Skip to main content
NCBI home page
Current Research in Structural Biology logoLink to Current Research in Structural Biology
. 2023 Feb 18;5:100097. doi: 10.1016/j.crstbi.2023.100097

Structural modelling and dynamics of full-length of TLR10 sheds light on possible modes of dimerization, ligand binding and mechanism of action

Vikas Tiwari 1, R Sowdhamini 1,∗
PMCID: PMC9996232  PMID: 36911652

Abstract

Toll like receptors (TLRs) play a pivotal role in innate and adaptive immunity. There are 10 TLRs in the human genome, of which TLR10 is the least characterized. Genetic polymorphism of TLR10 has been shown to be associated with multiple diseases including tuberculosis and rheumatoid arthritis. TLR10 consists of an extracellular domain (ECD), a single-pass transmembrane (TM) helix and intracellular TIR (Toll/Interleukin-1 receptor) domain. ECD is employed for ligand recognition and the intracellular domain interacts with other TIR domain-containing adapter proteins for signal transduction. Experimental structure of ECD or TM domain is not available for TLR10. In this study, we have modelled multiple forms of TLR10-ECD dimers, such as closed and open forms, starting from available structures of homologues. Subsequently, multiple full-length TLR10 homodimer models were generated by utilizing homology modelling and protein-protein docking. The dynamics of these models in membrane-aqueous environment revealed the global motion of ECD and TIR domain towards membrane bilayer. The TIR domain residues exhibited high root mean square fluctuation compared to ECD. The ‘closed form’ model was observed to be energetically more favorable than ‘open form’ model. The evaluation of persistent interchain interactions, along with their conservation score, unveiled critical residues for each model. Further, the binding of dsRNA to TLR10 was modelled by defined and blind docking approaches. Differential binding of dsRNA to the protomers of TLR10 was observed upon simulation that could provide clues on ligand disassociation. Dynamic network analysis revealed that the ‘open form’ model can be the functional form while ‘closed form’ model can be the apo form of TLR10.

Graphical abstract

Image 1

Open in a new tab

Highlights

  • •

    Full-length model of TLR10 is generated using a series of homology modeling and protein-protein docking exercises.

  • •

    Different forms of dimers, such as closed, semi-closed, semi-open and open forms have been modelled.

  • •

    The closed form of TLR10 dimer is observed to be energetically more favorable than the open form.

  • •

    Docking of double-stranded RNA and dynamics of the different forms provide clues on possible modes of ligand disassociation.

  • •

    Dynamic residue network analyses suggest that closed form, albeit energetically more stable, might be the apo form of TLR10.

1. Introduction

Toll like receptors (TLRs) play an important role in innate and adaptive immunity. These are type I transmembrane proteins with extracellular multiple LRR (Leucine rich repeat) motif containing domain (ECD) used for ligand recognition, a single transmembrane (TM) helix and a C-terminal TIR (Toll/Interleukin-1 receptor) domain which interacts with similar TIR domain containing adapter proteins for signal transduction. The TLRs can be grouped based on their localization to plasma membrane or endosome. TLR1, 2, 4, 5, 6 and 10 are plasma membrane TLRs, while TLR3, 7, 8 and 9 are endosomal TLRs (Gay et al., 2014; Kawai and Akira, 2010; Moresco et al., 2011; Su et al., 2021). Plasma membrane TLRs recognize wide variety of pathogen-associated molecular patterns (PAMPs) (TLR1-TLR2: Triacylated lipoprotein; TLR2-TLR6: Diacylated lipoprotein; TLR4: Lipopolysaccharide; TLR5: Flagellin) while the endosomal TLRs recognize the nucleic acids (TLR3: dsRNA; TLR7, TLR8: ssRNA; TLR9: dsDNA) (Akira et al., 2006; Kawasaki and Kawai, 2014; Rout et al., 2021). Upon binding of cognate ligands, TLRs homo or hetero-dimerize, followed by interactions with the adaptor proteins. The major adapter proteins include MyD88 (myeloid differentiation primary response protein 88), TRIF (TIR domain-containing adaptor protein inducing IFNβ), MAL (MYD88 adaptor-like protein) and TRAM (TRIF-related adaptor molecule) (Luo et al., 2019; O'Neill and Bowie, 2007). Except TLR3, all other TLRs interact with MyD88, while TLR3 interacts directly with TRIF. TLR4 interacts with all four adapter proteins. TLR4 interacts with MyD88 through MAL at plasma membrane and with TRIF through TRAM upon internalization (Kagan et al., 2008; Yamamoto et al., 2003).

In humans, there are 10 members of TLRs (TLR1-TLR10). TLR10 is the least understood human TLR and it is a pseudogene in mice (Fore et al., 2020; Hasan et al., 2005; Oosting et al., 2014). It belongs to TLR1 subfamily of TLRs, which includes TLR1, 2 and 6. As expected, it shares high sequence identity with TLR1 and TLR6 (Chuang and Ulevitch, 2001; Roach et al., 2005). TLR10 is expressed in a variety of immune cells, with its highest expression in B-cells and no detectable expression in T-cells (Hornung et al., 2002; Su et al., 2021). TLR10 can form homodimer as well as heterodimerize with TLR1, TLR2 and TLR6 (Hasan et al., 2005). Various ligands have been proposed for TLR10 including TLR2 ligands (FSL-1, Pam3Cys and Borrelia burgdorferi), lipopolysaccharide and HIV-gp41 (Henrick et al., 2019; Nagashima et al., 2015; Oosting et al., 2014; Su et al., 2021). Recently, TLR10 has been shown to localize to endosome and bind to double stranded RNA (dsRNA) resulting in the recruitment of MyD88 (Lee et al., 2018). TLR10 genetic polymorphisms are associated with several diseases like rheumatoid arthritis (Torices et al., 2016), tuberculosis (Bulat-Kardum et al., 2015), post bronchiolitis asthma (Törmänen et al., 2018), Crohn's disease (Morgan et al., 2012), chronic gastritis (Tongtawee et al., 2017), Hashimoto's disease (Li et al., 2019), complicated skin and skin structure infections (Stappers et al., 2015), Nasopharyngeal carcinoma (Zhou et al., 2006), papillary thyroid carcinoma (Kim et al., 2012), urothelial cancer (Guirado et al., 2012) and Crimean Congo hemorrhagic fever disease (Kızıldağ et al., 2018).

The experimental structure of extracellular domain of all human TLRs is available except for TLR10. In contrast, the full-length structure has not been solved for any of the TLRs (Zhang et al., 2017). Computational modelling has been employed to study the full-length structure and dynamics of TLR3 and TLR4 recently (Patra et al., 2020, 2018). In case of TLR10, only the crystal structure of the intracellular TIR domain (dimer) is available (Nyman et al., 2008). There have been computational efforts to model the TLR10-ECD monomer and dimer models in order to predict the dimerization region and possible ligands (Govindaraj et al., 2010). The heterodimer of TLR2-TLR10 ECD has also been modelled using TLR1-TLR2 complex (PDB: 2Z7X) as template (Guan et al., 2010). The lack of full-length structure of TLR10 obscures the study of domain organization and dynamics of individual TLR10 domains. Also, recent availability of homodimer structure of ECD of TLR1 and TLR2 which belong to the same subfamily as TLR10, provides an avenue to explore the TLR10-ECD dimer modelling in a template based manner (Su et al., 2019). These structures also point to the possibility of capturing TLRs in different dimer orientations – namely the closed, half-closed, half-open and open forms. There is little or no information available on the functional relevance and structural transitions of these states.

We have modelled the TLR10-ECD dimer in different orientations based on template structures of TLR1 homodimer, TLR2 homodimer, TLR1-TLR2 heterodimer and TLR3 homodimer corresponding to ‘closed-model, ‘semi-open-model, ‘semi-closed-model and ‘open-model’ respectively. The different TLR10-ECD models were combined with modelled TM helix dimer and known TIR dimer structures to give different full-length structure models of TLR10. Dynamic study of these full-length models in membrane environment provides key interacting residues in different models and overall domain dynamics with respect to the membrane. Some of the full-length models were further checked for their interaction with dsRNA owing to recent evidences of TLR10 binding to dsRNA. The full-length model of TLR10, along with key residues required for dimer formation, can be utilized for ligand binding site prediction and drug development in order to modulate TLR10 function.

