Developing a comprehensive solution aimed to disrupt LARS1/RagD protein-protein interaction

Abstract Aminoacyl-tRNA synthetases are crucial enzymes involved in protein synthesis and various cellular physiological reactions. Aside from their standard role in linking amino acids to the corresponding tRNAs, they also impact protein homeostasis by controlling the level of soluble amino acids within the cell. For instance, leucyl-tRNA synthetase (LARS1) acts as a leucine sensor for the mammalian target of rapamycin complex 1 (mTORC1), and may also function as a probable GTPase-activating protein (GAP) for the RagD subunit of the heteromeric activator of mTORC1. In turn, mTORC1 regulates cellular processes, such as protein synthesis, autophagy, and cell growth, and is implicated in various human diseases including cancer, obesity, diabetes, and neurodegeneration. Hence, inhibitors of mTORC1 or a deregulated mTORC1 pathway may offer potential cancer therapies. In this study, we investigated the structural requirements for preventing the sensing and signal transmission from LARS to mTORC1. Building upon recent studies on mTORC1 regulation activation by leucine, we lay the foundation for the development of chemotherapeutic agents against mTORC1 that can overcome resistance to rapamycin. Using a combination of in-silico approaches to develop and validate an alternative interaction model, discussing its benefits and advancements. Finally, we identified a set of compounds ready for testing to prevent LARS1/RagD protein-protein interactions. We establish a basis for creating chemotherapeutic drugs targeting mTORC1, which can conquer resistance to rapamycin. We utilize in-silico methods to generate and confirm an alternative interaction model, outlining its advantages and improvements, and pinpoint a group of novel substances that can prevent LARS1/RagD interactions. Communicated by Ramaswamy H. Sarma


Introduction
The coordination of seemingly unrelated protein complexes inside the cell is mind-boggling and complicates any medical task.For example, the mTORC1 (a master regulator of cell growth and the mammalian target of rapamycin) pathway simultaneously functions as a trigger for launching protein synthesis and autophagy.mTORC1 is an important target in drug discovery, especially in cancer research, but, in fact, there are only a few drugs designed for it.Human cytosolic leucyl-tRNA synthetase (LARS1) is overexpressed in many cancers, such as colorectal cancer, oral cavity floor carcinoma, squamous cell skin cancer, or acute myeloid leukemia (Yoon et al., 2018).In this light, LARS may become a promising anti-cancer target and a complementary pathway to the inactivation of the mTORC1 complex (Figure 1).One of the working hypotheses regarding the involvement of leucyl-tRNA synthetase in this pathway, along with its role in translation (aminoacylation of the corresponding tRNA molecules), is related to determining the presence of leucine (sensory function) in the cell (Bar-Peled et al., 2012;Han et al., 2012).
Functioning of LARS1 as a leucine sensor for the mTORC1 cascade involves direct interaction with the RagD subunit (a part of the RagB/RagD dimer) and modulates its condition.LARS1-regulated GTP-hydrolysis of RagD is the initiating controller of the GTPase Rag cycle during the leucine signaling cascade.It is suggested that LARS1 is able to physically interact with the RagD protein through the RagD-binding subdomain (RBD), one of the two subdomains of the C-terminal domain (CTD) (Zhao et al., 2021).At the same time, it is supposed that this interaction does not prevent RagB, similarly to the RagA homologue, from contacting the mTOR complex via the RAPTOR molecule (Lee et al., 2018).
In general, the literature analysis raises questions about the involvement of LARS1 in this process, and the specifics of the interaction and impact, as well as the three-dimensional complex topography, remain unclear (Sancak et al., 2008).On one hand, some sources claim that leucine concentration does not affect the relationship between LARS1 and RagD (Kim et al., 2021;Suryawan et al., 2020).On the other hand, experimental data suggest that the interaction between LARS1 and Rags depends directly on the amount of leucine and AMP (Kim et al., 2017;2021).Despite data on inhibitors suggesting a direct disruption of the interaction interface, there is no structural or mutational evidence for this.In particular, some alteration of the shape and accessibility of the RBD subdomain, which is responsible for binding to the Rag dimer molecules is believed to occur.It is thought that a conformational change in the length of the linker site and the number of hydrogen bonds with the amino acids of the surface of the catalytic domain may lead to the tilt and rotation of the CTD (Kim et al., 2017).
