Lysosomes are membrane-bound organelles best known for their capacity to degrade macromolecules and recycle their constituent metabolites and for their dysfunction in a group of rare metabolic disorders known as lysosomal storage diseases (
1,
2). Lysosomes also participate in signal transduction (
3), particularly in nutrient sensing by the mechanistic target of rapamycin complex 1 (mTORC1) pathway (
4,
5), and are often deregulated in common diseases such as cancer (
6). Given the critical roles of lysosomes in producing and sensing many metabolites, a better understanding of lysosomal function requires uncovering its metabolite content and its regulation in diverse cell states.
Traditional techniques for purifying lysosomes, such as density-based centrifugation, are too slow to preserve what is likely a labile lysosomal metabolome (“lysobolome”). To overcome this issue, we used insights from a recently reported method for the rapid isolation of mitochondria (
7) to develop an analogous approach for lysosomes. Our lysosome immunoprecipitation (LysoIP) method uses antibody to human influenza virus hemagglutinin (HA) conjugated to magnetic beads to immunopurify lysosomes from human embryonic kidney (HEK) 293T cells expressing transmembrane protein 192 (TMEM192) fused to three tandem HA epitopes (HA-Lyso cells) (
Fig. 1, A and B). TMEM192 is a transmembrane protein (
8) that we find retains its lysosomal localization upon overexpression better than other such proteins, such as lysosomal-associated membrane protein 1 (LAMP1). Starting with live cells, it takes ~10 min to isolate lysosomes that are highly pure and intact, as judged by the absence of markers for other cellular compartments (
Fig. 1C), retention of cathepsin D activity (
Fig. 1D), and capacity to take up radiolabeled arginine in vitro (
Fig. 1E). Moreover, tracking of either a lysosomal membrane protein (LAMP2), a luminal protein (cathepsin D), or a small molecule (LysoTracker Red), yielded the same value for the fraction of total cellular lysosomes purified (
Fig. 1F), indicating that the lysosomes do not leak soluble contents during the purification. Importantly, the LysoIP method uses buffers compatible with subsequent analyses of the lysosomal metabolome by liquid chromatography and mass spectrometry (LC-MS).
Because the metabolite content of human lysosomes is not established, we used LC-MS to determine the relative abundances of ~150 polar small molecules in lysosomes versus control anti-HA immunoprecipitates from cells stably expressing Flag-tagged TMEM192 (Control-Lyso cells) (fig. S1A and table S1). Of these, 57 were twice as abundant in the isolated lysosomes and thus deemed lysosomal metabolites (fig. S1A and table S1). The lysosomes did not contain metabolites characteristic of other compartments, such as the cytosolic glycolytic intermediates fructose 1,6-bisphosphate and lactate or the mitochondrially enriched coenzyme A (
7) (fig. S1B).
We quantified the concentrations of the 57 metabolites in lysosomal and whole-cell samples using standard curves for each and the volumes of lysosomes and intact cells (see the supplementary materials). Lysosomal metabolite concentrations correlated highly across biological replicates (
r2 = 0.95) (
Fig. 1G) and even with those obtained using the less preferable LAMP1-RFP-3xHA as the lysosomal antigen tag (
r2 = 0.95) (fig. S1C), mitigating concerns that expression of TMEM192, whose function is unknown, might have effects on the lysosomal metabolome. In the proliferating cells used in these experiments, the concentrations of metabolites tended to be, with a few exceptions, lower in lysosomes than in whole cells (
Fig. 1, H and I, and table S2). Two molecules previously predicted to be stored in lysosomes, cystine (the oxidized dimeric form of cysteine) and glucuronic acid (
9,
10), were indeed enriched in lysosomes, with concentrations 28- and 5.5-fold greater than those of whole cells, respectively (
Fig. 1, H and I, and table S2). All nucleosides (guanosine, adenine, cytidine, uridine, and inosine) were lysosomally enriched (9- to 25-fold), consistent with the lysosome also being a depot for these metabolites, at least in HEK-293T cells (
Fig. 1, H and I). The lysosomal concentrations of proteinogenic amino acids varied widely and did not correlate well with those in whole cells (
Fig. 1I), suggesting that although some lysosomal amino acids are in equilibrium with the rest of the cell, others are either sequestered in a different compartment or undergo preferred transport out of the lysosome and thus show higher concentrations in the whole-cell samples. Lysosomes also contained metabolites that are not thought to result from the degradation of macromolecules and thus are likely transported into lysosomes (
Fig. 1I). These include nonproteinogenic amino acids, such as beta-alanine (20 μM), taurine (11 μM), and hypotaurine (12 μM); cofactors and vitamins, such as choline (7 μM) and phosphocholine (94 μM); creatine (274 μM) and phosphocreatine (111 μM); and multiple species of carnitines (
Fig. 1I). The metabolomic landscape of the human lysosome is consistent with its role as a recycling center but also indicates that the transport of metabolites into lysosomes may influence lysosomal biology more than is widely recognized.