2. Materials and methods

2.1. Molecular modelling of full-length TLR10

The full-length sequence was retrieved from Uniprot (ID: Q9BXR5). The monomeric TLR10-ECD structure was modelled using MODELLER (Šali and Blundell, 1993) and TLR1-ECD (PDB: 2z7xB) was used as a template structure (Jin et al., 2007). The sequence identity between the template (TLR1-ECD) and TLR10-ECD is 41.3% and BLASTp was used to search against PDB.

For monomeric TLR10-TM modelling, the secondary structure was predicted for extended TM helix region (residues 577–621) using PSIPRED (McGuffin et al., 2000) followed by modelling the sequence as alpha-helix using Maestro, Schrodinger (Maestro, Schrödinger, LLC, New York, NY). The modelled TM-helix was minimized, using chloroform as implicit solvent and macromodel minimization of Schrodinger. Two residues within the extended region were predicted to have non-helical propensity and hence the minimized TM helix was used as template to model the discontinuous TM helix using MODELLER. The final TM monomer was considered for protein-protein docking using HDOCK to generate TM dimer model (Yan et al., 2017). The best dimer model was selected based upon the orientation of helices and energetics as calculated using PPCheck (Sukhwal and Sowdhamini, 2015). For TLR10-TIR, the crystal structure is available and was used as template for modelling (PDB: 2j67) as there are missing residues in the crystal structure (Nyman et al., 2008). The best model as per the DOPE score was considered for full-length modelling.

Multiple TLR10-ECD dimers were generated because of the difference in the dimerization of relevant templates and in the absence of experimental data. The ECD dimers were built by superposition of monomeric TLR10-ECD on different dimer templates. The dimer templates used for modelling include human TLR1 homodimer (PDB: 6NIH) (Su et al., 2019), human TLR2 homodimer (PDB: 6NIG) (Su et al., 2019), human TLR2-TLR1 heterodimer (PDB: 2Z7X) (Jin et al., 2007) and mouse TLR3 homodimer (PDB: 3CIY) (Liu et al., 2008). The sequence identity between human and mouse TLR3 as assessed by EMBOSS Needle is 79.1% while the similarity is 87.4% (Madeira et al., 2022). Four TLR10-ECD dimers were generated using these four templates.

The TM-dimer and TIR-dimer were aligned to different ECD-dimers to obtain different orientations of TLR10 full-length dimer. The individual domains were geometrically fixed in a straight line by aligning individual domains. The full-length model was built starting from the individual aligned domains using MODELLER. Finally, there are four different TLR10 full-length models (‘closed-model’ (based on TLR1 homodimer), ‘semi-open-model’ (based on TLR2 homodimer), ‘semi-closed-model’ (based on TLR2-TLR1 heterodimer) and ‘open-model’ (based on TLR3 homodimer)).

2.2. Model refinement and assessment

The C-terminal tail region, after TIR domain, was truncated and the final models were rebuilt using MODELLER. PROCHECK was used to assess the final full-length model (Laskowski et al., 1993). Further, loop refinement was done using ‘ModLoop’ (Fiser and Sali, 2003). The resulting model were assessed using ProSA-web and PPCheck (Sukhwal and Sowdhamini, 2015; Wiederstein and Sippl, 2007). The interactions between protomers were checked using PDBsum for each model (Laskowski et al., 2018) (Supplementary text for closed, semi-open, semi-closed and open models). The effect of mutations were studied using FoldX (Schymkowitz et al., 2005). Binding free energy was calculated using MMGBSA tool of Prime module of Schrodinger.

2.3. Docking of dsRNA to TLR10

Open-model (Fig. 2) was used for docking of dsRNA which was retrieved from the complex of TLR3-dsRNA (PDB: 3CIY). The possible interacting residues between RNA and TLR10 were obtained by superimposing the TLR10-ECD dimer to TLR3-ECD dimer. These residues were given as constraints for docking using HDOCK and HADDOCK (Dominguez et al., 2003; Van Zundert et al., 2016; Yan et al., 2017). The top models for TLR10-dsRNA complex were assessed for compliance of residue constraints. For the top models, binding energy (ΔG) was calculated using Prime-MMGBSA module of Schrodinger. Later the blind dsRNA docking was also carried out using closed-model and open-model without specifying any residues using HDOCK. Such models are referred as “closed-model-dsRNA-blind” and “open-model-dsRNA-blind”. The ECD of docked complexes were protonated at acidic pH (pH ​= ​5.5) using protein preparation wizard module of Schrodinger. The interactions between protein and dsRNA were checked using PLIP (Adasme et al., 2021) after protein preparation.

Fig. 2.

Fig. 2

Open in a new tab

Full-length TLR10 dimer models (A) closed-model (B) Interactions between protomer A (Green) and protomer B (Cyan) of closed-model. Select H-bonds have been shown as yellow dashes (C) semi-open-model (D) semi-closed-model (E) open-model. Color codes denotes the two chains and are maintained in all cases. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

2.4. Molecular dynamics simulation

The full-length dimer of TLR10 was used for molecular dynamics (MD) simulation using Desmond package of Schrodinger (Bowers et al., 2006). The model was initially split into ECD ​+ ​TM and TIR. In order to mimic the physiological state in the cell, ECD ​+ ​TM parts were prepared at 5.5 pH while the TIR domain was prepared at a pH of 7.2 using protein preparation wizard (Madhavi Sastry et al., 2013). Residues 579–599 were specified as transmembrane residues for the membrane (POPC) placement. The transmembrane residues were predicted using PPM web server, TOPCONS and CCTOP (Dobson et al., 2015; Lomize et al., 2012; Tsirigos et al., 2015). In system builder, TIP3P water model was selected for solvation of the system along with POPC lipid bilayer. The membrane position was adjusted manually. System was neutralized and ionic concentration of 0.15 ​mM using NaCl was specified. The solvated system was relaxed using default relaxation protocol followed by production MD run for 200 ​ns in NPgT ensemble at 300K, 1 ​atm pressure and 0.0 ​bar ​Å surface tension. This default relaxation protocol involved following steps (1) 50 ps simulation in NVT ensemble at 10 ​K temperature with restraints on solute heavy atoms (2) 20 ps simulation in NPT ensemble at 100 ​K with membrane restrained in Z-direction and protein restrained (3) 100 ps simulation in NPgT ensemble at 100 ​K with membrane restrained in Z-direction and protein restrained (4) 150 ps simulation in NPgT ensemble with gradual temperature increase from 100K to 300K and gradual release of restraints (5) 50 ps simulation in NVT ensemble at 300K with restraints on backbone or ligand (6) 50 ps simulation with no restraints in NVT ensemble at 300K. For replicate runs, random seeds were used.

For TLR10-dsRNA complexes, the dsRNA bound ECD was protonated at acidic pH (pH ​= ​5.5) while the TIR domain was prepared at pH 7.2 in the protein preparation step followed by minimization and MD simulation.

2.5. MD simulation analysis

The simulation results were analyzed using simulation interaction diagram and simulation event analysis tools of Desmond. For protein-protein interactions, trajectory analysis script ‘analyze_trajectory_ppi.py’ was used. For Network analysis, ‘Networkview’ plugin of VMD was used and suboptimal paths were calculated with edge length offset of 1 ​Å (Eargle and Luthey-Schulten, 2012). Hydrophobic interactions and interactions between protein-dsRNA were calculated using ‘Pycontact’ (Scheurer et al., 2018). Figures were generated using custom python scripts, Pymol and Maestro.

2.6. Conservation analysis

The orthologs of TLR10 were obtained from NCBI (383 sequences). CD-hit was used to remove identical sequences at a cutoff of 100% sequence identity giving rise to final dataset of 163 sequences (Fu et al., 2012). These sequences were aligned using MUSCLE (Madeira et al., 2022). Conservation analysis was performed using ConSurf and this alignment (Ashkenazy et al., 2016).

Residue numbering is the three letter code, followed by residue number and protomer chain identifier. For example, Asn416A refers to Asn residue at 416 position of ‘A’ chain. Where the chain identifier is not explicitly mentioned, it means that residue from both protomers is under question. Likewise, particular mutations are referred by the single letter code in the wild type, residue number followed by single letter code of the amino acid to which it is mutated (for example, M326T refers to Methionine at 326 mutated to Threonine).

