mTOR Inhibition via Displacement of Phosphatidic Acid Induces Enhanced Cytotoxicity Specifically in Cancer Cells

The serine/threonine kinase mTOR is a master regulator of cell growth, highly conserved among eukaryotes (1, 2). mTOR is organized in two structurally and functionally different complexes: the rapamycin-sensitive mTORC1 (mTOR complex 1) and the rapamycin-insensitive mTORC2 (mTOR complex 2; refs. 3–6). mTORC1 is mostly activated by the presence of amino acids, by growth factors, by the bioenergetics status of the cell, and by oxygen availability. In the control of mTORC1 by growth factors, the tuberous sclerosis complex (TSC) and the mTORC1 coactivator Rheb play a crucial role (7, 8). One of the mechanisms by which the TSC/Rheb pathway controls mTORC1 involves the production of phosphatidic acid (PA), which binds directly to mTOR at the FRB domain and activates mTORC1 downstream of TSC/Rheb. Indeed, the downregulation of PA production is sufficient to decrease mTORC1 activity (9, 10).

As a major cell growth regulator, mTORC1 is recurrently upregulated in cancer cells to allow rapid growth of tumors (11). Indeed, the use of rapamycin analogues has been approved as anticancer therapy for certain types of cancer. However, the results of these treatments are very modest with respect to patient survival and quality of life (12). Several reasons have been invoked for these modest results in the clinics, including the reactivation of a negative feedback loop downstream of mTORC1 that activates the PI3K pathway (13), the absence of mTORC2 inactivation upon rapamycin treatment (5), and recently, the potential features of mTORC1 as a tumor suppressor (14, 15). Still, inhibition of mTOR and the design of new compounds that increase cancer cytotoxicity upon mTOR inhibition are active fields of research, with recent reports proposing new-generation mTOR inhibitors that overcome resistance to mTOR inhibition in tumors and effectively induce tumor regression (16). However, to date, most of these mTOR inhibitors tested either showed a limited cytotoxicity toward cancer cells (having mostly a cytostatic effect) or showed an excessive cytotoxicity toward noncancer cells, thus increasing adverse side effects.

Recently, we reported the synthesis and cytotoxicity of ICSN3250, an analogue of the cytotoxic marine alkaloid halitulin (see Fig. 1A for the chemical structure of this compound; ref. 17). Halitulin was firstly reported in 1998 as a bisquinolinylpyrrole isolated from the sponge Haliclona tulearensis, showing cytotoxicity against several tumor cell lines (18). Our previous work concluded with the synthesis of a panel of halitulin analogues through the formation of N-substituted 3,4-diarylpyrroles. Among them, ICSN3250 (also called compound 25) was selected as a very potent derivative, presenting a high cytotoxicity at a nanomolar concentration in a caspase-independent cell death mechanism (17). Our preliminary results indicated an increased autophagy in cancer cells treated with ICSN3250. However, the exact mechanism of action of ICSN3250 underlying its toxicity and the specificity of this cytotoxicity toward highly proliferative (cancer) cells were not examined previously.

In this report, we investigated the molecular mechanism by which ICSN3250 induces toxicity in cancer cells. Starting from a targeted screening analyzing several signaling pathways, we identified the mTORC1 pathway as a main target inhibited by ICSN3250 in the nanomolar range. Our results indicated that ICSN3250 inhibited mTORC1 by following an unprecedented mechanism that involved its competition with PA at the FRB domain of mTOR to overcome the TSC-negative regulation of mTORC1. This particular mechanism of mTORC1 inhibition led to a potent and selective cytotoxicity observed in cancer cells upon ICSN3250 treatment, which was not observed in noncancer cells. Our results thus defined ICSN3250 as a new class of mTORC1 inhibitor and validated ICSN3250 as a potential anticancer drug for future clinical assays.

ICSN3250 (5,5′(1-(3-(azacyclotridecan-1-yl)propyl)-1H-pyrrole-3,4-diyl)bis(3-nitrobenzene-1,2-diol)) was synthesized as described previously (17) and in a published patent application WO2014/060366 (19). Briefly, a new efficient “one-pot” method of unsymmetrically substituted pyrroles synthesis was applied. It includes the condensation of an α-haloketone, first with a primary amine and then with an aldehyde. Subsequent intramolecular cyclization of this in situ–generated β-ketoenamine (enamine onto a ketone) results in formation of pyrrole-based ICSN3250 molecule.

Antibodies against mTOR (#2983, dilution 1:150), S6 (#2217, dilution 1:1,000), phospho-S6 (Ser235/236; #4856, dilution 1:1,000), S6K (#2708, dilution 1:1,000), phospho-S6K(T389) (#9205, dilution 1:1,000), 4EBP1 (#9452, dilution 1:1,000), phospho-4EBP1(T37/46) (#2855, dilution 1:1,000), AKT (#4691, dilution 1:1,000), phospho-AKT(Ser473) (#4060, dilution 1:1,000), phospho-AKT(Thr308) (#13038, dilution 1:1,000), AMPKα (#5832, dilution 1:1,000), phospho-AMPKα(Thr172) (#2535, dilution 1:1,000), p53 (#2524, dilution 1:1,000), phospho-p53(Ser15) (#9284, dilution 1:1,000), p44/42 MAPK (#4695, dilution 1:1,000), phospho-p44/42 MAPK(Thr202/Tyr204) (#9106, dilution 1:1,000), p90RSK (#8408, dilution 1:1,000), phosphor-p90RSK(Thr359/Ser363) (#9344, dilution 1:1,000), phospho-p65(Ser536) (#3033, dilution 1:1,000), p62 (#5114, dilution 1:1,000), LC3 AB (#12741, dilution 1:1,000), b-actin (#4967, dilution 1:1,000), RAPTOR (#2280, dilution 1:1,000), TSC2 (#4308, dilution 1:1,000), and Flag (#14793, dilution 1:1,000) were obtained from Cell Signaling Technology. Antibodies against p65 (#sc-8008, dilution 1:1,000) were obtained from Santa Cruz Biotechnology Inc. Antibody against CD63 (SAB4700215, dilution 1:400) was obtained from Sigma. The secondary antibodies anti-mouse (#7076, dilution 1:1,000) and anti-rabbit (#7074, dilution 1:1,000) were obtained from Cell Signaling Technology. PA, rapamycin, and paraformaldehyde were obtained from Sigma. pcDNA3-FLAG-Rheb plasmid (Addgene #19996) was a gift from Fuyuhiko Tamanoi (UCLA, Los Angeles, CA).

