Crystal structure of the human lysosomal mTORC1 scaffold complex and its impact on signaling

See allHide authors and affiliations

Science  20 Oct 2017:
Vol. 358, Issue 6361, pp. 377-381
DOI: 10.1126/science.aao1583

Structure of human mTORC1 components

The mTORC1 (mechanistic target of rapamycin complex 1) complex garners much attention as a signaling hub that coordinates input from growth-factor receptors and nutrient availability with metabolism and cell growth and proliferation. de Araujo et al. report the crystal structure of the LAMTOR (or “Ragulator”) complex that helps assemble mTORC1 at the lysosomal membrane for activation. The structure and functional studies reveal how LAMTOR1 wraps around the other subunits to hold them in place and interacts with the Rag guanosine triphosphatases in the complex.

Science, this issue p. 377


The LAMTOR [late endosomal and lysosomal adaptor and MAPK (mitogen-activated protein kinase) and mTOR (mechanistic target of rapamycin) activator] complex, also known as “Ragulator,” controls the activity of mTOR complex 1 (mTORC1) on the lysosome. The crystal structure of LAMTOR consists of two roadblock/LC7 domain–folded heterodimers wrapped and apparently held together by LAMTOR1, which assembles the complex on lysosomes. In addition, the Rag guanosine triphosphatases (GTPases) associated with the pentamer through their carboxyl-terminal domains, predefining the orientation for interaction with mTORC1. In vitro reconstitution and experiments with site-directed mutagenesis defined the physiological importance of LAMTOR1 in assembling the remaining components to ensure fidelity of mTORC1 signaling. Functional data validated the effect of two short LAMTOR1 amino acid regions in recruitment and stabilization of the Rag GTPases.

The LAMTOR [late endosomal and lysosomal adaptor and MAPK (mitogen-activated protein kinase) and mTOR (mechanistic target of rapamycin) activator] complex is a pentameric complex [LAMTOR1 to LAMTOR5; also known as p18, p14, MP1 (MEK binding partner 1), C7orf59, and HBXIP (hepatitis B virus X-interacting protein)] on late endosomes and lysosomes (1). LAMTOR2 and LAMTOR3 scaffold MEK1 (MAPK kinase 1) and ERK1 or ERK2 (Ras-dependent extracellular signal–regulated kinase) to lysosomes, providing spatial and temporal specificity in the MAPK pathway (2). Furthermore, LAMTOR anchors the Rag guanosine triphosphatases (GTPases) to the lysosomal surface (3) and may serve as a guanine nucleotide exchange factor (GEF) toward the Rag proteins (i.e., the “Rags”), contributing to mTOR complex 1 (mTORC1) activation. Hence, it was also named “Ragulator” (1). Lysosomal mTORC1 activation depends on concomitant inputs controlled by Rheb and Rag GTPases (4). Nutrient availability leads to GTP loading of RagA or RagB, which then recruit Raptor to target mTORC1 to lysosomes (5, 6). LAMTOR controls many cellular processes: embryonic development; tissue homeostasis; cell cycle progression; receptor trafficking; focal adhesion turnover; migration; maturation and biogenesis of lysosomes; and growth factor and amino acid signaling (711). To explore how these functions are executed and coordinated, we solved the crystal structures of the pentameric LAMTOR and of its complex with the C-terminal domains (CTDs) of the Rags at resolutions of 2.3 and 2.9 Å, respectively.

We crystallized the pentameric LAMTOR complex (Fig. 1, figs. S1 to S3, and supplementary text). The crystal structure (Fig. 1A and table S1) comprises two heterodimers (the first containing LAMTOR2 and LAMTOR3, the second encompassing LAMTOR4 and LAMTOR5) surrounded by LAMTOR1. Despite lacking the C-terminal helix, LAMTOR4 adopts a roadblock fold and interacts with LAMTOR5 through β-sheet augmentation, displaying the same dimerization mode as LAMTOR2 and LAMTOR3 (1214). The tilted arrangement between two neighboring proteins of each heterodimer is similar to that of the related trimeric TORC1-recruiting Ego complex in yeast (fig. S4), indicating structural conservation of the roadblock fold within eukaryotes (15, 16). Although interactions within the heterodimers are prominent and numerous, few contacts could be detected between heterodimers. Thus, the heterotetramer appears to be held together largely by its surrounding interactions with LAMTOR1.

Fig. 1 The crystal structure of the pentameric LAMTOR complex.

(A) The overall structure consists of two roadblock heterodimers formed by LAMTOR2 (purple) and LAMTOR3 (blue) or LAMTOR4 (orange) and LAMTOR5 (yellow), respectively, surrounded by LAMTOR1 (red). N and C termini are labeled. (B) The side view of the complex shows the complementation of helices absent in the roadblock domains of LAMTOR4 and LAMTOR5 by helices of LAMTOR1.

Although the version of LAMTOR1 that was crystallized included residues 21 to 161, electron density was only resolved for residues 80 to 149. We used electrophoresis and mass spectrometry analysis to exclude major proteolytic degradation during the preparation and crystallization process (fig. S5). Thus, we could not see residues 21 to 80, probably because of flexibility in the crystal. LAMTOR1 discontinued at the N and C termini, leaving an apparently accessible space on the solvent-exposed side of the LAMTOR2 and LAMTOR3 heterodimer. The α2 helices of both LAMTOR2 and LAMTOR3 are thought to mediate interaction with other proteins (12, 13). The pentameric structure showed that the absent C-terminal helices of LAMTOR4 and LAMTOR5 were spatially complemented by helices α4 and α5 of LAMTOR1 (Fig. 1B), closely resembling the interaction mode observed for Ego2 and Ego1 (15) (fig. S4). The structurally resolved region of LAMTOR1 is largely helical and stabilizes the complex by forming a U-shaped belt around the two heterodimers, providing additional contacts that may contribute to increased affinity between the different subunits (Fig. 2). We identified three areas of contact between LAMTOR1 and LAMTOR3, LAMTOR4, or LAMTOR5 (Fig. 2, figs. S6 and S7, and supplementary text).

Fig. 2 Interactions mediated by LAMTOR1 within the LAMTOR complex.

Chains are colored as in Fig. 1 [LAMTOR1 (red), LAMTOR2 (purple), LAMTOR3 (blue), LAMTOR4 (orange), and LAMTOR5 (yellow)]; N and C termini are labeled. Amino acid residues are labeled using single-letter codes [A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr]. (A) Overview of the pentameric LAMTOR complex showing Cα of LAMTOR1 residues mutated in this study as spheres. Regions encircled in gray are shown in the close-up views in (C) to (F). (B) Cartoon representation of the LAMTOR complex. Interacting residues (distance cutoff: 3.6 Å) are connected by black lines. The flexible N- and C-terminal regions of LAMTOR1 are symbolized by red dashed lines. (C to F) Close-up views of LAMTOR1 interacting with LAMTOR5 (C), LAMTOR4 (D and F), and LAMTOR3 (E). Polar contacts are indicated by dashed lines.

To investigate the physiological relevance of the extended LAMTOR1 conformation in the pentameric complex, we generated truncations and alanine mutants to specifically abolish the interaction to each of the remaining LAMTOR subunits (Fig. 3A). The choice of mutated residues was based on structurally identified contacts (Fig. 2 and fig. S1). For instance, in LAMTOR1_LT2*.SH, we mutated V148 and D149 of LAMTOR1 to alanine. LAMTOR1 variants were tagged with a streptavidin-binding peptide and a hemagglutinin (HA) epitope, hereafter designated as an SH tag. LAMTOR1WT.SH and mutants, as well as SH.GFP (GFP, green fluorescent protein) as a control, were expressed in modified human embryonic kidney (HEK) 293 Flp-In T-REx cells (hereafter HEK293) (Fig. 3B). LAMTOR1WT.SH coimmunoprecipitated all other LAMTOR proteins, Rags, SLC38A9 (the lysosomal amino acid transporter), and components of the BORC (for BLOC1–related complex): namely, Snapin, C10orf32, and C17orf59 (1721). The majority of mutants neither formed pentameric complexes nor interacted with their known partners. LAMTOR1_1-147.SH associated weakly with the remaining LAMTOR subunits. LAMTOR1_LT2*.SH mutations interfered with complex stability but permitted association with the Rags. In contrast, LAMTOR1_1-150.SH associated with the remaining LAMTOR and BORC components but did not recruit either Rags or SLC38A9. Thus, the belt-like function of LAMTOR1 appears to be fundamental for LAMTOR stability, and the C terminus (residues 150 to 161) of LAMTOR1 is essential for association with the Rags.

Fig. 3 Requirement of the C terminus of LAMTOR1 for recruitment of Rags to endomembranes and amino acid–dependent activation of mTORC1.

(A) Schematic representation of LAMTOR1 mutants with annotated positions at which LAMTOR1 was truncated. Asterisks indicate mutants in which the structurally identified contact sites between LAMTOR1 and each of the remaining LAMTOR proteins were exchanged to alanine. LT, LAMTOR. (B) Expression of SH.GFP, the LAMTOR1.SH wild type, and LAMTOR1.SH mutants was induced by incubating the corresponding HEK293 cell lines with tetracycline. Streptavidin (STREP) immunoprecipitates were analyzed by immunoblotting. SH, StrepII-HA-tag; GFP, green fluorescent protein; HA, hemagglutinin. n = 2 independent biological experiments. (C) Indirect immunofluorescence images of HeLa wild type (WT), LAMTOR1HM, and rescue cell lines kept under normal growth conditions. Merged and single-channel images of endogenous lysosomal marker LAMP1 (red) and endogenous RagC (green) are indicated. Representative cells are shown. Scale bars, 10 μm. (D) HeLa wild type, LAMTOR1HM, and rescue cell lines were starved for amino acids for 5 hours and then stimulated with amino acids for 10 min. Obtained lysates were analyzed by immunoblotting. n = 2. (E) Expression of SH.GFP, LAMTOR1.SH wild type, LAMTOR1 truncation mutants, and mutants in which the KEE (residues 151 to 153) or LVV (residues 154 to 156) of LAMTOR1 were mutated to alanines was induced by incubating the corresponding HEK293 stable cell lines with tetracycline. STREP immunoprecipitates were analyzed by immunoblotting. n = 2.

We then generated a LAMTOR1 hypomorph cell line (hereafter LAMTOR1HM) (fig. S8 and supplementary text) and transiently transfected it with LAMTOR1WT.SH and the same LAMTOR1 mutants previously tested in HEK293 cells. They all colocalized with endogenous LAMP1 (lysosomal-associated membrane protein 1) (fig. S8C), indicating that differences observed in the interactome analysis were not due to mislocalization. To functionally address the role of the C terminus of LAMTOR1 in anchoring the Rags to the lysosomal surface (3), we established LAMTOR1HM cell lines stably expressing LAMTOR1WT.HA, LAMTOR1_1-106.HA, or LAMTOR1_1-150.HA (fig. S9A). LAMTOR1WT.HA restored the interaction with the Rags and SLC38A9 (fig. S9B), whereas LAMTOR1_1-150.HA did not. In wild-type cells, RagC was recruited to LAMP1 structures (Fig. 3C). LAMTOR1 deletion resulted in impaired recruitment of RagC that could be restored by expression of LAMTOR1WT.HA but not by LAMTOR1_1-106.HA or LAMTOR1_1-150.HA (Fig. 3C). Complementarily, we observed colocalization of endogenous RagC with LAMTOR1WT.HA but not with LAMTOR1_1-106.HA or LAMTOR1_1-150.HA (fig. S9, C and D). Next, we tested the cell lines for their signaling properties (Fig. 3D). Control cells responded to withdrawal of amino acids with low phosphorylation of both p70S6K (ribosomal protein S6 kinase beta-1) and downstream S6. Reexposure of control cells to all essential amino acids and glutamine readily increased phosphorylation of p70S6K and S6, indicating mTORC1 activation. LAMTOR1HM cells showed impaired p70S6K and S6 phosphorylation when reexposed to amino acids, and this defect was rescued by expression of LAMTOR1WT.HA but not by LAMTOR1_1-150.HA. Although defective in mTORC1 signaling, LAMTOR1_1-150.HA cells contained stable pentameric LAMTOR complexes (Fig. 3D). Thus, the C terminus of LAMTOR1 appears to be required to recruit the Rags, thereby controlling amino acid–dependent activation of mTORC1. The 12 C-terminal residues of LAMTOR1 are highly conserved and contain two prominent charged (KEE) and hydrophobic (LVV) clusters (Fig. 3 and fig. S1), which we next mutated to alanine triplets. LAMTOR1_KEE.SH and LAMTOR1_LVV.SH interacted with LAMTOR3 and LAMTOR2, but mutation of KEE impaired the association with the Rags, and LVV mutation completely abolished it (Fig. 3E).

The CTD of Gtr1 associates with Ego1 and Ego2 (22), and dimerized CTDs of the Rags are necessary for LAMTOR1 binding (23). The termini of the resolved LAMTOR1 portions are in proximity to the LAMTOR2 and LAMTOR3 regions that were previously thought to be potential effector interaction sites (Fig. 1A) (12). Therefore, we tested whether the Rags might directly bind LAMTOR through their CTDs. We performed mass spectrometry after cross-linking to analyze the heptameric LAMTOR-RagA T21N-RagC Q120L complex (supplementary text and figs. S3 and S10). We detected intra- and intermolecular contacts among the LAMTOR subunits coherent with the crystal structure. Consistent with a previous model (23), we detected cross-linked peptides between the RagA CTD and LAMTOR2 and between the RagC CTD and LAMTOR3 but not between the G domains of the Rags and LAMTOR components.

As deduced from the crystal structure of heterodimeric Gtr1 and Gtr2, the CTDs of the Rags are predicted to form a stable roadblock heterodimer (23, 24) and coimmunoprecipitate LAMTOR1 in vivo (23). Purified Rag CTDs and LAMTOR eluted as a stable heptameric complex, as confirmed by size exclusion chromatography and mass spectrometry (fig. S11). Thus, in vitro, the Rags CTDs were sufficient to interact with the LAMTOR complex.

