Retractile lysyl-tRNA synthetase-AIMP2 assembly in the human multi-aminoacyl-tRNA synthetase complex

Multi-aminoacyl-tRNA synthetase complex (MSC) is the second largest machinery for protein synthesis in human cells and also regulates multiple nontranslational functions through its components. Previous studies have shown that the MSC can respond to external signals by releasing its components to function outside it. The internal assembly is fundamental to MSC regulation. Here, using crystal structural analyses (at 1.88 Å resolution) along with molecular modeling, gel-filtration chromatography, and co-immunoprecipitation, we report that human lysyl-tRNA synthetase (LysRS) forms a tighter assembly with the scaffold protein aminoacyl-tRNA synthetase complex-interacting multifunctional protein 2 (AIMP2) than previously observed. We found that two AIMP2 N-terminal peptides form an antiparallel scaffold and hold two LysRS dimers through four binding motifs and additional interactions. Of note, the four catalytic subunits of LysRS in the tightly assembled complex were all accessible for tRNA recognition. We further noted that two recently reported human disease-associated mutations conflict with this tighter assembly, cause LysRS release from the MSC, and inactivate the enzyme. These findings reveal a previously unknown dimension of MSC subcomplex assembly and suggest that the retractility of this complex may be critical for its physiological functions.

Assembly or disassembly of the complex is essential to regulate the functions of involved MSC components in cellular homeostasis. For example, GluProRS was reported to have roles in INF-␥-related inflammation induced gene-specific silencing of translation and antiviral immunity when released from MSC (7,8). IFN-␥ induces sequential phosphorylation of Ser 886 and Ser 999 in the noncatalytic linker of GluProRS, which triggers GluProRS release from the MSC. Phosphorylation of GluProRS is also required for the interaction with NS1-associated protein-1 (NSAP1), ribosomal protein L13a, and glyceraldehyde-3-phosphate dehydrogenase to form functional GAIT complex (9). The GAIT complex binds GAIT element in specific mRNA and eventually represses translation (10). Upon DNA damage response, the general control nonrepressed-2 (GCN2) kinase phosphorylates MetRS at Ser 662 . The phosphorylation induces a conformational change of MetRS and releases the scaffold protein AIMP3, resulting in AIMP3 translocation to nucleus to activate ataxia telangiectasia mutated/ATM-and Rad3-related pathways (11). Another scaffold protein AIMP1 has several proteolytic forms, including EMAP II and p43ARF (12)(13)(14)(15). EMAP II, p43ARF, and the full-length AIMP1 can all be secreted out of cells to play cytokine functions, including procoagulant, proinflammatory, proapoptotic, and angiogenic activities (12, 16 -18). Outside of MSC, AIMP2 binds and regulates the protein stability of far upstream element-binding protein 1 (FBP1) through the central region of AIMP2. AIMP2 promotes FBP1 degradation by Smurf 2-dependent post-translational ubiquitination and decreases the transcription level of c-myc (19).
Human LysRS plays noncanonical functions in several cellular processes, including HIV reverse transcription, viral packaging, neuropathy, immune response, cancer metastasis, etc. (20). Upon  cro ARTICLE could be selectively packed into the HIV-1 virion. LysRS, tRNA Lys3 , viral precursor proteins Gag, and GagPol form a cytoplasmic nucleoprotein complex to facilitate the HIV viral reverse transcription after infecting new host cells (21)(22)(23). In mast cells, IgE-antigen stimulation activates specific phosphorylation of LysRS on a Ser 207 residue (24). The Ser 207 phosphorylation causes a structural opening of LysRS that disrupts its interaction with AIMP2 and leads to its dissociation from the MSC. The phosphorylated LysRS further enters into the nuclear, binds a transcription factor MITF, and eventually activates the MITF-targeted genes transcription (25). Human LysRS also has a cytokine activity (26). When secreted from macrophage-like differentiated THP-1 cells, LysRS mediates the transduction of inflammatory signals in the Shiga toxinproducing Escherichia coli-infected host cell (27). In addition, through a different structural mechanism, LysRS binds to cell membrane through the laminin receptor 67LR to promote epithelial cell migration (28,29).
LysRS's recruitment to fulfill noncanonical functions involves its dissociation from MSC and association with new binding partners. Characterization of LysRS assembly within the MSC helps to understand these cellular mechanisms. Our previous studies showed one N terminus of the scaffold protein AIMP2 was able to hold one LysRS dimer in MSC (25,30), so that two LysRS dimers and two AIMP2 molecules per MSC exist in redundancy (25,30,31). The two LysRS dimers and the two copies of AIMP2 are packed with a V shape in solution (V form) (Fig. S1). In this form, each of the two LysRS dimers is connected by the N-terminal 32 residues from one AIMP2, so that two LysRS dimers do not directly contact. In this loose form, each LysRS dimer could move flexibly in solution, function for aminoacylation, and diffuse from AIMP2 for functions outside of MSC (30). Here we report a new crystal structure of LysRS-AIMP2 subcomplex, in which the two LysRS dimers are retracted by AIMP2 into a tighter assembly. Two human disease-related mutations disturbed LysRS's incorporation into MSC in cells and are conflicted with this tight assembly. This finding reveals a previously unknown dimension of MSC subcomplex assembly and suggests that the retractility of the complex may be critical for its diverse physiological functions.