HCT116, U2OS, U87, and K562 cells were obtained from the ATCC. GFP-LC3–expressing U2OS cells were kindly provided by Eyal Gottlieb (Cancer Research UK). WT and TSC2^–/– mouse embryonic fibroblasts (MEF) were kindly provided by David J. Kwiatkowski (Harvard Medical School, Boston, MA). HCT116, U2OS, and U87 cells were grown in DMEM high glucose (4.5 g/L; GIBCO) and K562 cells in RPMI (GIBCO), both supplemented with 10% of FBS (Dominique Dutscher), glutamine (2 mmol/L), penicillin (Sigma, 100 U/mL), and streptomycin (Sigma, 100 mg/mL), at 37°C with 5% CO₂ in humidified atmosphere. Human umbilical vein endothelial cells (HUVEC) were obtained from Promocell (Germany) and cultured according to the supplier's instructions in endothelial cell growth medium 2 containing growth factors and 2% FCS. Primary normal human dermal fibroblasts (NHDF) derived from adult skin tissue were purchased from Lonza and cultured according to the supplier's instructions in fibroblast growth medium containing human basic fibroblast growth factor (bFGF), insulin, and 2% FCS. Human follicle dermal papilla cells (HFDPC) isolated from human dermis originating from lateral scalp were purchased fromTebu-Bio and grown in Follicle Dermal Papilla Cells Medium containing 4% FCS, 0.4% bovine pituitary extract, 1 ng/mL bFGF, and 5 μg/mL of insulin (Tebu-Bio). The cells were maintained at 37°C in a humidified atmosphere containing 5% CO₂. Mycoplasma contamination check was carried out using the VenorGeM Kit (Minerva Biolabs GmbH). When indicated, ICSN3250 (dissolved in DMSO before further dilution in assay mixture) was added at the indicated concentration. PA was added to a final concentration of 1, 10, or 100 μmol/L.

Resected colon tumor tissue was obtained from Bergonié Institute (Bordeaux, France) following written-informed consent approved by Bergonié Institute. Patient consent forms were obtained at the time of tissue acquisition. Biopsies were deidentified and processed for cell culture. Briefly, to obtain a single cell suspension, tumor tissue was cut into small fragments and enzymatically/mechanically dissociated with the gentle MACS Octodissociator and the tumor dissociation kit for human (Miltenyi Biotec). Single cells were then plated at 4.5 × 10⁵ cell/mL in 96-well or 6-well plates previously coated with collagen I at 300 μg/mL (Rat tail, Gibco). ICSN 3250 (100 nmol/L) was immediately added to the cells. ICSN3250 did not affect the adhesion of primary patient cells in culture (data not shown). Proliferation was measured after 3 days of culture in 96-well plates by cell counting and trypan blue exclusion with at least 3 replicates per condition. Photomicrographs of control and treated cells were taken at 3 days from 6-well plates (Nikon Eclipse TS100, Archimed software from Microvision Instruments).

GFP-expressing HCT116 or U2OS cancer cells were seeded with GFP-negative HUVEC or NHDF cells in a ratio 1:1 and then treated with ICSN3250 at different concentrations during 72 hours. Cells were then collected and analyzed by flow cytometry. The percentage of GFP-positive and -negative cells was calculated.

Plasmid and siRNA transfections were carried out using Jetpei and Interferin@ (Polyplus Transfection), respectively, according to the manufacturer's instructions. Briefly, 70% confluent cells were transfected with 5 μg of plasmid. Twenty-four hours later, cells were treated with ICSN3250 for 24 more hours. Cells at 50% of confluence were transfected with siRNA (final concentration, 10 nmol/L) for 48 hours and then treated with ICSN3250 for 24 hours. All siRNAs were obtained from Dharmacon (on-target plus smartpool siRNA). Sequences of nontargeting control siRNAs (D-001810-02-05) were (1) UGGUUUACAUGUCGACUAA, (2) UGGUUUACAUGUUGUGUGA, (3) UGGUUUACAUGUUUUCUGA, and (4) UGGUUUACAUGUUUUCCUA, and sequences of TSC2 siRNAs (L-003029-00-0005) were (1) GCAUUAAUCUCUUACCAUA, (2) CGAACGAGGUGGUGUCCUA, (3) GGAAUGUGGCCUCAACAAU, and (4) GGAUUACCCUUCCAACGAA.

HCT116 cells, U2OS cells, NHDF, HUVEC, TSC2^+/+ MEFs, and TSC2^−/− MEFs were seeded in 10 cm plates. After the treatment, cells were washed with PBS (1X) and lysed on ice using home-made RIPA buffer (Tris-HCl 50 mmol/L pH 7.5, NaCl 150 mmol/L, NP-40 1%, sodium deoxycholate 0.5%, EDTA 2 mmol/L, and NaF 10 mmol/L) supplemented with protease inhibitors (Sigma), phosphatase inhibitors (Sigma), and PMSF (1 mmol/L; AppliChem). Protein quantification was performed with the BCA assay Kit (Thermo Fisher). After electrophoresis, the proteins were transferred to a nitrocellulose membrane (BioRad) with Trans-Blot Turbo Transfer System (Bio-Rad). The membranes were incubated for 30 minutes in PBS 1X with 0.01% Tween-20 and 5% BSA. Primary antibodies were incubated overnight at 4°C, and secondary antibodies were incubated for 2 hours at room temperature. Finally, membranes were imaged using the Chemi Doc MP Imager (Bio-Rad).

