Phosphatidic acid drives mTORC1 lysosomal translocation in the absence of amino acids

Mammalian target of rapamycin complex 1 (mTORC1) promotes cell growth and proliferation in response to nutrients and growth factors. Amino acids induce lysosomal translocation of mTORC1 via the Rag GTPases. Growth factors activate Ras homolog enriched in brain (Rheb), which in turn activates mTORC1 at the lysosome. Amino acids and growth factors also induce the phospholipase D (PLD)–phosphatidic acid (PA) pathway, required for mTORC1 signaling through mechanisms that are not fully understood. Here, using human and murine cell lines, along with immunofluorescence, confocal microscopy, endocytosis, PLD activity, and cell viability assays, we show that exogenously supplied PA vesicles deliver mTORC1 to the lysosome in the absence of amino acids, Rag GTPases, growth factors, and Rheb. Of note, pharmacological or genetic inhibition of endogenous PLD prevented mTORC1 lysosomal translocation. We observed that precancerous cells with constitutive Rheb activation through loss of tuberous sclerosis complex subunit 2 (TSC2) exploit the PLD–PA pathway and thereby sustain mTORC1 activation at the lysosome in the absence of amino acids. Our findings indicate that sequential inputs from amino acids and growth factors trigger PA production required for mTORC1 translocation and activation at the lysosome.


PA with a monounsaturated fatty acid drives mTOR to the lysosome in the absence of amino acids
We previously reported that the amino acid-induced activation of mTORC1 depends on PLD (13). Amino acids induce the translocation of mTOR from the cytoplasm to the lysosome. PLD and PA play critical roles in cellular vesicle trafficking (14). Thus, PLD and PA could play a role in the translocation of mTOR from the cytoplasm to the lysosome. To test this hypothesis, we performed amino acid depletion and restimulation experiments in the presence of PA, followed by confocal microscopy detection of LAMP2 (lysosomal marker) and mTOR (Fig. 1A). In DU145 prostate cancer cells we saw a pronounced amino acid-induced mTOR lysosomal translocation. We previously reported that activation of mTOR by PA generated by exogenously supplied fatty acids was achieved with oleic acid (15). We therefore used a PA species with the unsaturated fatty acid oleic acid, PA-18:1, and a PA species without unsaturated fatty acids, PA-16:0 (Fig. 1B). PA-18:1 is the most abun-dant species of PA in mouse brain, whereas PA-16:0 has low abundance (16). PA-18:1 induced mTOR lysosomal translocation similar to amino acids (Fig. 1A). If we added back both amino acids and PA-18:1, LAMP2/mTOR colocalization was even stronger, indicating a synergy between amino acids and PA. In contrast, PA-16:0 had no effect on mTOR translocation (Fig. 1A). Fig. 1C depicts the quantification of two parameters of Fig. 1A: the percentage of cells with LAMP2/mTOR colocalization and the Pearson's correlations of LAMP2/mTOR-labeled pixels. Next, we determined the levels of mTORC1 activity toward the targets ribosomal subunit S6 kinase (S6K), at Thr-389, and eukaryotic initiation factor 4E-binding protein 1 (4E-BP1), at Thr-37/46 (Fig. 1D). PA-18:1, but not PA-16:0, increased the phosphorylation of S6K (P-S6K) and 4E-BP1 (P-4E-BP1) to levels similar to amino acid stimulation. These results suggest that PA-18:1 drives mTOR to the lysosome in the absence of amino acids and promotes mTORC1 activity.
In our experiments with PA, we prepared small unilamellar vesicles (SUVs). To form SUVs, the concentration of PA has to be above the critical micelle concentration (CMC). Because different CMCs could account for the differences between PA species, we determined the CMC (17) of PA-18:1 (63.6 M) and PA-16:0 (78.8 M) (Fig. S1). These indicate that, at a final concentration of 300 M, both PA species form vesicles. To show that PA-18:1 can only activate mTORC1 in a vesicle format, we performed a dose-curve experiment in the absence of amino acids (Fig. 1E). We observed mTORC1 activation if the concentration of PA-18:1 was above 50 M. 50 M is very close to the CMC determined here: 63.6 M. The CMC was determined in PBS, so it is possible that PA will form vesicles at a slightly lower concentration in culture medium. Our findings suggest that PA-18:1 can only activate mTORC1 once it forms vesicles.