3. Results

3.1. Full-length structural models of TLR10 reveal different dimerization interface and key residues

In case of TLR10, there is no structure available for ECD or TM domain. Multiple TLR10-ECD models were built using homology modelling (Fig. 1A). For TLR10-TM, monomer models with kink (predicted disruption in helix) were generated. The top 10 models were inspected for kinked nature of helix. The angle was measured between CA atom of 1st residue, residue at kink and the last residue (Supplementary Table 1).

Fig. 1.

Fig. 1

Open in a new tab

Modelling of TLR10 individual domains. (A) Different models of TLR10-ECD dimers based on different dimer templates (B) Modelled Monomeric TM helix and TM helix dimer obtained by protein-protein docking (C) Modelled TIR dimer.

Model 7 was observed to be one of the good models as per ranking based on MODELLER DOPE score and hence was used to generate dimer model of TLR10-TM (Fig. 1B). For dimeric TLR10-TM, top 100 models were screened from HDOCK for parallel orientation and energetics. Model 16 was found to be the best model with −88.69 ​kJ/mol of total stabilizing energy and −1.21 ​kJ/mol of normalized energy per residue but it had positive electrostatic energy (+4.06 ​kJ/mol) while second best model was model 4 with −60.75 ​kJ/mol of total stabilizing energy, −0.74 ​kJ/mol of normalized energy per residue and −7.16 ​kJ/mol H-bond energy. It had second best electrostatic energy (−12.57 ​kJ/mol) therefore model 4 was considered for full-length modelling (Fig. 1B).

Crystal structure of cytoplasmic TIR domain alone of TLR10 is available as a dimer. This has been used as template to model the dimeric TIR as there are missing residues in the crystal structure (Fig. 1C). The dimeric TLR10-TM and TLR10-TIR were rotated and translated with respect to TLR10-ECD dimer. The aligned models were used as template to build the full-length dimer of TLR10 using MODELLER.

Finally, there are four different TLR10 models (closed-model, semi-open-model, semi-closed-model and open-model [V-shaped]). Structure validation was performed for these models by examining their backbone torsion angles and projecting them on the Ramachandran map. For closed-model, five residues (Asn416A, Lys629A, Asn416B, Glu541B, Thr623B) were in disallowed region of Ramachandran plot and after loop refinement there is one residue Thr623B in the disallowed region. For all other models, there are no residues in the disallowed region of Ramachandran map after loop refinement. These refined models were considered for further analysis (Fig. 2). The energetics of these models were checked using PPCheck after preparing the models using protein preparation wizard of Schrodinger. Closed-model was found to have better total energy (−1057.44 ​kJ/mol) and this could be attributed to the elaborate protein-protein interaction surface. Open-model is the last ranked model in terms of total energy (Supplementary Table 2).

Further, the binding affinity was calculated between the protomers of full-length dimer using MM-GBSA and ΔG of binding was found to corroborate with the PPCheck results. The ΔG of closed-model is −229.31 ​kcal/mol while that of open-model is −62.88 ​kcal/mol supporting that closed-model is more energetically favorable for TLR10 homodimer (Supplementary Table 3). Hotspot residues were predicted for each of the five models using SpotOn and PPCheck (Moreira et al., 2017; Sukhwal and Sowdhamini, 2015). Hotspot residues are interface residues, which upon mutation to Alanine, leads to binding free energy difference (ΔΔG) of more than or equal to 2 ​kcal/mol and are critical for stability of that protein-protein interaction. Significantly, both the tools predicted His707 (TIR domain) to be hotspot residue for all the five models. For ECD, different residues were predicted to be hotspot for each model (Supplementary Table 4). Glu26B and Glu27B were predicted to be hotspot for closed-model by both tools. Unique hotspot residues from ECD for each model (Glu26B for closed-model; Lys366B for semi-closed-model; Gln315A, Gln315B for semi-open-model and Gln509A for open-model) were identified (Supplementary Table 4). Our TLR10 full-length model was compared with the model of TLR10 generated by AlphaFold (https://alphafold.ebi.ac.uk/entry/Q9BXR5) (Supplementary Fig. 1A). It was observed that, the overall domain organization is different even though individual domains (ECD and TIR) are more similar (Supplementary Figs. 1B and 1C). The RMSD of ECD, TM and TIR are 2.070 ​Å, 5.852 ​Å and 0.812 ​Å respectively, between the two models. The superposition of AlphaFold model and the monomer of closed-model with respect to TM helix (RMSD ​= ​0.519 ​Å) leads to several clashes between the AlphaFold model and the membrane (Supplementary Fig. 1D). The multimer modelling was performed using AlphaFold2 Colab notebook. The top five dimer models were generated for ECD-TM domain (residues 20–604). Among these, only model 5 was observed to have satisfactory orientation of protomers and it resembles the open-model (Supplementary Fig. 2).

There are various mutations reported in TLR10 and have been shown to be associated with several diseases. These disease mutations were mapped on the surface of all the models and it was observed that only in case of closed-model, some of the disease variants are located near the dimer interface (Fig. 3).

Fig. 3.

Fig. 3

Open in a new tab

Mapping of disease (Red) and natural (Green) variants on the surface of TLR10 models. (A-D) TLR10 Models (E) FoldX results for the TLR10 variants. Positive “ΔΔG” indicates destabilization (F-G) Microenvironment of residue Met326 and its mutant Thr326 from chain A of closed-model (H–I) Microenvironment of residue Ile473 and its mutant Thr473 from chain A of closed model. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

For closed-model, the effect of these variants on the structural stability was assessed using FoldX and the variants M326T and I473T were found be highly destabilizing (Fig. 3E). The natural variants were also found to be destabilizing or neutral with L167P and G381D being the highly destabilizing while V298I being the neutral variant. Variants M326T and I473T are destabilizing in case of other TLR10 models as well (Supplementary Fig. 3). M326T and I473T were analyzed structurally and the wild type variant of both the residues were observed to be interacting with many hydrophobic residues. Therefore, mutation of these hydrophobic wild type residues to a polar residue (Threonine) leads to structural destabilization (Fig. 3F–I). Closed-model was used to predict possible binding sites using sitemap tool of Schrodinger and it was found that one of the top five sites has large volume and it is at the dimer interface (Supplementary Fig. 4). One of the disease-associated variants (N241H) is part of this site. This site also has good DScore indicating the site to be druggable and hence can be used for drug design.

3.2. TLR10 closed-model shows global movement of ECD and TIR domains across simulation time

The closed-model of full-length TIR10 model was subjected to all-atom MD simulation in aqueous-membrane (POPC) environment for 200 ​ns in three replicates to assess the interaction stability and dynamics of TLR10 dimer. Upon simulation, the ECD and TIR domains bend towards the membrane which can be attributed to the electrostatic interactions between the membrane and ECD or TIR (Fig. 4A and E). Similar movement of ECD towards membrane bilayer has been shown for in-silico full-length models of TLR3 and TLR4. The root mean square deviation (RMSD) of each frame was calculated with 0th frame as reference and it was found that RMSD remains stable after 100 ns for replicate 1 and 3 (Fig. 5A). The root mean square fluctuations (RMSF) was calculated for each residue with 0th frame as reference and the TIR domain was found to be more flexible than the ECD domain in all three replicates of closed-model (Fig. 5B). Further to examine the association of two protomers, radius of gyration (ROG) was monitored. It was observed that the protomers remain associated throughout the simulation time as indicated by ROG (Supplementary Fig. 5).

Fig. 4.

Fig. 4

Open in a new tab

TLR10 models (A, B, C, D) Models at 0th ns of simulation (E, F, G, H) Models at 200th ns of simulation. Replicate 1 of closed-model is shown. Pink dots represent the membrane bilayer. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 5.

Fig. 5

Open in a new tab

RMSD and RMSF of closed-model (A) RMSD of three replicates of closed-model (B) RMSF of three replicates of closed-model. r1: Replicate 1; r2: Replicate 2; r3: Replicate 3.

