|Home | About | Journals | Submit | Contact Us | Français|
Many multicomponent protein complexes mediating diverse cellular processes are assembled through scaffolds with specialized protein interaction modules. The multi-tRNA synthetase complex (MSC), consisting of nine different aminoacyl-tRNA synthetases and three non-enzymatic factors (AIMP1–3), serves as a hub for many signaling pathways in addition to its role in protein synthesis. However, the assembly process and structural arrangement of the MSC components are not well understood. Here we show the heterotetrameric complex structure of the glutathione transferase (GST) domains shared among the four MSC components, methionyl-tRNA synthetase (MRS), glutaminyl-prolyl-tRNA synthetase (EPRS), AIMP2 and AIMP3. The MRS-AIMP3 and EPRS-AIMP2 using interface 1 are bridged via interface 2 of AIMP3 and EPRS to generate a unique linear complex of MRS-AIMP3:EPRS-AIMP2 at the molar ratio of (1:1):(1:1). Interestingly, the affinity at interface 2 of AIMP3:EPRS can be varied depending on the occupancy of interface 1, suggesting the dynamic nature of the linear GST tetramer. The four components are optimally arranged for maximal accommodation of additional domains and proteins. These characteristics suggest the GST tetramer as a unique and dynamic structural platform from which the MSC components are assembled. Considering prevalence of the GST-like domains, this tetramer can also provide a tool for the communication of the MSC with other GST-containing cellular factors.
Because many cellular proteins exert and regulate their activities via diverse protein-protein interactions and macromolecular structures, it is important to identify the functional domains that are used for the assembly of proteins and to understand the process through which they are assembled. In many cellular pathways, scaffolding proteins bring together multiple signaling proteins through modules specialized for protein-protein interaction (1).
Glutathione transferase (GST) conjugates the sulfhydryl group of glutathione to xenobiotic substrates for detoxification and its homologs and isoforms have been described from bacteria to humans (2). Throughout evolution, GST structures have spread into different proteins (3), including aminoacyl-tRNA synthetases (ARSs)3 and other translational factors (4, 5). Among ARSs, GST homologs are found in methionyl-tRNA synthetase (MRS) from yeast to human and are involved in MRS catalysis (6). Mammalian valyl-tRNA synthetase, glutaminyl-prolyl-tRNA synthetase (EPRS), the largest isoform of human cysteinly tRNA synthetase, and glutamyl-tRNA synthetase (ERS) in some species, also contain a GST domain located in their N-terminal regions (5, 7). Although the functional implications of these embedded GST domains vary, they appear to play roles in protein assembly and folding.
Human multi-tRNA synthetase complex (MSC) is a macromolecular protein complex consisting of nine different ARSs and three ARS-interacting multifunctional proteins (AIMPs) (8). ARSs catalyze covalent bond formation between specific amino acids and tRNAs for protein synthesis, and AIMPs are non-enzymatic factors. The association and dissociation of MSC components is not only critical for protein synthesis but also for the control of diverse cellular signaling pathways (9). As system complexity is accreted, several different domains have been recruited at one or both ends of the ARS catalytic domains (10). These acquired domains appear to provide flexibility and efficiency not only in catalysis but also in protein synthesis and other signaling pathways. In MSC, two ARS-associated trans-acting factors, AIMP2 and AIMP3, also contain the inserted GST domains in addition to MRS and EPRS. However, the roles of the GST domains in catalysis and complex formation remain poorly understood.
In this work, we selected the four components of MSC containing different GST domains and determined their interactions with the crystal structures of the binary GST domain complexes and electron microscopic structures of tetrameric complexes to understand how they are assembled. Here we present the heterotetrameric GST complex as a novel dynamic molecular framework to bring MSC components together.
We cloned the full-length human MARS and AIMP2 genes into pET30a, the full-length EPRS into pET28a, and EEFIE1 (for AIMP3) into the pQE80L and pProEX vectors. We subcloned the genes for the GST domains of MRS (MRSGST1,2; 1–207 and 1–224), EPRS (EPRSGST1–3; 1–164, 1–175, and 1–196), and AIMP2 (AIMPGST1,2; 90–320 and 111–320), the GST-ERS region of EPRS (EPRSGST4; 1–769) and AIMP3C5(1–169) into the expression vectors containing His tag. For MRSGST, pET30a; for AIMP3, pQE80L, pET30a, and pProEX; for EPRSGST, pET30a; and for AIMP2GST, pQE80L and pET30a vectors were used. We generated the various sizes of constructs to facilitate purification and crystallization of proteins and to distinguish well the protein bends in SDS-PAGE. Mutations for pulldown assay were introduced to MRSGST2, AIMP3, EPRSGST2, and AIMP2GST2 using the QuikChange site-directed mutagenesis method and confirmed by sequencing. To generate proteins without a His tag, a stop codon (TAA) was introduced at the C terminus of the protein in the pET30a vector by mutagenesis as described above.