Overall structure of the LysRS-AIMP2 subcomplex in a tight assembly
We had previously determined a LysRS-AIMP2 complex at a resolution of 2.86 Å and discovered that one AIMP2 N terminus was able to bind one LysRS dimer through two LysRSbinding motifs (25,30). Recently, we solved a second LysRS-AIMP2 complex structure with 1.88 Å resolution in a different condition (Table 1 and Fig. S2). In the new crystal, only motif 1 (MYQVKPYH) of AIMP2 interacts with LysRS in the same asymmetric unit (ASU1) (Fig. 1A), and motif 2 (MYRLPNVH) extends to a nearby asymmetric unit (ASU2), where it interacts with another LysRS dimer (Fig. 1, B and C). Symmetrically, motif 2 from the AIMP2 in the ASU2 extends back to ASU1 and interacts with the other AIMP2-binding pocket on LysRS dimer 1. Thus, two nearby ASUs form one biological unit that each of the AIMP2s concatenates two LysRS dimers (Fig. 1, B and C). This new crystal structure reveals a tightly packed X-shaped assembly of the LysRS-AIMP2 subcomplex (X form) (Fig. 1C), which is different from the previously discovered V-form assembly (Fig. S1). Four tRNA molecules can be docked on the X-form complex at the same time without any clash with other

Retractile LysRS-AIMP2 assembly in human MSC
protein or tRNA atoms (Fig. 1D), indicating that all four LysRS catalytic subunits are functional in this form. It implies that the X-form complex might not only exist in crystal but also reflect physiological assembly.

LysRS-AIMP2 interactions in the X-form complex
Consistent with previous reports, motifs 1 and 2 of AIMP2 contribute to the major interactions between AIMP2 and LysRS ( In addition, the two LysRS dimers in the X-from complex are mainly held by the two AIMP2s. There are only a few direct interactions between the two LysRS dimers, such as a hydrophobic interaction between Pro 390 and His 348 or a van der Waals interaction between Thr 388 and Ser 478 (Fig. S4).