ROG was calculated for ECD and TIR domains separately as well and TIR domain remains more compact throughout the simulation time compared to ECD (Supplementary Fig. 5). The persistence of different types of non-covalent interactions between the protomers were monitored as the fraction of simulation time. Residues involved in the interactions that persist for more than 40% of simulation time were identified. For closed-model replicate1, Tyr668A-Lys622B form Cation-pi interaction for more than 45% of simulation time (Fig. 6A). Many residues are involved in H-bond interactions and the most prominent H-bond interactions involve residues Tyr668A-Val620B and Pro22A-Asn330B as these interactions are maintained for more than 60% of simulation time. Residues Glu642A-Lys622B, Arg120A-Glu272B, Arg741A-Glu709B, Glu27A-Arg300B and Glu704A-Lys622B form salt-bridge interaction for more than 90% of simulation time (Fig. 6A). There are many hydrophobic interactions and the most prominent interactions that last for more than 90% of simulation time have been summarized in Fig. 6B. The closed-model can be considered stable owing to these interactions. Further, the conservation scores were mapped on these interactions and some of the residues involved in these interactions (like Asn330 and Leu569) are also conserved across TLR10 orthologs.

Fig. 6.

Fig. 6

Open in a new tab

Non-covalent interactions between protomers of closed-model. The interactions that persist for more than 40% of simulation time have been summarized. Numbers in bracket indicate the ConSurf color grades (1–3: Variable; 4–6: Average; 7–9: Conserved). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.3. Dynamics of other TLR10 models recapitulate the domain motion with few persistent interactions

Similar to closed-model, other TLR10 models were simulated for 200 ns. Semi-open-model, semi-closed-model and open-model show similar ECD and TIR domains movement towards POPC membrane (Fig. 4). Initial RMSD of these models were high compared to closed-model but it plateaued after 150 ns (Supplementary Fig. 6). The RMSF analysis indicates high RMSF of the TIR domain residues compared to ECD or TM domains. Open-model showed highest RMSF compared to other models while closed-model shows the least. (Supplementary Fig. 7). Further, chain B of open-model has highest RMSF values for TIR domain. ROG analysis indicates that open-model has high fluctuations in ROG values throughout simulation time. The TIR domain of all the models remain compact while the ECD domain of open-model has high ROG values compared to other models (Supplementary Fig. 8).

Interactions between protomers for other models indicate that there are a smaller number of persistent interactions compared to closed-model (Supplementary Fig. 9). In case of semi-closed-model, 4 residues Ser677A-Glu704B, Glu704A-Ser679B, Ser670A-Val620B and Ile678A-Glu704B retain H-bond interaction for more than 60% of simulation time. There are no pi-cat or pi-pi interactions for more than 40% of simulation time. There are only three residue pairs (Glu363A-Arg310B, Arg741A-Glu709B and Glu680A-Arg741B) with salt-bridge interaction for more than 90% of simulation time. In case of semi-open-model, there are four H-bond interactions and four salt bridge interactions that last for more than 70% of simulation time, while there are no durable cation-pi or pi-pi interactions. For open-model, there is one pi-pi interaction (Phe672A-Phe672B), three salt-bridge and four H-bond interactions maintained for more than 40% simulation time (Supplementary Fig. 9). Similar to closed-model, there are many hydrophobic interactions between the protomers in all the models (Supplementary Fig. 10). Based on these interactions along with their conservation, the ECD of each model can be ascribed a key residue such that its interaction persists for more than 40% and ConSurf grade is more than 6. For closed-model, these residues are Asn330 and Leu569. Similarly, Pro339 and Pro556 are example of key residue for semi-closed-model and open-model respectively. For semi-open-model, there are no such residues from ECD domain. Further, ΔG was calculated between two protomers for all the dimers across simulation time and closed-model was found to have best binding energy (Supplementary Figs. 11 and 12). Through these dynamic studies, it can be suggested that the closed-model is more favorable TLR10 model than other models.

3.4. His83 and His308 of TLR10 interact with dsRNA upon defined docking

It has been reported that TLR10 can bind to dsRNA which is a cognate ligand for TLR3 (Lee et al., 2018). Initially open-model was used for docking with dsRNA as it was modelled using TLR3 as template. Defined docking was performed by specifying possible dsRNA binding residues obtained upon superposition of ECD dimer of open-model to TLR3-ECD dimer in complex with dsRNA. Top 10 models from HDOCK and top 4 models from top cluster of HADDOCK were inspected manually. Model 1 from HADDOCK (cluster1_1) was found to better satisfy the defined residue constraint compared to other models (Supplementary Table 5).

In case of TLR3, there are four histidine residues from each protomer that interact with dsRNA. Open-model-dsRNA complex involves two histidine residues (His 81 and His308) from each protomer in the vicinity of dsRNA (Supplementary Fig. 13). Residues Asn36, Ser38, Arg40, Tyr57, Arg83, Gln85, His308, Thr361, Glu385, Gln410 and Lys412 from Chain A form H-bond interactions with dsRNA while residues Tyr57, Arg83, Gln85, Arg107, His308 and Gln410 from chain B are involved in H-bond interaction with dsRNA. His81, Arg310, Lys383 and Lys432 of chain A form salt-bridge interaction with the phosphate of dsRNA. Salt bridge interactions from chain B include residues Arg40, Lys383 and Lys432.

3.5. Defined TLR10-dsRNA complex preferentially interacts with chain B of TLR10

The open-model-dsRNA complex was subjected to 200 ns MD simulation to assess the interaction stability of dsRNA with TLR10. Open-model-dsRNA complex shows similar behavior of ECD and TIR domain tilting towards the membrane (Fig. 7A and B). The protein RMSD indicates that the model attains stable conformation after around 100 ns (Fig. 7C). In case of RMSF calculations, it recapitulates the similar pattern of other models having high RMSF for TIR domain of chain B (Fig. 7D). The ROG calculations indicate that the ROG of Open model-dsRNA is less (average ROG ​= ​61.169 ​Å) compared to open-model without dsRNA (average ROG ​= ​64.856 ​Å) indicating that the open-model becomes more compact upon binding of dsRNA. The overall ROG of the model remains similar throughout simulation time indicating that the protomers do not dissociate (Fig. 7E).

Fig. 7.

Fig. 7

Open in a new tab

MD simulation results of open-model-dsRNA complex (A, B) Open-model-dsRNA complex at 0th and 200th ns (C) RMSD with 0th frame as reference (D) RMSF of chain A and chain B (E) ROG of entire dimer (top panel), ECD (middle panel) and TIR (bottom panel).

Further, H-bond interactions were monitored between dsRNA and TLR10. There is at least one H-bond between dsRNA and TLR10 throughout the simulation time (Supplementary Fig. 14A). However, one of the protomers (chain A) was observed to lose the H-bond interaction with dsRNA at multiple instances of simulation time (Supplementary Fig. 14B), while the other protomer (chain B) always maintains H-bond interaction (Supplementary Fig. 14C).

The number of all interactions between TLR10 and dsRNA were monitored and it was observed that salt-bridge interactions also persist for majority of the simulation period, while the Pi-cation and pi-pi interactions are very transient (Fig. 8A). His81B retains interaction for more than 60% of simulation time and it is conserved among the TLR10 orthologs. All of the persistent H-bond interactions are made by residues of chain B that include Arg83, Arg40, Arg107, Gln85, His81 and Arg107 (Fig. 8B). The key residues for the interaction with dsRNA are Gln85 and His81.

Fig. 8.

Fig. 8

Open in a new tab

Interactions between open-model and dsRNA (A) All Interactions between TLR10 and dsRNA (B) Residues involved in H-bond interactions with dsRNA. Numbers in bracket indicate the ConSurf color grades (1–3: Variable; 4–6: Average; 7–9: Conserved). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.6. TLR10 models form stable complex with dsRNA upon blind docking

To explore the possibility of other binding modes of dsRNA to TLR10, blind docking was performed (without specifying any residues) for both closed-model and open-model of dsRNA. Top 100 docked poses were evaluated after protein-preparation in terms of number of interactions between dsRNA and TLR10. In case of closed-model, the pose which maintained maximum interactions (20 ​H-bond, 5 salt-bridge, 6 Aromatic-H-bond and 1 Pi-cation calculated using interaction count of Maestro, Schrodinger) was retained. Similarly, in case of open-model, the best pose of bound RNA was found to have the highest number of contacts with TLR10 (14 ​H-bonds, 8 salt-bridges and 5 Aromatic-H-bonds). In both of these blind docked complexes, His81 was found to be interacting with the dsRNA. These two complexes were subjected to MD simulation. Simulation results indicated similar behavior for tilting of ECD domains (please see Supplementary Fig. 15 and Supplementary Fig. 16). The RMSD stabilized at relatively higher value than open-model-dsRNA complex and the ECD residues of chain B were found to have higher RMSF than corresponding TIR domain (Supplementary Fig. 15). In case of closed-model-dsRNA-blind complex, the RMSD stabilized at lower values than any of the open-model-dsRNA complexes (defined or blind) and the RMSF of TIR domains were found to be very high compared to the ECD domain (Supplementary Fig. 16).