Currently, only Rapalogs (rapamycin and its derivatives) are used as therapeutic allosteric mTORC1 inhibitors, but due to the discovery of a number of mutations, new therapeutic strategies are necessary to overcome mTORC1 resistance to rapamycin (Grabiner et al., 2014;Hardt et al., 2011;Sato et al., 2010;Wagle et al., 2014).More recently, with the advancement in the study of the mTOR pathway, researchers have been exploring the role of mTOR inhibitors in various tumors (Li et al., 2020;Park et al., 2019;Tang et al., 2019).A significant challenge in finding novel inhibitors is identifying the LARS1 and RagD inter-protein residue-residue contacts, the binding sites of potential inhibitors, and interpreting the effect of their binding to the protein.Thus, obtaining detailed data will aid medical chemists in formulating the task of finding potential drug candidates more closely.
The primary objective of this research was to simulate the effect of leucine/leucyladenylate on LARS1 on domain mobility and on the probability of interaction with the RagB/RagD dimer.Currently, the synthetase is thought to act as a GTPase-activating protein (GAP) for the RagB/RagD dimer, causing RagD to transition from the GTP-bound inactive form (RagD GTP ) to the GDP-bound activated form (RagD GDP ) thus initiating mTORC1 activation (Lehman & Abraham, 2020).In the second part of the research, the position of the LARS1/RagD interaction inhibitory site at the RBD site was reevaluated (Arkin et al., 2014).Previous attempts to mechanistically explain the action of substances at the level of in silico modeling solely focused on one compound that showed the necessary activity only at the phenotypic level (cellular experiment).This means that there are currently no evident 'direct' results on the interaction of inhibitors with leucyl-tRNA synthetase.Additionally, there is a lack of leucyl-tRNA synthetase inhibitors from all species (Garcia et al., 2018;Gudzera et al., 2016).One reason is related to the difficulty in creating an appropriate, cost-and time-efficient assay with easily observable readouts for the leucyl-tRNA synthetase.
Therefore, an in silico approach is perceptible for further investigation into the detailed interaction of novel and Figure 1.A schematic representation of the mTORC1 pathway.SLC7A5 forms a heterodimer with SLC3A2, which functions as a bidirectional antiporter to exchange intracellular branched-chain amino acids such as leucine.V-ATPase responds to the composition of lysosomes and, together with the KICSTOR (a negative regulator of the mTORC1 pathway), Ragulator structure, and Rheb (Ras homolog enriched in brain, ubiquitously expressed in mammalian cells) complexes, participates in the recruitment of the remaining mTORC cascade proteins.When interacting with leucine, Sestrin2 dissociates from GATOR2, which in turn stops inhibiting GATOR1 and the latter from Rag proteins.Rag proteins in the inactive state interact with the TSC (Tuberous Sclerosis Complex), inactivating Rheb.Probably, they are alternatively activated with LARS1, FLCN/FNIP and Ragulator by GAP or GEF mechanisms, and followed by interacting with Raptor and then with Rheb to trigger mTORC1 and protein synthesis.
'putative' inhibitors.Moreover, previous in silico predictions for the optimization of the inhibitor structure did not lead to success in phenotypic experimental validation.Therefore, we also decided to re-evaluate the previous in silico hypothesis about interaction site of the inhibitor with LARS1 to build more relevant models.

