FLCN is a novel Rab11A-interacting protein that is involved in the Rab11A-mediated recycling transport

Mutations of the tumor suppressor folliculin (FLCN) have been linked with the Birt–Hogg–Dubé (BHD) syndrome that is clinically characterized by benign skin tumors, spontaneous pneumothorax, and kidney cancer (Birt et al., 1977; Schmidt et al., 2005; Toro et al., 2008). FLCN was thought to be involved in a number of biological processes, including signal transduction, biogenesis of lysosome and mitochondria, cell-cell adhesion, membrane trafficking, energy and nutrient homeostasis, and so on (for a recent review, see Schmidt and Linehan, 2018), but its precise functions related to the BHD symptoms are not fully understood.

The Flcn gene has been highly conserved during evolution. FLCN orthologs have been characterized in yeast (Péli-Gulli et al., 2015; Roberg et al., 1997), nematode (Gharbi et al., 2013; Possik et al., 2014), fruit fly (Liu et al., 2013; Singh et al., 2006), zebrafish (Kenyon et al., 2016) and mouse (Baba et al., 2016; Chen et al., 2008). This implies that FLCN probably controls certain fundamental cellular processes that are not unique to higher organisms. Using both Drosophila (Liu et al., 2013) and the cultured human embryonic kidney HEK293 cells (Wu et al., 2016) as the model systems, we discovered that FLCN positively regulates the mTORC1 signaling pathway, in part, by maintaining the stimulatory amino acid signal within the lysosome. In addition, we found FLCN performed this function largely by suppressing the binding of the amino acid transporter proton-coupled amino acid transporter 1 (PAT1; also known as SLC36A1) on the lysosomal surface (Wu et al., 2016). These results can reveal a putative function of FLCN during the vesicular trafficking of PAT1.

PAT1 is called the proton-coupled amino acid transporter 1. It belongs to the solute carrier protein 36A subfamily that in mammals contains four members (PAT1-4, or SLC36A1-4). PAT1 is an eleven-pass transmembrane protein that can move amino acids, together with protons, from both the extracellular environment and the intracellular membrane compartments into the cytosol (Thwaites and Anderson, 2011). In many examined cell types, PAT1 is enriched on the lysosomal surface, suggesting it has a general function in the luminal nutrient homeostasis. Besides, PAT1 was also detected on several other cellular locations, including the plasma membrane (Anderson et al., 2004; Chen et al., 2003; Sagne et al., 2001; Wreden et al., 2003), axons (Wreden et al., 2003), podosomes (Cougoule et al., 2005), and even in the nuclei (Jensen et al., 2014), in different cell types. This indicates that PAT1 may have diverse physiological functions, which are related to its specific subcellular localizations. In addition to its well-characterized role in nutrient homeostasis, PAT1 was also found to regulate mTORC1, which is a multicomponent protein kinase complex that controls cell growth and metabolism by sensing amino acids in both the lysosome and the cytosol. Both Sabatini's group and ours have shown that PAT1 could negatively regulate mTORC1, probably through a mechanism by decreasing the lysosomal amino acid signal level (Tsun et al., 2013; Wu et al., 2016). More interestingly, we found the localization of PAT1 to the lysosome seemed to be dynamic and context-dependent (Wu et al., 2016).

Using the overexpressed PAT1 as a substrate, we found it could be transported between the plasma membrane and the lysosome in HEK293 cells. The distributions of PAT1 on these two sites were influenced by two factors, including amino acids and FLCN (Wu et al., 2016). We have recently shown that amino acids can inhibit the transport of overexpressed PAT1 to the lysosome by inducing the N-terminal protein cleavage, leading to the loss of a tyrosine-based lysosomal targeting signal (Ji et al., 2017). This discovery reveals a putative mechanism that amino acids can stimulate mTORC1 through modulating the localizations of amino acid transporters. So far, the mechanisms of FLCN in the targeting of PAT1 are unknown. Notably, a structural study identified a differentially expressed in normal and neoplastic cells (DENN)-like domain at the C-terminus of FLCN proteins (Nookala et al., 2012). Because some DENN-containing proteins can act as GTP exchange factors (GEF) to activate the Rab family of small GTPases, which are key regulators of vesicular trafficking, FLCN has been suspected to control the trafficking of membrane proteins (Nookala et al., 2012). However, it is only recently that the evidence of FLCN in protein transport has started to emerge, probably because the cargos transported by FLCN were largely unknown. During the ongoing of this research, two groups reported that FLCN promotes the transport of EGFR to the lysosome for degradation by modulating the activities of Rab7 (Laviolette et al., 2017) and Rab35 (Zheng et al., 2017), respectively.