All the interactions between protein and dsRNA were monitored across simulation time and it was observed that H-bond and salt-bridge interactions are maintained throughout the simulation time for open-model-dsRNA-blind complex, while for closed-model-dsRNA-blind complex, only the H-bond interaction is maintained at all time frames. Closed-model-dsRNA-blind has few transient cation-pi interactions compared to open-model-dsRNA-blind complex (Supplementary Figs. 17 and 18). Both the blind dock models (open-model-dsRNA-blind and closed-model-dsRNA-blind) have a greater number of persistent H-bond interactions than defined open-model-dsRNA complex and it includes residues from both protomers. The key residues are Ser56, Asp54, and Ser35 for closed-model-dsRNA-blind, while for open-model-dsRNA-blind, key residues are Ser436, Asn58, His81, Gln410, Leu559 and Gln85 (Supplementary Fig. 19).

3.7. Network analysis identifies open-model to be the functional form of TLR10 for dsRNA binding

In order to identify the critical residues for signal propagation and the optimal paths between two residues, dynamic network was constructed. In this network, nodes represent Cα atoms and the edge is drawn between two residues if they remain in contact (within 4.5 ​Å) for more than 75% of simulation time (Fig. 9A). The dynamic network was partitioned into many communities so that residues within a community have stronger connections among themselves than residues outside that community and residues within a community move in concerted manner. In case of open-model-dsRNA’, there are 18 communities (4 communities have less than 3 residues) (Fig. 9B). There are 14 communities (1 community has 1 residue) in open-model-dsRNA-blind complex and 24 communities (11 communities have less than 3 residues) in closed-model-dsRNA-blind complex. From these community structures, critical node information can be extracted. Critical nodes connect the two communities and hence are important residues for intra-protein communication.

Fig. 9.

Fig. 9

Open in a new tab

Dynamic network of open-model-dsRNA complex (A) Network partitioned into multiple communities colored differently (B) Critical nodes colored red (C) Optimal path between His308 and Pro674 of chain A (D) Difference in optimal path length between apo and dsRNA bound form of closed and open model. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Residue Met326 of protomer B, implicated in the disease variant, was noted as a ‘critical residue’ in terms of this residue network analysis during the dynamics of the open-model-dsRNA-blind model. Likewise, residue Ile473 of the same protomer, also implicated in the disease variant, was observed to be ‘critical’ in the RNA-bound closed form.

For optimal path analysis, source and sink residues are required. Initially, His308A was selected as source residue and Pro674A of BB-loop was selected as sink residue. The shorter length of optimal path and larger number of suboptimal paths indicate the efficiency of signal propagation. In case of open-model-dsRNA model, the optimal path length for His308A-Pro674A is 108 ​Å and involves 32 residues and total number of suboptimal paths with edge length offset of 1 ​Å is 1630 (Fig. 9C). Similarly, simulation of open-model-dsRNA-blind model shows optimal path length of 121 which includes 38 residues and there are 16174 suboptimal paths with edge length offset of 1. The optimal path length for closed-model-dsRNA-blind model is 140 and there are 32 residues in the optimal paths. There are less suboptimal paths (84) with edge length offset of 1 than open-model-dsRNA or open-model-dsRNA-blind model. This indicates that signal propagation upon dsRNA binding is more efficient in case of open-model. Similar analysis was also performed with source-and-sink residues from protomer B and combination of residues across protomers (source from one protomer and sink from other protomer). For protomer B, the closed-model-dsRNA was observed to have smaller path length (98) compared to open-model-dsRNA-blind (109) or open-model-dsRNA (156). Further, the optimal path lengths were calculated by considering other key residues as source residue (His81, Tyr57 and Asp54) and Pro674 as sink residue. The reduction in optimal path length was calculated for open and closed-models upon dsRNA binding. In all cases, open-model was found to have reduced path length upon dsRNA binding, compared to the closed form (Fig. 9D). Open-model also shows higher number of optimal paths in dsRNA bound form compared to the closed-form (in most combinations of His308-Pro674 or His81-Pro674).

4. Discussion

TLR10 is the least studied protein among human TLRs and also the only TLR with no ECD structure. In this study, we have attempted to model the full-length structure of TLR10 to understand domain organization and dynamics of TLR10. In the absence of known dimer form of ECD, multiple models were built for ECD-dimer utilizing different TLR templates. TLR10 belongs to TLR1 subfamily which involves TLR1, TLR2, TLR6 and TLR10. Experimental structures of TLR1-ECD homodimer and TLR2-ECD homodimer are available in different conformations (such as open and closed models) and these were used as template for TLR10-ECD dimer modelling. Further, there is a difference in the organization of protomers in the TLR1-TLR2 heterodimer structure compared to their homodimer forms and hence the heterodimer structure was also used as template. Finally, recent advancements in understanding the function of TLR10 have shown that TLR10 can localize to endosomes and bind to dsRNA which is a known ligand for TLR3. Another ECD-dimer was modelled using mouse TLR3-homodimer as template. The sequence identity between human and mouse TLR3 is 79.1% while the similarity is 87.4%. These ECD models were combined with TM-dimer (obtained by protein-protein docking of modelled TM helix monomer) and known TIR dimer to generate different full-length structures. The energetic calculations of these models suggest, closed-model to be the most favorable model. Several missense mutations in TLR10 have been associated with different diseases and these variants were also found to be in the close vicinity of the proposed interface of the closed-model. TLR1 homodimer has been suggested to be an inactive form of dimer as it needs to associate with TLR2 for function. Therefore, it can be hypothesized that the closed-model corresponds to the apo form of TLR10. Prediction of ligand binding resulted in an elaborate interface site in the closed-model, with good scores, indicating the possibility of ligand binding at this interface as well even though it might not be functional. For each model, unique hotspot residues could also be identified. The other endosomal TLRs were not considered as template due to low sequence identity. The sequence identity between TLR10 and other endosomal TLRs (TLR7, TLR8 and TLR9) is less than 25% (Supplementary Table 6). The endosomal TLR9 has been reported to undergo conformational change upon ligand binding by assessing the change in secondary structure through circular dichroism (Latz et al., 2007). However, the conformational changes involving closed and open form has not been studied for any other TLR homodimer.

In order to understand the dynamics of these models, MD simulation was performed in aqueous-membrane (POPC) environment. Each of the models show movement of ECD and TIR domain towards the membrane bilayer and this could be attributed to the electrostatic interactions between the TLR10 residues and lipid head groups. All the models exhibit high RMSF for TIR domain than the ECD domain. Even though the TIR domain move towards membrane, the protomers do not dissociate and maintain strong interactions throughout the simulation as evident by the ROG values. Apart from these common dynamics across models, the key interacting residues at the dimer interface (defined by persistent interaction and conservation) are different for different models. The key residues from the ECD of closed-model were found to be Asn330 and Leu569. Pro339 for semi-closed-model ECD and Pro556 for open-model ECD are example of key residue. Future studies can be done by mutating these residues and assessing the impact on dimer formation of TLR10 which will help in assessing which model corresponds to the native-like dimer.