Protein expression of the recombinant constructs in Escherichia coli BL21(DE3) strain was induced with 0.2 mm isopropyl 1-thio-β-d-galactopyranoside at 18 °C. His-tagged recombinant proteins were purified by affinity chromatography using a Ni-NTA column. The harvested cells were resuspended in Tris buffer (50 mm Tris-HCl, pH 8.0) and disrupted by sonication. After centrifugation, the crude extracts were loaded onto a Ni-NTA column (Qiagen) and washed with a buffer containing 300 mm NaCl and 15 mm imidazole. The proteins were eluted with a buffer containing 200 mm imidazole. AIMP3 expressed from a pProEX vector was digested by rTEV at 21 °C in the presence of 0.5 mm EDTA and 1 mm DTT, and reapplied to a Ni-NTA column after dialysis to remove the cleaved His6 tag. To obtain binary complexes, harvested cells that had expressed two different proteins were mixed and purified together as described above.
Interactions between four GST domains were examined by in vitro pulldown assays using the Ni-NTA affinity chromatography. EPRSGST2 and MRSGST2 with C-terminal His6-tagged and AIMP2GST2 and AIMP3 with N-terminal His6-tagged proteins were applied to the Ni-NTA column in the presence of other proteins without His6 tag and washed with 50 mm Tris-HCl buffer (pH 8.0) containing 150 mm NaCl and 5 mm imidazole to avoid nonspecific binding to the Ni-NTA beads. The samples were eluted with a buffer containing 150 mm NaCl and 200 mm imidazole. The eluted proteins were subjected to 12% SDS-PAGE. To analyze the interface between the GST domains (MRSGST2:His-AIMP3, His-AIMP3:EPRSGST2, and EPRSGST2:His-AIMP2GST2), the same protocol was used with the GST domain proteins containing the indicated mutations at interfaces 1 and 2.
To confirm the molecular weight of the different protein complexes, dimeric (MRSGST2-His:His-AIMP3, His-AIMP3:EPRSGST3-His, and EPRSGST3-His:AIMP2GST2-His), trimeric (MRSGST2-His:His-AIMP3:EPRSGST3-His, and His-AIMP3:EPRSGST3-His:AIMP2GST2-His), and tetrameric complexes (MRSGST2-His:His-AIMP3:EPRSGST3-His:AIMP2GST2-His and MRS-His:His-AIMP3:His-EPRSGST4:His-AIMP2GST2) were co-purified using Ni-NTA affinity chromatography. The purified proteins were injected onto a Superdex 75 10/300 column (dimeric complexes) and Superdex 200 10/300 column (trimeric and tetrameric complexes) (GE Healthcare) at a flow rate of 0.5 ml/min in a buffer containing 150 mm NaCl. The molecular weights of the eluted samples were calculated based on the calibration curve of standard samples.
We used the pair of CFP and YFP. The genes encoding AIMP2/histidyl-tRNA synthetase (HRS), AIMP3, and EPRS/MRS were inserted into the plasmids, pAmCyan1-N1 (SalI and SmaI), pAmCyan1-C1 (SalI and SmaI), and pEYFP-N1 (XhoI and BamHI), respectively. The indicated pairs of proteins were transfected into CHO-K1 cells that were cultivated in RPMI containing 10% FBS and 1% antibiotics, using Lipofectamine 3000 (Invitrogen). Cell imaging for FRET analysis was made using a confocal laser scanning microscope (Nikon, A1Rsi) with CFI Plan Apochromat VC 20X N.A. 0.75. The CFP- and YFP-tagged proteins were analyzed by excitation at 457 nm and emission at 464–499 nm, and excitation at 514 nm and emission at 525–555 nm, respectively. For FRET analysis, we defined the co-localized region of CFP- and YFP-proteins as return of intensity and FRET efficiency was determined using NIS-Elements AR 3.2 64-bit version 3.22 (Nikon imaging software).
Crystallization of the complex proteins was initially performed using sparse matrix screens (Hampton Research and Emerald Biostructures) and the sitting-drop vapor diffusion method at 21 °C. After optimization, the best crystals of the MRSGST1-His:AIMP3C5 complex were obtained under precipitant conditions of 18% PEG3350, 0.1 m HEPES (pH 7.7) in drops where 1.2 μl of protein solution (15 mg/ml) was mixed with 1.2 μl of precipitant solution. The protein crystals were soaked overnight in the precipitant solution containing an additional 0.2 m 5-amino-2,4,5-triiodoisophthalic acid (I3C, Hampton Research) for phasing (11). The crystal was transferred to a cryo-protectant solution containing an additional 13% glycerol in the crystallization solution, prior to x-ray diffraction data collection. A single anomalous x-ray dispersion dataset for the iodine from I3C was collected at a wavelength of 0.9793 Å using a Quantum 210 CCD detector at the 7A beamline of the Pohang Accelerator Laboratory (PAL, Korea). Data were integrated and scaled with HKL2000 (12). The MRSGST1-His:AIMP3C5 mutant crystal belonged to space group P22121 with dimensions a = 43.2 Å, b = 71.4 Å, c = 116.2 Å and contained one complex in the asymmetric unit.