In this study, we looked at overexpressed PAT1 as a substrate and discovered that FLCN regulates Rab11A-mediated recycling transport. Our study provides new insights into how FLCN can regulate mTORC1 and might link BHD to the dysregulation of Rab11A-regulated recycling pathways.

To understand how FLCN regulates the transport of overexpressed PAT1 to the lysosome, we generated HEK293 cells constitutively expressing Myc-PAT1, which has been shown to colocalize well with endogenous PAT1 (Ögmundsdottir et al., 2012). Theoretically, targeting of membrane proteins to the lysosomes can be affected by two opposing pathways – one that sends them to the lysosome and one that recycles them to the cell surface and/or the Golgi network. To study whether Myc-PAT1 travels via these two pathways, we performed fluorescence immunostaining and analyzed the colocalization of Myc-PAT1 with several important endosomal markers, including Rab5 (early endosome), Rab7 (late endosome), Rab11 (recycling endosome) and LAMP1 (lysosome).

To monitor the different types of endosome, we first used the mCherry-tagged endosomal markers. Myc-PAT1 was stained with a monoclonal anti-Myc antibody and the mCherry signals were captured directly without staining. As anticipated, Myc-PAT1 colocalized well with LAMP1-mCherry (Fig. 1A). In addition, we also observed many overlaps between Myc-PAT1 and overexpressed Rab5A, Rab7A or Rab11A (Fig. 1A). This indicates that Myc-PAT1 may be transported along both endocytosis and recycling pathways. Next, we examined the endogenous endosomal markers. Consistent with the previous reports (Ögmundsdottir et al., 2012; Zoncu et al., 2011), we found Myc-PAT1 was mainly localized to the LAMP1-marked endosomes. Importantly, we also observed some overlaps between Myc-PAT1 and endogenous Rab5, Rab7 or Rab11 (Fig. 1B). To further reveal its trafficking routes, we examined additional endosomal markers, including TGN38 (for the trans-Golgi network; TGN), EEA1 (for early endosomes), Chmp5 (for the multivesicular body) and the cation-dependent mannose-6-phosphate receptor (M6PR). We found that Myc-PAT1 colocalized more or less with all these endosomal markers (Fig. 1C). Of note, Myc-PAT1 showed many overlaps with M6PR, which can recognize and deliver many hydrolases from the Golgi to the lysosome, and M6PR itself will be sent back to the Golgi for recycling (Braulke and Bonifacino, 2009). Taken together, these results suggest that Myc-PAT1 can be transported along the classic endocytosis and recycling pathways in HEK293 cells.

From the TGN, the membrane proteins can be delivered to the lysosomes through both direct and indirect pathways. Using the cell fractionation method, we have shown before that overexpressed PAT1 can be localized to the plasma membrane in HEK293 cells (Wu et al., 2016). This implies PAT1 may first be sent to the cell surface, from where it can go into the cell through membrane invagination and then gradually move to the lysosome. We confirmed this hypothesis by using internalization assay of FLAG-tagged PAT1, containing 3×FLAG epitope at its C-terminus extending to the extracellular environment. When the living cells of PAT1-FLAG were incubated with a monoclonal anti-FLAG antibody on ice (to block endocytosis), we detected PAT1-FLAG on the cell surface (Fig. 1D, 0 min). After endocytosis was restarted by switching to high temperature (37°C), the pre-labeled PAT1-FLAG on the cell surface could move into the cell and eventually targeted the lysosomes (Fig. 1D, 45 min).

Of the various localizations of Myc-PAT1, we were particularly interested in Rab11A-marked endosomes (Fig. 1A,B). Rab11A is a key regulator of the slow recycling transport pathway (Welz et al., 2014) that, in theory, should inhibit the transport of PAT1 to the lysosome and promote its targeting on the plasma membrane and/or the Golgi network. To study whether Rab11A regulates PAT1 localization, we attempted to block Rab11A activity and then quantify PAT1 on both the lysosome and the plasma membrane. First, we used the RNA interference (RNAi) technique to knockdown Rab11A. The cells stably expressing PAT1-FLAG (60% confluence) were infected with lentiviruses carrying two different small hairpin RNAs (shRNAs) targeting Rab11A (shRab11A). Lysosomes were isolated 48 h later by cell fractionation (see Materials and Methods). We confirmed that lysosomes could be purified using this method by analyzing markers of different cell fractions. In addition, these assays also revealed that starvation could induce accumulation of PAT1-FLAG on the lysosome (Fig. S1). As shown in Fig. 2A, we found that suppression of Rab11A by either of the two shRab11As clearly promoted localization of PAT1-FLAG to the lysosomes. To measure PAT1-FLAG on the plasma membrane, we used biotinylation labeling to purify the total cell surface proteins. As a result, we found that Rab11A knockdown inhibited targeting of PAT1-FLAG to the plasma membrane (Fig. 2B). We also applied a different strategy to suppress the Rab11A activity by overexpressing a dominant negative S25N mutant of Rab11A (Rab11A-DN). Consistent with the results of shRab11A, we found that Rab11A-DN promoted localization of PAT1-FLAG to the lysosome and inhibited its localization to the plasma membrane (Fig. 2C,D).