In vitro kinase assays of mTOR, AKT1, EGFR, PDK1, SRC, PKCα, and PKCϵ were performed at CEREP Company (France). In vitro kinase assays of PI3Kα, PI3Kβ, PI3Kγ, and PI3Kδ were performed using the PI3 Kinase Activity/Inhibitor ELISA assay from Merck-Millipore. Detailed procedures of these in vitro kinase assays are described in Supplementary Material and Methods.

After treatment, cells were washed twice with cold PBS, and then they were lysed with lP lysis buffer (40 mmol/L Hepes, pH 7.5, 120 mmol/L NaCl, 1 mmol/L EDTA, 0.3% CHAPS) and supplemented with protease inhibitor cocktail and phosphatase inhibitor cocktail (Sigma). Protein extracts were incubated overnight at 4°C with anti-mTOR antibodies and then for 4 hours at 4°C with magnetic beads (Pierce Protein A/G Magnetic Beads, Thermo Fisher). Subsequently, beads were washed twice with cold PBS and eluted with Laemmli buffer for Western blot analysis.

To assess cell viability, 10,000 cells per well were seeded in triplicate in 96-well plates. The number of cells was determined using the TC20 Automated Cell Counter (Bio-Rad) according to the manufacturer's protocol. Briefly, after the respective treatments, cells were detached with trypsin/EDTA, and 10 μL of the cells suspension was mixed with 10 μL trypan blue 5% solution (Bio-Rad) and analyzed with the cell counter. Alternatively, cell viability was assessed using the CellTiter-Blue Cell Viability Assay (Promega). After the treatment, 20 μL of the reagent was added to each well, and the plate was incubated for 1 to 4 hours at 37°C with 5% CO₂ in humidified atmosphere. The fluorescence was recorded at 560/590 nm in a Tristar2 LB942 (Berthold) device to determine the cell viability.

Exponentially growing cancer cells (HCT116 and U2OS) were incubated with ICSN3250 or DMSO for 24 hours. Cell-cycle profiles were determined by flow cytometry using propidium iodide on a FC500 flow cytometer (Beckman–Coulter).

Surface plasmon resonance (SPR) experiments were performed at 25°C (sample rack at 10°C) on a Biacore T200 apparatus (Biacore, GE Healthcare) using CM5 sensor chips (Biacore). The surface was first activated with a 7-minute pulse of 50 mmol/L NHS/200 mmol/L EDC (GE Healthcare) aqueous mixture using a flow rate of 5 μL/min. Streptavidin (IBA Lifesciences), full-length (TP320457, Origene) and FRB-containing fragment (10012913, Thermo Fisher) mTOR recombinant proteins were prepared in 10 mmol/L sodium acetate buffer (pH 4.5 for mTOR recombinant proteins and pH 4.9 for streptavidin) and were injected in flow cells 2, 3, and 4, respectively. Note that 8,800 resonance units (RU), 20,300 RU, and 12,300 RU were immobilized, respectively. The surface was then deactivated with a 7-minute pulse of 1 mol/L ethanolamine HCl-NaOH, pH 8.5 (GE Healthcare), and washed with three 1-minute pulses of a mixture of 1 mol/L NaCl/50 mmol/L NaOH prepared in mq water. The ICSN3250 compound (10 μmol/L) was prepared in the running buffer (PBS containing 0.05% Tween-20) and was injected for 1 minute in duplicate at 100 μL/min. The regeneration of the surface was achieved with a 10-second pulse of SDS 0.05% at 30 μL/min followed by an extra wash with running buffer. One channel left blank was used for referencing of the sensorgrams. The SPR signals were normalized to the moles/mm² of streptavidin immobilized onto the sensor chip surface (flow cell 2).

Cells were grown on coverslips in 12-well plates. Subsequently, after the treatments, cells were rinsed with ice-cold PBS and fixed with 4% paraformaldehyde in PBS for 30 minutes at room temperature. After the fixation, cells were permeabilized using PBS with Triton-X 0.05% during 10 minutes and then blocked with BSA 5% in PBS for 30 minutes. When required, cells were incubated with primary antibody for 1 hour at 37°C. After three washes with PBS, the coverslip was incubated for 1 hour at 37°C with the appropriate secondary antibody (anti-rabbit Alexa488, dilution 1:400 or anti-mouse Alexa555, dilution 1:400, obtained from Invitrogen). Finally, coverslips were mounted with Prolong containing DAPI (Invitrogen). Fluorescence was detected using a Leica confocal microscopy. Images were analyzed using Image J software.