PA-18:1 vesicles enter the cell via macropinocytosis and rapidly reach the lysosome
To understand how exogenously supplied PA-18:1 vesicles enter the cell and promote mTORC1 translocation, we performed an endocytosis assay (18) of fluorescently labeled PA ( Fig. 2A). DU145 cells were given PA-18:1-NBD or PA-16:0 -NBD (50 M because of high background levels) for 30 min at 10°C, a temperature that inhibits endocytosis. The cells were then shifted to 37°C for 3 min to allow endocytosis to proceed. Both PA-18:1-NBD and PA-16:0 -NBD entered the cell through endocytosis (intracellular puncta at 37°C). Phospho-lipid SUVs are typically 50 nm in diameter (19) and enter the cell through pinocytosis (macropinocytosis, considering the size of the SUVs). Therefore, we investigated whether PA-18:1 enters the cell through macropinocytosis. For that, we depleted DU145 cells of amino acid for 1 h, after which we added 300 M of PA-18:1 in the absence or the presence of 75 M 5-(N-ethyl-N-isopropyl)-amiloride (EIPA), a macropinocytosis inhibitor (20). EIPA prevented mTOR translocation to the lysosome induced by PA-18:1 (Fig. 2B). These findings were confirmed by the quantification of LAMP2/mTOR percentage of colocalization and Pearson's correlation values (Fig. 2C). Considering that both PA-18:1 and PA-16:0 vesicles enter the cell but behave very differently regarding mTORC1 translocation and activation, we determined whether either of the PA species colocalized with mTOR at the lysosome (Fig. 2D). We observed a strong colocalization between PA-18:1-NBD and mTOR but not with PA-16:0 -NBD. These findings were supported by colocalization percentages and by Pearson's correlation values (Fig. 2E). These results suggest that even though both types of vesicles can enter the cell, only vesicles made of PA-18:1 drive mTOR to the lysosome. In addition, we generated evidence that rules out involvement of the lysophosphatidic acid receptor during mTORC1 translocation and activation by PA-18:1. In the presence of a short species of PA that ends right before the kink of PA-18:1 and that is known to inhibit several lysophos- 16:0 -18:1 PA drives mTORC1 to the lysosome phatidic acid receptors (21), we were still able to induce mTORC1 translocation and activation (Fig. S2).

Pharmacological and genetic inhibition of PLD prevents mTOR translocation to the lysosome
We propose that mTOR translocation and activation by PA-18:1 vesicles mimics an endogenous mechanism of translocation and activation of mTOR triggered by amino acids. To investigate this hypothesis, we used PLD inhibitors during amino acid stimulation. In Fig. 3A we show that PLD1/2 inhibitors (20 M each) (22) or 1-butanol (BuOH) (0.8%) (11) abolished mTOR translocation to the lysosome in response to amino acids. The doses of the PLD inhibitors were chosen based on a dose curve (Fig. S3). These results were supported by colocalization percentages and Pearson's correlations (Fig.  3B). Furthermore, biochemical analysis of mTORC1 activity revealed that both PLD inhibitors and 1-butanol inhibit P-S6K (Fig. 3C) in response to amino acids. These data suggest that endogenous PLD-generated PA is required for amino acidinduced mTOR transport to the lysosome.
Of the several PLD genes in mammalian cells, PLD1 and PLD2 are the most widely expressed PLD proteins (6). We have found that double inhibition of PLD1/2 leads to the greatest suppression of mTORC1 (11,13). In the context of amino acid induction of mTORC1, two other studies have shown a PLD1-specific requirement (4, 23). To assess a possible role of PLD1 in the amino acid-induced mTOR translocation to the lysosome, we performed PLD1 siRNA experiments in DU145 cells, after which we challenged the cells with amino acid deprivation/ restimulation (Fig. 4A). PLD1 knockdown (96 h) prevented mTOR translocation to the lysosome upon amino acid stimulation, PLD1 and mTOR colocalization at the lysosome upon amino acid stimulation (4), and lysosomal aggregation into a large structure. We used two different antibodies specific for PLD1 and saw the same results with both (Fig. 4A). These findings were supported by percentage of colocalization and Pearson's correlation quantifications (Fig. 4B). We could barely detect PLD1 protein after 96 h of siRNA treatment (Fig. 4C). Our data demonstrate that PLD1 is required for amino acidinduced translocation of mTOR to the lysosome.

