FLCN Maintains the Leucine Level in Lysosome to Stimulate mTORC1

The intracellular amino acid pool within lysosome is a signal that stimulates the nutrient-sensing mTORC1 signalling pathway. The signal transduction cascade has garnered much attention, but little is known about the sequestration of the signalling molecules within the lysosome. Using human HEK293 cells as a model, we found that suppression of the BHD syndrome gene FLCN reduced the leucine level in lysosome, which correlated with decreased mTORC1 activity. Both consequences could be reversed by supplementation with high levels of leucine, but not other tested amino acids. Conversely, overexpressed FLCN could sequester lysosomal leucine and stimulate mTORC1 in an amino acid limitation environment. These results identify a novel function of FLCN: it controls mTORC1 by modulating the leucine signal in lysosome. Furthermore, we provided evidence that FLCN exerted this role by inhibiting the accumulation of the amino acid transporter PAT1 on the lysosome surface, thereby maintaining the signal level within the organelle.


Introduction
The ancient mechanistic target of rapamycin (mTOR) signalling pathway is critical for cells to adjust their metabolism to the available nutrients. The central factor in this process is mTOR, a protein kinase whose targets include key regulators of protein synthesis. Not surprisingly, deregulated mTOR has been linked to multiple pathological conditions, including cancer and diabetes [1,2]. mTOR can be incorporated into two protein complexes, mTORC1 and mTORC2. In a current model, the amino acid pool within lysosome serves as a signal to sequentially recruit Ragulator and the heterodimeric Rag GTPase complex with the help of lysosome-associated vacuolar-type H+ ATPase (v-ATPase). In its active form, the latter anchors mTORC1 on the lysosome surface, where it encounters Rheb and is activated [3][4][5][6][7][8]. Rheb itself is inactivated by the tuberous sclerosis complex (TSC), which is excluded from lysosomes in the presence of growth factors [9]. Thus, lysosome plays at least two roles in mTORC1 signalling. First, it provides a physical space where activation takes place; second, its content of amino acids activates the signalling. To date, most studies have focused on the cascade that transfers the luminal signal to mTORC1, but the sequestration of upstream cues within the lysosomal lumen remains poorly understood.
The relationship between FLCN and mTOR is unclear yet because mTOR can be either upor down-regulated under loss-of-FLCN conditions in various model systems, suggesting that FLCN controls mTOR in a context-dependent manner [18,21]. Using Drosophila as an animal model, we previously found that the fly FLCN (DBHD) mutants grew slowly and died as small larvae, a consequence that is similar to that of protein starvation or impaired mTOR signalling. Surprisingly, supplementation with high level of leucine, but not other amino acids, in the medium markedly rescued the growth phenotypes. Furthermore, the addition of rapamycin, a specific inhibitor of mTOR, to the same leucine-rich medium failed to rescue the phenotype [28]. These results suggest that loss of DBHD sensitizes flies to the leucine signal that stimulates mTOR. However, the underlying mechanisms remain to be elucidated. More recently, two groups independently reported that in a mammalian cell system, FLCN bound and activated the Rag complex on the lysosome, which was necessary to anchor mTORC1 [21,29]. These works demonstrate that FLCN is involved in the signal transduction process, and lysosome is the site for FLCN to perform this function.
In this paper, we report a novel, more upstream role of FLCN in the leucine-stimulated mTORC1 signalling. Specifically, FLCN controls the leucine signal in lysosome. Using human embryonic kidney 293 (HEK293) cells as a model, we found that lysosomal leucine was reduced upon FLCN suppression, causing decreased mTORC1 activity. Supplementation with high leucine reversed both consequences. Conversely, elevated FLCN levels enable cells to sequester leucine within lysosome and activate mTORC1 under amino acids starvation conditions. We also provide evidence that FLCN exerts this function by preventing the accumulation of the transmembrane amino acid permease PAT1 (Proton-Assisted Amino Acid Transporter 1, or slc36a1) on the lysosomal surface. Our work demonstrates that the FLCN-mediated modulation of the leucine level in the lysosome is an important mechanism to regulate mTORC1 signalling.