TLR10 has remained an orphan receptor as the cognate ligand is not known. Many ligands have been proposed for TLR10 most of which are ligands of TLR2. It has been shown that TLR10 can be localized to endosome and can compete with TLR3 for binding of dsRNA. To understand the binding of dsRNA to TLR10, the docking of dsRNA to open-model of TLR10 was carried out in defined manner by specifying possible binding residues or in a generic blind manner without specifying any residue. It was found that blind docking of the dsRNA to open-model resulted in similar pose for dsRNA as obtained by defined docking. Several H-bond and salt bridge interactions could be identified between the dsRNA and both the chains of TLR10. Blind docking of dsRNA on TLR10-ECD monomer, results in many poses where the dsRNA bind in the central groove of ECD and therefore blind docking of dsRNA was also tried with closed-model. Top 100 docked poses were inspected and the pose with maximum interactions binds in the central groove of both protomers of closed-model. His81 was found to interact with the dsRNA in all three models and hence can be considered as critical residue for dsRNA binding to TLR10.

To explore the interaction stability of dsRNA binding to TLR10, all three docked complexes were subjected to MD simulation and the interactions were monitored throughout the simulation time. All the persistent interactions between dsRNA and TLR10 were from chain B in case of open-model-dsRNA while for other two blind complexes the interactions from both chains are maintained for significant simulation time. The set of key residues for the interaction with dsRNA based on persistence and conservation are different for each complex. For defined complex of open-model-dsRNA the key residues include Gln85 and His81 while for open-model-dsRNA-blind complex, key residues are Asn58, His81, Gln85, Gln410, Ser436 and Leu559. In case of blind complex of closed-model-dsRNA key residues are Ser35, Ser56, and Asp54. Future in vitro mutagenesis studies can be done by mutating these key residues and analyzing the binding of dsRNA to TLR10.

To gain insights into the signal propagation upon dsRNA binding, network analysis was performed. The network analysis revealed critical residues important for intra-protein communications and some of the disease variants were found to be the critical residues. It was observed that signal propagation will be efficient in open-model complexes compared to closed-model complex, through optimal path length and number of suboptimal paths.

5. Conclusion

In this study, full-length structure of TLR10 was modelled to gain insights into the important residues that might be required for homodimerization. Due to lack of knowledge about the dimer formation of ECD domain, multiple full-length models were generated by utilizing different and relevant templates. The dynamics of these models revealed that the closed-model is more stable and can be the apo form structure of TLR10. Persistent interacting residues across simulation time and their conservation among orthologs led to specific set of key residues for each model. These residues can be further taken for experimental validation in order to determine which of the full-length model is native model.

Docking of dsRNA in defined and blind manner to two different TLR10 models (Open and closed models) also revealed multiple interacting residues. His81 was found to be common among all three docked complexes. Interaction stability of dsRNA binding to TLR10 was studied through MD simulation and it was observed that many H-bond interactions are maintained throughout the simulation time. The suboptimal path analysis revealed that open-model can be the functional form for dsRNA binding owing to the efficiency of signal propagation from ECD to BB-loop of TIR domain. Thereby, we propose that TLR10 remains in closed form (closed model) which could be its inactive form, and binds to the dsRNA and is functional in the open form (open-model). Future experimental validation would help in addressing these proposals and the refinement of our models.

CRediT authorship contribution statement

Vikas Tiwari: Data curation, Formal analysis, Investigation, Validation, Visualization, Methodology, Resources, Writing – original draft. R. Sowdhamini: Conceptualization, Methodology, Supervision, Project administration, Funding acquisition, Resources, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

The authors would like to thank NCBS (TIFR) for infrastructural facilities. RS acknowledges funding and support provided by JC Bose Fellowship (JBR/2021/000006) from Science and Engineering Research Board, India and Bioinformatics Centre Grant funded by Department of Biotechnology, India (BT/PR40187/BTIS/137/9/2021). RS would also like to thank Institute of Bioinformatics and Applied Biotechnology for the funding through her Mazumdar-Shaw Chair in Computational Biology (IBAB/MSCB/182/2022).

Handling Editor: Prof S Ranganathan

Footnotes

Appendix A

Supplementary data to this article can be found online at https://doi.org/10.1016/j.crstbi.2023.100097.

Appendix A. Supplementary data

The following are the supplementary data to this article:

Multimedia component 1
mmc1.pdf (2.2MB, pdf)
Multimedia component 2
mmc2.txt (122.2KB, txt)

Data availability

Data will be made available on request.