AIMP3C5-EPRSGST2 complex crystals were grown in 1.75 m ammonium sulfate and 0.4 m sodium chloride. The crystals were flash-frozen in liquid nitrogen for data collection with the cryo-protectant solution containing 20% glycerol. Data sets for the complex were obtained at 0.9796 Å on beamline 5C at the PAL. Data were integrated and scaled with HKL2000. The crystal diffracted to 2.6 Å, and belongs to space group P31. The unit cell dimensions of the crystal were a = 92.1 Å, b = 92.1 Å, c = 186.0 Å, and γ = 120.0° with four complexes in the asymmetric unit.
After optimization, the best selenomethionine crystals of the EPRSGST2-His:AIMP2GST1-His complex were grown in 20% PEG3350, 0.2 m potassium citrate, whereas the crystals of the native EPRSGST2-AIMP2GST1-His complex were grown in 24% PEG3350, 0.2 m ammonium chloride with a protein concentration of 21.4 mg/ml. Crystals were transferred to a cryo-protectant solution containing 15% glycerol added to the crystallization solution, and the x-ray diffraction data for both crystals were collected at a wavelength of 0.9796 Å at the 5C beamline of PAL. The collected data were indexed, integrated, and scaled using the HKL-2000 software package. The EPRSGST2-His:AIMP2GST1-His and EPRSGST1:AIMP2GST2-His crystals diffracted to 3.3- and 2.6-Å resolution and belonged to space group P43 with dimensions of a = 84.3 Å, b = 84.3 Å, c = 147.3 Å and P212121 with dimensions of a = 94.7 Å, b = 111.8 Å, c = 181.3 Å, respectively. Each crystal contained four complexes in the asymmetric unit.
The structure of the I3C-soaked MRSGST1:AIMP3C5 complex was determined by the single anomalous dispersion method, at a resolution of 1.6 Å. Four iodine atoms were identified in the asymmetric unit using SOLVE (13), and density modification and subsequent automated model building was performed with RESOLVE (14). The RESOLVE-built partial model was used as a guide to build the remainder of the protein manually into density-modified electron density maps with the program COOT (15). Refinement with isotropic displacement parameters was performed with Refmac4 (16) in the CCP4 suite (17). The Rwork and Rfree values of the refined structure were 0.186 and 0.236, respectively.
The structure of AIMP3C5-EPRSGST2 was determined at a resolution of 2.6 Å by molecular replacement (MR) using the crystal structures of AIMP3 from the MRSGST1-His:AIMP3C5 complex and the yeast ERSGST structure (PDB code 2HRK) as search models. Model building and structure refinement were carried out using COOT and Phenix.Refine (18), respectively. The crystal structure was determined at a final resolution of 2.6 Å with Rwork = 0.138 and Rfree = 0.170.
The initial structure of EPRSGST-AIMP2GST was determined by MR-single anomalous dispersion using the x-ray diffraction data set from EPRSGST1-His:AIMP2GST2-His (SeMet) at 3.3 Å. A partial structure of EPRSGST1 was determined by MR using Phenix.Phaser (19) with EPRSGST2 coordinates, taken from the previously solved AIMP3-EPRSGST structure, used as the search model. The anomalous signal from selenium atoms was used to obtain initial phase information with Phenix.Autosol (20), and subsequently the complex structure was built using Phenix.Autobuild (21). With the partially built model of the EPRSGST1-His:AIMP2GST2-His complex, the EPRSGST2:AIMP2GST1-His structure was solved by MR at a resolution of 2.6 Å using Phenix.Phaser. The asymmetric unit contained four EPRSGST2-AIMP2GST1 complexes and the Matthews coefficient, Vm, was calculated to be 2.65 Å3/Da, corresponding to a solvent content of 53.6%. Model building and structure refinement were carried out using COOT and Phenix.Refine, respectively. The crystal structure was determined at a resolution of 2.6 Å with Rwork = 0.189 and Rfree = 0.247.
Data collection and model statistics for the structures of the three binary complexes are summarized in Table 1. The complex structures of MRSGST1:AIMP3C5, AIMP3C5-EPRSGST2, and EPRSGST2-AIMP2GST1 were deposited under PDB codes 4BVX, 5BMU, and 5A34, respectively.
The purified protein samples, MRSGST2-AIMP3-EPRSGST2-AIMP2GST1 and MRS-AIMP3-EPRSGST4-AIMP2GST2, were diluted 100-fold with 50 mm Tris buffer (pH 7.5) containing 150 mm NaCl and 2 mm DTT to a final concentration of 100 nm. The treated samples (5 μl) were immediately applied to carbon-coated grids that had been glow-discharged (Harrick Plasma, Ithaca, NY) for 3 min in air. Grids were negatively stained using 1% uranyl acetate and examined in a Technai G2 Spirit Twin transmission electron microscope fitted with anti-contaminator (FEI, U.S, used instrumentation in Korea Basic Science Institute) operated at 120 kV. Images were recorded on a 4Kx4K, Ultrascan 895 CCD camera (Gatan, USA) at a magnification of 30,000 (0.36 nm/pixel)(22). Single-particle three-dimensional reconstruction was carried out using the EMAN package approach with C1 symmetry applied (23). The protein particles in the micrographs were selected semi-automatically in 100 × 100 pixel boxes, a surface area slightly larger than the actual size of particles using the program BOXER. To eliminate any variations in the density of images, the selected particle images were masked and normalized. A set of 322 and 710 boxed particles from negatively stained molecules of the MRSGST2-AIMP3-EPRSGST2-AIMP2GST1 and MRS-AIMP3-EPRSGST4-AIMP2GST2 complexes, respectively, were classified into five classes to provide a sufficient number of the views. Chimera (24) was used for visualization and analysis of three-dimensional volumes.