Pharmacological and genetic inhibition of PLD prevents lysosomal retention of mTOR in cells with constitutive Rheb activation
It was previously shown that Rheb interacts with PLD1 in a GTP-dependent manner to promote PLD1 activity (5). Thus, we hypothesized that cancer cells with constitutive Rheb activation elevate PLD activity and increase production of PA, preventing mTOR from leaving the lysosome in the absence of amino acids. To test our hypothesis we used tuberous sclerosis 16:0 -18:1 PA drives mTORC1 to the lysosome complex 2 (TSC2)-null mouse embryonic fibroblasts (MEFs). TSC2-null MEFs have constitutive Rheb activation because they lack the TSC complex, which is a GTPase-activating protein for Rheb (24) (Rheb inhibitor). These cells represent benign human tumors (hamartomas) originated from loss of the TSC complex in the human syndrome TSC. However, to avoid senescence induced by the loss of TSC2 in culture conditions, p53 had to be deleted in these cells. In that sense, TSC2-null MEFs are closer to cancer cells than cells found in human hamartomas. We show that TSC2-null MEFs have residual retention of mTOR at the lysosome in the absence of amino acids (25) (Fig. 5A). We then tested whether this phenomenon depended on PLD and PA. We deprived cells of amino acids for 1 h and then added PLD inhibitors for 1 h. We used this time point because a 10-min amino acid treatment in TSC2-null MEFs does not elevate mTORC1 activity (Fig. 5B). PA-18:1 could also elevate P-S6K at 1 h (Fig. 5C). PLD inhibitors abolished the residual retention of mTOR at the lysosome (Fig. 5A). The addition of PA-18:1 simultaneously with PLD inhibitors completely rescued the lysosomal localization of mTOR ( Fig.  5A). Unlike PLD inhibitors, high-dose rapamycin did not remove mTOR from the lysosome (Fig. 5A). These findings were supported by the quantification of percentages of colocalization and Pearson's correlations (Fig. 5D). They were also supported by Western blotting analysis of P-S6K (Fig. 5E). The data show that TSC2-null MEFs fail to remove mTOR from the lysosome in the absence of amino acids because of PLD and PA. To show that PLD activity is increased in these cells, we performed PLD activity assays (26) (Fig. 5F). TSC2-null MEFs have significantly higher PLD activity compared with WT MEFs in complete medium (CM). Upon amino acid stimulation, PLD activity in WT MEFs elevated and reached the same levels of PLD activity of TSC2-null MEFs. Importantly, PLD activity in TSC2-null MEFs did not increase with amino acids, suggesting that the main event induced by amino acids-PLD translocation to the lysosome leading to Rheb activation of PLD (4, 5)-is already in place in these cells. Next, we wanted to show that TSC2-null MEFs exploit this mechanism through elevated PLD production of PA, to survive and thrive in nutrient-poor conditions. For that, we treated WT and TSC2-null MEFs with PLD1/2 inhibitors (20 M each) for 24 h (Fig. 5G). This treatment led to 60% of cell death in TSC2-null MEFs but only 20% in WT MEFs, confirming that TSC2-null MEFs require PLD and PA for survival.
To assess a possible specific role for PLD1 in the retention of mTORC1 at the lysosome in TSC2-null MEFs, we performed PLD1 knockdown experiments, after which we subjected the cells to amino acid deprivation for 1 h or kept them in CM (control) (Fig. S4A). PLD1 knockdown abolished LAMP1/ mTOR colocalization both in CM (large perinuclear rings) and in medium lacking amino acids (residual retention), lysosomal organization around the nucleus, and Rheb/mTOR colocalization both in CM and in medium without amino acids. We ver-ified PLD1/mTOR colocalization in large perinuclear rings if cells were grown in CM or in smaller perinuclear globules if cells were deprived of amino acids (residual retention; Fig.  S4A). We observed a similar PLD1/Rheb colocalization pattern (Fig. S4A). These results indicate that PLD1 fails to leave the lysosome upon amino acid withdrawal in cells with hyperactive Rheb, and this in turn is required for mTOR residual lysosomal retention. These results were supported by the quantification of percentages of colocalization and Pearson's correlations for each pair of markers (Fig. S4B).