We performed double-staining experiments of Myc-PAT1 and LAMP1. As shown in Fig. 2E, we found that colocalization of Myc-PAT1 and LAMP1 was increased by either Rab11A small interfering RNA (siRNA) or Rab11A-DN but decreased by overexpressing a constitutively active Q70L mutant of Rab11A (Rab11A-CA,). Interestingly, after Rab11A-CA overexpression, we observed many Myc-PAT1-positive and LAMP1-negative endosomes that were preferentially located at sub-plasma membrane regions (arrowheads in Fig. 2E). We suspect that these endosomes probably recycle to the cell surface. Taken together, our results demonstrate that Rab11A can modulate the distribution of overexpressed PAT1 on the lysosome and the plasma membrane, probably by promoting its recycling transport pathway.

During these experiments, we noticed that suppression of Rab11A often caused a mild decrease of total PAT1 levels. This could be reversed by bafilomycin A1 (BafA1), which is a specific inhibitor of the lysosome-dependent degradation pathway (Luo et al., 2017; Yoshimori et al., 1991; Fig. 2F. These data suggest that the level of lysosomal PAT1 is probably adjustable.

To reveal the significance of Rab11A-PAT1 relationships, we decided to focus on the mTORC1 signaling pathway. It has been shown previously that overexpression of PAT1 can inhibit mTORC1 when cells were synchronized by starvation, followed by a short time (10 min) of nutrient replenishment (Wu et al., 2016; Zoncu et al., 2011). In a recent study, we have demonstrated that this result was probably caused by incomplete nutrient-induced relocation of lysosomal PAT1, extending the nutrient stimulation time can reactivate mTORC1 in the PAT1-overexpressing cells (Zhao et al., 2018).

We starved cells for 50 min and then re-stimulated them with nutrient-containing medium for 10 min. As anticipated, the mTORC1 activity, as revealed by the presence of phosphorylated ribosomal protein S6 kinase β1 (pS6K1, at T389), was decreased in cells stably expressing PAT1-FLAG, compared to wild-type cells (Fig. 3A, lanes 1, 2). Interestingly, suppression of Rab11A by shRNAs caused further reductions of the mTORC1 activity (Fig. 3A, lanes 3, 4). A similar result was observed by overexpressing the Rab11A-DN mutant (Fig. 3B, lane 3). By contrast, overexpressing Rab11A-CA increased mTORC1 activity in PAT1-overexpressing cells (Fig. 3B, lane 4). These results suggest that Rab11A can antagonize the inhibitory effect of overexpressed PAT1 on mTORC1.

We then investigated the influence of Rab11A on mTORC1 in a wild-type cell background, and found that mTORC1 activity was increased by Rab11A-CA but decreased by Rab11A-DN or shRab11A (Fig. 3C,D). This raises a question whether Rab11A can regulate mTORC1 through PAT1. Knockdown of PAT1 has been shown to inhibit mTORC1 in HEK293 cells (Heublein et al., 2010). We suspected that this might be caused by a reduction of the cytosolic amino acid pool, which is another important signal source of mTORC1 (Zhao et al., 2018). We confirmed this result by using the same PAT1 siRNA (i.e. si160) as described before by Heublein et al., 2010 (Fig. 3E, compare lane 1 with 2). Importantly, the mTORC1 activity did not change upon co-expression of either Rab11A-DN and siPAT1, or Rab11A-CA and siPAT1 (Fig. 3E, lanes 2–6). Considering that Rab11A-CA can stimulate mTORC1 in wild-type cells (Fig. 3D, lane 2), these results suggest PAT1 can function downstream of Rab11A, possibly by being a Rab11A substrate. According to this mechanism, we predict that the effect of Rab11A can be overcome by blocking the transport of PAT1 to the lysosome. Indeed, as shown in Fig. 3E, suppression of Rab11A (by using shRab11As) decreased mTORC1 activity in PAT1-overexpressing cells (compare lane 1 with lanes 2 and 5); however, this result could be rescued by blocking endocytosis with either the dominant negative T22N mutant of Rab7A (Rab7A-DN) (Fig. 3E, lanes 3, 6) or the dominant negative S34N mutant of Rab5A (Rab5A-DN) (Fig. 3E, lanes 4, 7).