The multicomponent vacuolar H
+–adenosine triphosphatase (V-ATPase) maintains the lysosomal lumen at a pH of ~4.5 (
11), which is thought to be required for the optimal activity of lysosomal hydrolases and to set up a proton gradient with the cytosol that provides energy for transporters to move metabolites across the lysosomal membrane. To directly ask how loss of the acidic pH affects lysosomal metabolites, we profiled lysosomes from cells acutely treated with the V-ATPase inhibitors bafilomycin A1 (BafA1) or concanamycin A (ConA) (
12,
13) at concentrations that do not inhibit mTORC1 signaling (
4) (fig. S2A). Although neither had a major impact on the whole-cell metabolome, both caused large changes in the metabolome of the lysosome (
Fig. 2A and fig. S2B). This emphasizes the value of LysoIP for studying an organelle that in HEK-293T cells occupies only 2 to 3% of the total cell volume. V-ATPase inhibition caused the accumulation of many metabolites in lysosomes (
Fig. 2B and table S3), and only the concentration of cystine dropped significantly (
P ≤ 0.01 in either treatment; two-tailed
t test) (fig. S2C), consistent with in vitro work showing that the lysosomal entry of cysteine requires the pH gradient (
14). Although all the nonessential amino acids accumulated in lysosomes upon V-ATPase inhibition (
Fig. 2C)—with proline, alanine, and glycine being the most affected (
Fig. 2C)—seven of the nine essential amino acids did not, with histidine and threonine being the exceptions (
Fig. 2D). Given that lysosomes harbor several well-characterized proton-dependent amino acid transporters, such as lysosomal amino acid transporter 1 (LYAAT-1) (
15), lysosomal accumulation of the nonessential amino acids caused by V-ATPase inhibition may result from their decreased efflux. We therefore undertook pulse-chase experiments using
15N-labeled alanine, a representative pH-dependent amino acid, and isoleucine, a non–pH-dependent one (
Fig. 2, C to F). In live cells, both entered lysosomes, with isoleucine doing so more rapidly than alanine (
Fig. 2E). ConA treatment slowed the efflux of alanine, but not that of isoleucine, from lysosomes (
Fig. 2F), consistent with proton-dependent transporters mediating the efflux of this and other nonessential amino acids from lysosomes. Furthermore, the failure of V-ATPase inhibition to affect the lysosomal levels of most essential amino acids, particularly the nonpolar ones, raises the question of what, if anything, regulates their abundance.
To investigate this, we examined other conditions that might affect lysosomal metabolites, including nutrient starvation. We starved cells of all amino acids for 60 min and measured the concentrations of amino acids in lysosomes and in whole cells. Concentrations of most nonessential amino acids did not drop in either sample, consistent with the capacity of cells to synthesize them. In contrast, the concentrations of most essential amino acids, including those that were insensitive to V-ATPase inhibition, diminished in the whole-cell samples, but most showed little, if any, change in lysosomes (
Fig. 3A and table S4). Thus, amino acid starvation appears to inhibit the lysosomal egress of many essential amino acids.
Given that a major consequence of amino acid starvation is inhibition of mTORC1 (
Fig. 3B) (
16,
17), we asked whether mTORC1 regulates the abundance of amino acids in lysosomes. Consistent with this possibility, in cells that lack functional GATOR1 [DEPDC5 knockout (KO) cells] and thus have amino acid–insensitive mTORC1 signaling (
18), amino acid starvation did decrease the concentrations of lysosomal amino acids (
Fig. 3A and
3B). Moreover, inhibition of the kinase activity of mTOR with Torin1 (
19) increased the lysosomal concentrations of six of the seven V-ATPase–insensitive amino acids (leucine, phenylalanine, isoleucine, tryptophan, methionine, and valine) and of tyrosine, while having small effects on most other amino acids, including histidine and serine, as well as many additional metabolites (
Fig. 3, C and D; fig. S3A; and table S5). Torin1 also increased the lysosomal concentrations of nucleosides, although in this case the effect was also seen in whole cells (
Fig. 3E). Of the seven amino acids most strongly affected by Torin1, all are nonpolar and essential, with the exception of tyrosine, which is generated from the essential amino acid phenylalanine (
20). Importantly, other chemically distinct ATP-competitive inhibitors of mTOR, including AZD8055 and WYE-132 (
21,
22), also increased the concentration of these seven amino acids (fig. S3B), and mTOR inhibition had similar effects across multiple cell lines (fig. S3C). Although Torin1, AZD8055, and WYE-132 inhibit both mTORC1 and mTOR complex 2 (mTORC2), inhibition of only mTORC1 with the allosteric inhibitor rapamycin or with lower concentrations of Torin1 also increased the concentration of these amino acids, albeit to smaller extents (fig. S3, B and D). mTORC1 is essential for cell survival, but it is possible to generate cells lacking rictor, a critical mTORC2-specific component needed for phosphorylation of the protein kinase Akt (
23). Loss of rictor did not increase lysosomal amino acid concentrations, and, importantly, Torin1 increased the abundance of the seven amino acids in lysosomes even more in cells lacking rictor than in wild-type cells (fig. S3E). Thus, mTORC1 appears to mediate the effects of mTOR inhibition on lysosomal amino acids.
Because mTORC1 inhibits autophagy (
24–
28), a potential explanation for the effects of Torin1 is that it activates autophagic flux to such a degree that the production of metabolites by lysosomal macromolecular degradation exceeds the capacity of lysosomes to export them. We tested this possibility in cells lacking
ATG7 (fig. S3F), which encodes a key component of the autophagy machinery (
29). For most metabolites, loss of autophagy almost completely eliminated the Torin1-induced increases in their lysosomal concentrations, but it had only minor effects on those of the seven strongly affected amino acids (leucine, tyrosine, phenylalanine, isoleucine, tryptophan, methionine, and valine) (
Fig. 3, D and E; fig. S3G; and table S5). mTOR inhibition also activates the proteasome (
30), but bortezomib, a proteasomal inhibitor (
31), had no effect on the capacity of Torin1 to increase abundance of lysosomal amino acids (
Fig. 3F and fig. S3H). Lastly, mTORC1 inhibition suppresses mRNA translation (
32), but the protein synthesis inhibitor cycloheximide did not mimic the effects of the mTOR inhibitors on lysosomal amino acid levels, although it did mildly increase whole-cell and lysosomal pools of the mTOR-regulated amino acids (fig. S3L). Thus, mTORC1 regulates the lysosomal concentrations of a largely distinct set of amino acids from those affected by the V-ATPase (
Fig. 3G) through a mechanism that does not involve autophagy, the proteasome, or protein synthesis (
Fig. 3G).
Given that mTORC1 does not affect the seven amino acids through established downstream processes, we considered the possibility that it controls their flux across the lysosomal membrane. We used
15N-labeled amino acids to monitor the transport of four of the mTOR-regulated amino acids (leucine, tyrosine, phenylalanine, and isoleucine) and a control amino acid (serine) into lysosomes in live cells. When added to the culture media, the labeled amino acids rapidly exchanged with the
14N-containing amino acids already in lysosomes (
Fig. 2E and
Fig. 4A). In cells treated with or without cycloheximide, Torin1 caused lysosomal accumulation of
15N-labeled leucine, tyrosine, phenylalanine, and isoleucine, but not serine (fig. S4, A and B, and
Fig. 4B), demonstrating that mTOR regulates the movement of free amino acids across the lysosomal membrane independently of their incorporation into protein. Pulse-chase experiments revealed that mTOR inhibition slows the efflux of leucine, tyrosine, phenylalanine, and isoleucine, but not that of serine, from lysosomes but not from whole cells (
Fig. 4C and fig. S4C).
We recently identified the multipass protein SLC38A9 as a lysosomal effluxer of many essential nonpolar amino acids (
33). Its loss led to the accumulation in lysosomes of the seven amino acids most affected by mTOR inhibition and greatly reduced their efflux from lysosomes (
Fig. 4D and fig. S4D). In cells lacking SLC38A9, Torin1 did not boost the already high lysosomal concentrations of the seven amino acids (
Fig. 4D). Thus, mTOR inhibition and loss of SLC38A9 do not have additive effects on lysosomal amino acids, suggesting that mTORC1 regulates the lysosomal abundance of amino acids through a mechanism that involves SLC38A9. Loss of SLC38A9 greatly impaired the capacity of cells to survive amino acid starvation and of the GCN2 pathway, which senses uncharged tRNAs, to return to baseline activity levels upon prolonged starvation (
Fig. 4, E and F). Thus, the efflux of the mTORC1-regulated essential amino acids from lysosomes is important for the cellular response to starvation.
Our data show that mTORC1 has a previously unknown role in promoting the efflux of essential amino acids from the lysosome into the cytosol (
Fig. 4G). mTORC1 inhibition leads to the sequestration of these amino acids in the lysosome by slowing their movement across the lysosomal membrane, in effect converting it into a storage compartment for them. We speculate that this function of mTORC1 is important for preventing the inappropriate use of essential amino acids during amino acid starvation, a state in which lysosomal and proteasomal protein degradation are thought to be a major source of amino acids (
30,
34,
35). One can imagine the following scenario: Early in a starvation period, mTORC1 becomes profoundly inhibited, which, by suppressing SLC38A9 and perhaps other transporters, prevents the exit from lysosomes of essential amino acids. Over time, as proteolysis partially restores amino acid levels, mTORC1 becomes sufficiently reactivated so that essential amino acids are released into the cytosol at a faster rate to be used to execute the ongoing gene expression program that cells induce to adapt to starvation (
3,
34). In this regard, it is interesting that Torin1, which completely inhibits mTORC1, causes a greater accumulation of amino acids in lysosomes than rapamycin, which only partially inhibits it (
19,
36–
38). This pattern is also true for several other processes downstream of mTORC1, such as autophagy and protein synthesis (
19,
36–
38), and may indicate that the mechanisms through which mTORC1 regulates lysosomal amino acid efflux, such as through SLC38A9, are also sensitive to the exact amount of mTORC1 activity, allowing for distinct outcomes at different levels. How mTORC1 affects SLC38A9 function is unknown, and it may do so indirectly or directly. Activated mTORC1 resides on the lysosomal surface (
5,
39), so it has the correct localization to control SLC38A9 or its regulators. The fact that SLC38A9 also signals arginine levels to mTORC1 (
40–
42) suggests that SLC38A9 is part of a sophisticated system for coordinating mTORC1 activity and lysosomal amino acid efflux with the concentrations of cytosolic and lysosomal amino acids. Our findings provide an example of the utility of LysoIP for uncovering a new function for lysosomes—the sequestering of essential amino acids upon mTORC1 inhibition. The method that we described may be useful for studying the emerging roles of lysosomes and for probing the metabolic state of the lysosome in the various diseases in which it is implicated.
Acknowledgments
We thank all members of the Sabatini laboratory for helpful insights, particularly J. R. Cantor, W. W. Chen, and J. M. Orozco; C.A. Lewis and T. Kunchok from the Whitehead Institute Metabolite Profiling Core Facility; and N. S. Gray (Dana-Farber Cancer Institute) for Torin1. This work was supported by grants from NIH (R01 CA103866, R01 CA129105, and R37 AI47389) and Department of Defense (W81XWH-15-1-0230) to D.M.S., from Department of Defense (W81XWH-15-1-0337) to E.F., and from the European Molecular Biology Organization Long-Term Fellowship to M.A.-R.; a Saudi Aramco Ibn Khaldun Fellowship for Saudi Women to N.N.L.; and fellowship support from the National Defense Science and Engineering Graduate Fellowship (NDSEG) Program to G.A.W. D.M.S. is an investigator of the Howard Hughes Medical Institute.