Cell culture and transfection
Cells were cultured in DMEM (Life Tech.) supplemented with 8% fetal bovine serum (FBS), penicillin and streptomycin. FLCN-HA, EGFP-PAT1 or LAMP1-EGFP plasmid was introduced into HEK293 cells using the TurboFect transfection reagent (Life Tech.). Stable cell lines were selected on the bases of the neomycin resistance. For RNAi experiments, RNAiMax diluted in OptiMEM (Life Tech.) was used to deliver siRNA according to the manufacturer's instructions; shRNA sequences were cloned into the expressing plasmid (pCD513B-U6) and co-transfected with the helper plasmids (GAG, REV and VSV-G) into HEK293T cells using TurboFect transfection reagent (Life Tech.). Purified and concentrated virus was used to infect cells with puromycin selection. To confirm the data of the knockdown experiments, we tested at least two RNAi of each target gene following published works and obtained similar results, including two siRNAs of FLCN (siFLCN-1 and 2 in ref. 21), two shRNAs of FLCN (shFLCN-1 and 2 in ref. 29), two siR-NAs of PAT1 (si158 and si160 in ref. 30) and two siRNAs of PAT4 (si435 and si437 in ref. 30). After transfection, cells were allowed to recover for at least 36 hrs before the following treatment.

Western blotting and coimmunoprecipitation
Cell extracts were subjected to protein separation using sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and then electrophoretically transferred to PVDF membranes (Millipore). The membrane was blocked with Tris-buffered saline containing 0.5% Tween 20 (TBST) and 5% nonfat dry milk powder for 1 h. The incubation with primary antibody was carried out at 4°C overnight, and the secondary antibody incubation was performed at room temperature for 1 h. After extensive washes with TBST, the blot was visualized by enhanced chemiluminescence (Advansta, USA). The quantification analysis was performed using the Bio-Rad Quantity One software.
For the coimmunoprecipitation experiments, HEK293 cells were rinsed once with ice-cold PBS and lysed with NP40 lysis buffer (0.5% NP-40, 25 mM Tris, 200 mM NaCl, 200 mM KCl, 1.5 mM MgCl2, 0.5 mM PMSF, 1 mM EDTA, 5% glycerol, and protease inhibitor cocktail, Roche, pH7.4), followed by centrifugation at 10,000g for 10 mins at 4°C. The supernatant were immunoprecipitated with either anti-GFP polyclonal antibody (Life Tech.) or anti-FLCN. Protein A/G PLUS-Agarose (Santa Cruz) was then added to the above cell lysate with rotation at 4°C for 2 hrs. The immunoprecipitated proteins were analysed by western blotting.

Immunofluorescent staining
The cells were grown on glass coverslips. After fixation with 4% formaldehyde in PBS for 10 mins. The cells were rinsed once with PBS and permeabilized by incubation with PBST for 5 mins (0.1% Triton X-100 in PBS). Incubation with either primary or second antibody was performed at room temperature for 2 hours. Nuclei were counterstained with DAPI (1 μg/ml in PBS, Sigma). Images were captured using confocal microscope (Nikon A1R-si). To quantify the colocalizations of FLCN and LAMP1, two-channel stacks of unedited images were analyzed using the Nikon NIS-Element software and the Pearson's correlation coefficient was calculated. For each sample, at least 10 cells in each of three repeated experiments were analyzed. The ttest was used to assess the results, ±SEM.