Structural modeling procedure
The structural modeling procedures were carried out using several databases and computational tools.The Leucyl-tRNA synthetase (UniProt ID: Q9P2J5) and LARS1 sequences were taken from the UniProt and GtRNAdb 2.0 genomic databases (Chan & Lowe, 2016;Consortium, 2021Consortium, , 2021) ) (Supplement Figures 1 and 2).The LARS1 molecule was deposited in the AlphaFold 2.0 server model database (Jumper et al., 2021), while the tRNA Leu molecule was reconstructed through homology to the tRNA Leu structures of P.horikoshii and T.thermophilus [PDBIDs: 1WZ2, and 2BTE], using 3dRNA server.The tRNA Leu structure underwent a 50 ns relaxation period (Kalvari et al., 2021;Zhang et al., 2022).
To form the required synthetase-substrate complexes, the leucine/leucyladenylate was placed in the aminoacylating site by superposition of the chosen conformation of the apoenzyme with known crystal structures [PDBIDs: 6KQY, and 6KIE] and then performing a local optimization of the geometry.In the absence of resolved synthetase-tRNA complexes, we constructed an independent model in a tRNA-bound state using molecular docking.The protein-tRNA complex was obtained by FFT docking to the surface of the catalytic domain and human LARS1 CTD (Arnautova et al., 2018).Searches were performed after pre-generating 1,000 positions and sorted according to the scores of the estimated functions.Finally, relying on data on the interaction of the aminoacylated 3 0 -end of tRNA in thermophilic bacteria and archaeal systems, we reproduced a similar model with norvaline to anchor the CCA-end of the tRNA molecule (Boyarshin et al., 2017;Rayevsky et al., 2021;Rayevsky & Tukalo, 2018).

Molecular dynamics simulation
To model long-term molecular dynamics, we utilized the Amber 99 force field and the GROMACS 2018 package, which includes data analysis tools (P� all et al., 2015).We set input parameters for temperature, pressure, and environmental atom exposures, and maintained the system values at 310 K and 1 bar to regulate temperature and pressure using the Nose-Hoover thermo-and Parinello-Rahman barostat methods (Nos� e, 1984;Parrinello & Rahman, 1981).The radius of Coulomb and Van der Waals interactions was set at 1.2 nm to balance calculation quality and speed.To ensure equilibrium in the system after each iteration of atomic motion calculation, we employed the PME (Particle mesh Ewald) algorithm to compute the long electrostatic interactions between atoms (Hess et al., 2008).We also used the Antechamber module from AmberTools to parameterize synthetase substrates and inhibitors, which resulted in excellent results (Case et al., 2018;Ilchenko et al., 2019;Rybak et al., 2019;Wang et al., 2004).
The distribution of water molecules and counterion atoms to compensate for the charge of the protein molecule was achieved by gradually replacing NVE ensembles with NVT.The MD simulations of protein structures in complex with substrates was performed for 500 ns, while the MD of the protein-tRNA complex lasted for 100 ns.An analysis of truncated protein, complying CTD and core domain, its restructuring or stability, conformational distribution, and frequency of occurrence in the MD trajectory were evaluated using the RMSD data and RMSF indices.To identify stable conformations, ready for small-molecule and protein-protein docking we used Gromacs module for conformation clustering.Based on the RMSD data, we separated the trajectory files into chunks of similar conformations with a dissimilarity threshold of 0.15 nm between this clusters.The group of atoms analyzed was different for small-molecule docking (only CTD was analyzed) and the protein-protein interaction (a catalytic domain connected to CTD were analyzed).Correlation matrices were calculated using the CorrelatonPlus module (Tekpinar et al., 2021), and changes in the position of this domain across the catalytic part of the protein were determined by aligning 3D structures in VMD software (Humphrey et al., 1996).Hydrogen bonds between the aminoacylation site and the substrate were analyzed by a built-in module of the Gromacs package.