There are several pieces of evidence suggesting that FLCN can interact with Rab11A. First, FLCN contains a DENN-like domain, and researchers have long suspected that FLCN can bind Rab GTPases and modulate their activities (Nookala et al., 2012). Second, both FLCN (Wu et al., 2016) and Rab11A (Fig. 2A,C) can inhibit the localization of overexpressed PAT1 to the lysosome. Third, FLCN was found to bind the cell–cell adhesion protein PKP4 (Medvetz et al., 2012; Nahorski et al., 2012), which is known to be a Rab11A-ineracting factor (Keil and Hatzfeld, 2014). On the basis of these results, we decided to explore the potential FLCN-Rab11A interaction.

First, we wondered whether FLCN can be localized to the Rab11A-marked endosomes. We transiently expressed GFP-Rab5A, GFP-Rab7A and GFP-Rab11A in HEK293 cells, and performed double immunostaining experiments using antibodies against GFP and FLCN. We confirmed the specificity of the rabbit monoclonal anti-FLCN antibody by both western blotting and immunostaining assays (Fig. S2). Notably, this anti-FLCN antibody also recognized an unspecific nuclear signal, as it was still present in the FLCN−/− cells (Fig. S2). A similar result has been reported by another research group using the same anti-FLCN antibody (Tsun et al., 2013). In the double immunostaining experiments, we detected certain overlaps between FLCN and all three Rab proteins in the cytoplasm (Fig. 4A). A quantification assay (the nuclear signals were excluded) further revealed that FLCN signals overlap more with those of GFP-Rab7A and GFP-Rab11A than with those of GFP-Rab5.

Next, we performed co-immunoprecipitation experiments (co-IP) to study the physical interaction between FLCN and Rab11A. To this end, we transiently expressed GFP-Rab5A, GFP-Rab7A or GFP-Rab11A in HEK293 cells, and incubated the cell lysates with anti-FLCN antibody. Immunoprecipitates were then examined by western blotting. We found that FLCN interacts with GFP-Rab7A or GFP-Rab11A but not (or very weakly) with GFP-Rab5A (Fig. 4B). Similar results were obtained by using the mouse proteins in mouse 3T3 embryonic fibroblast cells (Fig. 4C). Because FLCN has recently been found to interact with Rab7A in several human tumor cell lines (Laviolette et al., 2017), we considered the FLCN-Rab7A interaction as a positive control in our co-IP experiments. It is probably worth noting that, on the basis of these co-IP data; it seems that only a small proportion of GFP-tagged Rab proteins co-immunoprecipitated with FLCN. To gain more evidence regarding the interaction between FLCN and Rab11A, we tried to map the protein motif(s) that mediate the binding of FLCN to Rab11A. We found that FLCN preferentially binds to Rab11A through the DENN domain located at its C-terminus (Fig. 4D). Together, these results suggested that FLCN physically interacts with Rab11A.

To examine the specificity of the interaction between FLCN and Rab11A, we performed co-IP experiments to study whether FLCN can interact with Rab11B – another Rab11 family protein that is mainly expressed in neuronal cells (Lai et al., 1994). Surprisingly, we found that FLCN co-immunoprecipitated with GFP-Rab11B when transiently expressed in HEK293 cells (Fig. 4E). This suggests that FLCN is a common effector of Rab11A and Rab11A (Horgan and McCaffrey, 2009).

The above results led us to speculate that FLCN can regulate the Rab11A-mediated transport pathway. To test this, we used overexpressed PAT1 as a substrate and asked whether FLCN can regulate PAT1 localization.

Consistent with the previous observations (Wu et al., 2016), we found that suppression of FLCN by siRNA (siFLCN) increased localization of PAT1-FLAG on the lysosome (Fig. 5A). Interestingly, this result was accompanied by a decrease of PAT1-FLAG on the plasma membrane (Fig. 5B). To confirm this finding, we generated FLCN-knockout HEK293T cell lines (FLCN−/−) by using CRISPR/Cas9-mediated gene-editing technique (Fig. S3). Similar to the siFLCN results, we found PAT1-FLAG was increased on the lysosome but decreased on the plasma membrane in the FLCN−/− cells (Fig. 5C,D). In double immunostaining experiments (Fig. 5E), we found the overlaps between Myc-PAT1 and LAMP1 were increased by FLCN siRNA, but were decreased by FLCN overexpression (FLCN-HA). Thus, FLCN and Rab11A have the similar influences on the subcellular distribution of overexpressed PAT1.