The binding affinity between His-AIMP3 C147S and EPRSGST2-His was measured by the ITC method using a MicroCal iTC200 titration calorimeter (GE Healthcare). To determine the effect of other partner proteins on the interaction of EPRS and AIMP3, complexes of MRSGST2-His:HIS-AIMP3 C147S and EPRSGST2-His:AIMP2GST2-His were also used for ITC measurements. The sample cell was filled with 250 μl of EPRSGST2-His either alone or in complex with AIMP2GST2-His, and the syringe was filled with 40 μl of His-AIMP3 C147S either alone or in complex with MRSGST2-His. Prior to the ITC experiment, the purified proteins were dialyzed overnight against 50 mm Tris-HCl buffer (pH 8.0) containing 150 mm NaCl and 2 mm DTT. His-AIMP3 C147S and His-AIMP3 C147S:MRSGST2-His complex were titrated into a solution of EPRSGST2-His and EPRSGST2-His with AIMP2GST2-His using the following concentrations: 331 μm His-AIMP3 C147S into 16 μm EPRSGST2-His, 378 μm MRSGST2-His:His-AIMP3 C147S into 14 μm EPRSGST2-His, 346 μm His-AIMP3 C147S into 15 μm EPRSGST2-His:AIMP2GST2-His, 345 μm MRSGST2-His:His-AIMP3 C147S into 15 μm EPRSGST2-His:AIMP2GST2-His. All experiments were conducted in 50 mm Tris buffer (pH 8.0) containing 150 mm NaCl and 2 mm DTT at 25 °C. Typically, an initial 0.4-μl injection was followed by 19 injections of 1.0 μl of syringe into the cell constantly stirred at 1,000 rpm, and data were recorded for 150 s between injections. The heat generated from dilution was determined in separate experiments by diluting proteins into the buffer alone and was taken as the blank value for each injection. The corrected heat values were fitted using a nonlinear least squares curve-fitting algorithm (MicroCal Origin 7.0) to obtain the stoichiometry (n), the dissociation constant (Kd), and the change in enthalpy for each enzyme-ligand interaction (ΔH).
GST is known to be a homodimeric protein with each monomer consisting of an N-terminal thioredoxin fold (GST-N) and a C-terminal α-helical subdomain (GST-C) (25). Among the MSC components, two ARSs (EPRS and MRS), and two AIMPs (AIMP2 and AIMP3) contain a GST domain (Fig. 1A). These GST domains display strong homology to the GST domain of elongation factor eEF1Bγ, which belongs to the theta class (4, 5). Sequence alignment of the domains reveals that the GST-C subdomains are similar in size and well conserved in sequence, especially the residues involved in stabilizing the helical bundle structure, whereas GST-N subdomains are less conserved in size and sequence (Fig. 2).
The potential interactions between the GST domains of the four MSC components were investigated by in vitro pulldown assays using the His-tagged EPRSGST, MRSGST, and AIMP2GST, and the full-length AIMP3 with their untagged equivalents (Fig. 1B). A specific interaction was observed between MRSGST-His and AIMP3, and between AIMP2GST-His and EPRSGST. His-AIMP3 pulled down both EPRSGST and MRSGST, whereas EPRSGST-His interacted with both AIMP3 and AIMP2GST. In contrast to previously known GST dimers, these GST domains appear to prefer heterodimer formation. Each of the mixtures of MRSGST2-AIMP3, AIMP3-EPRSGST2, and EPRSGST2-AIMP2GST2 was eluted from gel chromatography at the expected molecular size for a GST dimer (Fig. 1, C–E). The proteins in each peak were separated by gel electrophoresis at an equimolar ratio, further confirming the formation of 1:1 heterodimers. These results suggest that these GST domains form heterodimers rather than homodimers.
We investigated cellular interactions between the GST domains by FRET analysis using CFP and YFP (Fig. 3). We introduced the pairs of CFP-AIMP3:YFP-MRS, CFP-AIMP3:YFP-EPRS, CFP-AIMP2:YFP-EPRS, and CFP-HRS (histidyl-tRNA synthetase):YFP-MRS as negative control into CHO-K1 cells and monitored the FRET signal. The pairs of AIMP3:MRS, AIMP3:EPRS, and AIMP2:EPRS showed high FRET efficiencies with averages of 80.2 (n = 37), 75.8 (n = 34), and 80.8% (n = 40), respectively, whereas the non-interaction pair of HRS:MRS showed a FRET efficiency of only 0.9% (n = 27), further suggesting the formation of AIMP3-MRS, AIMP3-EPRS, and EIMP2-EPRS complexes.