Presence of X-form complex in human cells
To confirm the X-form complex exists in human cells, we designed an MSC incorporation assay based on gel-filtration chromatography. In this assay, we fused the LysRS binding region of AIMP2 (amino acids 1-36) to the N-terminal end of eGFP and expressed the fusion protein in HEK293T cells. Dis-tinct from the V form, only the X-form assembly allows two AIMP2 N-terminal peptides to bind corporately across two LysRS dimers. If the X form indeed exists in cells, the AIMP2N-eGFP protein can be incorporated into the MSC by forming a heterogeneous X-form complex together with one endogenous AIMP2 and two LysRS dimers (Fig. 3A). In the V-form complex, one AIMP2 N-terminal peptide binds one LysRS dimer. Competitive binding of AIMP2N-eGFP with the endogenous AIMP2 will lead to release of LysRS from the complex (Fig. 3B). As a result, when we loaded the AIMP2N-eGFP-expressing cell lysate onto a Superose 6 gel-filtration column, the AIMP2N-fused protein was detected in both low-molecularweight fractions and high-molecular-weight fractions (Ͼ1 MDa). Although the control eGFP only existed in low-molecular-weight fractions (Fig. 3C). Therefore, the presence of AIMP2N-eGFP in the high-molecular-weight fractions implies the existence of X-form complex in human cells. In addition, more LysRS in low-molecular-weight fractions was found from the AIMP2N-eGFP-expressing cells compared to the eGFP control cells (Fig. 3D), suggesting that the X and V forms may co-exist in HEK293T cells.

Two human disease-related mutations conflict with the X-form complex
A 14-year-old girl patient harboring two novel biallelic mutations in LysRS (L350H and P390R) was reported recently (33). The L350H mutation was inherited from her mother, whereas the P390R mutation was a de novo mutation. The patient manifests a severe form of cardiomyopathy associated with lactic acidosis, mild myopathy, and intellectual disability (33). The pathogenesis is unclear. We therefore analyzed the two mutations on the X-form complex. Leu 350 locates on the bottom side of LysRS, which is opposite to the tRNA-binding side, and is

Retractile LysRS-AIMP2 assembly in human MSC
ϳ6.5 Å away from the Met 3 residue of AIMP2 (Fig. 4A). The side chain of Leu 350 makes hydrophobic interactions with LysRS residues Phe 340 , Met 342 , Ala 345 , and Ala 545 (Fig. 4A). Among these residues, Met 342 and Ala 345 have direct hydrophobic interaction with Met 3 residue of AIMP2 (Fig. 4A). When Leu 350 is mutated to a bigger histidine residue, the Met 342 and Ala 345 residues are slightly pushed away for ϳ2.0 and 0.6 Å, which is negative for AIMP2 binding (Fig. 4B, and  Fig. S5). The Pro 390 residue locates 13 Å away from AIMP2 (Fig. 4C); thus, the P390R mutation should not affect the AIMP2 binding directly. However, Pro 390 closely faces the other LysRS dimer in the X-form complex and makes hydrophobic interaction with His 348 from the other LysRS dimer (Fig. S4). The P390R mutation causes clashes with the LysRS dimer 2 at residues such as Tyr 347 , His 348 , Met 351 , and Arg 393 (Fig. 4D). Because of the crystallographic symmetry, the P390R mutation from LysRS dimer 2 causes same clash with LysRS dimer 1. This indicates that the P390R mutation disturbs the X-form complex. On the other side, the two LysRS dimers in the previous V-form complex are distant from each other; thus, the P390R mutation might not affect the V-form complex assembly.

Mutations related to human disease intervene in MSC assembly and enzyme activity
To confirm the effect of the mutations upon MSC assembly in cell, we performed co-immunoprecipitation experiment to examine the ability of MSC association of LysRS WT and mutant proteins. L350H mutant protein co-precipitated with AIMP2 and MetRS as WT protein (Fig. 5A and Fig. S6), indicating that a single L350H mutation did not significantly affect the MSC association, consistent with the subtle change in the L350H crystal structure and the normal phenotype of the patient's mother. The P390R mutation was also able to interact with AIMP2 and MetRS as normal, whereas the L350H/P390R double mutant significantly lost the ability to interact with AIMP2 and MetRS in the co-immunoprecipitation experiment ( Fig. 5A and Fig. S6). These results indicate that the two mutations L350H and P390R might each weakly affect the association of LysRS within MSC. However, the dual mutations can aggravate the interruption of MSC assembly.
We then examined how the two mutants might affect the enzyme activity for supporting essential protein translation using a functional replacement assay in Saccharomyces cerevi-