Three-dimensional structures of ligands were generated using CORINA version 3.44 (http://www.molecular-networks.com). Molecular docking calculations were carried out using GOLD software (20) and GoldScore scoring function, with the protein 2NPU (representative conformer 1; ref. 21) as receptor. The binding site was defined as a sphere with 15 Å radius around a point with coordinates −6.449, 6.669, and −5.742. In agreement with our previous studies (22–26) showing that an enhanced conformational search is beneficial, especially for large molecules, a search efficiency of 200% was used to better explore the ligand conformational space. All other parameters were used with the default values. Molecular dynamics simulations were carried out with GROMACS version 4.6.5 (27) using the OPLS-AA (28) force field. Each system was energy-minimized until convergence using a steepest descents algorithm. Molecular dynamics with position restraints for 200 ps was then performed, followed by the production run of 100 ns. During the position restraints and production runs, the Berendsen method (29) was used for pressure and temperature coupling. Electrostatics was calculated with the particle mesh Ewald method (30). The P-LINCS algorithm (31) was used to constrain bond lengths, and a time step of 2 fs was used throughout. Ligand topologies for the OPLS-AA force field were generated using MOL2FF, an in-house developed script, and were deposited into the Ligandbook repository (32) with IDs 2929 (https://ligandbook.org/package/2929) and 2930 (https://ligandbook.org/package/2930). DFT calculations were carried out using Gaussian09, version D01 (33). Experimental pKa values were taken from Jencks and Regenstein (34). All calculations were performed using the high-performance computing facilities available at the ICSN (Gif-sur-Yvette, France). Images were generated with Pymol, version 1.8.6 (http://pymol.org).

The results are expressed as a mean ± SEM of at least three independent experiments. T test comparison was used to evaluate the statistical difference between two groups. One-way ANOVA followed by Bonferroni comparison as a post hoc test was used to evaluate the statistical difference between more than two groups. Statistical significance was estimated when P < 0.05.

The authors declare that all the data supporting the findings of this study are available within the article and its Supplementary Information files and from the corresponding author upon reasonable request.

To better understand the consequences at cell signaling level of ICSN3250 (Fig. 1A) in human cells, we treated two human cancer cell lines, the colorectal carcinoma cell line HCT116 and the osteosarcoma cell line U2OS, with increasing concentrations of ICSN3250, and we performed a targeted screening of different signaling pathways. These included AMPK pathway (phospho-Thr172 AMPK), p53 pathway (phospho-Ser15 p53), PI3K pathway (phospho-Thr308 AKT), ERK pathway (phospho-Thr202/Tyr204 p44/42 MAPK; phospho-Thr359/Ser363 RSK), NF-κB pathway (phospho-Ser536 p65), mTORC1 pathway (phospho-Thr389 S6K), and mTORC2 (phospho-Ser473 AKT). As shown in Fig. 1B and Supplementary Fig. S1A, the only pathway that was inhibited by ICSN3250 treatment in both cancer cells was mTORC1 pathway. Indeed, some other pathways, such as PI3K, ERK, and mTORC2, showed an increase in the phosphorylation of their respective downstream targets. This increase would be in agreement with a specific inhibition of mTORC1 pathway and the subsequent release of the negative feedback loop that leads to PI3K reactivation (13). To better evaluate the potential feedback role that these activations would have upon ICSN3250 treatment, we investigated the effect of ICSN3250 in both U2OS and HCT116 cancer cells through long time-course experiments (24, 48, and 72 hours). As shown in Supplementary Fig. S1B and S1C, we observed that ICSN3250 treatment efficiently inhibited mTORC1 activity even at that long term (72 hours). However, the upregulation of PI3K, ERK, and NF-kB pathways observed at short term was robustly attenuated at longer times. These results indeed suggested that the upregulation of these pathways is a transient compensatory response of the cell, probably due to the release of the negative feedback loop downstream of mTORC1, having no major contribution to the long-term effects of ICSN3250 in cancer cells.

Dose-dependent analysis showed a complete inhibition of mTORC1 (looking at the phosphorylation of 3 well-known targets of mTORC1 pathway: S6K, S6, and 4EBP1) at concentrations equal or higher than 50 nmol/L of ICSN3250 (Fig. 1C; Supplementary Fig. S1D). Time-course analysis showed a slow yet efficient inhibition of mTORC1 that reached a maximal inhibition upon 8 to 15 hours of treatment (Fig. 1D; Supplementary Fig. S1E). This is considerably slower than previously reported mTORC1 inhibitors, such as rapamycin or PP242. Confirming the capacity of ICSN3250 to inhibit mTORC1, we also observed that ICSN3250 treatment induced an increase in autophagy, negatively regulated by mTORC1 (35), as determined by increasing levels of LC3-II, by decreasing levels of the adaptor protein p62, and by the accumulation of GFP-LC3 puncta, all of them standard markers of autophagy (Fig. 1E–H; Supplementary Fig. S1F and S1G). Finally, ICSN3250 treatment caused cell-cycle arrest at G₀–G₁ phase both in HCT116 cells and U2OS cells, as expected upon mTORC1 inhibition (Fig. 1I; Supplementary Fig. S1H).

To validate ICSN3250 as a direct inhibitor of mTORC1, we investigated if ICSN3250 interacted directly with mTOR protein. For this purpose, we performed SPR experiments using a recombinant mTOR peptide, including the full length of the protein. As shown in Figure 1J, SPR analysis showed a direct interaction between ICSN3250 and mTOR protein, whereas no interaction with ICSN3250 was observed when a control protein (streptavidin) was used instead of mTOR, demonstrating the specific interaction between mTOR and ICSN3250. In addition, we repeated the analysis using only the FRB domain of mTOR. Again, SPR analysis showed a specific interaction of ICSN3250 with the FRB domain of mTOR (Fig. 1J). Altogether, our results indicated that ICSN3250 is an inhibitor of mTORC1 that directly interacts with mTOR at the FRB domain.

Rapamycin and its analogues, as well as dual mTORC1/mTORC2 inhibitors, act as kinase inhibitors of mTOR, with fast time-course kinetics. We analyzed if ICSN3250 is a kinase inhibitor of mTOR in vitro. The results shown in Fig. 2A indicated that, although ICSN3250 had a capacity to inhibit the mTOR kinase activity, this effect occurred at concentrations much higher (10 μmol/L) than the observed inhibition of mTORC1 in cells (50 nmol/L) and is 10⁵ times higher than the IC₅₀ reported for rapamycin (0.1 nmol/L; ref. 36). This result confirmed that ICSN3250 is not a kinase inhibitor of mTOR, suggesting that it operates differently than other mTORC1 inhibitors. Indeed, ICSN3250 did not show any inhibitory capacity neither toward PI3Kα, β, γ, or δ (Supplementary Fig. S2A), nor toward other kinases analyzed, such as PKCα, PKCϵ, SRC, AKT1, EGFR, and PDK1 (Supplementary Fig. S2B).