To understand the molecular details for the GST-mediated complex formation, we determined the heterodimeric complex structures of MRSGST-AIMP3, AIMP3-EPRSGST, and EPRSGST-AIMP2GST at a resolutions of 1.6, 2.6, and 2.6 Å, respectively (Table 1), and analyzed the monomeric structures first. Although all four GST domains adopt a GST-fold consisting of the N and C subdomains (Fig. 4, A–D), MRS and EPRS contain a canonical four-stranded β-sheet, corresponding to that of the canonical thioredoxin-fold, in their GST-N subdomains, whereas AIMP3 and AIMP2 have a three- and five-stranded β-sheet, respectively. Unlike human GSTθ, which has an α-helix that completes the α-β-α thioredoxin structure inserted between the β2 and β3 strands (27), MRS, EPRS, AIMP2, and AIMP3 lack the helix in the β-sheet, although MRS retains the long β2-β3 loop (Fig. 4, E–I).
The AIMP3 structures in the complexes with MRSGST and EPRSGST are essentially identical to the previously determined structure of AIMP3 alone (28), suggesting little conformational change from the complex formation. AIMP3 adopts an α-helical bundle structure with the central α5 helix surrounded by helices from GST-C and GST-N (Fig. 4A). Its α7 helix is perpendicular to the other helices and is followed by a long C-terminal tail. Compared with AIMP3, MRSGST contains a helix bundle with a long kinked α3 helix (α3A and α3B), and lacks the short α6 helix of AIMP3 situated between the central α5 and the next helix (Fig. 4B). MRS has two additional helices (α7 and α8) at its C terminus, whereas AIMP3 has a flexible C-terminal peptide. The last helix α8 is situated between GST-N and GST-C, as observed in human GSTθ. The structure of EPRSGST in the complex of AIMP3-EPRSGST2 is almost identical to that in the EPRSGST2-AIMP2GST1 complex. EPRSGST has a relatively short β2-β3 loop compared with MRSGST. GST-C of EPRSGST has a helix (α6) linking central α5 and perpendicular α7 helices and its C-terminal tail forms α8 helix (Fig. 4C). AIMP2 has a larger GST-N than others. Its GST-N is comprised of a five-stranded β-sheet and three helices with an additional helix-loop-strand motif inserted between the first helix and the following strand (Fig. 4I). It has a relatively short β3-β4 loop and no helix connecting the central (α6) and perpendicular helix (α7) in GST-C (Fig. 4D).
The structures of the MRSGST-AIMP3 and EPRSGST-AIMP2GST complexes reveal that the GST heterodimerization is formed by a helical bundle involving the α2 and α3 helices (α3 and α4 helices for AIMP2) (Fig. 5, A and B). Two GST domains in the complexes are related by 2-fold rotational pseudo-symmetry. It is similar to the canonical GST homodimer.
In the MRSGST-AIMP3 complex, α3A of MRS and α3 helices from AIMP3 come into close contact with one another through the interaction of small residues such as Ala-64 of MRS and Ala-69 of AIMP3 (Fig. 5D). There is an additional interaction between GST-N of MRS (β-sheet) and GST-C of AIMP3 (α4 helix) (Fig. 5A). MRS β4 strand is inserted between α3 and α4 helices of AIMP3, and the side chain of Gln-73 from the α3 helix has hydrogen bonds with the main chain of the β4 strand. The long β2-β3 loop of MRS contacts the α4 helix of AIMP3 (Figs. 5A and and66A), whereas limited interaction is seen between GST-N of AIMP3 and GST-C of MRS. The surface potential of the binding interfaces from the heterodimer shows that the binding is mainly mediated through polar interactions between helices α2 and α3, and hydrophobic contacts between GST-N of MRS and GST-C of AIMP3 (Fig. 6B). For example, charged residues, Arg-67, Asp-79, Glu-86 and Glu-91 of MRS and Arg-28, Lys-53, Glu-76, and Asp-97 of AIMP3 participate in the interaction (Fig. 5D). At the center of the 2-fold pseudo-symmetry a stacking of side chains was found between MRS Arg-67 and AIMP3 Gln-72 from both α3 helices. The A64R and E86R mutations of MRSGST2 and the Q73R mutation of AIMP3 ablated the interaction between the two proteins, validating the binding interface of MRS and AIMP3, as observed in the crystal structure (Fig. 5G).