Fly feeding experiment
Fly stocks: esg-Gal4, da-Gal4 and lsp-Gal4 were obtained from the Bloomington stock centre; UAS-DBHD was described previously [28]. Flies were routinely raised on normal medium (8% sugar, 10% corn flour, 1.5% baker's yeast, 1% agar, 0.4% Propionic Acid and 0.1% Nipagin) under standard conditions (25°C,~60% humidity). Each Gal4 line was individually crossed with UAS-DBHD. Once females started to lay eggs, they were transferred to fresh starvation food (0.3% yeast) or starvation food containing 1 μM rapamycin and allowed to lay eggs for 24 hours. The newborn flies were counted each day. Heterozygote flies eclosed from the same cross (no overexpression) were taken as the control. Typically, a wild-type embryo fed normal food requires approximately 7 days reaching adulthood.

Lysosome purification and amino acid quantification
Crude lysosomes were isolated using a Lysosome Isolation Kit (LYSISO1, Sigma-Aldrich, USA). For each sample, five dishes (90 mm) of cells grown to 90% confluence were collected by centrifugation at 600g for 5 min (~1.5 X 108 cells in total). The following steps were carried out on ice or at 4°C. The cells were resuspended in 200 μl extraction buffer and homogenized with 5 gentle strokes in a 2 ml Dounce glass tissue grinder. Normally, this treatment lyses approximately 80% of cells (checked under microscope). After centrifugation at 1000g for 10 min, the supernatant was collected, and the pellet was homogenized for another four rounds. The final pellet was considered the "nuclei" fraction, which contained mostly nuclei and un-homogenized cells. Supernatants from all five homogenizations were then collected (~1 ml) and centrifuged at 20,000g for 20 min. The resulting supernatant was referred to as the "cytosol" fraction. The pellet was resuspended in 200 μl of extraction buffer to yield the crude lysosome fraction (CLF), which was checked by immunoblotting of LAMP1 and LysoTracker Red staining.
The lysosome was further purified from the CLF through density gradient ultracentrifugation method following a previous report [31]. The density gradient was prepared by sequentially adding different concentration of Optiprep Density Gradient Medium Solution (Sigma) diluted with 2.3 M of sucrose (bottom to top: 27, 22.5, 19, 16, 12 and 8%) into a centrifuge tube. The CLF was firstly mixed with 300 μl of 19% Optiprep Density Gradient Medium Solution and carefully placed on top of the density gradient. Centrifugation was carried out at 150,000g for 4 hours at 4°C. According to the previous report [31], we collected the upper 50% (v) resulting solution, mixed it with an equal volume of PBS and centrifuged at 180, 000g for 30 min. The pellet (~20-25 μl) was washed once with PBS and resuspended in 20 μl of ice-cold extraction buffer containing proteinase inhibitor and 8 mM CaCl2 to destroy the residual mitochondria contaminant. After incubation on ice for 15 min, it was centrifuged at 5000g for 15 min. The supernatant was collected and centrifuged at 180, 000g for 30 min 4°C. The pellet was the purified lysosome and lysed with 50 μl of lysis buffer containing phosphatase inhibitor, proteinase inhibitor and PMSF. Cellular fractions were examined by immunoblotting for LAMP1. Calnexin (ER) and histone 4 (nucleus) were taken as two indicators of the purity of lysosomal fractionanation. The amount of lysosome (or cytosolic fraction) was normalized to the total protein content of the purified lysosomes (or cytosol). The free amino acid concentrations were measured using liquid chromatography-tandem mass spectrometry (LC-MS/MS) at the Center for DNA typing of the Fourth Military Medical University, China.

Results
High leucine supplementation reverses the down-regulated mTORC1 in FLCN-deficient HEK293 cells In our previous Drosophila work, we demonstrated that a high level of dietary leucine, but not other tested amino acids (including tryptophan, arginine and glutamine), considerably rescued the slow growth phenotype of DBHD mutants [28]. However, it is not known whether this mechanism is conserved in mammals. Recently, two groups reported that suppression of FLCN inhibited mTORC1 in both insect and mammalian cells [21,29]. We replicated these experiments and confirmed their discovery: in human HEK 293 cells, the knockdown of FLCN by expressing a virus-mediated short-hairpin RNAs targeting FLCN (shFLCN-1, ref. 29) suppressed the phosphorylation of S6K1 on threonine 389 (pS6K1), a marker of mTORC1 activity ( Fig 1A; lane 2). On the basis of this experiment, we increased the amount of leucine in the culture medium as a supplement, attempting to reverse the down-regulation of mTORC1.