We suspected that FLCN can interfere with overexpressed PAT1 to regulate mTORC1, as Rab11A does. As shown in Fig. 5F, mTORC1 was downregulated in starved PAT1-overexpressing cells after 10 min of nutrient stimulation treatment (Fig. 5F, lanes 1, 2). Interestingly, mTORC1 activity was further decreased when FLCN was also suppressed (lane 3) but increased when FLCN was overexpressed (lane 4). These results demonstrate that an increase of FLCN levels can antagonize overexpressed PAT1 to regulate mTORC1. Given that FLCN overexpression was unable to stimulate mTORC1 the wild-type cells (Fig. 5F, right panel), we suspected the influence of PAT1 on mTORC1 to be sensitive to the level of FLCN. Similar observations have been reported by us previously, when we found FLCN and overexpressed PAT1 antagonized each other during mTORC1 regulation (fig. 4 in Wu et al., 2016). On the basis of these results, we propose that FLCN promotes Rab11A-mediated recycling of PAT1. We also investigated how FLCN regulates mTORC1 in the absence of PAT1 (Fig. 5G) and found that compared with suppression of FLCN alone, double-suppression of PAT1 and FLCN (siPAT1+siFLCN) did not further decrease mTORC1 activity. However, we do not consider this result to be very surprising because FLCN is known to be involved in signal transduction through Rag GTPases (Petit et al., 2013; Tsun et al., 2013), the signal maintenance function (by FLCN and PAT1) must, therefore, lie upstream of the FLCN-mediated signal transduction step.

The next question was how FLCN promotes Rab11A-mediated recycling transport. We think this can be achieved by using one of the following two mechanisms: either FLCN can activate Rab11A by functioning as a GEF, or FLCN promotes the cargo loading process.

By using purified proteins, Tom Blundell's group previously reported that FLCN does not show clear in vitro GEF activity on Rab11A (Nookala et al., 2012). We performed the similar experiment using His-tagged FLCN and His-tagged Rab11A expressed in bacteria, and obtained the same result (our unpublished observation). Generally, GEF factors have high affinities with GDP-loaded Rab proteins. We checked the binding of FLCN to different forms of Rab11A and found that FLCN preferentially binds Rab11A-CA, i.e. GTP-loaded Rab11A (Fig. 6A). Thus, FLCN is probably not a strong GEF of Rab11A. However, FLCN might possess certain GEF activity on Rab11A under special conditions. More sensitive analyses, such as using FLCN-containing protein complexes isolated from mammalian cells, in both normal and starvation environments, should be carried out to clarify this issue.

We also carried out experiments to explore whether FLCN can interfere with loading of PAT1 on Rab11A, and first checked the interactions of overexpressed PAT1 with different Rab proteins. As a result, we found that PAT1-FLAG can interact with GFP-Rab5A, GFP-Rab7A and GFP-Rab11A, but not with the control GFP protein (Fig. 6B). These results support our previous immunostaining data, which showed that the overexpressed PAT1 localizes to endosomes marked with Rab5A, Rab7A or Rab11A (Fig. 1A,B). Next, we checked whether FLCN controls the interaction between PAT1-FLAG and the Rab proteins. Interestingly, we found that interaction between PAT1-FLAG and GFP-Rab11A was clearly decreased in the FLCN−/− cells. In contrast, FLCN deficiency did not disturb the interaction between PAT1-FLAG and either GFP-Rab5A or GFP-Rab7A (Fig. 6C). Consistent with a role of FLCN in cargo loading, FLCN preferentially bound to Rab11A-CA (Fig. 6A). In an attempt to check the specificity of this FLCN function, we performed co-IP experiments to examine the interaction of Rab11A with transferrin receptor (TfR), which is constitutively recycled by the Rab11A-mediated pathway. Surprisingly, we found that FLCN can interact with TfR. Moreover, the interaction between TfR and Rab11A was decreased in the absence of FLCN (Fig. 6D). Thus, FLCN promotes binding of Rab11A to its cargos, including PAT1 and TfR.

Previously, a structural study revealed a DENN-like domain at the C-terminus of FLCN. This led to the hypothesis that FLCN binds Rab GTPases and regulates vesicular trafficking processes (Nookala et al., 2012). In support of this view, three recent reports have shown that FLCN interacts through its DENN domain with Rab34, Rab7A and Rab35 (Starling et al., 2016; Laviolette et al., 2017; Zheng et al., 2017). In one study, Dodding's group discovered that the FLCN-Rab34 interaction regulates the intracellular movement of lysosomes in a nutrient-sensitive manner in HeLa cells (Starling et al., 2016). In another study, Iliopoulos's group reported that FLCN is a GAP of Rab7A and accelerates the lysosome-mediated degradation of the EGFR protein in several human tumor cell lines (Laviolette et al., 2017). Zheng et al. also found that FLCN promotes degradation of EGFR in both HeLa and HEK293T cells, although, in this case, FLCN was found to be a GEF of Rab35 (Zheng et al., 2017). It is not clear yet how FLCN promotes EGFR degradation in the lysosome through Rab7 and Rab35, because both proteins are known to promote the transport of membrane proteins to the late endosome/lysosome (Rab7) or to the fast recycling endosomes (Rab35). Theoretically, inactivation of Rab7A or activation of Rab35 (through FLCN) should inhibit transport of EGFP to the lysosome.