In the EPRSGST-AIMP2GST complex, an additional contact is made between the N subdomain (β-sheet of AIMP2GST1) and the C subdomain (α4 helix of EPRSGST2), similar to that observed in the MRSGST-AIMP3 complex (Fig. 5B). The long β4-β5 loop and β5 strand of AIMP2GST1 are inserted between the α3 and α4 helices of EPRSGST2, and the loop connecting strands β3 and β4 extends to the α4 helix of EPRSGST2 (Fig. 6C). The β4 to β5 motif of AIMP2GST1 and an additional peptide region of AIMP2 at the N terminus of the β1 strand make hydrophobic contacts with the N terminus of EPRSGST2 α3 helix. Surface potentials of the interaction interface between the two proteins show charged interactions of AIMP2GST1 α3 and α4 helices with the EPRSGST2 α2 and α3 helices (Fig. 6D). These involve residues Arg-56, Arg-60, and Asp-79 from EPRS, and Arg-215, Asp-234, and Asp-238 from AIMP3 (Figs. 5E and and66D). As observed in the MRSGST-AIMP3 complex, a stacking interaction of the side chains from two arginine residues, Arg-56 in EPRSGST2 α2 helix and Arg-215 in AIMP2GST1 α3 helix occurred at the center of pseudo 2-fold symmetrical axis in the heterodimer. The binding interfaces were further validated by an in vitro pulldown assay. Among the mutants tested, the R215A and D238R mutations of AIMP2GST1 and R56A mutation of EPRSGST2 abolished the interaction, whereas the mutations in other surfaces did not (Fig. 5H).
Four AIMP3-EPRSGST complexes were found in the crystal asymmetric unit. All four interactions between AIMP3 and EPRSGST are established in an identical manner via their GST-C subdomains. The binding interfaces of both AIMP3 and EPRSGST consist of α7 helices and the loop connecting α4 and α5 helices, and the two proteins are oriented in pseudo 2-fold symmetry (Figs. 5C and and66E). This is similar to the interaction between two AIMP3s, which form an asymmetric unit via their binding interface 2 found in the AIMP3 crystal structure (28). Surface potential analysis indicates low electric potentials at the binding area, suggesting that the interaction between two proteins takes place mainly through hydrophobic contacts (Fig. 6F). A few hydrogen bonds between polar residues help the positioning of some side chains, including Tyr-107, Tyr-111, Phe-164, Val-106, and Leu-140 of AIMP3, and Tyr-111, Phe-153, His-146, and Leu-108 of EPRS (Fig. 5F). There is stacking of arginine side chains from α7 helices, Arg-144 of AIMP3 and Arg-149 of EPRS, at the axis of the pseudo 2-fold symmetry (Fig. 6E). We introduced mutations at interfaces 1 and 2 of AIMP3 and EPRS, and tested their effect on the interaction. Among these mutants, only R144A of AIMP3 and R149A of EPRS ablated complex formation between the two proteins (Fig. 5I).
Because AIMP3 can use different binding interfaces to form heterodimers with MRSGST and EPRSGST, we examined whether AIMP3 could form a heterotrimer with both proteins. His-AIMP3 was mixed with MRSGST2 and EPRSGST2, and the mixture was passed through a Ni-NTA column. SDS-PAGE analysis of the His-AIMP3 eluted fraction showed the three proteins were present with comparable band densities, indicating that AIMP3 can bind to both MRSGST2 and EPRSGST1 simultaneously (Fig. 7A). Similarly, we checked whether EPRSGST could bind to both AIMP2GST and AIMP3 simultaneously. All three proteins were present in the eluent from the Ni-NTA column at an approximately equal ratio (Fig. 7B). To further elucidate the complexes, we subjected the mixture of the three purified proteins (MRSGST2-His, EPRSGST3-His, and His-AIMP3) to gel filtration chromatography. The three proteins eluted as a single peak at 112 kDa, a size compatible with that of the 2:2:2 heterotrimeric complex (Fig. 7D) and the peak fraction contained all three proteins at a similar molar ratio in gel electrophoresis. The mixture of EPRSGST3-His, AIMP2GST2-His, and His-AIMP3 also eluted as a single peak from gel filtration chromatography at 61 kDa, a size compatible with that of a 1:1:1 heterotrimeric complex (Fig. 7E).
Because the two heterotrimers have AIMP3 and EPRS in common, there is a possibility that the four GST proteins form a heterotetrameric complex. We tested this possibility by repeating the pulldown methods described above with a mixture of His-AIMP3, MRSGST2, EPRSGST2, and AIMP2GST2. All four proteins were co-purified as a complex, as shown by SDS-PAGE (Fig. 7C). In the gel filtration chromatography using a mixture of MRSGST2-His, AIMP2GST2-His, EPRSGST3-His, and His-AIMP3, the proteins eluted as a single peak at a volume corresponding to the size of a 1:1:1:1 heterotetramer (Fig. 7F).