Molecular docking
The obtained initial protein-ligand complexes with a reference compound were analyzed using 4D docking in ICM, a method that leverage the information on receptor flexibility by using an ensemble of protein conformations from a previously clustered trajectory (multiple associated receptor conformations) (Kufareva et al., 2012).A high-throughput virtual screening was conducted using a protocol of a Ligand and Receptor Complex Refinement (Bottegoni et al., 2008).The thoroughness value, which determines Monte Carlo run length, was set to 4.0.The maps were generated in a rectangular box with 0.5 Å grid spacing centered on the identified ligand-binding site.The molecules then were processed to generate the lowest energy spatial conformations before starting the grid-based docking.The resulting protein-ligand complexes were subjected to MD simulations of 200 ns.

Conformational and functional analysis of LARS1 structures
The results section describes the findings of the research on the conformational and functional analysis of LARS1 structures.This full synthetase structure is represented by four main domains: the aminoacylating domain, which contains a Rossmann fold (AD); the CP1 domain, or a connective polypeptide 1, which edits an incorrectly aminoacylated tRNA molecule; the characteristic anticodon-binding domain (ABD), responsible for the tRNA recognition; and the C-terminal (CTD) domain for tRNA and RagD binding.Additionally, there are eukaryotic leucyl-specific domains 1 (LSD1, residues 106-176) and 2 (LSD2, residues 606-659), the CP2 domain (residues 534-569), which are part of the aminoacylation domain, and the CP core and CP1 hairpin (residues 231-261, 510-535 and 570-575), which bind CP1, CP2 and AD; and the stem contact fold (SC fold, residues 708-776, red), which connects the AD and ABD domains.Considering the UNE-L domain, LARS1 contains a total of ten domains/motifs.
The three-dimensional alignment of LARS1 structures with respect to Ca atoms revealed that the position and geometry of CP1 and the catalytic synthetase domain in the complex with leucine or leucyl adenylate are almost indistinguishable (virtually the same) in the RCSB crystal structures [PDBIDs: 6LPF, 2Q3F, 6SB2, and 6EHR] ( Kim et al., 2021).The only notable differences are in the position and orientation of the CTD, as well as the electrostatic surface map generated by the Molsoft program (Figure 2).
To understand these differences and determine if the CTD positions during crystallization are a result of chance, we conducted a molecular dynamics simulation of the apoenzyme.The results then were used as the initial coordinates for further studies.Remarkably, the CTD in a thermophilic bacterium is represented by a very small domain that is extensively mobile in the absence of tRNA, that is not fixed by the methods of X-ray analysis.In addition, two sequence fragments are missing in all crystal structures, which suggests that these fragments are likely highly mobile as well.We used two models for our studies: a full model (presented in the AlphaFold 2.0 database with the completed fragments) and a truncated model (which adds 3-4 amino acid residues to each end of the sequence and stapling them together) (Figure 2A).
The results of the preliminary short 100 ns molecular dynamics were evaluated using RMSD, the number and occurrence of stable conformations detected throughout the dynamics trajectory, and a visual assessment of the full structure geometry (Figure 2).Overall, the modeled Une-L domain had no impact on the CTD position, being bound to the base of the VC-b subdomain.Also, the large loop at the catalytic site begins to shift toward the synthetase body, closing the aminoacylating site and overlapping the tRNA binding zone.One important point is that the presence of tRNA and the formation of a complex with LARS1 have been experimentally shown to prevent interaction with Rag proteins because of the likely competitive binding to the tRNA molecule.Considering that both models shared common sequences, we didn't find any notable changes (Figure 3).Therefore, we focused on the truncated protein structure without active site loop (118-155), loop from CTD (918-933) and UNE-L domain.Ultimately, based on the final state of the MD simulation, we generated a model suitable for further research on LARS1 interaction with substrates, tRNAs, and ligands.We clustered its trajectory and isolated centroid conformations from the largest clusters evenly distributed across the MD trajectory, and thus applicable for further manipulations (Figure 3).For convenience, the diagram shows only that include more than 25 repeating conformations.The folding of individual domains remained indistinguishable in the analyzed conformations from the clusters, and a major impact in cluster formation was made by interdomain motions.