References

  1. Adasme M.F., Linnemann K.L., Bolz S.N., Kaiser F., Salentin S., Haupt V.J., Schroeder M. PLIP 2021: expanding the scope of the protein–ligand interaction profiler to DNA and RNA. Nucleic Acids Res. 2021;49:W530–W534. doi: 10.1093/nar/gkab294. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Akira S., Uematsu S., Takeuchi O. Pathogen recognition and innate immunity. Cell. 2006;124:783–801. doi: 10.1016/j.cell.2006.02.015. [DOI] [PubMed] [Google Scholar]
  3. Ashkenazy H., Abadi S., Martz E., Chay O., Mayrose I., Pupko T., Ben-Tal N. ConSurf 2016: an improved methodology to estimate and visualize evolutionary conservation in macromolecules. Nucleic Acids Res. 2016;44:W344–W350. doi: 10.1093/nar/gkw408. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bowers K.J., Chow E., Xu H., Dror R.O., Eastwood M.P., Gregersen B.A., Klepeis J.L., Kolossvary I., Moraes M.A., Sacerdoti F.D., Salmon J.K., Shan Y., Shaw D.E. Proceedings of the 2006 ACM/IEEE Conference on Supercomputing. 2006. Scalable algorithms for molecular dynamics simulations on commodity clusters; p. SC’06. [Google Scholar]
  5. Bulat-Kardum L.J., Etokebe G.E., Lederer P., Balen S., Dembic Z. Genetic polymorphisms in the toll-like receptor 10, interleukin (IL)17A and IL17F genes differently affect the risk for tuberculosis in Croatian population. Scand. J. Immunol. 2015;82:63–69. doi: 10.1111/sji.12300. [DOI] [PubMed] [Google Scholar]
  6. Chuang T.H., Ulevitch R.J. Identification of hTLR10: a novel human Toll-like receptor preferentially expressed in immune cells. Biochim. Biophys. Acta Gene Struct. Expr. 2001;1518:157–161. doi: 10.1016/s0167-4781(00)00289-x. [DOI] [PubMed] [Google Scholar]
  7. Dobson L., Reményi I., Tusnády G.E. CCTOP: a Consensus Constrained TOPology prediction web server. Nucleic Acids Res. 2015;43:W408. doi: 10.1093/nar/gkv451. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Dominguez C., Boelens R., Bonvin A.M.J.J. HADDOCK: a protein-protein docking approach based on biochemical or biophysical information. J. Am. Chem. Soc. 2003;125:1731–1737. doi: 10.1021/ja026939x. [DOI] [PubMed] [Google Scholar]
  9. Eargle J., Luthey-Schulten Z. NetworkView: 3D display and analysis of protein·RNA interaction networks. Bioinformatics. 2012;28:3000. doi: 10.1093/bioinformatics/bts546. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Fiser A., Sali A. ModLoop: automated modeling of loops in protein structures. Bioinformatics. 2003;19(18):2500–2501. doi: 10.1093/bioinformatics/btg362. [DOI] [PubMed] [Google Scholar]
  11. Fore F., Indriputri C., Mamutse J., Nugraha J. TLR10 and its unique anti-inflammatory properties and potential use as a target in therapeutics. Immune Netw. 2020;20:1–10. doi: 10.4110/in.2020.20.e21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Fu L., Niu B., Zhu Z., Wu S., Li W. CD-HIT: accelerated for clustering the next-generation sequencing data. Bioinformatics. 2012;28:3150–3152. doi: 10.1093/bioinformatics/bts565. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Gay N.J., Symmons M.F., Gangloff M., Bryant C.E. Assembly and localization of Toll-like receptor signalling complexes. Nat. Rev. Immunol. 2014;14:546–558. doi: 10.1038/nri3713. [DOI] [PubMed] [Google Scholar]
  14. Govindaraj R.G., Manavalan B., Lee G., Choi S. Molecular modeling-based evaluation of hTLR10 and identification of potential ligands in toll-like receptor signaling. PLoS One. 2010;5:1–13. doi: 10.1371/journal.pone.0012713. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Guan Y., Ranoa D.R.E., Jiang S., Mutha S.K., Li X., Baudry J., Tapping R.I. Human TLRs 10 and 1 share common mechanisms of innate immune sensing but not signaling. J. Immunol. 2010;184:5094–5103. doi: 10.4049/jimmunol.0901888. [DOI] [PubMed] [Google Scholar]
  16. Guirado M., Gil H., Saenz-Lopez P., Reinboth J., Garrido F., Cozar J.M., Ruiz-Cabello F., Carretero R. Association between C13ORF31, NOD2, RIPK2 and TLR10 polymorphisms and urothelial bladder cancer. Hum. Immunol. 2012;73:668–672. doi: 10.1016/j.humimm.2012.03.006. [DOI] [PubMed] [Google Scholar]
  17. Hasan U., Chaffois C., Gaillard C., Saulnier V., Merck E., Tancredi S., Guiet C., Brière F., Vlach J., Lebecque S., Trinchieri G., Bates E.E.M. Human TLR10 is a functional receptor, expressed by B cells and plasmacytoid dendritic cells, which activates gene transcription through MyD88. J. Immunol. 2005;174:2942–2950. doi: 10.4049/jimmunol.174.5.2942. [DOI] [PubMed] [Google Scholar]
  18. Henrick B.M., Yao X.D., Zahoor M.A., Abimiku A., Osawe S., Rosenthal K.L. TLR10 senses HIV-1 proteins and significantly enhances HIV-1 infection. Front. Immunol. 2019;10 doi: 10.3389/fimmu.2019.00482. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Hornung V., Rothenfusser S., Britsch S., Krug A., Jahrsdörfer B., Giese T., Endres S., Hartmann G. Quantitative expression of toll-like receptor 1–10 mRNA in cellular subsets of human peripheral blood mononuclear cells and sensitivity to CpG oligodeoxynucleotides. J. Immunol. 2002;168:4531–4537. doi: 10.4049/jimmunol.168.9.4531. [DOI] [PubMed] [Google Scholar]
  20. Jin M.S., Kim S.E., Heo J.Y., Lee M.E., Kim H.M., Paik S.G., Lee H., Lee J.O. Cell; 2007. Crystal Structure of the TLR1-TLR2 Heterodimer Induced by Binding of a Tri-acylated Lipopeptide. [DOI] [PubMed] [Google Scholar]
  21. Kagan J.C., Su T., Horng T., Chow A., Akira S., Medzhitov R. TRAM couples endocytosis of Toll-like receptor 4 to the induction of interferon-β. Nat. Immunol. 2008;94 9:361–368. doi: 10.1038/ni1569. 2008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Kawai T., Akira S. The role of pattern-recognition receptors in innate immunity: update on Toll-like receptors. Nat. Immunol. 2010;11(5):373–384. doi: 10.1038/ni.1863. [DOI] [PubMed] [Google Scholar]
  23. Kawasaki T., Kawai T. Toll-like receptor signaling pathways. Front. Immunol. 2014;5:461. doi: 10.3389/fimmu.2014.00461. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Kim S.K., Park H.J., Hong I.K., Chung J.H., Eun Y.G. A missense polymorphism (rs11466653, Met326Thr) of toll-like receptor 10 (TLR10) is associated with tumor size of papillary thyroid carcinoma in the Korean population. Endocr. 2012;431 43:161–169. doi: 10.1007/s12020-012-9783-z. 2012. [DOI] [PubMed] [Google Scholar]
  25. Kızıldağ S., Arslan S., Özbilüm N., Engin A., Bakır M. Effect of TLR10 (2322A/G, 720A/C, and 992T/A) polymorphisms on the pathogenesis of Crimean Congo hemorrhagic fever disease. J. Med. Virol. 2018;90:19–25. doi: 10.1002/jmv.24924. [DOI] [PubMed] [Google Scholar]
  26. Laskowski R.A., Jabłońska J., Pravda L., Vařeková R.S., Thornton J.M. PDBsum: structural summaries of PDB entries. Protein Sci. 2018;27:129. doi: 10.1002/pro.3289. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Laskowski R.A., MacArthur M.W., Moss D.S., Thornton J.M. PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 1993;26:283–291. [Google Scholar]
  28. Latz E., Verma A., Visintin A., Gong M., Sirois C.M., Klein D.C.G., Monks B.G., McKnight J.C., Lamphier M.S., Duprex P.W., Espevik T., Golenbock D.T. Ligand-induced conformational changes allosterically activate Toll-like receptor 9. Nat. Immunol. 2007;8:772–779. doi: 10.1038/ni1479. [DOI] [PubMed] [Google Scholar]
  29. Lee S.M.Y., Yip T.F., Yan S., Jin D.Y., Wei H.L., Guo R.T., Peiris J.S.M. Recognition of double-stranded RNA and regulation of interferon pathway by toll-like receptor 10. Front. Immunol. 2018;9:516. doi: 10.3389/fimmu.2018.00516. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Li M., Han W., Zhu L., Jiang J., Qu W., Zhang L., Jia L., Zhou Q. IRAK2 and TLR10 confer risk of Hashimoto's disease: a genetic association study based on the Han Chinese population. J. Hum. Genet. 2019;647 64:617–623. doi: 10.1038/s10038-019-0613-5. 2019. [DOI] [PubMed] [Google Scholar]
  31. Liu L., Botos I., Wang Y., Leonard J.N., Shiloach J., Segal D.M., Davies D.R. Structural basis of toll-like receptor 3 signaling with double-stranded RNA. Science (80-.) 2008;320:379–381. doi: 10.1126/science.1155406. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Lomize M.A., Pogozheva I.D., Joo H., Mosberg H.I., Lomize A.L. OPM database and PPM web server: Resources for positioning of proteins in membranes. Nucleic Acids Res. 2012;40(Database issue):D370–6. doi: 10.1093/nar/gkr703. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Luo L., Lucas R.M., Liu L., Stow J.L. Signalling, sorting and scaffolding adaptors for Toll-like receptors. J. Cell Sci. 2019;133 doi: 10.1242/jcs.239194. [DOI] [PubMed] [Google Scholar]
