|Home | About | Journals | Submit | Contact Us | Français|
REDD1 is a conserved stress-response protein that regulates mTORC1, a critical regulator of cell growth and proliferation that is implicated in cancer. REDD1 is induced by hypoxia and REDD1 overexpression is sufficient to inhibit mTORC1. mTORC1 is regulated by the small GTPase Rheb, which in turn is regulated by the GTPase-activating protein complex, TSC1/TSC2. REDD1 induced-mTORC1 inhibition requires the TSC1/TSC2 complex, and REDD1 has been proposed to act by directly binding to and sequestering 14-3-3 proteins away from TSC2 leading to TSC2-depedent inhibition of mTORC1. Structure/function analyses have led us to identify two segments in REDD1 that are essential for function, which act in an interdependent manner. We have determined a crystal structure of REDD1 at 2.0 Å resolution, which shows that these two segments fold together to form an intact domain with a novel fold. This domain is characterized by an α/β sandwich consisting of two antiparallel α-helices and a mixed β-sheet encompassing an uncommon psi-loop motif. Structure-based docking and functional analyses suggest that REDD1 does not directly bind to 14-3-3 proteins. Sequence conservation mapping to the surface of the structure and mutagenesis studies demarcated a hotspot likely to interact with effector proteins that is essential for REDD1-mediated mTORC1 inhibition.
mTORC1 (mammalian target of rapamycin complex 1) is a protein complex that plays an important role in the regulation of anabolic processes and cell growth. The best characterized function of mTORC1 is in promoting protein synthesis. mTORC1 promotes mRNA translation initiation, at least in part, by facilitating the assembly of a preinitiation complex at the 5’ end of nuclear-encoded mRNAs (at a m7GTP moiety referred to as the cap) (1). mTORC1 is composed of a core complex consisting of two subunits, the atypical protein kinase mTOR (mammalian target of rapamycin), and the adaptor protein raptor (regulatory-associated protein of mTOR) (2, 3). Several additional mTORC1 partners have been described including mLST8 (mammalian lethal with Sec13 protein 8) (4, 5), which at least during development appears to be dispensable for mTORC1 function (6), PRAS40 (proline-rich Akt substrate of 40 kDa) (7, 8) and DEPTOR (DEP domain TOR binding protein) (9). mTORC1 phosphorylates 4E-BP1 (eukaryotic initiation factor 4E [eIF4E]-binding protein 1) releasing it thereby from eIF4E at the 5’ cap and allowing eIF4E interaction with eIF4G and the nucleation of a preinitiation complex (1). mTORC1 also phosphorylates S6K1 (S6 kinase 1), and this phosphorylation event [at Thr389; (10)] primes S6K1 for further phosphorylation and activation. S6K1 in turn phosphorylates, among others, eIF4B, a component of the preinitiation complex, and the small ribosomal subunit protein S6 (11).
mTORC1 activity is tightly regulated by a variety of cues including the availability of nutrients and energy resources as well as oxygen and growth factors (12). mTORC1 interacts with the small G protein Rheb (Ras homologue enriched in brain) (13), and in vitro, GTP-loaded Rheb has been shown to activate mTORC1 (7). The levels of Rheb·GTP are in turn regulated by a protein complex with GAP (GTPase-activating protein) activity formed by the proteins tuberous sclerosis complex 1 (TSC1) and 2 (TSC2). The TSC1/TSC2 complex, and in particular TSC2 is extensively phosphorylated in response to upstream signals and these phosphorylation events play a very important role in the relay of signals to mTORC1 (11). Among other kinases, TSC2 is phosphorylated in response to growth factors by Akt (14-16), and Akt phosphorylation is thought to result in TSC2 binding to 14-3-3 (17), a process that may be important for TSC2 inactivation and consequent activation of mTORC1 (18).