Next, we investigated if ICSN3250 prevents the translocation of mTORC1 to the surface of the lysosome, a well-known mechanism involved in the activation of mTORC1 by nutritional inputs (37). As shown in Fig. 2B, and quantified in Fig. 2C, the addition of 100 nmol/L of ICSN3250 (a concentration at which mTORC1 was completely inhibited, see Fig. 1B and C) to HCT116 cells did not prevent the colocalization of mTOR with the lysosomal marker CD63, clearly indicating that lysosomal localization of mTORC1 was not impaired by ICSN3250. Similar results were obtained in U2OS cells (Supplementary Fig. S2C and S2D). Furthermore, even when ICSN3250 was not able to prevent the amino acid–induced lysosomal translocation of mTORC1, still ICSN3250 was able to prevent the activation of mTORC1 mediated by amino acid in both cell lines (Fig. 2D; Supplementary Fig. S2E), again suggesting that the inhibition of mTORC1 occurred once mTORC1 is at the lysosomal surface. To finally discard that lysosomal translocation is involved in the mechanism of action of ICSN3250, we overexpressed a delocalized Flag-Rheb, which renders mTORC1 activation outside the lysosome. As expected, Flag-Rheb overexpression induced mTORC1 activation in the absence of amino acids (Supplementary Fig. S2F and S2G). However, delocalized Flag-Rheb did not prevent the inhibitory effect of ICSN3250 toward mTORC1 activity (Fig. 2E; Supplementary Fig. S2H), finally confirming that ICSN3250 operates after the translocation of mTORC1 to the lysosome.

mTORC1 destabilization has been proposed as a mechanism of mTORC1 inhibition upon certain metabolic stresses (38). Thus, we investigated if ICSN3250 destabilizes mTORC1 as an inhibitory mechanism. For this purpose, we immunoprecipitated mTOR and analyzed the presence of the specific mTORC1 component Raptor in the immunoprecipitates. As expected, in the absence of the compound, Raptor was observed upon mTOR immunoprecipitation (Fig. 2F). Our results showed that ICSN3250 treatment (at 24 hours) was not able to prevent the interaction of mTOR with Raptor (Fig. 2F). Hence, we concluded that the mechanism of action of ICSN3250 does not affect the integrity of the mTORC1, localized at the lysosome.

We next investigated the mechanisms that allow mTORC1 activation at the lysosome. These mechanisms are controlled by the Tuberous Sclerosis Protein 1/2 complex (TSC complex) that exerts a negative regulation toward mTORC1 (7). To investigate if TSC complex plays a role in the mechanism of action of ICSN3250, we treated TSC2^+/+ MEFs and TSC2^−/− MEFs with increasing concentrations of ICSN3250. Similarly to what we observed in cancer cell lines, ICSN3250 induced a complete inhibition of mTORC1 at concentrations higher than 50 nmol/L in TSC2^+/+ MEFs (Fig. 3A). Concomitantly, we observed an activation of autophagy (as determined by increasing LC3II levels), as expected upon mTORC1 inhibition. However, the inactivation of TSC complex in TSC2^−/− MEFs induced a complete recovery of mTORC1 activity and a lack of autophagy activation even in the presence of ICSN3250 at 100 nmol/L (Fig. 3A). These results were confirmed by knocking down TSC2 in HCT116 and U2OS cells. As shown in Fig. 3B and Supplementary Fig. S3A, the efficient silencing of TSC2 resulted in the reactivation of mTORC1 in ICSN3250-treated cells. We concluded that the regulation of mTORC1 by TSC complex might be involved in the mechanism of action of ICSN3250.

The production of PA by phospholipase D1 has been previously invoked as a mechanism of the regulation of mTORC1 by TSC complex (39), and it is largely known that PA binds to and activates mTORC1 (9). In addition, our previous results indicating that ICSN3250 interacts with the FRB domain of mTOR, where the pocket for PA is located, supported the hypothesis that ICSN3250 could compete with PA in mTORC1 binding, thus displacing PA from its binding site, leading to mTORC1 inhibition downstream of TSC complex. To test this possibility, we first performed a competitive analysis of mTORC1 activation between PA and ICSN3250. For this purpose, we treated HCT116 cells with ICSN3250 (100 nmol/L) in the presence of increasing concentrations of PA (0 to 100 μmol/L). As we observed previously, ICSN3250 alone induced the inhibition of mTORC1. However, coincubation of cells with PA induced a dose-dependent mTORC1 reactivation and autophagy inhibition even in the presence of ICSN3250 (Fig. 3C–F; Supplementary Fig. S3B and S3C). Conversely, increasing concentrations of ICSN3250 limited the activation of mTORC1 and the inhibition of autophagy induced by PA (Fig. 3G and H; Supplementary Fig. S3D and S3E). These results strongly suggest that ICSN3250 antagonizes with PA to inhibit mTORC1.

To confirm the previous conclusion that ICSN3250 antagonizes with PA, we performed molecular docking calculations to identify the binding modes of ICSN3250 and PA within the FRB domain of mTOR. Three different protonation states of the catechol group in ICSN3250 were considered during the docking process (i.e., neutral and deprotonated on either OH group), and the strongest interactions and the best protein-ligand shape complementarity were obtained with the form deprotonated on the OH-situated ortho from the NO₂ substituent. We computed the pKa of this OH group using the protocol described by Muckerman and colleagues (DFT calculations on a simplified analogue of ICSN3250 with implicit solvent and removal of the systematic error; ref. 40), and we found a value of 5.93 ± 0.55, meaning that this group is negatively charged at physiologic pH. This is in strong agreement with the docking results, showing interactions between this group and the positively charged side chains of Lys2095 on the one side and of Arg2042 on the other (Fig. 4A and B).