  34. Madeira F., Pearce M., Tivey A.R.N., Basutkar P., Lee J., Edbali O., Madhusoodanan N., Kolesnikov A., Lopez R. Search and sequence analysis tools services from EMBL-EBI in 2022. Nucleic Acids Res. 2022;50(W1) doi: 10.1093/nar/gkac240. W276–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Madhavi Sastry G., Adzhigirey M., Day T., Annabhimoju R., Sherman W. Protein and ligand preparation: parameters, protocols, and influence on virtual screening enrichments. J. Comput. Aided Mol. Des. 2013;27:221–234. doi: 10.1007/s10822-013-9644-8. [DOI] [PubMed] [Google Scholar]
  36. McGuffin L.J., Bryson K., Jones D.T. The PSIPRED protein structure prediction server. Bioinformatics. 2000;16:404–405. doi: 10.1093/bioinformatics/16.4.404. [DOI] [PubMed] [Google Scholar]
  37. Moreira I.S., Koukos P.I., Melo R., Almeida J.G., Preto A.J., Schaarschmidt J., Trellet M., Gümüş Z.H., Costa J., Bonvin A.M.J.J. SpotOn: high accuracy identification of protein-protein interface hot-spots. Sci. Rep. 2017;7(1):8007. doi: 10.1038/s41598-017-08321-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Moresco E.M.Y., LaVine D., Beutler B. Toll-like receptors. Curr. Biol. 2011;21:R488–R493. doi: 10.1016/j.cub.2011.05.039. [DOI] [PubMed] [Google Scholar]
  39. Morgan A.R., Lam W.J., Han D.Y., Fraser A.G., Ferguson L.R. Genetic variation within TLR10 is associated with Crohn's disease in a New Zealand population. Hum. Immunol. 2012;73:416–420. doi: 10.1016/j.humimm.2012.01.015. [DOI] [PubMed] [Google Scholar]
  40. Nagashima H., Iwatani S., Cruz M., Abreu J.A.J., Uchida T., Mahachai V., Vilaichone R.K., Graham D.Y., Yamaoka Y. Toll-like receptor 10 in Helicobacter pylori infection. J. Infect. Dis. 2015;212:1666–1676. doi: 10.1093/infdis/jiv270. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Nyman T., Stenmark P., Flodin S., Johansson I., Hammarström M., Nordlund P.R. The crystal structure of the human toll-like receptor 10 cytoplasmic domain reveals a putative signaling dimer. J. Biol. Chem. 2008;283(18):11861–11865. doi: 10.1074/jbc.C800001200. [DOI] [PubMed] [Google Scholar]
  42. O'Neill L.A.J., Bowie A.G. The family of five: TIR-domain-containing adaptors in Toll-like receptor signalling. Nat. Rev. Immunol. 2007;75(7):353–364. doi: 10.1038/nri2079. [DOI] [PubMed] [Google Scholar]
  43. Oosting M., Cheng S.C., Bolscher J.M., Vestering-Stenger R., Plantinga T.S., Verschueren I.C., Arts P., Garritsen A., Van Eenennaam H., Sturm P., Kullberg B.J., Hoischen A., Adema G.J., Van Der Meer J.W.M., Netea M.G., Joosten L.A.B. Human TLR10 is an anti-inflammatory pattern-recognition receptor. Proc. Natl. Acad. Sci. U.S.A. 2014;111:E4478–E4484. doi: 10.1073/pnas.1410293111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Patra M.C., Batool M., Haseeb M., Choi S. A computational probe into the structure and dynamics of the full-length toll-like receptor 3 in a phospholipid bilayer. Int. J. Mol. Sci. 2020;21:1–25. doi: 10.3390/ijms21082857. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Patra M.C., Kwon H.K., Batool M., Choi S. Computational insight into the structural organization of full-length toll-like receptor 4 Dimer in a model Phospholipid Bilayer. Front. Immunol. 2018;9:1–15. doi: 10.3389/fimmu.2018.00489. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Roach J.C., Glusman G., Rowen L., Kaur A., Purcell M.K., Smith K.D., Hood L.E., Aderem A. The evolution of vertebrate Toll-like receptors. Proc. Natl. Acad. Sci. U.S.A. 2005;102:9577–9582. doi: 10.1073/pnas.0502272102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Rout A.K., Acharya V., Maharana D., Dehury B., Udgata S.R., Jena R., Behera B., Parida P.K., Behera B.K. Insights into structure and dynamics of extracellular domain of Toll-like receptor 5 in Cirrhinus mrigala (mrigala): a molecular dynamics simulation approach. PLoS One. 2021;16 doi: 10.1371/journal.pone.0245358. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Šali A., Blundell T.L. Comparative protein modelling by satisfaction of spatial restraints. J. Mol. Biol. 1993;234:779–815. doi: 10.1006/jmbi.1993.1626. [DOI] [PubMed] [Google Scholar]
  49. Scheurer M., Rodenkirch P., Siggel M., Bernardi R.C., Schulten K., Tajkhorshid E., Rudack T. PyContact: rapid, customizable, and visual analysis of noncovalent interactions in MD simulations. Biophys. J. 2018;114:577–583. doi: 10.1016/j.bpj.2017.12.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Schymkowitz J., Borg J., Stricher F., Nys R., Rousseau F., Serrano L. The FoldX web server: an online force field. Nucleic Acids Res. 2005;33:W382–W388. doi: 10.1093/nar/gki387. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Stappers M.H.T., Oosting M., Ioana M., Reimnitz P., Mouton J.W., Netea M.G., Gyssens I.C., Joosten L.A.B. Genetic variation in TLR10, an inhibitory toll-like receptor, influences susceptibility to complicated skin and skin structure infections. J. Infect. Dis. 2015;212:1491–1499. doi: 10.1093/infdis/jiv229. [DOI] [PubMed] [Google Scholar]
  52. Su L., Wang Y., Wang J., Mifune Y., Morin M.D., Jones B.T., Moresco E.M.Y., Boger D.L., Beutler B., Zhang H. Structural basis of TLR2/TLR1 activation by the synthetic agonist diprovocim. J. Med. Chem. 2019;62(6):2938–2949. doi: 10.1021/acs.jmedchem.8b01583. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Su S.B., Tao L., Deng Z.P., Chen W., Qin S.Y., Jiang H.X. TLR10: insights, controversies and potential utility as a therapeutic target. Scand. J. Immunol. 2021;93:1–15. doi: 10.1111/sji.12988. [DOI] [PubMed] [Google Scholar]
  54. Sukhwal A., Sowdhamini R. PPcheck: a webserver for the quantitative analysis of protein-protein interfaces and prediction of residue hotspots. Bioinf. Biol. Insights. 2015;9:141–151. doi: 10.4137/BBI.S25928. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Tongtawee T., Bartpho T., Wattanawongdon W., Dechsukhum C., Leeanansaksiri W., Matrakool L., Panpimanmas S. Role of toll-like receptor 10 gene polymorphism and gastric mucosal pattern in patients with chronic gastritis. Turk. J. Gastroenterol. 2017;28:243–247. doi: 10.5152/tjg.2017.16673. [DOI] [PubMed] [Google Scholar]
  56. Torices S., Julia A., Muñoz P., Varela I., Balsa A., Marsal S., Fernández-Nebro A., Blanco F., López-Hoyos M., Martinez-Taboada V., Fernández-Luna J.L. A functional variant of TLR10 modifies the activity of NFkB and may help predict a worse prognosis in patients with rheumatoid arthritis. Arthritis Res. Ther. 2016;18 doi: 10.1186/s13075-016-1113-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Törmänen S., Korppi M., Lauhkonen E., Koponen P., Teräsjärvi J., Vuononvirta J., Helminen M., He Q., Nuolivirta K. Toll-like receptor 1 and 10 gene polymorphisms are linked to postbronchiolitis asthma in adolescence. Acta Paediatr. 2018;107:134–139. doi: 10.1111/apa.13984. [DOI] [PubMed] [Google Scholar]
  58. Tsirigos K.D., Peters C., Shu N., Käll L., Elofsson A. The TOPCONS web server for consensus prediction of membrane protein topology and signal peptides. Nucleic Acids Res. 2015;43:W401–W407. doi: 10.1093/nar/gkv485. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Van Zundert G.C.P., Rodrigues J.P.G.L.M., Trellet M., Schmitz C., Kastritis P.L., Karaca E., Melquiond A.S.J., Van Dijk M., De Vries S.J., Bonvin A.M.J.J. The HADDOCK2.2 web server: user-friendly integrative modeling of biomolecular complexes. J. Mol. Biol. 2016;428(4):720–725. doi: 10.1016/j.jmb.2015.09.014. [DOI] [PubMed] [Google Scholar]
  60. Wiederstein M., Sippl M.J. ProSA-web: interactive web service for the recognition of errors in three-dimensional structures of proteins. Nucleic Acids Res. 2007;35:W407–W410. doi: 10.1093/nar/gkm290. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Yamamoto M., Sato S., Hemmi H., Hoshino K., Kaisho T., Sanjo H., Takeuchi O., Sugiyama M., Okabe M., Takeda K., Akira S. Role of adaptor TRIF in the MyD88-independent toll-like receptor signaling pathway. Science (80-.) 2003;301:640–643. doi: 10.1126/science.1087262. [DOI] [PubMed] [Google Scholar]
  62. Yan Y., Zhang D., Zhou P., Li B., Huang S.Y. HDOCK: a web server for protein-protein and protein-DNA/RNA docking based on a hybrid strategy. Nucleic Acids Res. 2017;45(W1):W365–W373. doi: 10.1093/nar/gkx407. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Zhang Z., Ohto U., Shimizu T. Toward a structural understanding of nucleic acid-sensing Toll-like receptors in the innate immune system. FEBS Lett. 2017;591:3167–3181. doi: 10.1002/1873-3468.12749. [DOI] [PubMed] [Google Scholar]
  64. Zhou X.X., Jia W.H., Shen G.P., Qin H. De, Yu X.J., Chen L.Z., Feng Q.S., Shugart Y.Y., Zeng Y.X. Sequence variants in toll-like receptor 10 are associated with nasopharyngeal carcinoma risk. Cancer Epidemiol. Biomarkers Prev. 2006;15:862–866. doi: 10.1158/1055-9965.EPI-05-0874. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Multimedia component 1
mmc1.pdf (2.2MB, pdf)
Multimedia component 2
mmc2.txt (122.2KB, txt)

Data Availability Statement

Data will be made available on request.

Articles from Current Research in Structural Biology are provided here courtesy of Elsevier

RESOURCES