Retractile LysRS-AIMP2 assembly in human MSC
siae yeast (25). WT LysRS could substitute for the yeast cytoplasmic LysRS (which is controlled by a tetracycline-induced promoter and can be suppressed by doxycycline) and sustain normal cell growth (Fig. 5B). The inactive LysRS S207D mutant was used as a negative control (25). Both L350H and P390R single mutants were as active as WT in supporting cell growth. However, the L350H/P390R double mutant was completely inactive (Fig. 5B). These results show the two mutations have a synergetic effect in disrupting both the MSC assembly and the enzyme activity.
Notably, Leu 350 locates at the ␣-helix (amino acids 346 -366) right below the seven-strand central ␤-sheet of the active center, and Pro 390 locates at a long loop connecting the same helix to the ␤3 of the seven-strand ␤-sheet (Fig. S7). The double mutation might have a synergistic affect in altering the local conformation of the helix (amino acids 346 -366) and the central ␤-sheet of the enzyme, which disturbed both X-form complex assembly and enzyme activity. Either of the dual defects or both might play a role in the pathogenesis.

A pseudo-disulfide bond stabilizes the X-form complex
The two AIMP2 N-terminal peptides bind antiparallelly across the two LysRS dimers (Fig. 6A). Interestingly, Cys 23 next to motif 2 formed a disulfide bond in the crystal structure, which stabilizes the compact X-form complex (Fig. 6B and Fig.  S8). Following the LysRS binding sequence, AIMP2 has a leucine-zipper region at residues 48 -81 and a GST domain at the C terminus. The leucine-zipper region dimerizes AIMP2 and interacts with AIMP1, ArgRS, and GlnRS, whereas the GST domain interacts with AspRS and GluProRS (34,35). The distance between the two His 31 residues of AIMP2 in the X-form complex is ϳ41 Å (Fig. 6C), agreeable to the dimerization of the C-terminal part of the protein for further assembly of the whole MSC complex (Fig. 6D). If the disulfide bond formed in vivo, it would stabilize not only the LysRS-AIMP2 subcomplex but also the whole MSC.
Interestingly, the C23S mutant of the AIMP2N-eGFP protein was still capable of forming X-form complex with endogenous AIMP2 and LysRS (Fig. 3C), suggesting that the disulfide bond is not a prerequisite of the X-form assembly. On the other side, the Cys 23 together with residues involved in the X-form complex formation (such as Leu 16 -Thr 22 ) are strictly conserved from zebrafish to human (Fig. S9). In addition, LysRS from the cells expressing AIMP2N-eGFP C23S mutant showed slightly different distribution comparing to the cells expressing WT fusion protein (Fig. 3D). These results imply the Cys 23 might play a role in controlling the MSC assembly at certain physiological conditions, for example, when intracellular redox potential fails under oxidative stress conditions (Fig. 6E).

Potential advantage for the retractile assembly of MSC subcomplex
From our previously solved LysRS-AIMP2 crystal structure, one LysRS dimer forms two symmetric AIMP2-binding pockets by both N-terminal anticodon binding domain and C-terminal catalytic domain (25). Phosphorylation of Ser 207 on the N-C domain interface triggers significant conformational change to LysRS and disrupts the AIMP2-binding pocket, thus releasing LysRS from MSC for nontranslational function in mast cell activation (25). In the current crystal structure, we show that LysRS could also form tight assembly with AIMP2 ( Fig. 1). In this assembly, the major interactions between LysRS and AIMP2 are consistent with the previous result (Fig. 2), which double ensures the molecular mechanisms for the phosphorylation triggered LysRS release from MSC (25). These two complex forms could both represent the assembly of two LysRS dimers and two AIMP2 proteins in the human MSC. The two forms might co-exist in an equilibrium in cells.
Complementary to the previously solved V-form complex, the X-form setup may have a merit of orderliness. The more organized status can avoid potential clash of tRNAs and ensure efficient aminoacylation catalysis (Fig. 7A), although the V-form assembly has an advantage in release one LysRS dimer from the scaffold for nontranslational functions upon cellular stimuli while retaining the other LysRS dimer for fundamental protein translation (Fig. 7, B and C). The two forms of assembly may reflect different stages of LysRS function (Fig. 7). The mechanisms for controlling the switch of the two forms yet need to be explored through further research.
In summary, this work solved a tighter LysRS-AIMP2 subcomplex in a compact X form. The study reveals a previously