Additionally, we explored which CTD areas can be shielded by the tRNA molecule, preventing the interaction of the synthetase with RagD but not hindering the binding of the Leucyl-tRNA synthetase inhibitor BC-LI-0186.The tRNA molecule after homology modeling also underwent a relaxation step by molecular dynamics of 50 ns (Supplement Figure 3).To obtain a conformation with the 3 0 -tRNA end oriented toward the CP1, we generated a tRNA with a bound 'erroneous' substrate, norvaline, to provide additional nucleic acid interaction with synthetase (Rayevsky et al., 2017;2018).At the same time, the main part of the tRNA (anticodon and variable loops) remained stable during 150 ns of molecular dynamics.In all cases, two sites on the CTD surface always remained free from interaction with tRNA (showed no overlap with tRNA) (Figure 3).
Next, the prepared protein structures with substrates or in the apoenzyme state underwent 500 ns of molecular dynamics simulation (Figure 4).Taking the most stable conformations from each trajectory, we performed a three-dimensional alignment of LARS1 these slices in the presence and absence of substrates (leucine, leucyl adenylate).
Using the same initial coordinates of the crystal structure as a reference, we observed that the outputs from these three trajectories exhibit unequal positions of the CTD at the end of the dynamics.In the absence of substrates, the CTD approaches the catalytic domain.This directional and gradual movement begins after 250 ns of MD.However, there were no clear reasons for such conformational changes.Conversely, the synthetase with leucyladenylate at the aminoacylating site remained stable, as did the leucylbound protein.Interestingly, when the clustered conformations were co-aligned, we noticed clear retention of the CTD slope angle with respect to the catalytic domain (depending on the substrate).The most stable complex with leucyladenylate had the highest CTD tilt angle, the free synthase had the lowest one, and the CTD of the synthase complex with leucine occupied the middle position.The linker peptide and 'swing helix' failed to form additional hydrogen bonds, whereas rotation of the helix itself wasn't observed either; however, the lowest mobility for this element occurred for the leucine-bound complex, whereas this index was the same for the apoenzyme and leucyladenylate-bound complexes (Figure 4).Another piece of evidence supporting the realistic tendency toward substrate-associated mobility is that the strongly associated leucine lost its position after 400 ns and moved into the aqueous solution.Then, the trajectory of CTD movement changed as well -it initially deviated at the beginning taking the position of the CTD complex with leucyladenylate, but after 450 ns, it curved in the direction of the catalytic domain, as in the case of a ligand-free synthetase.
The results of the simulation were visualized using a trajectory clustering approach that focused on the ABD/CTD fragment (Figure 4).To trace any trends in conformational transitions, these conformations were ranked based on time intervals on the RMSD plot scale of the protein structure.The number of alternate mutual dispositions for ABD and CTD, especially its LV-b subdomain, differs in all three cases.For the leucine-bound state, there was only one cluster containing stable and close conformations, while for the other two, we detected 4 and 5 large clusters.The mobility of CTD subdomains, appeared to depend on the presence or absence of the substrate in the active site.
The flexibility of domains and individual secondary structures was analyzed by means of the correlation mapping method.The results showed that the elements of the secondary structures that form the active center of the catalytic domain of LARS1, including the leucine-sensitive motifs (LSD1 and LSD2) were stable exactly in the apoenzyme and the leucine-bound form.By contrast, the mobility of the linker and the 'swing helix' amino acids increased in the leucine-bound shape and approximated to those of the adenylate-bound form.The same was true for the Anticodon Binding Domain (ABD), which indirectly confirms the assumption about the influence of the substrate on the behavior of some synthetase domains.Notably, the increased amplitude of CP1 movement relative to CTD in both substratebound forms compared to the apoenzyme levels is a significant finding.