According to the manufacturer, the culture medium that we were using contained 0.8 mM of leucine (DMEM, Life Tech.). When this concentration was gradually increased up to 40 mM (50X) and combined with an extended incubation time (up to 2 hrs), the signal of pS6K1 was markedly maintained in the FLCN-depleted cells (Fig 1A). We confirmed this finding by using two different sets of shFLCN following high leucine stimulation treatment (Fig 1B).
During these experiments, we did not find notable phenotypes of the cells following high leucine stimulations (less than 2 hours of stimulation before harvesting the cells). Therefore, we used 40 mM (50X) of leucine in the following related experiments.
Leucine is potent stimulator of mTORC1. Excess leucine may induce strong mTORC1 activity to compensate for the reduction caused by FLCN-loss or mTORC1 can be triggered by a high leucine signal via an FLCN-independent pathway. However, both cases are somewhat unlikely because the pS6K1 signal in wild-type cells did not apparently increase, even in response to high leucine supplies (Fig 1B; compare lane 1, 2). Considering the similar rescues in the fly mutants that the FLCN gene was completely missing [28], it is probably a conserved mechanism that leucine can substitute for at least some FLCN functions to activate mTORC1. We also tested high levels of three other amino acids (50X), arginine, glutamine and lysine, and found that the rescue effect was specific to leucine ( Fig 1C). This result was not completely unexpected because the FLCN-containing Rag complex has been shown to transfer signals from leucine, but not glutamine, to mTORC1 [32].
We also performed another dosage experiment by starving cells with amino acid-free medium and then added either basal or high doses of leucine. In wild-type cells, leucine at a basal dose (0.8 mM) has almost maximized the activation (Fig 1D, lane 3, 5). However, in FLCN-deficient cells, a basal dose resulted in much weaker signalling than a high dose (Fig 1D; lane 4, 6). Thus, the suppression of FLCN sensitizes cells to the available leucine for mTORC1 induction.
Upstream of the Rag complex, the lysosome-anchored v-ATPase maintains the acidic environment in the lysosome and is required to sense the amino acid signal to mTORC1 [6]. We found that bafilomycine A1 (BafA1), a specific inhibitor of v-ATPase [33], blocked mTORC1, even in the presence of a high leucine supply (Fig 1E). This finding implies that the stimulation of mTORC1 by high levels of leucine requires functional v-ATPase and probably also intact lysosomes.

FLCN regulates the level of leucine in lysosome
The amino acid pool within the lysosome has been shown to be a direct signal to activate mTORC1 [6,34]. Combined with the feeding experiment data in Drosophila, we proposed before that FLCN might modulate intracellular leucine, especially those in the lysosome, to control mTORC1 [28].
This hypothesis implies that FLCN resides on the lysosome. Previously, two groups have provided immunofuorescent staining data showing that FLCN is localized on lysosome in a nutrient-dependent manner: FLCN was accumulated on lysosome upon starvation and was dispersed into the cytoplasm by restimulation with amino acids [21,29]. We repeated this experiment and obtained similar observations: FLCN seemed to be evenly distributed within the cytoplasm in complete medium and became accumulated on the lysosome when amino acids were depleted (Fig 2A). Based on the immunostaining data, it is not clear if FLCN is located on the lysosome under the normal culture conditions. Therefore, we applied another approach by purifying the lysosome from cell lysate (see Materials and methods) and measured FLCN through western blotting. Consistent with the immunostaining data, there was a strong signal of the lysosome-anchored FLCN when cells were starved with amino acids, (Fig 2B; lane  4). Notably, we could also readily detect a relatively weaker signal even when cells were cultured in complete medium (Fig 2B; lane 3). This result supports that a small amount of FLCN on the lysosome is probably sufficient to activate mTORC1 (via Rags) in the presence of amino acid [21,29].