In our study here, we have demonstrated that FLCN is a so-far-unknown Rab11A-interacting protein and that this interaction also requires the FLCN DENN domain. To explore the molecular mechanism(s) of this interaction, we carried out in vitro GEF activity assays by using bacterium-expressed proteins – but no clear GEF activity of FLCN on Rab11A was detected. The same result has also been reported by Blundell's group in a previous study (Nookala et al., 2012). On the basis of these results, it seems that FLCN does not modify Rab11A activity directly. By using overexpressed PAT1 as a substrate, we found that FLCN promotes the loading of PAT1 on Rab11A and, consistent with this, FLCN preferentially binds Rab11A-CA (Fig. 6A). These results suggest that FLCN promotes recycling of PAT1 by mediating the interaction between Rab11A and PAT1. This mechanism would also explain the FLCN overexpression results. As shown in Fig. 5F, overexpression of FLCN displayed a stimulatory effect on mTORC1 in PAT1-overexpressing cells (lane 4) but not in wild-type cells (lane 7). We think that, when cargo (i.e. PAT1) is increased, more FLCN proteins are needed to promote transport. The same mechanism has also been observed for Rab34-regulated localization of lysosomes (Starling et al., 2016). In that study, Dodding's group reported that FLCN did not act as a Rab34 GEF but, instead, FLCN promoted the binding of active Rab34 to its effector RILP. We carried out the in vitro pull-down assay by using purified FLCN protein and overexpressed Rab11A, but no direct interaction was observed (our unpublished observation). We, therefore, suspect that FLCN is likely to require other effectors in order to interact with Rab11A.

Since FLCN can interact with several Rab proteins, FLCN might inhibit localization of PAT1 to the lysosome through several mechanisms. For example, FLCN has been found to be a GAP of Rab7A (Laviolette et al., 2017). Thus, FLCN might inhibit the transport of PAT1 to the lysosome through inactivating Rab7. In support of this, we found that overexpressed PAT1 has close relationships with Rab7A – as revealed by both immunostaining (Fig. 1A,B) and co-immunoprecipitation experiments (Fig. 6B). Interestingly, loss of FLCN did not affect the interaction between PAT1 and Rab7A (Fig. 6C). This result, indeed, supports the proposed mechanism that FLCN directly modulates Rab7A activity (Laviolette et al., 2017). In another possible mechanism, because FLCN is a GEF of Rab35 (Zheng et al., 2017) that, in turn, is a regulator of the fast recycling pathway, FLCN promotes the transport of PAT1 from early endosomes directly to the recycling endosomes. As a result, the amount of PAT1 destined to the late endosome/lysosome will be reduced.

How PAT1 regulates mTORC1 is still a mystery. For example, overexpression of PAT1 has been shown to either activate or inactivate mTORC1 in HEK293 cells. We have recently discovered a mechanism to explain these controversial observations when finding that lysosomal PAT1 is increased in response of starvation and decreased when nutrients are replenished (Wu et al., 2016; Zhao et al., 2018). After a short term of nutrient exposure (10 min), mTORC1 can be reactivated in wild-type cells but not in PAT1-overexpressing cells. However, by extending the nutrient stimulation, mTORC1 can be reactivated in PAT1-overexpressing cells (Zhao et al., 2018). In this current study, we mainly used the PAT1 overexpression system to display the significance of the interaction between Rab11A and FLCN. Consistent with a role of FLCN during the Rab11A-mediated recycling transport, we found that both FLCN and Rab11A can counteract the inhibitory effect of PAT1 overexpression on mTORC1. Although our data support a negative role of lysosomal PAT1 on mTORC1, we think the mTORC1 readout is determined by the relative level of lysosomal PAT1 (compared with that of total PAT1), because the PAT1 localizing elsewhere, for example, to the plasma membrane, may have different influences. This view is supported by the finding that overexpression of PAT1 does not necessarily inhibit mTORC1 under the normal culture conditions or following extended nutrient replenishment (Zhao et al., 2018). Notably, knockdown of PAT1 by using RNAi can also inhibit mTORC1. To explain this, we suspect that PAT1 knockdown decreases the amino acids input into the cytosol – from either the environment or the lysosomal lumen, or both. Eventually, this will lead to a reduction of the cytosolic amino acid pool, which is another important signal of mTORC1 (Zhao et al., 2018). However, we were unable to exclude another possibility, i.e. that PAT1 can positively regulate mTORC1 as s signal transducer (through Rags) – as suggested by Ögmundsdottir et al., 2012. To test this hypothesis, more careful analyses should be carried out. Besides, due to the lack of a good anti-PAT1 antibody, it is still unclear whether endogenous PAT1 exhibits dynamic lysosomal localizations in response to nutrient fluctuation and alterations of Rab11A and FLCN activities. Therefore, it should be noticed that the PAT1 overexpression system does not necessarily reflect the in vivo situation.