Exogenously supplied PA-18:1 drives mTOR to the lysosome in the absence of Rheb or Rag GTPases
To fully validate the Rheb-PLD-PA-mTOR pathway, we investigated whether the residual retention of mTOR at the lysosome in TSC2-null MEFs depends on hyperactive Rheb. For, that we performed Rheb siRNA experiments, after which we deprived cells of amino acids for 1 h and then added PA-18:1 for 1 h (Fig. 6A). Rheb knockdown induced loss of residual retention of mTOR at the lysosome (25), which was rescued by exogenously supplied PA-18:1. Rheb knockdown, similar to PLD1 knockdown, induced lysosomal dispersal. Considering that the amino acid-induced Rag pathway of mTOR lysosomal translocation is very well-established (3), and residual lysosomal retention of mTOR happens in the absence of amino acids, we knocked down the Rag GTPases as a control. To prevent compensation between redundant Rag proteins (27), we knocked down both RagC and RagD simultaneously. We did not expect Rag knockdown to abolish mTOR residual lysosomal retention. However, that was not the case. In the absence of both RagC and D, mTOR lysosomal retention was lost, but PA-18:1 could rescue this effect and drive mTOR to the lysosome (Fig. 6A). This suggests that PA can bring mTOR to the lysosome regardless of the Rag GTPases. These results were supported by quantifications of marker colocalizations and Pearson's correlations (Fig. 6B). In addition, they were also supported by biochemical analysis of mTORC1 activity toward S6K (Fig. 6C). Next, we wanted to show that the residual retention of mTOR at the lysosome lost with Rheb knockdown was due to loss of PLD activity. In the absence of Rheb, PLD activity in TSC2-null cells deprived of amino acids suffered a 60% reduction (Fig. 6D), showing that Rheb activates PLD and production of PA. Upon Rag C ϩ D knockdown, we observed a 20% reduction in PLD activity (Fig. 6D), suggesting that PLD and the Rags are not directly linked. To further dissect the role of PLD and PA relative to the Rags, we used RagA GTP/GTP MEFs, which have constitutive mTOR activity refractory to amino acid depletion. The levels of P-S6K, however, are lower than the ones seen in RagA ϩ/ϩ MEFs upon amino acid induction (28). We confirmed these findings in our experiments, where we challenged RagA ϩ/ϩ and RagA GTP/GTP MEFs to inhibition of PLD1, PLD2, or both (Fig. 6E). In RagA ϩ/ϩ cells, we observed increased P-S6K levels upon amino acid stimulation. This was  RagA GTP/GTP behaved differently. PLD1 or PLD2 inhibition reduced the levels of P-S6K. However, only simultaneous PLD1/2 inhibition completely abrogated P-S6K, suggesting that mTORC1 is not accessible to Rheb activation in the complete absence of PLD activity, even if the Rags are constitutively activated.

Discussion
Our data support a model (Fig. 7) where PLD1, similar to mTORC1, acts like a coincidence detector and effector of both amino acids and growth factors. Amino acids induce PLD activity and production of PA. Exogenously supplied PA vesicles enter the cell through endocytosis and drive mTOR to the lysosome, suggesting that amino acids induce production of PAcontaining endosomes that carry mTOR to the lysosome. In agreement, inhibition of endogenous PLD prevented mTOR translocation to the lysosome in response to amino

16:0 -18:1 PA drives mTORC1 to the lysosome
acids. Amino acid-induced RagA/B-GTP RagC/D-GDP heterodimers provide a parallel pathway that locks mTOR on the lysosome. Amino acids also induce the translocation of PLD1 from cytoplasmic puncta to the lysosome. Once on the lysosome, PLD1 binds to Rheb. Growth factors activate Rheb, which then activates PLD1. Lysosomal PA production promotes further binding of PA to mTOR to allow complex stability and activation.
Previous studies showed that exogenously supplied PA induced mTORC1 activation in the presence but not in the absence of amino acids (4). Here, we were able to induce mTORC1 translocation to the lysosome and mTORC1 activity with exogenously supplied PA-18:1 vesicles in the absence of amino acids. The main difference between the two studies is that we performed amino acid and serum deprivation (to prevent contamination of amino acids present in serum) for 1 h, followed by amino acid stimulation for 10 min. In contrast, Yoon et al. (4) performed amino acid starvation for 2 h after overnight serum starvation, followed by amino acid stimulation for 30 min.
Previous findings suggest that mTORC1 assembles before reaching the lysosome because binding of mTOR to the Rag GTPases requires the mTORC1 component raptor (29). PA promotes mTORC1 assembly and stability (11). PA-containing endosomes carrying mTORC1 to the lysosome is therefore an attractive model in which PA would allow mTORC1 formation and stability. We show that exogenously supplied PA can drive mTOR to the lysosome and induce mTORC1 activity in TSC2null MEFs where RagC and D were genetically ablated. Therefore, we propose that delivery of mTORC1 to the lysosome does not require the Rags. However, we did find that residual reten-tion of mTOR at the lysosome in TSC2-null MEFs was lost upon RagC and D knockdown. This suggests that the Rags operate in parallel to PA to lock mTORC1 on the lysosome, after mTORC1 delivery by PA. We show that genetic ablation of PLD1 induced lysosomal scattering. This favors the idea that PA-containing endosomes carry mTORC1 to the lysosome, fuse with the lysosome, and increase its size.
Exogenously supplied PA vesicles induce mTORC1 translocation and activation in the absence of Rheb, indicating that the key step in Rheb activation of mTORC1 is increased PLD1 activity and PA production. Consistent with this observation, the Rheb association with mTOR is independent of GTP loading (30), whereas the Rheb association with PLD1 depends on GTP loading (5). Additionally, we found that PLD1/2 inhibitors, in combination, abolished mTORC1 activity in RagA GTP/GTP MEFs, indicating that PA production is required for complex assembly and stability. If mTORC1 is not intact, then constitutive Rag activation is lost. Thus, the effect of PA on mTORC1 is downstream of Rheb and parallel to Rag GTPases.
Genetic deletion of mTOR in mice is embryonic lethal (31). Unlike mTOR, mice with genetic deletion of PLD1 (32), PLD2 (33), or both (34) are viable, suggesting that PLD and PA may not be required for steady state but rather acute activation of mTOR. PLD inhibitors preferentially killed TSC-null MEFs, whereas rapamycin did not. This suggests an advantage in terms of cancer therapeutics because targeting PLD may selectively target cancer cells, with minimal side effects. PLD inhibitors were developed from halopemide (22), a psychotropic drug extensively used in humans without toxicities. Thus, PLD inhibition might be a viable alternative to current therapies in cancers with mTORC1 hyperactivation that requires PLD-generated PA.

Amino acid depletion and replenishment experiments
The cells were seeded 24 h prior to experiments in 6-cm dishes for cell lysis or 3-cm dishes containing fibronectincoated microscope coverslips (BioCoat 354088) for immunofluorescence. We used 1 ϫ 10 6 DU145 cells, 0.5 ϫ 10 6 Tsc2 Ϫ/Ϫ p53 Ϫ/Ϫ cells, 0.7 ϫ 10 6 Tsc2 ϩ/ϩ p53 Ϫ/Ϫ cells, 1 ϫ 10 6 RagA GTP/ GTP cells, and 1 ϫ 10 6 RagA ϩ/ϩ cells, per 6-cm dish in 3 ml of medium. We seeded 0.5 ϫ 10 6 DU145 cells and 0.25 ϫ 10 6 Tsc2 Ϫ/Ϫ p53 Ϫ/Ϫ cells per 3-cm dish in 2 ml of medium. 24 h later we discarded the growth medium, washed the cells once in RPMI 1640 without amino acids (US Biologicals R8999-04A), and then added RPMI 1640 without amino acids for 1 h (3 ml  (Figs. 1 and 2). It is not clear how exogenously supplied PA recruits mTORC1 to the lysosome as indicated by the question marks, but the PLD requirement for amino acid-induced mTORC1 translocation (Fig. 3) suggests a role for PA in both the trafficking and activation of mTORC1. On the lysosome, mTORC1 binds to amino acid-activated Rags, which lock mTORC1 at the subcellular compartment. Amino acids also induce translocation of PLD 1 from the cytoplasm to the lysosome where it colocalizes with mTOR and Rheb (Fig. 4). Upon growth factor stimulation, Rheb is activated, binds to, and elevates PLD activity. Increased PA production at the lysosome prevents mTOR from leaving the compartment and allows complete activation of mTORC1 ( Fig. 5 and Fig. 6). Precancerous cells caused by loss of the TSC complex exploit permanent activation of the pathway Rheb-PLD-PA-mTORC1 at the lysosome to thrive in nutrient-poor conditions (Fig. 5 and Fig. 6). Full activation of mTORC1 in response to amino acids and growth factors happens when both Rags and PA operate simultaneously (Fig. 6).
16:0 -18:1 PA drives mTORC1 to the lysosome per 6-cm dish or 2 ml per 3-cm dish). 1 h later, the cells were stimulated with amino acids for 10 min (DU145 cells) or for 1 h (Tsc2 Ϫ/Ϫ p53 Ϫ/Ϫ cells) by adding a liquid solution of RPMI 50ϫ amino acids (Sigma, R7131) to a final concentration of 1ϫ (20 l of 50ϫ amino acid solution per ml). RPMI 1640 without amino acids was bought as a powder and made fresh before experiments. After adjusting the pH to 7.4, the medium was filter-sterilized (0.22 m). Serum was absent in all amino acid depletion and replenishment experiments to prevent serumcontaining amino acid contamination. These experimental conditions are identical to the ones described by Sancak et al. (29), with the exception that we used a 50ϫ RPMI amino acid solution instead of a 10ϫ amino acid mixture.

Preparation of PA SUVs
We bought all our phospholipids from Avanti Polar Lipids: PA-18:1, 1-palmitoyl-2-oleoyl-sn-glycero-3P, catalog no. After opening the vial, lipids in chloroform were transferred to a sterile glass vial with a lid (scintillation vial). Before closing the lid, vial atmosphere was replaced with N 2 (g) to prevent lipid oxidation. We sealed vials with parafilm and immediately stored them at Ϫ20°C. We prepared 6.9 mM PA small unilamellar vesicles in DPBS without Ca 2ϩ and Mg 2ϩ (Gibco 14190 -144). For that, we placed 63 l of PA-18:1, 0.6 mg of PA-16:0 and 40 l of diC8-PA into Eppendorf tubes and dried PA-18:1 and diC8-PA under a N 2 gas thin outlet to remove chloroform. Then we added 130 l of DPBS and vortexed the tube until a cloudy suspension was formed. To make small unilamellar vesicles, we sonicated the lipid suspensions with an ultrasonicator (QSonica Q700) with the following settings: 3 ϫ 30 s at amplitude 60, with 30-s intervals between pulses. After sonication, lipid suspensions were kept at room temperature and used immediately. We added 130 l of SUVs per 3 ml of growth medium to reach a final concentration of 300 M. For PA-18:1-NBD and PA-16:0 -NBD, we used a final concentration of 50 M in 100 l of growth medium in a coverslip previously labeled with a hydrophobic pen (Dako, S2002). For that we dried 4 l of each lipid, after which we added 25 l of DPBS and sonicated the suspensions. We added 25 l of 0.2 mM SUVs directly to a coverslip with 75 l of medium.

Immunofluorescence
The cells were grown in 3-cm dishes containing fibronectincoated glass microscope coverslips (BD Biocoat 354088) where we previously placed a hydrophobic barrier with a Dako pen (Dako S2002). For that, we made a circle around the coverslip and allowed it to dry for 5 min. After the experiments were complete, the cells were rinsed in PBS and fixed in 4% parafor-maldehyde (Alfa Aesar, 43368) for 15 min. Immunofluorescence labeling took place at room temperature. After fixation, the cells were washed in PBS and permeabilized for 5 min in PBS with 0.05% Triton X-100, (Fisher Scientific, BP151). The cells were then washed three times in PBS or until reconstitution of the hydrophobic barrier. After that cells were incubated for 1 h in PBS with 5% BSA with primary antibodies. The cells were rinsed four times in PBS before the addition of secondary antibodies in a PBS with 5% BSA solution. Incubation with secondary antibodies took place in the dark for 30 min. Then the cells were washed four times in PBS, and coverslips were mounted onto microscope slides with a drop of Prolong gold antifade mounting medium with DAPI (Molecular Probes P36931).

Confocal microscopy
Image acquisition was done with a PerkinElmer Ultra View Spinning disk confocal microscope coupled to a Nikon camera and equipped with Volocity 3D image capture and analysis software. We used a 60ϫ oil immersion objective to acquire 0.5-m interval stacks (z-stacks). After manually defining the center of the stack (center of the cell), we set the range Ϫ5 to ϩ5 m for DU145 cells and Ϫ3 to ϩ3 m for Tsc2 Ϫ/Ϫ p53 Ϫ/Ϫ MEFs because these cells are very flat. For each stack, we acquired the green channel followed by simultaneous acquisition of the red and blue channels. We captured images with 1344 ϫ 1024 pixels and 155.78 ϫ 118.69 microns. Five random samples of each experimental condition were acquired. We used the same exposure time and magnification for all our samples. For each sample, a cell density of 35-65 cells per field was achieved, unless the cells were treated with drugs known to lift cells off the plates or kill cells. The images were analyzed with ImageJ. File format conversion was done with the LOCI bioformats importer plugin. Colocalization analysis was performed in two different ways. First, we visually scored the percentage of cells with marker colocalization relative to the total number of cells. In all our images, marker colocalization is shown in yellow and orange. Then we computed colocalization correlation coefficients (Pearson's correlation) in ImageJ with the plugin JACoP. Pearson's correlations were based on pixel intensity after applying Costes's thresholds. We used threshold intensities to exclude the background signal in siRNA-treated samples. In all our figures, we show representative cells and the respective cell population percentage of colocalization and Pearson's correlations. In Fig. 7, we show whole images with the purpose of allowing comparability with (25).

Endocytosis assay
To study endocytosis (18) of exogenously supplied PA vesicles, the cells were seeded in DMEM with 10% FBS 1ϫ pen/ strep onto a microscope coverslip in a 3-cm dish. 24 h later we changed the growth medium to 2 ml of DMEM and added SUVs of fluorescently labeled PA analogs (PA-18:1-NBD and PA-16: 0 -NBD) to a final concentration of 50 M. The cells were then incubated at 10°C for 30 min to label the plasma membrane, after which the medium was replaced with 2 ml DMEM. The cells were then shifted to 37°C for 3 min to allow endocytosis to proceed. After that, we washed cells with PBS and then 16:0 -18:1 PA drives mTORC1 to the lysosome removed excess fluorescent PA remaining at the outer leaflet of the plasma membrane by BSA back exchange, in which cells were washed five times with 2 ml of a 5% fatty acid-free BSA (Sigma, A9205) solution made with PBS at 10°C for 10 min.