To generate a model for the tetrameric complex of MRSGST-AIMP3-EPRSGST-AIMP2GST, two heteroternary complexes, MRSGST-AIMP3-EPRSGST and AIMP3-EPRSGST-AIMP2GST, were initially modeled by superimposition of AIMP3 and EPRSGST, respectively. No contact between MRSGST and EPRSGST was observed in the model of MRSGST-AIMP3-EPRSGST, whereas AIMP3 and AIMP2GST were predicted to have extra contacts between each other in the model of AIMP3-EPRSGST-AIMP2GST (Fig. 8, A and B). The C-terminal tail of AIMP3 meets the β4-β5 loop of AIMP2GST in the vicinity of the C terminus of the α4 helices of EPRSGST. The N-terminal peptide region located before the β1 strand of AIMP2GST extends to the α7 helix and following the tail of AIMP3 over the α3 helix of EPRS. This region could provide further specificity for the formation of this ternary complex. For the tetrameric complex, three heterodimeric complex structures were placed by simultaneous superimpositions of AIMP3 and EPRS (Fig. 8C). This produced a linear MPSGST-AIMP3-EPRSGST-AIMP2GST complex without clashes between MRSGST and AIMP2GST. Thus, the tetrameric GST complex was formed by alternative usage of the binding interfaces of GST domains. The linear arrangements of four GST domains through their binding interface 1 and 2 were observed in the asymmetric units of the AIMP3-EPRSGST and EPRSGST2-AIMP2GST1 complex crystals.
Multimeric assembly of GST domains would be affected in the full-length proteins depending on the orientation of the other peptides attached to them. The catalytic domains of MRS and EPRS are several times bigger in size than GST domains. However, in the heterotetrameric complex of MRSGST-AIMP3-EPRSGST-AIMP2GST, the N- and C-terminal ends of the GST domains are oriented so that the attached catalytic and functional domains can be positioned without steric hindrance (Fig. 8D). To determine whether this tetrameric GST complex could serve as a scaffold, full-length MRS and EPRSGST4 containing the ERS catalytic domain were used to form a tetrameric complex with AIMP3 and AIMP2GST2. The mixture of full-length MRS, His-EPRSGST4, His-AIMP3, and His-AIMP2GST2 was eluted from a gel filtration column as a single peak at a molecular size of around 290 kDa, the size predicted for a 1:1:1:1 tetrameric complex of the four proteins (Fig. 8E), suggesting that the GST tetramer can be formed with full-length proteins.
To observe the three-dimensional morphology of these complexes, we subjected the MRSGST2-AIMP3-EPRSGST2-AIMP2GST1 and MRS-AIMP3-EPRSGST4-AIMP2GST2 tetramers to electron microscopic analysis followed by three-dimensional single particle analysis. Using selected individual particles of the tetrameric complexes from negatively stained fields, two-dimensional class averages and three-dimensional reconstructions of both complexes were generated (Fig. 9). When surface views of the reconstructed three-dimensional density map were compared with the two-dimensional class averages and raw particle images, three-dimensional density maps from both tetrameric complexes were consistent with the raw data, rendering the generated three-dimensional density maps reliable for further analysis (Fig. 9, C and D). Based on the two-dimensional class averages and reconstructed three-dimensional electron density models, the GST tetramers appeared to be rod shaped with a size of about 100 × 50 Å2, which is compatible with the modeled GST tetramer structure (Figs. 9A and and1010A). The modeled tetramer based on the crystal structures of the heterodimer fits well in the density map generated by three-dimensional reconstruction (Fig. 10C). In the case of the MRS-AIMP3-EPRSGST4-AIMP2GST2 complex, the larger size of the observed particles suggests that the particle would encompass the catalytic domains of MRS and ERS (Figs. 9B and and1010B). When the GST tetrameric complex was fitted into the three-dimensional reconstruction maps, sufficient densities remained on both sides to place the catalytic domains of MRS and ERS. The structure of the MRS-AIMP3-EPRSGST4-AIMP2GST2 complex was modeled by fitting homolog structures of MRS and ERS from Pyrococcus abyssi (29) and Methanothermobacter thermoautotrophicus (30), respectively, into the reconstructed three-dimensional map along with the structures of the GST tetramer (Fig. 10D).
Although the sequential assembly of the GST domains of the four MSC components was demonstrated, dissociation of the MSC complex is a necessary step in the control of diverse cellular signaling pathways. For example, AIMP3 is a tumor suppressor that dissociates from MRS and translocates to the nucleus upon UV radiation (31). For AIMP3 to be released from the tetrameric complex, it also needs to dissociate from EPRS. The surface area at the binding interface of the AIMP3ΔC5-EPRSGST2 complex is 760 Å2, whereas that of the MRSGST1-AIMP3ΔC5 and ERPSGST2-AIMP2GST1 complexes is ~1,200 Å2 (Fig. 11). There are fewer residues involved in the interaction between AIMP3ΔC5 and EPRSGST2 than in the MRSGST1-AIMP3ΔC5 and EPRSGST2-AIMP2GST1 complexes. Thus, the affinity between AIMP3 and EPRSGST, which occurs via binding interface 2, appears to be weaker than those between the subunits in MRSGST-AIMP3 and EPRSGST-AIMP2GST pairs, which employ binding interface 1.