FLCN has been found to promote the amino acid signal transduction through Rags (Petit et al., 2013; Tsun et al., 2013). On the basis of our previous findings (Wu et al., 2016) and the results shown in this study, we propose that FLCN has another function by maintaining the amino acid signal level within the lysosome through amino acid transporters, such as PAT1. Interestingly, this transport function of FLCN seems to be conserved during evolution. For example, in yeast, the transport of the general amino acid permease Gap1p to the vacuole (lysosome) is inhibited by Lst7 (Roberg et al., 1997), an ortholog of FLCN. In addition, Gap1p can be transported between the cell surface and the lysosome, and this translocation is controlled by several components of the mTORC1 signaling pathway, including LST7 (FLCN), LST4 (FNIP1), LST8 (a component of mTORC1/2), Gtr1/2 (Rag GTPases) and SEC13 (a component of the GAP complex targeting the Rag GTPases) (Bar-Peled et al., 2013; Roberg et al., 1997). The relationship between trafficking of amino acid transporters and signal transduction processes deserves further investigations.

Antibodies against pS6K1 (1:1000; T389, #9205), S6K1 (1:1000; #9202), FLCN (1:1000; #3697) and Rab5 (1:1000; #3547) were obtained from Cell Signaling Technology (Danvers, MA); against Rab7 (1:200; B-3), TGN38 (1:200; B-6), Chmp5 (1:200; F-7), pan-cadherin (1:200; CH-19) and TfR (1:1000; 3B82A1) from Santa Cruz Biotechnology (Santa Cruz, CA); against EEA1 (1:200; 1FB), M6PR (1:200; 22d4) and LAMP1 (1:200; H4A3) from Developmental Studies Hybridoma Bank (Iowa City, IA); and against rabbit polyclonal HA (1:500; #9110) and Myc (1:300; #9106) from Abcam (Cambridge, UK). Mouse monoclonal anti-GFP (1:200; HT801) and anti-Myc (1:200, HT101) antibodies were from Transgen Biotech (Beijing, China); mouse monoclonal anti-β-actin (1:1000; KM9001T) antibody was from Sungene Biotech (Tianjin, China); mouse monoclonal anti-FLAG (M2) antibody was from Sigma-Aldrich (1:1000; St Louis, MO); mouse monoclonal anti-Rab11 antibody (1:400; #610656) was from BD Biosciences (Franklin Lakes, NJ); rabbit polyclonal anti-Rab11 (1:400; #71-5300) and anti-GFP (1:1000; #A11122) antibodies, and all fluorescent secondary antibodies were purchased from Life Technology (Carlsbad, CA).

The human FLCN cDNA was kindly provided by Dr L. S. Schmidt (NIH at Bethesda, MD). The cDNAs of PAT1, Rab5A, Rab7A, Rab11A and LAMP1 were derived from HEK293 or mouse 3T3 cells. For ectopic expression, the cDNAs were cloned into either pEGFP-C1 (with N-terminal EGFP tag) or pcDNA3.1(+). The resulting constructs were confirmed by sequencing.

Cells were normally cultured in complete medium: Dulbecco's modified Eagle's medium (DMEM) supplemented with 8% fetal bovine serum (FBS), 4 mM L-glutamine, 4500 mg/l glucose, and sodium pyruvate.

For starvation, cells were cultured in complete medium and allowed to grow up to ∼80% confluence. Then, cells were washed with PBS twice, followed by incubation with RPMI1640 medium lacking both amino acids and serum (US Biological, R8999-04A) for the indicated period of time.