The FRB domain of mTOR (apo form) and the docking complexes with ICSN3250 and PA were used for molecular dynamics (MD) simulations (100 ns each), to take into account two factors that were missing in the docking process: protein flexibility and the presence of explicit aqueous solvent. As expected, the apo simulation reached very quickly an equilibrium conformation that is conserved until the end. In contrast, the two complexes evolved slowly toward an equilibrium structure, which is attained only after 75 to 80 ns (Supplementary Fig. S4A–S4C), highlighting the need for relatively long MD simulations in the study of flexible proteins. The representative equilibrium structures from these simulations showed a number of interesting elements. The protein surface is very flexible, changing the shape according to the interaction partner. Consequently, a very good protein-ligand surface complementarity was observed for the two complexes, bringing an important contribution to the ligand affinity, which is complemented by strong ionic interactions between nitrocatechol groups and Lys2095 and Arg2042 in the case of ICSN3250 and between the phosphate group and Arg2109 in the case of PA (Fig. 4C and F). The interaction between ICSN3250 and its binding site showed three distinct regions: (i) the ionic interaction between the nitrocatechol groups and Lys2095 and Arg2042 that was already mentioned; (ii) a π-stacking interaction between the pyrrole ring and Phe2039, and (iii) the interaction between the macrocycle and a hydrophobic subpocket composed of residues Trp2101, Tyr2105, Phe2108, Leu2031, and Tyr2104. Ser2035, which was shown to be important for the interaction of mTOR with rapamycin (41), is also part of the binding site (Fig. 5A and B). ICSN3250 is relatively flat on the protein surface, whereas PA is deeply buried with its two hydrophobic tails that interact with a subpocket containing Trp2101, Tyr2105, Phe2108, Leu2031, Leu2054, Tyr2104, Ser2035, Phe2039, Leu2051, Tyr2038, Val2044, and Met2047. Only the phosphate head is solvent-exposed and interacts with Arg2109 (Fig. 5C and D). This orientation is similar to the one previously observed by NMR (21), with the exception of the tail chains that are more deeply buried in our case.

Overall, the residues involved in the interaction between mTOR and the two ligands studied in this work clearly show a significant overlapping of the two binding sites. Our results supported that ICSN3250 binds to the FRB domain of mTOR and displaces PA, leading to mTORC1 inhibition. This mechanism defines ICSN3250 as a new class of mTORC1 inhibitor.

Previously, we reported that ICSN3250 showed an increased cytotoxicity in human cells (17). We have now confirmed that no radical intermediate can be observed by electron spin resonance in vitro with ICSN3250 in the presence of superoxide anion, the main ROS in cells (Supplementary Fig. S5A). We also observed that ICSN3250 did not react with superoxide (Supplementary Fig. S5B), confirming the redox stability of ICSN3250. In addition, as shown in Supplementary Fig. S5C and S5D, ICSN3250 did not induce an increase of ROS levels, neither at the cytosol nor at the mitochondria. In agreement with these results, we also observed that the cytotoxicity of ICSN3250 in HCT116 was not reversed (neither partially reduced) by treating the cells with N-acetylcysteine, a classical antioxidant (Supplementary Fig. S5E). These results indicated that the cytotoxicity exerted by ICSN3250 in cancer cells is not mediated by a potential increase in ROS levels.

Our results demonstrating that ICSN3250 acts as a new class of mTOR inhibitor led us to investigate if the inhibition of mTORC1 was the primary reason for the cytotoxicity induced by ICSN3250. For this purpose, we investigated if the reactivation of mTORC1 mediated by TSC ablation in TSC2^−/− MEFs protected from the cytotoxic effect of ICSN3250. As shown in Fig. 6A and Supplementary Fig. S6A, TSC2^−/− MEFs showed an increased protection against cytotoxicity induced by ICSN3250 with respect to TSC2^+/+ MEFs (as control, TSC2^−/− MEFs did not show any increased viability with respect to TSC2^+/+ in the absence of ICSN3250). Similar results were obtained when TSC2 was knocked down in HCT116 or U2OS cells: the efficient silencing of TSC2 was sufficient to restore (at least partially) cell viability in ICSN3250-treated cells (Fig. 6B; Supplementary Fig. S6B).

We previously showed that treatment of HCT116 cells with PA (100 μmol/L) was sufficient to reactivate mTORC1 (see Fig. 3B). Now, we confirmed that PA treatment also prevented the cytotoxic effect of ICSN3250 in a dose-dependent manner in HCT116 cells (Fig. 6C–E). This result clearly suggested that the inhibition of mTORC1 by ICSN3250 is responsible for its cytotoxicity. Furthermore, the particular mechanism of mTORC1 inhibition induced by ICSN3250 is likely the reason of the increased cytotoxicity showed by this compound with respect to other mTORC1 inhibitors, such as rapamycin. Indeed, although rapamycin induced a stronger inhibition of mTORC1 than ICSN3250 (Fig. 6F), it did not cause the cytotoxic effect that we observed upon ICSN3250 treatment (Fig. 6G). Compared with a panel of mTOR inhibitors, ICSN3250 was not the most potent mTORC1 inhibitor among them as determined by the dephosphorylation of mTORC1-downstream targets (Supplementary Fig. S6C and S6D), but yet it ranked among the most cytotoxic compounds for cancer cells, showing one of the lowest IC₅₀ values (Fig. 6H; Supplementary Fig. S6E). Hence, we concluded that the qualitative (and not quantitative) differences between the inhibition exerted by ICSN3250 with respect to other mTOR inhibitors are key for the marked cytotoxicity induced by ICSN3250.

To validate the potential applicability of ICSN3250 as an anticancer drug, we compared the cytotoxicity of ICSN3250 in a panel of cells including both cancer cells and noncancer cells. As shown in Fig. 7A, ICSN3250 showed a cytotoxicity in cancer cells (HCT116, U2OS, U87, and K562) that was 10 to 100 times more potent than its cytotoxicity in human noncancer cells (NHDF, HUVEC, and HFDPC; MEFs, as they are highly proliferating, do not really behave as normal cell in culture, and they were sensitive to ICSN3250). Lack of toxicity in noncancer cells (NHDF and HUVEC) was confirmed in long time-course experiments, at 72 hours of treatment, in clear contrast with cancer cells (HCT116 and U2OS; Fig. 7B and C; Supplementary Fig. S7A and S7B). Importantly, the reduced cytotoxicity of ICSN3250 observed in noncancer cells correlated with its reduced capacity to inhibit mTORC1 in these cells: the inhibition of mTORC1 in HUVEC and NHDF cells was only reached at concentrations higher than 500 nmol/L (Fig. 7D; Supplementary Fig. S7C), whereas full mTORC1 inactivation in U2OS and HCT116 cells was observed at 50 nmol/L (see Fig. 1C; Supplementary Fig. S1D).

To further sustain the notion that ICSN3250 exhibits selectivity for cancer cells over noncancer cells, we performed coculture experiments of GFP-labeled cancer cells (HCT116 or U2OS) together with unlabeled noncancer cells (HUVEC and NHDF) treated with increasing concentrations of ICSN3250 for 72 hours. Flow cytometry analysis showed a clear decrease in the GFP-positive (cancer cells) population with respect to the GFP-negative (noncancer cells) population (Fig. 7E–G; Supplementary Fig. S7D–S7F). These results corroborated that ICSN3250 exhibits selective cytotoxicity toward cancer cells. Next, in order to validate the capacity of ICSN3250 to target primary cancer cells, we performed ex vivo treatment of cancer cells from a patient with colorectal cancer. As shown in Fig. 7H–J, we observed a substantial decrease in the viability of these primary cancer cells after 72 hours of treatment with ICSN3250 ex vivo. Importantly, primary fibroblasts obtained from the same patient failed to show any decrease in viability upon ICSN3250 treatment. These results confirmed the validation and specificity of ICSN3250 to kill primary cancer cells.

Finally, compared with other mTOR inhibitors that showed cytotoxicity in cancer cells (such as INK 128, gedatolisib, or VS-5584, among others), ICSN3250 was substantially less toxic in human primary normal cells (Fig. 7K), supporting the concept that ICSN3250 presents an action mechanism that makes it particularly interesting to develop anticancer strategies.

The results shown herein presented ICSN3250 as a new class of mTORC1 inhibitor that acts through a mechanism that differs from those described by other mTOR inhibitors. ICSN3250 is an analogue of the cytotoxic marine alkaloid halitulin, previously reported to present an increased cytotoxicity (17). However, the mechanism of action underlying this cytotoxicity was not known. Our results showed a specificity of ICSN3250 targeting mTORC1, without inhibiting other signaling pathways, such as AMPK, p53, PI3K, ERK, NF-κB, or even mTORC2. Surprisingly, ICSN3250 did not affect the kinase activity of mTOR, neither the stability of mTOR complex. Instead, our results showed that ICSN3250 binds to the FRB domain of mTOR, displacing PA to overcome the TSC-negative regulation of mTORC1 as a mechanism for mTORC1 inhibition. Indeed, increasing amounts of exogenously added PA or TSC ablation restored mTORC1 activity. This competition with PA seems to be key for the cytotoxicity of ICSN3250, as exogenously added PA not only restored mTORC1, but also restored cell viability. Of note, ICSN3250 did not show an increased capacity to inhibit mTORC1 with respect to previously reported mTOR inhibitors, but yet it showed a particularly high cytotoxic effect in cancer cells, showing a lower IC₅₀ than typical inhibitors such as temsirolimus, accepted by FDA as a treatment against renal cell carcinoma. Importantly, the cytotoxicity of ICSN3250 toward noncancer cells is substantially lower than the most potent of the other inhibitors of mTOR, placing ICSN3250 as a good candidate for future clinical assays.

mTOR inhibition has been approved as a cancer therapy for several types of tumor (42). Yet, the efficiency of those treatments is very modest. Rapamycin and analogues showed mostly cytostatic effect, which in the patient results in a mild delay of tumor growth, with little effect (although statistically significant) in patient survival. These modest results have been explained by the reactivation of the PI3K pathway as a consequence of the release of negative feedback loop downstream of mTORC1 (13). This is why a new generation of dual mTORC1/mTORC2 inhibitors and PI3K/mTOR inhibitors are being proposed and tested. However, these inhibitors still show increased cytotoxicity in noncancer cells. Besides, the use of monotherapies targeting single signaling pathways to treat cancer is under reconsideration. Due to the intrinsic genetic heterogeneity of tumors and the rapid evolution and adaptation of tumor cells during the progression of the disease (43), developing drug resistance is a recurrent problem during treatment, particularly when monotherapies have been used. Still, the efficacy of ICSN3250 to selectively target tumor cells in vivo remains to be elucidated.