Isothermal titration calorimetry (ITC) gave a Kd value of 0.515 μm for equimolar binding between AIMP3 and EPRSGST2 (Fig. 12A). There are possible interactions between AIMP3 and AIMP2GST in the model of the AIMP3-EPRSGST-AIMP2GST ternary complex (see Fig. 8B). To evaluate whether the interaction between AIMP3 and EPRSGST is affected by the status of EPRSGST, we measured the binding affinity between AIMP3 and EPRSGST2 in the presence of AIMP2GST2 and found that EPRSGST2 remained practically unchanged (Kd = 0.442 μm) (Fig. 12B). In the model of the MRSGST-AIMP3-EPRSGST ternary complex, no additional interactions were expected to occur between MRSGST and EPRSGST (see Fig. 8A). However, the binding affinity of the AIMP3-MRSGST2 complex for EPRSGST2 increased ~3-fold (Kd = 0.172 μm) (Fig. 12C). The AIMP3-MRSGST2 complex also enhanced the binding affinity of AIMP3 for EPRSGST2 complexed with AIMP2GST2 (Kd = 0.135 μm) (Fig. 12D). Although the conformation of AIMP3 was not significantly affected by its interaction with MRSGST, B-factor (temperature factor) analysis suggests that the AIMP3 α4-α5 loop and α7 helices, i.e. binding interface 2 for EPRSGST, is stabilized when AIMP3 is complexed with MRSGST (Fig. 12E). However, binding of AIMP2GST does not seem to stabilize the binding interface 2 of EPRSGST for AIMP3 (Fig. 12F). These results indicate that MRS may positively affect complex formation between AIMP3 and EPRS. Conversely, dissociation of AIMP3 from EPRS could be facilitated in the absence of MRS.
We showed that the four different GST domains of MSC components can assemble together to form multisubunit complexes in an orderly fashion. Thus, this tetrameric complex can serve as a nucleation platform to which multiple proteins can be integrated. Eight different human ARSs form a MSC with three AIMPs. Although the structure of the whole complex is not yet solved, much progress has been made toward determining not only the structures of the individual components and subcomplexes, but also the overall shape of MSC and the stoichiometry of its components. Based on the tetrameric complex of the GST-containing proteins described in this work, the assembly of nine proteins, RRS, QRS, AIMP1, KRS, DRS, AIMP2, EPRS, AIMP3, and MRS in the MSC can be modeled (Fig. 13). The N-terminal peptide of AIMP2 complexed with the KRS homodimer (35) can be added to the N-terminal end of the AIMP2 GST domain via a polypeptide region containing an α-helix. The ternary complex of QRS-AIMP1-RRS (32) can be attached to the KRS-AIMP2 complex via the leucine zipper between the N-terminal helices of AIMP1 and AIMP2 (33). Because PRS is covalently linked in EPRS and DRS tightly binds AIMP2 (36, 37), the presence of the two homodimers, PRS and DRS, suggests duplication of the MRS-AIMP3-ERPS-AIMP2 complex in an MSC. The ternary complex of QRS-AIMP1-RRS also has the potential to form a hexamer consisting of homodimers of each component (32). This subcomplex could be symmetrically duplicated and the stoichiometry of KRS:(AIMP2:AIMP1:DRS:ERPS:RRS):(MRS:AIMP3:QRS) would be 4:(2):(2). However, an MSC with a stoichiometry of 4:(2):(1) has also been suggested (38). Based on our own model, MRS and AIMP3 can be released from EPRS without significantly affecting the assembly of other MSC components because they are located at the periphery of the MSC. The dynamic affinity between AIMP3 and EPRSGST supports the release of the two components from the MSC. QRS is also weakly bound to MSC via its N-terminal helical interaction with the long helix of AIMP1 (32). Thus, linkage of the MSC subcomplex appears to accommodate both possible stoichiometries of the components. At this moment, it is not clear the exact location and stoichiometry of the remaining MSC components, leucyl-tRNA synthetase and isoleucyl-tRNA synthetase.
Considering that the protein synthesis machinery involves diverse cellular enzymes and regulatory factors, cells appear to have recruited GST domains as a tool for communication among translational components. In this work, we reveal the structure, dynamics, and mode of assembly of GST domains involved in protein-protein interactions. The novel heterotetrameric complex formed by GST domains in the MSC subcomplex provides an example of how GST domains can be used as a platform for the assembly of multicomponent protein complexes. The GST domains found in valyl-tRNA synthetase and cysteinly-tRNA synthetase mediates its association with elongation factors (7, 39). As shown here, there is a potential for complex formation via multiple GST domain assembly between the eEF1B complex and associated ARSs, and for communication between MSC and eEF1B mediated by the dynamic interactions of their GST domains.
H. Y. C., S. J. M., H. J. C., Y. S. C., J. M. C., S. L., H. K. K., J. J. K., and C-Y. E. worked on collection and analysis of data, Y-G. K., M. G., and H. S. J. worked on conception and design; B. S. K. and S. K. provided conception and design, financial support, manuscript writing, and final approval of manuscript.
*This work was supported by Global Frontier Project Grant NRF-2014M3A6A4062857 of the National Research Foundation funded by the Ministry of Science, ICT & Future Planning (MSIP) of Korea. The authors declare that they have no conflicts of interest with the contents of this article.
3The abbreviations used are: