Abstract
mTORC1 is a central signal hub that integrates multiple environmental cues, such as cellular stresses, energy levels, nutrients and certain amino acids, to modulate metabolic status and cellular responses. Recently, SLC38A9, a lysosomal amino acid transporter, has emerged as a sensor for luminal arginine levels and as an activator of mTOCRC1. The activation of mTORC1 occurs through the N-terminal domain of SLC38A9. Here, we determined the crystal structure of SLC38A9 and surprisingly found its N-terminal fragment inserted deep into the transporter, bound in the substrate binding pocket where normally arginine would bind. Compared with our recent arginine bound structure of SLC38A9, a significant conformational change of the N-terminal domain was observed. A ball-and-chain model is proposed for mTORC1 activation where in the starved state the N-terminal domain of SLC38A9 is buried deep in the transporter but in the fed state the N-terminal domain could be released becoming free to bind the Rag GTPase complex and to activate mTORC1. This work provides important new insights into how SLC38A9 senses the fed state and activates the mTORC1 pathways in response to dietary amino acids.
One Sentence Summary N-plug inserted state of SLC38A9 reveals mechanisms of mTORC1 activation and arginine-enhanced luminal amino acids efflux.
Main Text
The mechanistic target of rapamycin complex 1 (mTORC1) protein kinase acts as a central signaling hub to control cell growth and balance the products from anabolism and catabolism (1–3). Not surprisingly this pathway is dysregulated in many diseases (4, 5). Activation of the mTORCl is mediated by a variety of environmental cues such as nutrients, cellular stresses and energy levels (6, 7). Specifically, certain amino acids signal to mTORCl through two Ras-related guanosine triphosphatases (GTPases) (8, 9). When amino acids are abundant, the heterodimeric Rag GTPases adopt an active state and promote the recruitment of mTORCl to the lysosomal surface (10), which is now recognized as a key subcellular organelle involved in mTORCl regulation (11). Several essential amino acids in the lysosomal lumen including arginine, leucine and glutamine have been identified as effective activators of mTORCl (12–15). However, the molecular basis of the amino acids sensing mechanism has remained, by and large, elusive. Recently, SLC38A9, a low-affinity arginine transporter on lysosome vesicles, was identified as a direct sensor of lumen arginine levels for the mTORCl pathway (16–18). SLC38A9 also mediates the efflux of essential amino acids from lysosomes, such as leucine, in an arginine regulated manner (19), to drive cell growth by modulating cytosolic sensors (20, 21). Moreover, SLC38A9 senses the presence of luminal cholesterol and activates mTORCl independently of its arginine transport (22).
SLC38A9 is a transceptor. Studies showed that two parts of SLC38A9, its N-terminal domain and its transmembrane bundle, are responsible for two distinct functions. The bulk of SLC38A9 are ll alpha helices that pack against one another forming a transmembrane bundle that transports amino acids and function as an amino acid transporter (23). The N-terminus of SLC38A9, om the other hand, was previously shown to interact directly with the Rag-Regulator complex to activate mTORCl (16). Collectively, these results suggest that SLC38A9 is a “transceptor”, which is membrane protein that embodies the functions of both a transporter and a receptor (23–27).
Recently we solved the crystal structure of N-terminally truncated SLC38A9 from Danio rerio (ΔN-drSLC38A9) with arginine bound (23). The substrate arginine was observed deep in the transporter at a binding pocket consisting of residues from TM1a, TM3 and TM8 of SLC38A9. Because the N-terminally truncated form of SLC38A9 was used that initial study focused solely on the transporter function of SLC38A9 and resulting structures could not inform on the signaling function of SLC38A9.
Here we report a new crystal structure of drSLC38A9 with its N-terminus but without the substrate arginine. Surprisingly, we found that part of the N-terminus formed a beta hairpin that lodged itself deep in the transporter occupying the arginine binding site and blocking the transport path. These new results suggest that in the fed state the N-terminal domain would be released from within SLC38A9 and freed to interact with the Rag GTPase and activate mTORCl. We propose a ball-and-chain model to describe this mechanism of amino acid sensation and signaling by SLC38A9.
In the present study we used the antibody fragment llD3 to facilitate crystallization of SLC38A9 in the absence of substrate. Well-ordered crystals were diffracted to ~3.4 Å with high completeness and acceptable refinement statistics (Table. S1). Each asymmetric unit contained two copies of drSLC38A9-Fab complex, arranged in a propeller-like head-to-head fashion (fig. Sl). As with the recently determined structure (23), the transmembrane domain of drSLC38A9 was captured in the cytosol-open state and was folded into the same inverted topology repeats made of TMs 1-5 and TMs 6-10 with TMll wrapping around the transceptor (Fig. 1A and 1B). The two structures shared an overall similar fold with an r.m.s.d. of 0.8 Å. However, instead of an arginine molecule bound, this time an unexpected electron density was observed, which extended along the solvent accessible tunnel leading from the substrate binding site to the cytosolic side of SLC38A9 (fig. S2). The density was of sufficient quality to allow an unambiguous assignment of drSLC38A9 N-terminal section from Asp 75 to Leu 9l (fig. S3). This fragment formed a folded domain, resembling a beta hairpin, filling the entire path from the cytosolic side of SLC38A9 to the substrate binding site (Fig. 1C). Electrostatic potential analysis indicated that the transport pathway in SLC38A9 is generally positively charged, while the N-terminal fragment (referred to as the “N-plug” from this point on) is largely negatively charged (fig. S4), suggesting that the interaction is electrostatically driven. Deletion of the N-terminal domain of drSLC38A9 did not affect arginine transport (Fig. 1D), indicating that the N-plug does not directly participate in arginine translocation.
(A) Stereo view in the plane of the membrane. TMs are rainbow colored as blue to red from N- to C-terminus. The N-plug is shown in magenta. (B) Two-dimensional topology model of drSLC38A9, which is folded into a characteristic 2-fold LeuT-like pseudo-symmetry (five transmembrane-helix inverted-topology repeat). N-plug is marked by a filled pink triangle, next to the TM1a helix. (C) The N-plug blocks an otherwise cytosol-open state of drSLC38A9. (D) Truncation or mutation of the N-plug does not affect arginine transport. Shown here is the time course of [3H]-arginine uptake in proteoliposomes reconstituted with purified wild-type drSLC38A9 and its mutants. Error bars represent standard error of the mean (s.e.m.) of triplicate experiments.
We captured SLC38A9 in a new state that we term the “N-plug inserted state”. TMs 1, 5, 6 and 8 of SLC38A9 form a V-shaped cavity into which the N-plug inserts and is stabilized by several bonds (Fig. 2). At the tapered tip on the N-plug, Ser 80 and His 81 bound to the main-chain carbonyl oxygens of Thr 117, Met 118 and Met 119 in the unwound region of TM1 (Fig. 2A). His 81 further stabilizes the tip region of the N-plug through a hydrogen bond between its imidazole side chain and Thr 121 (Fig. 2A). Likewise, the main-chain carbonyl oxygen of Ile 84 is bound to Cys 363 on TM6 (Fig. 2B). At this juncture, the N-plug is jammed in between the two essential TMs 1 and 6 where it would probably prevent the transmembrane domain from transitioning to an alternate state for transport. At the N-terminus of the N-plug, the flanking residues are anchored against TM5 through a hydrogen bond formed between the main-chain carbonyl oxygens of Val 77 and the side-chain hydroxyl group of Thr 303 (Fig. 2C). At the C-terminus of the N-plug, the Tyr-Ser pairs involving Tyr 87, Tyr 448 and Ser 88, Ser 297 also stabilize the interaction by hydrogen bonds (Fig. 2D). All residues that participate in the inter-domain interactions are conserved across species as indicated in the sequence alignments (fig. S3), suggesting that this interaction is evolutionarily conserved and likely plays an important functional role. The beta-hairpin structure of the N-plug is also self-stabilized by several hydrogen bonds between Ser 80 and Glu 82, His76 and Tyr 85 which fasten the two ends of the N-plug together (Fig. 2E). Structural modeling by PEP-FOLD (28, 29) indicated that the beta hairpin motif would be converted to an alpha helical fragment should these residues were changed to alanines (fig. S5).
(A to D) The N-plug interacts with transmembrane bundle though multiple inter-domain hydrogen bonds. Residues that contribute interactions between the N-plug and TMs are highlighted in sticks, of which hydrogen bonds are depicted as dashed lines. (E) The folded conformation of N-plug as a beta hairpin is complementarily stabilized by several intra-domain interactions.
SLC38A9 has higher affinity toward leucine than arginine although the transport of leucine is largely facilitated by the presence of arginine (19). Uptake studies performed here with drSLC38A9 corroborate the previous findings using the human protein (Fig. 3B). Leucine uptake was significantly higher in the presence of supplemented arginine than without. An overlay of the N-plug bound structure and the arginine bound structure indicated that the same set of backbone atoms are used for binding the N-plug and the arginine molecule (Fig. 3A). This superposition suggests that in the presence of arginine the N-terminal plug may not occupy the binding site, but that in the absence of arginine it would be free to insert and bind. Is it possible, therefore, that in the presence of arginine the released N-terminal plug could play an important role in facilitating leucine transport?
(A) Superposition of substrate binding site of arginine-bound state (PDB ID: 6C08) with N-plug inserted state of drSLC38A9. TM1 of two different states are shown in gold and blue, respectively. Atoms of arginine molecule are depicted as spheres while the N-plug in magenta. (B) Adding 200μm unlabeled arginine boosts leucine transport by wild-type drSLC38A9 in proteoliposomes. (C and D) Either deletion or mutation of N-plug interferes the arginine enhancement of leucine transport. Without adding supplemented arginine, the mutant proteins show similar transport capacity for leucine regardless whether arginine was supplemented.
To examine whether the N-terminal plug plays an important role in facilitating leucine transport, two drSLC38A9 variants were generated. One has residues 1-96 of N-terminus deleted (called N-truncated). The other has 5 key residues mutated (P79A, S80A, H81A, E82A, and Y85A) in the N-plug (named 5A mutant) which would lead to a disrupted secondary structure of the N-plug (fig. S5). As observed in the uptake study, both variants could transport arginine like the wild-type drSLC38A9 even with the dramatic structural changes at the N-plug (Fig. 1D). From the results of leucine uptake by wild-type drSLC38A9, the arginine-enhanced transport of leucine is reflected as increased uptake of [3H]-leucine when the buffer was supplemented with arginine. This characteristic of arginine-enhanced leucine transport was lost when the N-plug was eliminated or its structure altered by mutation (Fig. 3C and 3D). Only the SLC38A9 with intact N-terminal plug in its native beta hairpin like structure showed the characteristic enhanced leucine uptake in the presence of supplemented arginine (Fig. 3B).
It is known that the N-terminal domain of SLC38A9 can bind to, and activate, the Rag GTPases complex (16). Moreover, it was shown that the N-terminal fragment of human SLC38A9 (hSLC38A9) was sufficient and required to bind the Ragulator-Rag GTPases complex (16). The binding of Rag GTPases and the human SLC38A9 involves the 85PDH87 motif (17), Pro 85 and Pro 90 (16), corresponding to a conserved region on the N-plug in drSLC38A9 (fig. S3). To probe the N-plug interaction with the Rag GTPases in drSLC38A9, we co-purified zebrafish Rag GTPases complex (drRagA and drRagC) with two N-terminal fragments of drSLC38A9 by size-exclusion chromatography. The first fragment (residues 1-96) contained the N terminus in its entirety (called drSLC38A9-N.1) while in the second fragment (residues 1-70) the N-plug was deleted (called drSLC38A9-N.2). Fractions from size exclusion chromatography were collected and analyzed by SDS-PAGE (fig. S6). Contrary to fragment drSLC38A9-N.1 which maintains the N-terminal domain in its entirety, the N-plug deleted construct, drSLC38A9-N.2 did not associate with Rag GTPases complex (fig. S6). These results clearly demonstrated that the interaction between the zebrafish SLC38A9 N-terminus and the zebrafish Rag-GTPase recapitulate the experiments reported previously using the human proteins (17, 16): the same region of the N-plug of drSLC38A9 is essential for binding with Rag GTPases complex.
In considering our recently determined structure of SLC38A9 with arginine bound, and the current structure without arginine but with the N-plug inserted into the arginine binding site, we now captured SLC38A9 is at least two distinct conformations of the N-terminus. The first is when the N-plug is bound snuggly in the arginine binding site (in the absence of arginine, starved state) and the second where the N-terminal plug was released and the substrate binding site was occupied by arginine (in the presence of arginine, fed state). The vestibule into which the N-terminal plug inserts measures ~20Å in diameter. A recently determined crystal structure of Rag GTPases-Ragulator (30–32) indicated that the GTPases-regulator is far too large to fit inside the vestibule of SLC38A9 suggesting that the N-plug must exit the tranceptor for binding the Rag GTPase. Together, these data suggest a mechanism by which SLC38A9 can act as a receptor to signal the activation of Rag GTPase and therefore of mTORC1 in the presence of arginine.
We thus propose a ball-and-chain model (Fig. 4). At lower arginine concentrations, two conformational states could be at an equilibrium where the N-terminal plug is equally inserted or released from the arginine binding site of SLC38A9. When the equilibrium shifts to the right in the fed state with elevated arginine levels, an arginine molecule will occupy the binding site of SLC38A9 for transport and the N-terminal plug would remain released as long as arginine flows. As a result, the N-terminal plug becomes available for binding to the Rag GTPases complex which in turn could activate the mTORC1. Moreover, the release of the N-terminal plug from the helical bundle of SLC38A9 will also facilitate the efflux of other essential amino acids, which simultaneously increases the cytosolic concentration of amino acids and synergistically activates mTORC1 through other cytosolic sensors.
At low luminal arginine, N-plug domain naturally samples both the inserted and released state as an equilibrium. As the concentration of luminal arginine increase in the fed state, arginine molecules enter the substrate binding site and the N-plug remains in the released state while arginine transport takes place. In the released state the N-plug could both trigger the efflux of other luminal amino acids such as leucine and interact with the Rag-GTPases to activate the mTORC1 signaling pathway.
While the present study provides the first line of evidence on the function of SLC38A9 as a transporter and sensor for amino acids it remains unclear how the N-terminal domain associates with the Rag GTPase complex. Likewise, it is still not known what the open-to-lumen conformation of the transporter looks like and whether the N-plug remains inserted or not. Future studies must delve into these important open questions but with the newly proposed ball-and-chain model for signaling new biochemical assays can be designed and tested.
Acknowledgements
We thank D. Cawley for development and production of monoclonal antibodies; K. Rajashankar and the staffs in Northeastern Collaborative Access Team (NE-CAT) for their support with X-ray data collection; J. Hattne for discussions over structural determination; L. Shao for careful review and scientific feedback on the manuscript. This work is based upon research conducted at the NE-CAT beamlines, which are funded by the National Institute of General Medical Sciences from the National Institutes of Health (P41 GMl03403). The Pilatus 6M detector on 24-ID-C beam line is funded by a NIH-ORIP HEI grant (S10 RR029205). This research used resources of the Advanced Photon Source, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CHll357. Research in the Gonen laboratory is funded by the Howard Hughes Medical Institute. The coordinates and the structure factors have been deposited in the Protein Data Bank (PDB) under accession codes 6DCI.