Single, bound leucine is a relatively short-lived substrate under natural conditions that must quickly form aminoacyladenylate from ATP in a tRNA dependent or tRNA-independent scenario.This includes the formation of aminoacyladenylate, which can either be hydrolyzed or quickly transfer the amino acid to the CCA-end of the tRNA.Therefore, the most common states are the apoenzyme and the enzyme binding the cognate amino acid.However, in our case, each substrate or product of the aminoacylation reaction can influence the initiation of the cascade and the transmission of the leucine-induced signal.

Virtual screening and investigation on the identification of novel LARS1-RagD inhibitors
Experimental studies shown that a specific LARS1 area (951-971 amino acids) and the last 30 residues of the RagD segment (230-400 amino acids) are involved in the interaction, and it is driven by the GTP-bound state of RagD (Han et al., 2012).One can distinguish the 'long variable' motif LVb (or VC-a), which interacts with tRNA Leu , and the RagD-binding subdomain RBD (or VC-b) on the CTD (Figure 5).Moreover, there is an add-on domain UNE-L, which is probably involved in the interaction of LARS1 with other components of the mammalian multi-synthetase complex.Note that it was previously reported that tRNA binds to the synthetase (regardless of the presence of the inhibitor) (Kim et al., 2017).Some experiments evaluated the effects of noncognate aminoacyladenylate and BC-LI-0186 separately on the synthetase structure and demonstrated that the inhibitor stabilizes the structure without aminoacyladenylate at the aminoacylating binding site (Kim et al., 2017).However, the cooperative effect of these substrates on the formation of the synthetase-tRNA complex has yet to be reported.
The presumed binding of a highly active inhibitor BC-LI-0186 was originally thought to involve residues H958, E960, K965, D968, and K970 on the CTD surface (image ⵘ shown in Figure 6).Recent investigations provided further support for the correct choice of the binding site, novel compounds also disrupted the interaction with the D-subunit of the Rag dimer (Kim et al., 2021).However, our molecular dynamics simulations accidentally demonstrated that the proposed binding site does not possess the necessary ligand-binding features (stiffness, depth, number of donors/acceptors).Evaluation of the fluctuations in amino acid residue mobility (RMSF) of the domain, especially around mentioned residues, showed the instability of the binding site (Liu et al., 2020).MD trajectories revealed that this surface area undergoes deformation regardless of the substrate presence.Apparently, only the presence of tRNA reduces the mobility of the structural elements forming the VC-b subdomain.
Next, we performed additional clustering of both groups of atoms, a single CTD and an APB/CTD segment, through the Gromacs indexing tools.We found that there is no correlation between conformational changes of synthetase (mutual geometry of protein domains) and flexibility of a single CTD subdomain, which is quite stable in the case of apoenzyme and leucine-bound synthetase.As a result, the clusters chosen in the previous step were considered to be quite suitable for the subsequent research tasks.
To identify additional binding sites on the CTD subdomain, we analyzed the surface with the icmPocketFinder module in the software (www.molsoft.com)where the module was run against each converted ICM object.We  determined the most likely ligand interaction sites for all three complexes (leucyl-, adenylate-bound, and ligand-free) from the selected stable conformations after molecular dynamics using the values of estimation functions.Finally, molecular docking of the BC-LI-0186 molecule was performed in each of the selected binding sites (Figure 6A).
The highest value of the scoring functions demonstrated the ligand position in the cleft between LV-a and LV-b of the LARS1 in complex with leucyladenylate; the superposition of the tRNA molecule from the MD (Figure 7) shows, that there is no overlap of this binding site.The next top pose of the inhibitor is located in the same site, but in the LARS1-leucine complex.The lowest rates were shown for the inhibitor at the RBD site in the ligand-free complex, and in the aforementioned site, which was observed in the crystal structure of 6KQY (Kim et al., 2017).