PLD activity assay
Our PLD activity assays were slightly modified from (37) and (26). DU145 and Tsc2 Ϫ/Ϫ p53 Ϫ/Ϫ cells were seeded in 6-well plates as described above for 3-cm dishes. 24 h later cell growth medium was replaced with 2 ml of fresh medium per well, and cellular phosphatidylcholine was labeled at 37°C for 4 h with 3.0 Ci (1.5 l/ml) of [ 3 H]myristic acid (PerkinElmer, NET830005MC). 1 h before labeling completion (3 h), we shifted the cells to RPMI without amino acids with 3.0 Ci (1.5 l/ml) of [ 3 H]myristic acid. 20 min later (3 h, 40 min labeling), we added 1-butanol (Sigma-Aldrich, B7906) to a final concentration of 0.8%. 20 min later (total 4 h labeling), we placed the plates on ice, discarded the radioactive media appropriately, and washed cells twice with 2 ml of ice-cold PBS. To collect cells we placed 1 ml of medium from plate (with cells) in an Eppendorf tube and spun 5 min at 3000 rpm at 4°C. We discarded the supernatant, and then we added the remaining 1 ml left in the plate after scraping the plate. We spun the tube again for 5 min at 3000 rpm at 4°C. After discarding the supernatant, we washed the pellet in ice-cold PBS twice and centrifuged for 5 min at 3000 rpm at 4°C after each wash. Finally, we removed the supernatant and resuspended the pellet in 500 l of acidified methanol (methanol with 6 N HCl, 50:2). We then transferred the cells into polypropylene tubes kept on ice with the first extraction buffer: 155 l of 1 M NaCl and 500 l of chloroform. We vortexed the tubes and then centrifuged them for 3 min at 13300 rpm at 4°C, after which we discarded the top layer. We placed the bottom layer (avoiding middle layer) into polypropylene tubes kept on ice with the second extraction buffer: 350 l of distilled H 2 O, 115 l of methanol, and 115 l of 1 M NaCl. The tubes were then centrifuged for 3 min at 13,300 rpm at 4°C. These tubes were stored at Ϫ80°C for no longer than 1 week. Upon thawing, the tubes were spun again to reseparate the layers. Equal amounts of volume from the bottom layer were added to a new tube along with 10 l of phosphatidylbutanol standard solution (Avanti Polar Lipids, 860203C, 10 mg/ml). We dried the samples with nitrogen for 10 min. All of the dried samples were resuspended in 30 ml of spotting solution (9:1 chloroform:methanol) and applied to a TLC plate (Millipore Sigma, 1.11798.0001). The plates were run in a chamber containing 100 ml of the upper mobile-phase solution of ethyl acetate:isooctane:glacial acetic acid:water, 88:40:16:80, for 2 h 30 min. The plates were air-dried for 10 min and then placed in an iodine chamber for 5 min to detect the presence of lipids. We marked the spots with a pencil and let the plate sit for an hour to allow iodine to leave. At this stage, the plates can be left at room temperature overnight before scraping. Lipid bands corresponding to phosphatidylbutanol were scraped and placed in 3 ml of scintillation fluid (PerkinElmer, 6013681) and quantified (cpm) in a scintillation counter. All our experiments were performed in triplicate and normalized to protein levels determined with extra plates subjected to identical conditions until cell lysis.

Cell viability assay
We seeded three 6-cm dishes with 0.5 ϫ 10 6 Tsc2 Ϫ/Ϫ p53 Ϫ/Ϫ cells, and three 6-cm dishes with 0.7 ϫ 10 6 Tsc2 ϩ/ϩ p53 Ϫ/Ϫ cells. 24 h later we started drug treatments and allowed the drugs to act for 24 h, after which we collected medium into a 15-ml Falcon tube (to include dead cells). We collected cells with 0.5 ml of 1ϫ trypsin (Sigma T4174) into the respective 15 ml Falcon tube and added DMEM 10% FBS 1ϫ pen/strep to each tube to a total of 14 ml. We centrifuged cells at 1500 rpm for 5 min at 4°C. We discarded supernatants and resuspended cellular pellets in 2 ml of DMEM 10% FBS 1ϫ pen/strep. We prepared Eppendorf tubes for each condition with 60 l of DMEM 10% FBS 1ϫ pen/strep, 20 l of cellular suspension (1/4 dilution factor), and 8 l of a 0.4% trypan blue solution (Sigma T8154). We allowed trypan blue incorporation for 5 min. Then we loaded 10 l into a cell-counting hemocytometer chamber and counted nonviable cells (blue) and viable cells (not blue) per condition. We performed three cellular counts per condition.

Quantification and statistical analysis
Statistical comparisons of two different treatments in the same cell type were performed with Student's t test (nonpaired, two tails, equal variance) with R programming in R studio or Excel. All figures describe significant differences found in t test comparisons performed with means of triplicate experiments per condition: ***, p Ͻ 0.001; **, p Ͻ 0.01; and *, p Ͻ 0.05.

Supporting "Experimental procedures"
Reagents, antibodies, CMC determination, siRNAs, transient transfection of siRNAs, cell lysis, and Western blotting can be found in the supporting text.