After confirming its presence on the lysosome, we explored the influence of FLCN on the upstream cue, namely the amount of amino acids within the lysosome. To this end, we purified lysosomes and measured the luminal concentrations of free leucine. In three independent experiments, leucine was consistently reduced in the lysosomes from cells expressing shFLCN, which was efficiently blocked by supplementation with high levels of leucine in the medium (Fig 2C). These results suggest that FLCN maintains the leucine level in lysosome, which can be alternatively achieved by increasing the leucine supply in the environment.
The transport of leucine across cellular membranes was thought to be mediated through facilitated diffusion, meaning it moves down the concentration gradient with the aid of special transmembrane carriers. Under certain conditions, such as when the extracellular supply is high, the transport can also occur via low-affinity systems or passive diffusion [35][36][37]. Consistent with these alternative mechanisms, incubation with high amounts of leucine significantly increased its level in the cytosol of both wild-type and FLCN-deficient cells (Fig 2D). In contrast, the level of leucine in the lysosomes of the same population of wild-type cells only slightly increased. Accordingly, the level of pS6K1 was not markedly increased ( Fig 1A). We conclude that the suppression of FLCN leads to decreased leucine in the lysosome, which is insufficient to fully activate mTORC1. However, high leucine supplementation results in its accumulation in both the cytosol and lysosome due to saturation mechanisms. As a result, the leucine signal in the lysosome strengthens sufficiently to activate mTORC1.

Elevated FLCN renders cells resistant to starvation to activate mTORC1
We attempted to test this FLCN function from a different perspective by investigating the ability of ectopic FLCN to sequester more leucine within the lysosome. To this end, we generated several stable cell lines expressing FLCN-HA controlled by the ubiquitous CMV promoter, which were considered to be FLCN-overexpression systems in the following experiments.
In complete culture medium, the pS6K1 signal was not markedly increased in cells expressing ectopic FLCN-HA ( Fig 3A; lane 1, 3, 5). However, in amino acid-free medium, pS6K1 was decreased in both wild-type and two stable FLCN-HA cell lines, but the decrease was less pronounced in FLCN-HA cells (Fig 3A; lane 2, 4, 6). Thus, elevated FLCN enables cells to resist the shortage of amino acids for mTORC1 induction. In a dosage experiment, we re-stimulated the starved cells with different doses of leucine. In wild-type cells, the signal of pS6K1 parallels the concentration of supplemented leucine when it is below basal doses (<0.8 mM), implying that mTORC1 is sensitive to the leucine level under this culture condition. However, this sensitivity was clearly decreased in the FLCN-HA stable cell lines ( Fig  3B). In addition, BafA1 inhibited mTORC1 in the FLCN-HA cells in a dosage-dependent manner, suggesting v-ATPase acts downstream of or in parallel with the overexpressed FLCN ( Fig 3C).
To examine the correlation of the above data with the levels of lysosomal leucine, we measured the abundance of leucine in purified lysosomes. Taking the value in wild-type cells cultured in complete medium as a reference, we found about 41% reduction in lysosomal leucine after amino acid deprivation. In contrast, this number decreased to 2% and 20% in the two stable FLCN-HA lines (Fig 3D). Even cultured with complete medium, the FLCN-HA cells contained more lysosomal leucine than wild-type cells (Fig 3E). However, this increase seemed to be more prominent in amino acid-free medium (Fig 3D), probably because there are more FLCN proteins on the lysosome under starvation.