Determining five potential localizations of the inhibitor binding site and possible differences in the CTD position, we performed several MD studies to estimate the probability of inhibitor dissociation.As demonstrated by the 150 ns molecular dynamics trajectories, only one complex was so strong that the ligand managed without losing all interactions, but also strengthened in this position.Herewith, the hydrogen bonds were quite stable, but the main contribution to the binding force was made by the stacking interactions.Overall, we identified three binding sites based on the analysis of the interaction forces.However, despite satisfactory energy values (arrangement of lines in the plots), only one state was stable throughout the MD and therefore of some interest for virtual screening.In addition, it is particularly noteworthy that this specific ligand interaction explicitly affects the conformation of the LV-b subdomain, as can be seen when superimposed on the crystal structures.
With the results from MD simulations, we determined a potential binding site for the known inhibitor BC-LI-0186 and formulated a theory about its most likely localization of the inhibitor.Further, to clarify the main interaction points with CTD, we showed how the most active BC-LI-0186 analogues could bind to it (5j, 7b, 8a) (Kim et al., 2021).Although the positions of the substances were similar, we observed a correlation between the values of the scoring function and the rank of corresponding activities from the biochemical experiments as shown in Figure 7A.
These data were sufficient to select a few pieces from the entire pool of conformations.The pharmacophore groups covered the aromatic systems of the substances, while hydrogen bonds could be formed with several residues, Glu904, Arg1006, Lys1056, Asp908, and Asn999 with the latter two being the most prominent (Figure 7A).The formation of the specified interactions with the protein and preservation of the diversity of the chemical space was ensured by the partial matching algorithm.According to the studied complexes with reference compounds, the inhibitors should have two aromatic ring structures in their composition and either form two hydrogen bonds with any of mentioned amino acid residues, or from one hydrogen bond and possess a hydrophobic group in the designated area (Figure 7B) Then we performed a virtual screening of Enamine chemical compounds and obtained an output of about 150 compounds as a milestone for further Hit-to-lead optimization.The detected substances have different structural properties Positions of donors, acceptors, and hydrophobic pockets were selected based on the overlay of the docking results (B).Several leader chemotypes that formed interactions similar to the analogues but had higher estimated functions (C).On the 3-dimensional plot, the structures of BC-LI-0186 and the analogues (5j, 7b, 8a) are circled and highlighted.
and form a greater number of interactions compared to the reference compounds, As seen in the increase in statistics of Pi-stacking and Pi-cation interactions, as well as the total energy and hydrogen bond contribution values.This increase can be seen both in the images with some promising compounds and in the 3x-dimensional plot denoting the position of the substances based on their interaction energies with the protein environment (Figure 6C).Along with this, we used important descriptors such as the number of rotated bonds and the values of the scoring functions represented by the color spectrum.
Finally, we conducted MD simulations on three randomly selected compounds from the hit list, shown in Figure 6, via MD simulation, like we did with BC-LI-0186 inhibitor (Supplement Figure 4).Trajectory analysis revealed that these compounds showed higher binding strength compared to BC-LI-0186 and its analogues.In addition, our hit compounds exhibited a greater impact on the CTD structure.

Conclusions
The results of this study provide evidence for a direct interaction between LARS1 and RagD, with the RagD-binding subdomain of LARS1 making contact with the GTPase domain of RagD.MD simulations used to investigate the structural basis of the interaction between LARS1 and its substrates, particularly on how the binding of different substrates (leucine, and leucyl adenylate) affects the position and mobility of the C-terminal domain (CTD) of LARS1.While carrying out our task, we tried to study in detail the behavior of LARS1 depending on the bound substrate and what structural changes can be observed in this case.For this purpose, we prepared three separate complexes.We then ran a series of prolonged simulations of the synthetase in different substrate binding states but under the same conditions.Further, we examined how the binding of tRNA to the synthetase affects the interaction with the RagD protein, which is involved in the regulation of the mTORC1 pathway.