As mTORC1 is not the only protein activated by PA, it could be envisioned that other mechanisms or pathways could be involved in ICSN3250-induced cytotoxicity. However, our results showing that mTORC1 reactivation in TSC2^−/− cells restored cell viability indicated that mTORC1 inhibition is at the basis of ICSN3250-induced cytotoxicity. The unprecedented mechanism of action of ICSN3250, displacing PA to overcome TSC-negative regulation of mTORC1, without affecting mTOR kinase activity, seems to be key to explain the specific cytotoxicity for cancer cells showed by this type of mTORC1 inhibitor. Why this mechanism of action would be more cytotoxic than mTOR kinase inhibition mediated by ATP-competitive inhibitors would require further investigations. As ICSN3250-induced PA displacement from the FRB domain of mTOR would likely occur at the surface of the lysosome (where mTORC1 is located upon activation), it could be hypothesized that this displacement causes a collapse in the lysosomal surface, perturbing lysosomal function and leading to cell death, as proposed for other types of stress (44). Alternatively, the slower inactivation of mTORC1 mediated by ICSN3250 as compared with other mTOR inhibitors that we observed could be playing in favor of its cytotoxicity, as our recent results showed that a fast and complete inhibition of mTORC1 upon rapamycin treatment prevents apoptotic cell death during nutritional imbalance (14).

Finally, our results make particular emphasis in the control of mTORC1 activity by PA, a regulation that has not received as much attention as the regulation exerted by amino acids or by PI3K signaling. However, our results clearly indicated that interfering with PA binding in the FRB domain of mTOR is indeed an effective approach to inhibit mTORC1 even in the presence of amino acids and growth factors, underscoring the importance of PA for mTORC1 activity. Besides, as mentioned above, the regulation of mTORC1 by PA seems to be particularly important at the cell physiology level, as the interference with the mTOR–PA interaction resulted in cell death. How exogenously added PA results in mTORC1 activation is not clear (9, 45, 46). In our experiments, we needed a substantially higher concentration of PA to compete with ICSN3250, probably reflecting that the concentration of PA at the lysosomal surface do not reach the same concentration than exogenously added PA.

In conclusion, ICSN3250 defines a new class of mTORC1 inhibitors that, due to its particular mechanism of action, induces cell death specifically in tumor cells but not in noncancer cells. Additional researches will determine the applicability of this type of compound for anticancer therapy.

P. Collin has an ownership interest (including stock, patents, etc.) in a patent. No potential conflicts of interest were disclosed by the other authors.

Conception and design: B.I. Iorga, R.V. Durán, P. Collin

Development of methodology: T.-L. Nguyen, M. Egorov, F. Peyrot, K. Abbas, S. Terés, B.I. Iorga, R.V. Durán, P. Collin

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.-L. Nguyen, M.-J. Nokin, M. Tomé, C. Bodineau, L. Minder, M.C. Garcia-Alvarez, J. Bignon, Y.-M. Frapart, F. Peyrot, K. Abbas, S. Evrard, A.-M. Khatib, B.I. Iorga, R.V. Durán, P. Collin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.-L. Nguyen, M.-J. Nokin, M. Tomé, C. Bodineau, C.D. Primo, L. Minder, M.C. Garcia-Alvarez, J. Bignon, G. Surpateanu, Y.-M. Frapart, F. Peyrot, K. Abbas, B.I. Iorga, R.V. Durán, P. Collin

Writing, review, and/or revision of the manuscript: M.-J. Nokin, M. Egorov, L. Minder, A.-M. Khatib, P. Soubeyran, B.I. Iorga, R.V. Durán, P. Collin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Minder, O. Thoison, F. Peyrot, K. Abbas, S. Evrard, P. Soubeyran, B.I. Iorga, R.V. Durán, P. Collin

Study supervision: R.V. Durán

Other (concept, the design, and the synthesis of the ICSN3250 molecule): M. Egorov

Other (technical manager of SPR platform): L. Minder

Other (contributed to the biological and biochemical studies as well as the UPLC-ESI-MS analysis): M.C. Garcia-Alvarez

Other (performed the synthesis of ICSN 3250 compound): O. Thoison

Other (contributed to chemical synthesis): B. Delpech

Other (EPR studies): Y.-M. Frapart

This work was supported by funds from the following institutions: Centre National de la Recherche Scientifique—CNRS, Institut National de la Santé et de la Recherche Médicale—INSERM, Fondation pour la Recherche Médicale, the Conseil Régional d'Aquitaine, Fondation ARC pour la Recherche sur le Cancer, Ligue Contre le Cancer - Gironde, SIRIC-BRIO, Institut de Chimie des Substances Naturelles—ICSN, Institut Européen de Chimie et Biologie, Université Paris-Descartes, Société d'Accélération de Transfert de Technologie d'Ile de France-SATT IDF-innov, Institut Bergonie, and National Fund for Scientific Research of Belgium. M.-J. Nokin is Télévie Post-Doctoral Fellow from the National Fund for Scientific Research (FRNS, Belgium). pcDNA3-FLAG-Rheb plasmid (Addgene #19996) was a gift from Fuyuhiko Tamanoi. GFP-LC3–expressing U2OS cells were kindly provided by Eyal Gottlieb (Cancer Research UK). We thank Professor J.Y. Lallemand, director of the ICSN (2000–2009). We extend our thanks to A. Pinault for skillful technical assistance. We would like to remember Dr. C. Marazano (†11/12/2008), who initiated and supervised the initial biomimetic synthesis of ICSN3250, one of the halitulin's analogues. We are grateful to the structural biology facility (UMS 3033/US001) of the Institut Européen de Chimie et Biologie (Pessac, France) for access to the T200 Biacore instrument, which was acquired with the support of the Conseil Régional d'Aquitaine, the GIS-IBiSA, and the Cellule Hôtels à Projets of the Centre National de la Recherche Scientifique. We thank Vincent Pitard and the Flow Cytometry Platform (Université de Bordeaux, France) for technical assistance in flow cytometry experiments.

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