Oxygen signals are relayed to mTORC1 through a process that involves REDD1 (regulated in development and DNA damage response 1; also called DDIT4). REDD1 is a 25 kDa protein that is transcriptionally upregulated in response to hypoxia (19). REDD1 is necessary for hypoxia-induced mTORC1 inhibition in fibroblasts and REDD1 overexpression is sufficient to inhibit mTORC1 (20). In addition, other stress conditions upregulate REDD1 (21), including ER stress (22) and DNA damage (23). The inhibition of mTORC1 by REDD1, or its paralogue REDD2 (also called DDIT4L), requires the TSC1/TSC2 complex (20) and can be blocked by Rheb (24, 25). It has been reported that in response to hypoxia, TSC2 dissociates from 14-3-3 proteins and REDD1 overexpression appeared to similarly disrupt TSC2 binding to 14-3-3 (26). While TSC2 binding to 14-3-3 is thought to be phosphorylation dependent, TSC2 phosphorylation at 14-3-3 binding sites appeared to be unaffected by hypoxia (26). It has been proposed that REDD1 acts by directly binding to 14-3-3 proteins and sequestering them away from TSC2 leading thereby to TSC2 activation (26). A putative 14-3-3 binding site was identified in REDD1 whose disruption impaired REDD1 function, and REDD1 was found to interact with 14-3-3 proteins in overexpression studies in cell lysates.
Here we report the crystal structure of REDD1 and evaluate its functional implications in the context of the reported model for REDD1 action.
Antibodies were obtained from the following sources: Covance, HA.11; Invitrogen, V5; Sigma, α−tubulin (clone B-5-1-2) and FoxO3; Cell Signaling: phosphorylated S6K1 (T389), phosphorylated S6 (S235/S236), S6, S6K1, eIF4E, 4E-BP1 and GST antibodies; Bethyl: HIF1α, TSC1 and REDD1; Santa Cruz: TSC2, HA (Y-11), REDD1 (N-20) and normal mouse IgG; Biosource, PRAS40; Lab Vision, pan-14-3-3. The REDD1 monoclonal antibody was described (Kucejova et al., submitted). The following were gifts: pRK7-HA-S6K (rat αII isoform) (J Blenis, Harvard Medical School), pcDNA3-Flag-Rheb (B Manning, Harvard School of Public Health), pDONR223-REDD1 (human) (M. Vidal, Dana-Farber Cancer Institute). pRK5-Flag-PRAS40 plasmid was obtained from Addgene. siRNA oligonucleotides for TSC2 (20) and REDD1 (27) were purchased from Dharmacon. HRP-conjugated secondary antibodies were purchased from the following sources: Pierce, goat anti rabbit IgG and goat anti-mouse IgG; Jackson Immunolabs, goat anti-mouse IgG light chain; Santa Cruz, donkey anti-goat IgG. TnT Quick Coupled transcription-translation kit was purchased from Promega and used according to manufacturer's instructions.
Plasmids were generated by site-directed mutagenesis (Quikchange mutagenesis kit, Stratagene) according to manufacturer's instructions with minor modifications or by PCR of human REDD1 cDNA in pcDNA3 (20). All cDNAs were validated by sequencing. Plasmids (Database ID): HA-REDD1 N-terminal deletion series [48-232 (#182), 77-232 (#426), 85-232 (#427), 101-232 (#183)]; C-terminal deletion series [1-165 (#187), 1-192 (#421), 1-209 (#430), 1-225 (#431)]; internal deletions [Δ96-157 (#188), Δ96-110 (#189), Δ111-135 (#190), Δ136-153 (#191), Δ200-204 (#597)]; “NAAIRS” substitutions [N109-119 (#432), N132-145 (#433), N153-165 (#434), N166-178 (#435), N179-193 (#437), N194-206 (#436), N207-225 (#439), N181-211 (#595), N187-207 (#596)]; point mutations [S103W (#623), S103L (#624), R133A (#564), S137A (#627), S137D (#628), P139A (#625), R133A/P139A (#626), C140S (#622), C150S (#516), C157S (#517), K219A (#565), L221A (#567), Y222A (#566), K219A/Y222A (#568)]. V5-REDD1 (#272B) was shuttled from pDONR223-REDD1 using the Gateway cloning system (Invitrogen) to a pDEST3.2 vector. REDD185-193 (#465) and REDD1207-225 (#442) expression constructs were cloned in a pCMV-GST vector.
For REDD1 crystallization the human REDD1 cDNA sequence coding for amino acids 89-226 was amplified by PCR and cloned using BamHI and XhoI sites into a modified pET28 vector (Novagen) expressing the target protein with an N-terminal His6-Sumo tag. The mutant with residues 200-204 deleted was generated by Quikchange reaction (Stratagene). Expression vectors were transformed into the bacteria strain BL21(DE3) and protein expression was induced in TB medium with 0.5 mM IPTG at 18 °C overnight. Seleno-methionine replaced protein was expressed in the same bacteria strain by using the protocol described elsewhere (28). The protein was first purified using a 1 ml HisTrap column (GE Healthcare). The N-terminal His6-Sumo tag was then removed by treatment by a Sumo-specific protease Ulp1 at 4 °C overnight. The protein was further purified by a UnoS cation exchange column (Bio-Rad), concentrated to ~10 mg/ml and stored at −80 °C.
Crystals of both native and selenomethionine replaced protein were grown in 0.1 M NaF, 20-26% PEG3350, and 0.05 mM C12E9 (polyoxyethelene-9-lauryl ether, Hampton Research) in 4 °C. Diffraction data were collected at the beamline 19BM at the advanced photon source (Argonne, IL). The structure was solved by selenomethionine single-wavelength anomalous dispersion (SAD) using the program suite Phenix. The initial SAD electron density map was of sufficient quality and a partial model was built by the Autobuild module in Phenix (29). Subsequent iterative manual model building and refinement were performed using the programs Coot and Phenix, respectively (30). Statistics for data collection and refinement are summarized in Table 1. The sequence alignment and similarity scores were calculated using the program Clustal X (31). All the molecular graphics were rendered in Pymol (http://www.pymol.org).
HeLa cells and MEFs were grown at 21% O2 and 5% CO2 in a humidified incubator in high glucose DMEM (Dubelcco) supplemented with 10% heat-inactivated fetal bovine serum (Atlanta Biologicals) and 1% Penicillin/Streptomycin (Gibco). U2OS HA-REDD1 cells were grown under similar conditions with Puromycin (1 μg/ml) and Hygromycin (50 μg/ml) and induction was performed with 1 μg/ml tetracycline (Sigma) for 3 hours unless otherwise specified. Cells were exposed to hypoxia conditions (1% O2, 5% CO2) for 3 hours in a hypoxia chamber (Coy Laboratory Products).
Plasmid DNA was transfected using the MirusIT-LT1 transfection reagent (Mirus BioCorporation) according to the manufacturer's recommendations. The relative amounts of expression vector DNA to be transfected were adjusted (when necessary) to obtain similar protein levels and the amount of transfected DNA was kept the same in all reactions by supplementing transfection mixtures with the corresponding empty vector. Cells were harvested 48 hours after transfection. siRNA oligonucleotides were transfected with Oligofectamine (Invitrogen) according to manufacturer's instructions and cells were harvested 48 hours after transfection.
Cells were washed with PBS and lysed in lysis buffer [50 mM Tris-HCl (pH 7.4), 250 mM NaCl, 0.5% Igepal] supplemented with protease inhibitors [0.1 μM aprotinin (USB), 0.02 mM leupeptin (USB), 0.01 mM pepstatin (USB), 0.5 mM benzamidine (Sigma), 0.5 mM PMSF (Sigma), 0.01 M NaF (Sigma)] and phosphatase inhibitors [2 mM imidazole (Sigma), 1.15 mM sodium molybdate (Sigma), 1 mM sodium orthovanadate (Sigma), 5 nM microcystin (Calbiochem)] for 10 minutes at 4°C. Lysates were cleared by centrifugation at 16000 g for 10 minutes, and protein concentration was measured by Bradford's method (BioRad). Protein lysates were supplemented with 3x SDS-loading buffer (6.7% SDS, 33.3% glycerol, 300 mM DTT, bromophenol blue) and denatured by boiling for 10 minutes. Similar amounts of protein were resolved by SDS-PAGE, transferred to a nitrocellulose membrane (BioRad), blocked with 5% milk in TBST (10 mM Tris-HCl, 15 mM NaCl, 0.1% Tween-20) and probed for the desired primary antibodies followed by appropriate secondary antibodies conjugated to HRP and the signal was detected by chemiluminiscence [mixing 1:1 solution 1 (2.5 mM luminol, 0.4 mM pCoumaric acid, 0.1 M Tris-HCl) and solution 2 (0.015% H2O2, 0.1 M Tris-HCl)].
Cells were rinsed with ice-cold PBS twice and lysed with IP buffer [50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.5% Igepal)] supplemented with protease and phosphatase inhibitors for 10 minutes at 4°C. Cell lysates were cleared by centrifugation at 16000 g for 10 minutes, and pre-cleared with protein A (or protein G) sepharose beads (Amersham; 50% slurry in IP buffer) for 1 hour at 4°C. Samples were normalized for protein content and rocked in the presence of ~1 μg of antibody/mg of protein overnight at 4°C. Protein A or protein G sepharose was added for 1 hour at 4°C and immunoprecipitates were recovered by centrifugation. After washing with IP buffer 3 times, 1x loading buffer was added, samples were boiled, resolved by SDS-PAGE and evaluated by western blot.
Cells were washed with ice-cold PBS and lysed in IP buffer containing protease and phosphatase inhibitors. Cell lysates were cleared by centrifugation and incubated with m7GTP sepharose beads (Amersham; 50% in IP buffer) for 2 hours at 4°C. Beads were recovered by centrifugation, washed 3 times with IP buffer, resuspended in 1x loading buffer, and protein was eluted by boiling in SDS loading buffer. Protein samples were resolved by SDS-PAGE and evaluated by western blot.
REDD1 sequences were collected using PSI-BLAST (32) and aligned with PROMALS-3D (33). Distances were calculated from the alignment using an amino acid transition probability matrix (34). Initial tree topologies were built using the Njdist program of the MOLPHY package (35, 36). Maximum likelihood trees were built using a local rearrangement search of tree topology (-R option) of PROTML (35). The reliability of the resulting tree topologies were assessed by the resampling of estimated log-likelihood method of MOLPHY (37).
The REDD proteins (REDD1 and REDD2) arose through a gene duplication event that occurred independently in humans and insects (see Figure S1 in the supplemental material) and regulate mTORC1 in response to different stimuli (21). REDD1 is induced by hypoxia (Figure 1A), and REDD1 overexpression is sufficient to inhibit mTORC1 (Figure 1B-D). mTORC1 inhibition by REDD1 requires the TSC1/TSC2 complex (Figure 1E), but REDD1 does not appear to interact with TSC1/TSC2; in reciprocal immunoprecipitation experiments, and under conditions in which TSC1 was recovered bound to TSC2, REDD1 was not found in the complex (Figure 1F). REDD1-induced mTORC1 inhibition can also be blocked by overexpression of Rheb (Figure 1G).
To obtain insight into the mechanism of REDD1 action, structure/function analyses were performed in HeLa cervical carcinoma cells, a cell type in which (1) hypoxia results in a REDD1-dependent inhibition of mTORC1 (Figure 1H), and (2) REDD1 overexpression is sufficient to inhibit mTORC1 (Figure 2A). REDD1 mutants were transfected into HeLa cells along with an expression vector for the p70 isoform of S6K1 that would serve as a readout. Consistent with sequence conservation studies, which revealed that the N-terminus of REDD1 was poorly preserved, the N-terminal 84 amino acids of REDD1 were dispensable for function (Figure 2A). However, deletion into a sequence predicted to form a α helix (beyond Glu90) disrupted REDD1 activity (Figure 2A). In contrast, the C-terminus of REDD1 is well conserved and only a few residues could be deleted without disrupting function (Figure 2B). Internal deletions were poorly tolerated (Figures 2C). Because internal deletion may disrupt the structure by juxtaposing residues that force the protein into an unphysiological conformation, experiments were also conducted using a flexible linker previously shown to be able to adopt both a α-helix or a β-strand conformation (NAAIRS, single amino acid code; (38)). As for internal deletions, NAAIRS substitutions were poorly tolerated (Figures 2D-E). In one instance, however, a NAAIRS substitution was tolerated encompassing a poorly conserved region predicted not to adopt any secondary structure (REDD1194-206; Figure 2D). These data indicated that there were two linear segments in the REDD1 sequence required for function (REDD185-193 and REDD1207-225) separated by a dispensable region (Figure S2).
To determine whether the isolated segments, REDD185-193 and REDD1207-225, retained some functionality, their ability to act as dominant negative (i.e., inhibit wild-type REDD1) was individually evaluated. Hemmaglutinin (HA)-tagged versions of REDD185-193 and REDD1207-225 were subcloned into a GST expression vector and transfected into HA-REDD1-inducible cells. Expression of REDD185-193 or REDD1207-225 at levels higher than full length inducible REDD1 failed to block wild-type REDD1 function (Figure 2F, G). These data indicate that REDD185-193 and REDD1207-225 do not function as dominant negative, suggesting that neither of the two functionally important regions when expressed separately is able to interact with (and titrate out) binding partners of REDD1 required for its function.
Failure of REDD185-193 and REDD1207-225 to act as dominant negative suggests that they function in an interdependent manner. To explore this possibility further and to obtain additional insight into the molecular mechanism of REDD1 action we sought to solve its crystal structure. Based on the functional studies above and secondary structure predictions, we decided to crystallize a segment of human REDD1 encompassing amino acids 89-226, which preserves the putative N-terminus α helix and all essential C-terminal residues. REDD189-226 was prone to precipitation, however, and despite extensive attempts it did not crystallize. An assessment of factors contributing to the lack of solubility led to the discovery of a hydrophobic region (200FLPGF204) which was absent in orthologues in other species, such as Drosophila (Figure S3), which have been previously shown to similarly inhibit dTor (39). This hydrophobic stretch corresponded to a region between the two essential REDD1 segments that was predicted to be unstructured and could be substituted for a NAAIRS sequence without affecting function (Figure 2D). In addition, simply deleting 200FLPGF204 did not affect REDD1 activity (Figure 2E). REDD189-226 with a deletion of 200FLPGF204 (REDD189-226Δϕ) was more soluble and yielded crystals that diffracted beyond 2.0 Å resolution.
The overall structure of REDD1 is characterized by a two-layered sandwich with one layer made strands ordered 2134 (Figure 3B). Strands β1-β3 form a rare structural motif known as a psi-loop, two antiparallel strands separated by an intervening strand that makes hydrogen bonds with both flanking strands (40). Of note, residues flanking the 200FLPGF204 deletion formed a loop that, as predicted, did not adopt any secondary structure. In keeping with the idea that the deletion does not impose any restrains on the structure, this loop represented the most flexible region in REDD189-226Δϕ (highest B-factors). There were two REDD189-226Δϕ molecules (A and B) in the unit cell of the P1 space group, which were essentially identical to each other; these two proteins did not, however, form a symmetric dimer and no evidence for dimerization was observed by gel filtration chromatography of the purified protein (Figure 3C). In addition, REDD1 did not appear to oligomerize in cells as determined by immunoprecipitation experiments of cells expressing REDD1 fused to two different epitope tags (Figure 3D).
At a glance the topology of REDD1 seemed unusual. To determine whether other proteins existed with a similar topology, structure searches were conducted on the PDB (Protein data bank, www.pdb.org). A PDB search using Dali (41) did not reveal any significant hits (Z-score > 5). The top Dali hit (Z-score 3.8) was a structure with two similar antiparallel α-helices packed against a β-sheet (PDB ID: 1i2l). However, the β-strands were ordered differently and the structure lacked the psi-loop motif (compare Figure 4B with 4A). A less stringent structure topology search using ProSMoS (42) identified a portion of the YrdC-like hypothetical protein (PDB ID: 1K7J) as a top hit (Figure 4C). Although the identified YrdC-like region includes two antiparallel α-helices packed against a β-sheet of similar strand order to REDD1, the cross connecting “loop” of the psi-loop motif forms the main structural core of the protein. Inspection of evolutionarily related YrdC-like structures classified by the database SCOP (Structure classification of proteins) (43) revealed that the identified YrdC-like substructure was not conserved and that it was unlikely to form independently of the core, and thus it was unlikely to be related to REDD1. Two other candidates identified by the ProSMoS topology search were representative structures of two all-β class families: a pua domain-like pseudobarrel (PDB ID: 3d79) and a Lexa-related family C-terminal domain (PDB ID: 2fjr). While these hits contained the unusual psi-loop motif, their overall topology was unrelated to REDD1 (Figure 4D, E). Taken together these data show that the topology of REDD1 is unique.
The uniqueness of REDD1 structure precluded inferences about structure/function relationships. Mapping of REDD1 sequence conservation to the crystal structure revealed a surface patch composed of highly conserved residues (Figure 5A and B) formed largely by two segments that are not contiguous in the primary sequence but cluster together in the three-dimensional structure. The first stretch was made up by a loop between helix α2 and strand β1 (138EPCG141) and the second by the C-terminal portion of strand β4 (218KKKLYSSE225) (Figure 5A). To evaluate the functional significance of this conserved surface patch, single amino acid substitutions were introduced. A conservative substitution of Cys140 by a serine residue mildly impaired REDD1 function (Figure S4B). In addition, individual mutation of Lys219, Leu221 and Tyr222 to alanine residues resulted in similarly modest functional impairments (Figures 5C and D). However, simultaneous alanine substitutions of Lys219 and Tyr222 completely abrogated REDD1 function (Figure 5D). Since all of the mutated residues are surface exposed and do not make structurally significant interactions, it is unlikely that the effects of these mutations are due to destabilization of the REDD1 structure. Thus, these results suggest that this conserved surface patch is a functional hotspot likely responsible for interacting with binding partners and essential for signaling.
We observed in the REDD1 structure a large hydrophobic pocket located between helix α1 and strand β2. In the crystal, this pocket accommodated a leucine residue (Leu192) from the second REDD1 molecule (molecule B; Figure S4A). We found that residues forming the base were conserved (Leu96, Leu100, Leu104, Leu147, Val149 and Val160) but those forming the ridge were not. Substitutions intended to block access to the pocket (S103L and S103W) did not however appreciably affect REDD1 function (Figure S4B).
REDD1 was proposed to act by directly interacting with 14-3-3 proteins through 133RLAYSEP139, a sequence that conforms to a putative 14-3-3 binding motif (26). Canonical binding to 14-3-3 proteins typically involves an unstructured motif including a phosphorylated serine residue, which largely determines the binding affinity and specificity by interacting with a positively charged patch in the ligand-binding groove of 14-3-3 proteins (44). The serine residue is typically preceded by an arginine at position -3 (mode I) or -4 (mode II) and followed by a proline at position +2 (45, 46). The 133RLAYSEP139 motif conforms to a mode II 14-3-3 binding peptide, which typically interact in a largely linear conformation. However, the 133RLAYSEP139 motif in the REDD1 structure forms part of helix α2 and the subsequent loop (Figure 6A). Binding of this segment to 14-3-3 would require a dramatic conformational change with at least partial unfolding of helix α2, and this would be predicted to destabilize the structure and is therefore unlikely to occur. Furthermore, while phosphorylation of the conserved serine residue in mode II peptides plays a critical role in stabilizing their interaction with 14-3-3 proteins (46), Ser137 in REDD1 is not well conserved (Figure S3) and Ser137 mutation to alanine does not affect REDD1 function (Figure 6C).
While most proteins bind to 14-3-3 proteins through a linear peptide motif, Exoenzyme S interacts with 14-3-3β through a motif in a α helical conformation (Figure 6B and (44)). In this instance, the helical structure lies in the 14-3-3 substrate-binding groove in the opposite orientation of mode I and II peptides. We attempted manually docking REDD1 onto 14-3-3β based on this binding mode by superimposing the helical portion of the binding motif in REDD1 onto the Exoenzyme S peptide. REDD1 could not be docked without many inevitable steric clashes (Figure 6B). The narrow substrate-binding groove in 14-3-3 was not able to accommodate the neighboring α1 helix and β4 strand. In addition, the potential binding motif in REDD1 did not show the amphipathic property of the exoenzyme S peptide that established binding specificity for 14-3-3 (Figure 6B). These observations strongly suggest that REDD1 does not bind to 14-3-3 in this rare helical mode either.
We find that simultaneous mutation of Arg133 and Pro139 to alanine residues, as previously shown (26), disrupts REDD1 function (Figure 6C), but mutation of Pro139 alone is sufficient for this effect (Figure 6C). Notably, Pro139 is also part of the conserved surface patch and that may explain its importance for REDD1 function (Figure 5A).
By gel filtration studies, purified REDD189-226Δϕ did not interact with 14-3-3β (Figure 7A). Similarly in vivo experiments failed to show an interaction between REDD1 and 14-3-3 proteins. Under conditions in which 14-3-3 proteins were found to bind known interacting partners like PRAS40 (47) and FoxO3 (Forkhead box O3; (48)), we could not detect binding of endogenous 14-3-3 proteins to REDD1 (either overexpressed or endogenous) using a variety of experimental systems (Figures 7B, C and S5A, B). These data suggest that REDD1 does not interact with 14-3-3 proteins directly (or perhaps even indirectly). While REDD1 induction appeared to decrease the amount of TSC2 bound to 14-3-3 in pan-14-3-3 immunoprecipitations, the significance of these findings remains to be fully elucidated and REDD1 was not found to bind to 14-3-3 (Figure 7D and E).
Herein we report structure/function analyses of the REDD1 protein, a protein that plays a critical role in the regulation of mTORC1 in response to stress. The crystal structure of the C-terminal functional domain of REDD1 shows a novel topology consisting of an independently folded α/β sandwich with a psi-loop motif. We have identified a conserved surface patch formed primarily by the carboxy-terminal end of strand β4 and the loop between helix α2 and strand β1 as a functional hotspot that is essential for REDD1 signaling.
It was previously proposed that REDD1 functions by directly interacting with and sequestering 14-3-3 proteins from TSC2 (26). Several lines of evidence suggest that this model may not be correct. First, the alleged 14-3-3 binding motif in REDD1 (133RLAYSEP139) is not conserved (Figure 5B); and Arg133 and Ser137 in the motif, which are key determinants for 14-3-3 binding, are replaced by other residues in orthologues in Xenopus and Drosophila, respectively (Figure S3). Second, mutation of either Ser137 or Arg133 does not affect REDD1-induced inhibition of mTORC1 (Figure 6C). Third, the REDD1 structure does not conform to any mode known of 14-3-3 interaction (Figure 6A, B). Fourth, no evidence for interaction was observed in vitro (Figure 7A), or in vivo when endogenous proteins were evaluated and under conditions in which binding of 14-3-3 proteins to other interacting proteins was appreciated (Figure 7B and C). Fifth, the finding that mutations in Pro139 disrupt function may be explained by its forming part of the conserved surface hotspot we identified (rather than by disrupting binding to 14-3-3 proteins directly) (Figure 5A). Finally, given that 14-3-3 proteins are rather abundant and are involved in interacting with over 70 proteins in cells (49), it is not clear how they would be sequestered effectively and specifically from TSC2 by REDD1 under hypoxia. Taken together these data strongly suggest that REDD1 does not function by directly interacting with 14-3-3 proteins.
REDD1-induced mTORC1 inhibition requires the TSC1/TSC2 complex, but REDD1 does not appear to interact with TSC1/TSC2. While proteins that interact with REDD1 and relay signals to TSC1/TSC2 remain to be identified, the functional hotspot in REDD1 is likely the binding site for these proteins. The crystal structure will allow us to design mutations that specifically disrupt this hotspot without affecting the overall structure of REDD1, which will prove to be valuable in distinguishing authentic and nonspecific hits in futures efforts on identifying REDD1 binding proteins.
Another potentially interesting site that emerged from analyses of the structure was a large hydrophobic pocket located between helix α1 and strand β2. In the crystal, this pocket in one molecule (molecule A) accommodated a leucine residue (Leu192) from the second REDD1 molecule (molecule B; Figure S4A). However, this pocket in molecule B did not make such an interaction and gel filtration chromatography of the purified protein failed to show evidence for oligomerization. In addition, REDD1 did not seem to oligomerize in cells as determined by immunoprecipitation experiments of cells expressing REDD1 labeled with two epitope tags. While an interaction between two REDD1 molecules involving the hydrophobic pocket is unlikely to occur in vivo, this pocket may be involved in interacting with other proteins. Due to the close proximity of the hydrophobic pocket to the functional hotspot, we speculate that they together form a bipartite binding site with improved affinity for the REDD1 binding partners. Detailed analyses of the pocket show that residues at the base are conserved, but those at the ridge are not. We tested two mutations (S103L and S103W) at the edge of the pocket, which did not appreciably affect REDD1 function. However, the direct connection of this pocket with the hydrophobic core of the protein makes it difficult to effectively block it without altering the overall structure, and the functional role of this pocket requires further investigation.
We thank members of the Brugarolas laboratory and the Zhang laboratory for helpful discussions and Dr. Kaelin for his encouragement and helpful discussions. The content is solely the responsibility of the authors and does not represent official views from any of the granting agencies.
† This work was supported by a Fundacion Caja Madrid fellowship to S. V. and the following grants to J. B.: K08NS051843 and RO1CA129387, as well as a Basil O'Connor Starter Scholar Research Award (5-FY06-582) from the March of Dimes Foundation and a V Scholar Award from The V Foundation for Cancer Research. J. B. and X. Z. are Virginia Murchison Linthicum Scholars in Medical Research at UT Southwestern.
††The coordinates of the crystal structure of REDD1 have been deposited in the PDB with the PDB ID 3LQ9.
Supporting information is available, including supplemental results and five supplemental figures, free of charge via Internet at http://pubs.acs.org/journal/bichaw.