An array of data characterizing both interdomain shifts and intradomain rearrangements was processed by molecular dynamics trajectory analysis.The obtained data were used to select the most stable, and thus energetically more profitable, conformations of CTD from the computer modeling standpoint.These protein structures were then used to search for binding sites of small molecules on the CTD surface, followed by molecular docking of the reference compound BC-LI-0186 into the detected sites.Molecular dynamics calculations were performed to estimate the dissociation probability of each complex, suggested a novel, more preferable binding site for RagD interaction inhibitors located in a cleft at the boundary of the two CTD subdomains.We also found that the presence of tRNA and the formation of a complex with LARS1 prevents interaction with Rag proteins due to the likely competitive binding to the tRNA molecule.Additionally, the presence of tRNA in the complex resulted in a further change in the orientation of the CTD and a decrease in its mobility.This suggests that the binding of tRNA to LARS1 may play a role in the regulation of its interaction with other proteins, such as RagD.
On the basis of additional data on the activity of the analogues of the reference compound, we developed a pharmacophore model that showed a successful result during a high-throughput screening of the Enamine chemical database.Thus, the goal of the study was to conduct an in silico search for new potential inhibitors of the mTOR signaling cascade, which is a target for the development of new cancer therapies.This goal has now has been achieved, and the project now can be further extended with the investigation of their biological activity.

Figure 2 .
Figure 2. The verified models excluded the presence of the Une-L domain but contained labile, unstructured loops of different lengths at the points indicated by the spheres (A).The electrostatic maps of the two crystallographic structures, with an alternative position of the CTD domain, showed little difference in the charge distribution on the surface of this domain (B).These differences in orientation are likely due to the variations in hydrogen bonding network and the slope of the 'swing helix' (C).

Figure 3 .
Figure 3. Graphics of RMSD and RMSF (A and B, respectively) showed the common trend with respect to the stability of the truncated form (red color) and fulllength (dark blue color) of LARS1.The conformations of the initial (dark green) and final (light green) loop disposition and the trimmed form (black) of synthase without Une-L (C).The resulting distribution of structures in conformational clusters (D).Multiple alignments of synthetase and tRNA complexes corresponding to different slices of the MD simulation (E).

Figure 4 .
Figure 4. Distribution of stable ABD/CTD conformations (A) over 500 ns corresponds to the time scale of the RMSD plot (C).The color scheme for the graphs reflects the conditions: apoenzyme (gray), leucyl-bound (green), and leucyladenylate-bound (blue).The reference molecule 6LR6 (human leucyl-tRNA synthetase) is shown with red ribbons, and the studied complex is colored according to the color scheme.Correlation maps for synthetase without substrate (I), with leucine (II), and with leucyladenylate (III) demonstrate the mobility of the domains of each complex (B), and the RMSF plot shows changes in the corresponding structural elements (D).

Figure 5 .
Figure 5. Structural details of human LARS1.CTD (yellow) connected to the catalytic domain (green) with a 35 Å long 'swing helix' a-helix (blue) and linker peptide (pink), and the editing domain (CP1) with a b-fold structure (red).Putative site of interaction with RagD and inhibitor on RBD surface.

Figure 6 .
Figure 6.The most favorable positions of BC-LI-0186 at putative binding sites without overlapping with the tRNA molecule on the schematic surface of LARS1 and the corresponding values of the scoring functions for the poses (A).For the positions that remained in contact with the CTD, the effect on the structure of the CTD at the end of the MD is shown with ribbons (A).MD simulation analysis for each of the complexes after 200 ns are represented as a graph of the electrostatic (blue) and van der Waals (red) interaction energies (B).

Figure 7 .
Figure7.The pharmacophore model and screening results.Docking results of the published analogues in the novel binding site (A).Positions of donors, acceptors, and hydrophobic pockets were selected based on the overlay of the docking results (B).Several leader chemotypes that formed interactions similar to the analogues but had higher estimated functions (C).On the 3-dimensional plot, the structures of BC-LI-0186 and the analogues (5j, 7b, 8a) are circled and highlighted.