Plasmids were introduced into cells using TurboFect transfection reagent (Life Technology). Stable cell lines were selected with G418 (Life Technology). RNAiMax diluted in OptiMEM (Life Technology) was used to deliver the siRNA 5′-GAUAAAGAGACCUCCAUUATT-3′ targeting FLCN or the siRNA 5′-AUGCUCCCAUCUUCAUCAATT-3′ targeting PAT1, both of which have been described previously (Petit et al., 2013; Wu et al., 2016), into the cells according to the manufacturer's instructions; shRNAs were cloned into pCD513B-U6 vector and co-transfected into the cells with the helper plasmids (GAG, REV and VSV-G). Purified and concentrated virus was used to infect cells, which were then selected by using their resistance against puromycin. The following two RNAi target sequences of Rab11A were used: #1, 5′-GTAACCTCCTGTCTCGATTTAC-3′ and #2, 5′-GGAGTAGAGTTTGCAACAAGA-3′.

We designed two different CRISPR sites on the coding region of the human FLCN genomic locus. The double-strand CRISPR fragments were cloned into pX330 plasmid. By using one of the targets 5′-GCGTCGAGTCCAGCAGCCCG-3′, we obtained two independent mutant clones in a HEK293T background. The mutations were confirmed by both genome sequencing and western blotting assays (Fig. S3).

Cells were fixed with 4% formaldehyde in PBS for 20 min, rinsed once with PBS and permeabilized by incubation with PBST for 5 min (0.1% Triton X-100 in PBS). Incubation with either primary or secondary antibody was performed at room temperature for 2 h or at 4°C overnight. Nuclei were counterstained with DAPI. Images were captured by using a confocal microscope (Nikon A1R-si). For colocalization assays, two-channel stacks of each image were analyzed by using the Nikon NIS-Element confocal microscope program and the Pearson's correlation coefficient was calculated. At least 30 cells from three repeated experiments were analyzed in each assay with one way ANOVA followed by Fisher's least significant difference test (Fisher's LSD) and using SPSS software (20.0, SPSS, Inc., Chicago, IL).

Lysosomes were isolated using LYSISO1 (Sigma-Aldrich). Briefly, the cells were collected by centrifugation at 600 g for 5 min (∼1.5×10⁸ cells in total). The following steps were carried out on ice: cells were suspended in 200 µl of extraction buffer and homogenized with five gentle strokes in a 2 ml dounce glass tissue grinder. After centrifugation at 1000 g for 10 min, the supernatant was collected, and the pellet was homogenized in four more rounds. The supernatants from all homogenates were then collected and pooled (∼1 ml) and centrifuged at 20,000 g for 20 min. The pellet was re-suspended in 50 µl of RIPA lysis buffer, yielding the lysosome fraction. Labeling of membrane proteins with biotin was performed as described elsewhere (Okimoto et al., 2004; Tsun et al., 2013); ∼1.5×10⁸ cells (3×90 mm dishes) were used.

To measure the PAT1 signal intensities on the lysosome and cell surface, we used the following two calculation methods (1) [lysosome (PAT1)/lysosome (LAMP1)]/[lysate (PAT1)/lysate (LAMP1)] and (2) [surface (PAT1)/surface (cadherin)]/[lysate (PAT1)/lysate (cadherin)].

Cells were washed once in ice-cold 1×PBS, harvested and lysed with the RIPA lysis buffer. The lysates were cleared by centrifugation at 12,000 g for 15 min at 4°C and then were mixed with SDS loading buffer. After that, samples were either boiled for 5 min or kept at 4°C overnight. The protein concentration was measured by BCA assay. For western blotting, the samples were separated by SDS-PAGE and transferred to PVDF membrane, blocked in 5% non-fat milk, and incubated with primary antibodies followed by incubation with the HRP-conjugated secondary antibodies. Immunoreactivity was detected by using ECL and chemiluminescence reagents (Bio-Rad, Hercules, CA). Western blot data were quantified by using the Bio-Rad Quantity One software. For quantification assays, we used Student's t-test with Statistical Package for the Social Sciences (SPSS) software (Version 20.0; SPSS, Inc., Chicago, IL). P<0.05 was considered significant. All error bars represent the s.e.m. of at least three repeated experiments.

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, pH 7.4), supplemented with protease inhibitor cocktail (Roche, Mannheim, Germany), followed by centrifugation at 12,000 g for 20 min at 4°C. Supernatants were added to either anti-HA polyclonal, anti-FLCN monoclonal, anti-Tfr monoclonal or anti-FLAG monoclonal antibody, and immunoprecipitated with Protein A/G Agarose beads (GE Healthcare, Pittsburgh, PA) at 4°C for 2 h. In the quantification assays, the following calculation method was used: [IP(Rabs)/IP(PAT1 or TfR)]/[lysate(Rabs)/ lysate (PAT1 or TfR)], n=3 repeated experiments.

We thank all the lab members for discussions about the manuscript.