These data suggest that ectopic FLCN promotes the accumulation of leucine within the lysosome, which decreases the sensitivity of these cells to leucine shortages in the environment and facilitates mTORC1 induction. Previously, Baba et al. observed that a patient-derived, FLCNmutated cell line stably expressing ectopic FLCN was resistant to amino acid deprivation to activate mTORC1 (Fig 7B in ref. 24). Because they also expressed FLCN using a constitutive promoter, we suspect that their cell line might act similarly with our FLCN-overexpression systems.

Overexpression of DBHD promotes the growth of Drosophila in starved environment
To test the discovery in multicellular organisms, we used Drosophila as an animal model and overexpressed the fly FLCN gene (DBHD) using three different tissue-specific Gal4 driver lines, including da-Gal4 (ubiquitous), lsp-Gal4 (fat body, the nutrient storage organ) and esg-Gal4 (intestinal progenitor cells). When fed with standard medium, most embryos developed into adults within one week without apparent abnormalities, indicating that excessive DBHD does not affect general morphogenesis and growth. When fed a low-protein food (20% of the level in the normal recipe), all newly eclosed flies shank in size ( Fig 3F) and required an extended period of time to reach adulthood (nearly doubled). Both were typical responses to starvation or impaired mTOR signaling [38]. Interestingly, flies expressing a high level of FLCN ubiquitously (da-DBHD) showed a growth advantage over other flies, as exhibited by a shorter developmental time (Fig 3G). Nevertheless, this effect was suppressed by administration of rapamycin (Fig 3G), suggesting that the growth advantage is caused by relatively high mTOR activity.

FLCN and PAT1 antagonize each other to control mTORC1
Lysosomal leucine levels can be reduced by either impaired protein breakdown or accelerated loss. As FLCN is located on the lysosome, it may regulate the efflux of leucine, a process that normally requires it to be loaded onto membrane-anchored transporters. We searched the literatures and considered PAT1 as a good candidate of this transporter. PAT1 was initially identified as a lysosome-associated membrane protein whose function was to export amino acids into the cytosol [39]. In fact, previous work has shown that the overexpression of PAT1 promoted the release of amino acids from lysosome and inhibited mTORC1 [6]. Based on these results, we investigated the involvement of PAT1 in the FLCN-mediated mTORC1 signalling.
We first confirmed that the overexpression of PAT1 inhibited mTORC1. HEK293 cells either stably or transiently expressing an EGFP-PAT1 fusion protein under the control of the CMV promoter were starved of amino acids for 50 minutes, followed by re-stimulation with amino acids for 10 minutes. Following this treatment, the level of pS6K1 was decreased in all PAT1-overexpressing cell populations (Fig 4A). Next, we added in the medium with high levels of leucine. We found that this treatment prevented the decrease of pS6K1 in the PAT1-overexpressing cells (Fig 4B; lane 4), indicating that leucine is probably the major cargo of PAT1 to stimulate mTORC1. Interestingly, a similar rescue effect was obtained by the overexpression of FLCN (Fig 4B; lane 5). In a reciprocal fashion, the pS6K1 remained stable upon starvation in cells expressing ectopic FLCN-HA but not in the same cells co-expressing EGFP-PAT1 ( Fig  4C). Together, these results suggest that FLCN and PAT1 antagonize each other to regulate mTORC1. Consistent with this conclusion, the knockdown of PAT1 by two different sets of siRNA somehow inhibited the starvation-induced decrease of pS6K1 (Fig 4D; lane 2, 3, 5), which could be due to attenuated leucine loss; nevertheless, this effect was counteracted by the co-suppression of FLCN (Fig 4D; lane 4, 6).
The PAT gene family contains four member (PAT1-4). Among them, PAT1 is mostly related to PAT4 based on sequence similarities. Moreover, PAT4 has been shown to positively regulate mTORC1 [30]. In a similar experiment of Fig 4D, we found that FLCN did not counteract PAT4 in the regulation of mTORC1 (S1 Fig, lane 4, 6), confirming certain specificity of FLCN on the PAT genes.
FLCN inhibits the accumulation of PAT1 on lysosome FLCN may either inhibit the transport activity or abundance of PAT1, especially on the lysosomal surface. We selected one stable line expressing a relatively low level of EGFP-PAT1 to monitor PAT1 and checked its abundance on the purified lysosome by western blotting. We found that the lysosome-anchored EGFP-PAT1 in the FLCN-depleted cells was significantly increased compared with that in wild-type cells (Fig 5A). In contrast, the total EGFP-PAT1 prepared from the whole lysate of FLCN-depleted cells was slightly elevated. Similar results were obtained by using a different set of siFLCN (siFLCN-2 in S2 Fig). These results suggest that FLCN inhibits the lysosomal localization of PAT1 and may also affect the stability of PAT1 protein. Consistent with these findings, the budding yeast LST7 gene, a homologue of FLCN, has been shown to block the translocation of the general amino acid permease Gap1p to yeast lysosome-like vacuole and suppress the expression of several amino acid permeases [40,41]. Because the intracellular localizations of several related genes, including FLCN, yeast LST7 and Gap1p, are all responsive to the nutrient status [21,29,40], we suspect that PAT1 may exhibit a similar dynamic pattern within the cells. In support of this hypothesis, we found that starvation induced the accumulation of EGFP-PAT1 on the lysosomal fractionation ( Fig 5B;  lane 1, 2). Thus, the binding of PAT1 to the lysosome correlates with the nutrient status in a manner similar to that of FLCN or the yeast counterpart Gap1p. Interestingly, ectopic FLCN efficiently inhibited the starvation-induced accumulation of EGFP-PAT1 on the lysosome ( Fig  5B; lane 3, 4), implying that FLCN inhibits the starvation-induced recruitment of PAT1 on lysosome. These findings could explain the previous data of the vibrations of the lysosomal leucine levels. When the amino acid supply is limited, cells recruit PAT1 to the lysosomal surface to recycle the luminal storage, resulting in a decreased leucine level (Fig 3D). However, this movement of PAT1 can be inhibited by ectopic FLCN. As a result, the leucine level in the lysosome was maintained (Fig 3D).
As a further evidence of the close relations between FLCN and PAT1, we found that EGFP-PAT1 co-immunoprecipitated with FLCN in the cell lysate in a reciprocal manner (Fig 5C). Moreover, such interaction seemed not to be affected by the nutrient status (S3 Fig). These results suggest that FLCN may regulate the localization of PAT1 through direct interactions. Currently, we are investigating the mechanisms that govern the dynamic distribution of PAT1 by FLCN. Because FLCN has been shown to contain a distant DENN domain and thus was suspected to be a modulator of the Rab family proteins [19], we suspect that FLCN may be involved in the vesicular trafficking process of PAT1. Alternatively, FLCN may affect the transport activity of PAT1. We could not exclude this possibility yet.

Discussion
Previously, people have shown that the depletion of FLCN failed to activate mTORC1 in the presence of amino acids (1X) and consequently regarded FLCN as a signal transducer. We herein describe a new more upstream role of FLCN in maintaining the leucine level in the lysosome. In the absence of FLCN, the concentration difference across the lysosomal membrane tends to drive leucine out of the lumen. This trend can be stopped by supplementation with high levels of leucine, and this effect seems to be facilitated by low-affinity systems and/or passive diffusion. In support of the low-affinity systems, we found that incubation with high levels of leucine requires longer time than the common amino acid stimulation treatment to activate mTORC1 in FLCN-deficient cells (up to 2 hrs vs. less than 30 mins, Figs 1A and 2C). In contrast, 30 mins of incubation is sufficient to activate mTORC1 in the PAT1-overexpressing cells (Fig 4B). This observation also indirectly supports that PAT1 is a transporter of leucine. PAT1 has been found to on both the lysosome and cell surface [39,[42][43][44]. The lysosomeanchored PAT1 inhibits mTORC1 by attenuating the leucine signal, whereas at other sites, including the plasma membrane, it may promote mTORC1 via the bulk uptake of amino acids into the cell. Our data reveal that the intracellular localization of PAT1 is dependent on the nutrient status and FLCN expression levels. When the amino acids supply is limited, PAT1 translocates to lysosome to export luminal storage into the cytosol to meet the cell's nutrient requirements. Conversely, decreased leucine fails to stimulate mTORC1. In this way, cells can adjust their metabolism to the nutrient conditions. The mechanism by which amino acids guide the translocation of PAT1 and FLCN remains unexplored.
Previously, one research group reported a different observation in which elevated PAT1 promoted mTORC1 activity in HEK293 cells [30,43]. We noticed that the author stimulated the starved cells with complete medium for 30 mins. In contrast, we and others [6] performed the re-stimulation for 10 mins. With this shorter stimulation treatment, a significant amount of PAT1 may still be anchored on the lysosome, which is sufficient to maintain a low leucine signal.
A number of budding yeast genes, their mammalian homologues have been characterized to be components of mTOR signalling, are involved in a common pathway that regulates the anchorage of the general amino acid transporter Gap1p on the yeast lysosome-like vacuole. These genes include LST7 (FLCN), LST4 (FNIP1), LST8 (component of mTORC1/2), Gtr1/2 (Rag GTPases,) and SEC13 (component of the GAP complex towards Rag GTPases) [40,[45][46][47]. Thus, certain ancient machinery may control the nutrient signal in the lysosome by adjusting the spatial location of amino acid transporters. To elucidate this point, several questions require further investigations. For example, do these mammalian members regulate the amino acid storage in lysosome? Do they control the amino acid transporters and how about the specificities of their targets? Two recent works demonstrate that the transfer of the arginine signal to mTORC1 requires the lysosome-associated arginine transporter SLC38A9 [48,49]. In contrast, we found that a high dose of either arginine or glutamine failed to rescue the FLCN-loss phenotype in either mammalian cells or Drosophila, confirming the existence of distinct mechanisms in the transduction of amino acid signals [32].
In summary, our work reveals a new role of FLCN and demonstrates that the modulation of the leucine level in the lysosome is an important mechanism to regulate mTORC1 signaling. In our provisional model, multiple factors, including the intracellular localizations, the expression levels of FLCN and PAT1, and the local nutrient conditions, can all affect the leucine signal in the lysosome and thereby the mTORC1 activity. This work extends our knowledge of how cells sense the available nutrients and also helps us to appreciate the context-dependent relationships between FLCN and mTORC1.
Supporting Information S1 Fig. FLCN does not counteract PAT4 in the regulation of mTORC1. The experiment was performed similarly as that in Fig 4D. The two different sets of siPAT4, including si435 and si437, have been described before [30]. The knockdown efficiencies of both siPAT1s were analyzed by qRT-PCR and shown to the right. (TIF) S2 Fig. Suppression of FLCN (using a second siFLCN) promotes the accumulation of PAT1 on the lysosome. siFLCN-2 is different with the one used in Fig 5A. Both siFLCN-1 and siFLCN-2 have been described before [21]. (TIF) S3 Fig. Nutrient status has little effects to influence the interactions between the endogenous FLCN and EGFP-PAT1. Left: Cell lysates of EGFP-PAT1 stable cells were co-immunoprecipitated with the antibody against either calnexin (negative control) or FLCN; Right: the opposite direction of the co-IP experiment, the 293 cells transfected with an empty EGFP plasmid was taken as a negative control. Note with or without amino acids, the strength of interactions (marked with stars) did not show significant differences. (TIF) stocks. This work was supported in whole or part by a NWAF grant (Z111021005) and a National Natural Science Foundation of China (31372256) to WL.