Retractile LysRS-AIMP2 assembly in human MSC
unknown dimension of MSC subcomplex assembly and suggests that the retractility of the complex may be critical for its diverse physiological functions.

Structure determination
The LysRS-AIMP2 1-36 complex crystal diffraction data were obtained from Beamline LS-CAT at Advanced Photon Source of Argonne National Laboratory. The LysRS L350H crystals diffraction data were obtained from Beamline 17U1 at Shanghai Synchrotron Radiation Facility (36). All data sets were processed with HKL2000 (37). The structures were solved by molecular replacement using human LysRS structure (Protein Data Bank code 3BJU) with the program MOLREP (38). Iterative model building and refinement were performed using Coot and Phenix (39,40). The data collection and refinement statistics are given in Table 1.

Modeling LysRS-AIMP2-tRNA complex structure
Crystal structure of the AspRS-tRNA asp complex (Protein Data Bank code 1IL2) was used to generate the LysRS-AIMP2-tRNA structure model used in Fig. 1D, because no LysRS-tRNA crystal structure is currently available. The chain A (AspRS) and chain C (tRNA asp ) were extracted from 1IL2 and aligned onto the LysRS-AIMP2 complex structure based on the protein C␣ atoms in PyMOL (root mean square deviation, 2.317 Å). The tRNA molecules could then be merged into the LysRS-AIMP2 structure with reasonable coordinates. The LysRS-AIMP2-tRNA model is generated by repeating this operation for all four LysRS subunits in the X-form complex.

Yeast viability assay
The cDNA sequences encoding full-length human LysRS and its mutations were constructed in the p413GPD vector multicloning site. The plasmids were then transformed into the yeast Tet-Promoters Hughes Collection (yTHC) mutant strain from Open Biosystems (Dharmacon). The endogenous promoter of yeast cytoplasmic LysRS gene (krs1) has been replaced with a TET-titratable promoter in the yTHC genome. Thus, the expression of the gene can be switched off by the addition of doxycycline to the growth medium. 10-fold serial dilutions of freshly grown yeast cells were spotted onto selective medium SCM-HIS with or without doxycycline. The plates were incubated at 30°C for 3 days and then photographed.

MSC incorporation assay
Human AIMP2/p38 (amino acids 1-36) and the C23S mutant were fused to the N terminus of eGFP, and expressed in HEK293T cells using a pMSCV-puro vector. EGFP alone was also constructed in the same vector and expressed in HEK293T cells as control. After 48 h of transfection, the cells were collected and lysed in lysis buffer (150 mM NaCl, 20 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 1% Nonidet P-40) and then loaded onto a Superose 6 gel-filtration column (GE Healthcare, 10/300 GL) with 20 mM Tris-HCl, pH 8.0, 150 mM NaCl. The fractions   A, the compact and ordered X-form complex ensures efficient tRNA aminoacylation. B, Two LysRS dimers are held by the two AIMP2 N terminus separately in a V-form assembly, which is ready to release one LysRS dimer for nontranslational functions. C, a potential third assembly form of LysRS-AIMP2 complex, in which one LysRS dimer was released for nontranslational function and one LysRS dimer was retained for fundamental translational function.