The crystal structure of LAMTOR with Rag CTDs revealed CTDs binding to the predicted region of LAMTOR2 and LAMTOR3 with additional contact surfaces provided by LAMTOR1 (Fig. 4, fig. S12, and table S1) (23). In contrast to the pentameric LAMTOR structure, LAMTOR1 residues 47 to 64 form a helix in the heptameric complex (Fig. 4 and figs. S13 and S14). We have reconstituted the LAMTOR1HM cell line with a version of LAMTOR1 in which N64, I66, and V68 (hereafter NIV) were mutated to alanines (fig. S1). Immunoprecipitated LAMTOR1_NIV.HA failed to recruit the Rags or SLC38A9 despite restoring the assembly of the LAMTOR complex (fig. S9B), and colocalization of endogenous RagC with LAMTOR1_NIV.HA (fig. S9C) was abolished.

Fig. 4 Interaction of Rag GTPases with the LAMTOR complex.

(A) Crystal structure of the heptamer. Dark green, RagA CTD; light green, RagC CTD; red, LAMTOR1; purple, LAMTOR2; blue, LAMTOR3; orange, LAMTOR4; yellow, LAMTOR5. N and C termini are labeled. (B) Rotated view of the structure shown in (A) with the G domains, modeled on the basis of the Gtr1 and Gtr2 structures [PDB: 4ARZ (24)] to define their positional orientation in the heptameric complex. Experimental cross-links are indicated by black dotted lines. A cartoon representation is included (bottom right). The antiparallel arrows between two roadblock domains indicate the central β-sheet augmentation of roadblock heterodimers.

The additional density observed at the C terminus of LAMTOR1 (residues 150 to 158) revealed that the LVV motif contacts the N terminus of LAMTOR2 (unstructured in the pentamer crystal) that mediates most of the interactions between LAMTOR and the RagA CTD. Residues 157 to 160, adjacent to the LVV motif of LAMTOR1, come close to the RagA CTD and may further contribute to this interaction (figs. S12 to S14). Together with the functional data and the cross-linking analysis (fig. S10C), where three different pairs of cross-linked peptides identified the N terminus of LAMTOR2 linked to the C terminus of RagA (K295 and K299), a second essential region for Rags recruitment in the LAMTOR was identified.

We superimposed the structure of the CTDs of yeast Gtr1 and Gtr2 to the heptameric complex (fig. S15). With some minor deviations, the CTDs of Rag and Gtr proteins adopted a similar fold. This defined the approximate orientation of the G domains of the Rags for interaction with mTORC1 (Fig. 4B). Targeting of Gtr1 and Gtr2 to the membrane in the absence of the Ego complex is sufficient to promote TORC1 activity (15), indicating that in yeast the Ego complex serves as scaffold for the Gtr proteins. In contrast, mammalian LAMTOR exhibits GEF activity toward RagA and RagB (1). Extrapolating from our heptameric complex, the G domains and associated nucleotide binding sites would be far away from the LAMTOR components, raising the possibility of an unusual or allosteric mechanism of nucleotide exchange, if no other cellular components are required for GEF activity.

Altering the Rag-LAMTOR interaction might represent a previously unknown mechanism for specific inhibition of mTORC1 (25). The identification of two small structural motifs (LVV and NIV) in LAMTOR1, necessary for Rag recruitment, may be used to design compounds interfering with this interaction.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S16

Tables S1 to S3

References (2660)

References and Notes

  1. Acknowledgments: We thank J. Frankel and E. M. Nelsen for depositing 12G10 anti–α-tubulin to the Developmental Studies Hybridoma Bank. We also thank M. Gstaiger for providing the pTO-HA-STREPIIc-GW-FRT and pCDNA5/FRT/TO/SH/GW constructs, I. Berger (European Molecular Biology Laboratory Grenoble) for support on the design of the synthetic LAMTOR gene construct and for the acceptor vector, M. Nanao and D. Sanctis at the European Synchrotron Radiation Facility for support with data acquisition, C. Herrmann for technical support, K. Pansi for technical support and maintenance of the insect cell culture, and W. Kabsch (Max Planck Institute Heidelberg, Germany) and F. McCormick (University of California, San Francisco) for valuable comments and discussion. The obtained data sets from LAMTOR pentamer and LAMTOR with RagA and RagC CTDs have been deposited under Protein Data Bank (PDB) IDs 6EHP and 6EHR, respectively. The work presented in this manuscript was supported by grants from the Austrian Science Funds (FWF): P26682 to L.A.H. and P28975 to K.S.
View Abstract

Navigate This Article