTOR/S6K to the level of the mitochondria. Indeed, inhibition of apoptosis at the mitochondrial level would account for the observed resistance to a wide array of apoptotic stimuli involving the intrinsic and extrinsic pathways (see Fig. 8). mTOR has also been shown to associate with mitochondria (55) possibly serving as a modulator of stress signals (56) and cell fate (57). Although we were able to localize the resistance to the mitochondria, we were not able to identify a specific molecule that transduced the survival function of mTOR/S6K in spheroids. In particular, with the development of a three-dimensional structure, there was no clear change in the abundance of many pro-/anti-apoptotic proteins thought to be important for mesothelioma apoptotic resistance (45) or for molecules known to be regulated by S6K such as phospho-Bad (supplemental Fig. S4) (18). Anti-apoptotic Bcl-2 and FLIPs, which were up-regulated in the transition from two-dimensional to three-dimensional, were found to be independent of mTOR/ S6K activity, suggesting that they may account for resistance that is not controlled by mTOR/S6K, at least in two of the cell lines. mTOR may also contribute to survival in general by its support of metabolism and energy (56) or of protein translation, key functions to which the cells may become reliant or "addicted" (58).
The survival function of mTOR became evident only in the three-dimensional setting. This may represent a redirection of the function of mTOR in three-dimensional, as has been noted for other signaling pathways that, with a transition from two-to three-dimensional, can be spatially reorganized (59,60), coupled to other pathways (61), or redirected downstream to different functions (59). In fact, mTOR has previously been noted to have a greater effect on survival in three-dimensional than in two-dimensional breast cancer cell cultures (62). The Akt/ mTOR pathway was clearly altered by the three-dimensional environment as seen in the reduction in phosphorylation of Akt and S6K (see Fig. 4, A and B). Because the signals can respond to stresses (see supplemental Fig. S2), we consider the reduction in phosphorylation to represent a down-regulation, instead of a suppression, of the pathway. Indeed, down-regulation may provide an improved signal to noise ratio, as can be achieved with modulation from feedback loops (63). An increase in PTEN activity in spheroids, as suggested by our finding of a decrease in the inhibitory phosphorylation at Ser-380 (see supplemental Fig. S3), might account for down-regulation by providing a brake on unrestrained PI3K activity. Such an increase in PTEN may parallel a general increase in phosphatases that has been described in spheroids compared with monolayers (64). Like other pathways, the Akt/mTOR pathway may thus be differ-ently regulated and assume different functions in the threedimensional setting.
Limited diffusion of macromolecules and the presence of a hypoxic core have been proposed to drive resistance of threedimensional cell models (35,36). We did not find evidence that the spheroids utilized in this study were affected by an impaired diffusion of agents. They had no baseline apoptosis, they showed homogeneous TRAIL diffusion and P-Akt/P-S6K staining, and they responded similarly as microspheroids of a different size and shape. Of note, the estimated maximal diffusion distance of the spheroids was 25-75 m, below the distance of 100 -200 m described as limiting for the diffusion of oxygen in vivo (31). Clearly though, there will be diffusion gradients in three-dimensional models that may be relevant to gradients existing in avascular units of tumor units (36).
In this study of three-dimensional spheroids, we have discovered a role for mTOR in survival that could provide a therapeutic rationale for the use of mTOR inhibitors against mesothelioma, probably as an adjunct to current therapies. Inhibitors of mTOR are currently in use or in clinical trials for several tumors (65). If inhibition of mTOR is found to be useful in mesothelioma, it will underscore the value of three-dimensional studies for revealing underlying causes of acquired multicellular resistance in tumors.