Bioinformatic Analysis of Human TLR4 Coding Variations Associated with Ocular Infection: A Structural Prediction and Molecular Docking Studies

Conjunctivitis is an inflammatory condition impacting the conjunctiva, the transparent membrane that covers the white of the eye and the inner eyelids. This condition is marked by symptoms such as excessive discharge, sensitivity to light, redness of the conjunctiva, and itching. Conjunctivitis can be triggered by multiple factors, including allergens, viral agents, and bacterial infections [1]. A recent ophthalmology study in the USA highlights that eye diseases treated in emergency departments incur an annual cost of $2 billion. Conjunctivitis alone contributes around 28% of this total, underscoring the substantial financial strain eye conditions place on healthcare resources [2].  Conjunctivitis can be caused by several viral pathogens, including the varicella-zoster virus (VZV), herpes simplex virus (HSV) and adenovirus [3]. Fungi are significant contributors to eye infections, especially in tropical and developing regions. They can cause severe ocular candidiasis and keratitis, which may result in permanent vision loss [4].Various bacteria, including Gram-positive types (Staphylococcus aureus, Staphylococcus epidermidis, Streptococcus, and Bacillus species), as well as Gram-negative (Pseudomonas aeruginosa, Moraxella, and Haemophilus spp.,) often cause conjunctivitis and keratitis. They may enter the eye through skin contact, the respiratory tract, or external infection from another person[4, 5]. Genetic variants influencing angiogenic responsiveness, such as the pink-eyed mutation in the oca2 gene, impact pink eye development in mice by altering angiogenesis. This mutation modifies blood vessel formation within the eye, contributing to changes associated with pink eye [6]. Allergic conjunctivitis is widespread, affecting between 15% and 40% of various populations. A study in Saudi Arabia found an age- and sex-adjusted prevalence of 70.5% among adults in the western region [7]. Pakistan reported 400,000 cases of viral conjunctivitis, with a particularly high surge in cases observed in Karachi [8]. Moreover, variants in the interleukin-4 gene (rs2243250) and its receptor (rs1805010) have been associated with viral conjunctivitis. Research has revealed significant differences in genotype distributions between viral conjunctivitis patients and healthy controls, suggesting these genetic variants may influence susceptibility to viral infections leading to conjunctivitis [9].

The genes associated with Herpes Simplex Virus Type 1 ocular infection include UL24, UL29 (ICP8), UL41 (VHS), UL53 (gK), UL54 (ICP27), UL56, ICP4, US1 (ICP22), US3, and gG, play critical roles in the virus’s replication, immune evasion, and ocular tissue damage during infection [3]. Toll-like receptor 4 (TLR4) is significant in the innate immune response by recognizing pathogen-associated molecular patterns (PAMPs) and initiating inflammatory signaling pathways. In allergic conjunctivitis, TLR4 can be activated by environmental allergens, such as pollen, dust mites, or pollutants, which are perceived as threats by the immune system [10]. The TLR4 (ARMD10 or CD284) gene is located on chromosome 9q32 in humans and is composed of several exons and introns, which are transcribed into mRNA and subsequently translated into the TLR4 protein. It spans multiple kilobases and is classified as a type I transmembrane protein, characterized by a single transmembrane domain that anchors it within the cell membrane. The extracellular domain of TLR4 plays a key role in recognizing PAMPs, such as lipopolysaccharides (LPS), while the intracellular domain is involved in triggering signaling pathways [11]. The extracellular region of TLR4 includes a leucine-rich repeat (LRR) motif, essential for binding ligands. This allows TLR4 to identify specific molecular signatures from pathogens, including bacterial LPS, which activates an immune response. The intracellular portion contains the TIR domain, which is responsible for initiating downstream signaling after activation. The TIR domain interacts with adaptor proteins, including MyD88 and TRIF, leading to the activation of inflammatory pathways such as NF-?B and MAPK, which ultimately results in the release of pro-inflammatory cytokines [12, 13]. Upon ligand binding, TLR4 typically dimerizes, often in conjunction with the co-receptor MD-2, which is necessary for LPS and other PAMP recognition. The TLR4 gene also undergoes alternative splicing, producing various isoforms that may have different functional characteristics or tissue-specific expression. Genetic variations in the TLR4 gene can alter its function and are linked to a range of conditions, including sepsis, autoimmune diseases, and allergic reactions [14].

Furthermore, TLR4 expression was detected in circulating CD⁴⁺ T cells from individuals with chronic allergic conjunctivitis, and in conjunctival epithelial cells. The D299G (rs4986790) and T399I have been studied and 18% TLR4 mutations, with a significantly higher incidence of gram-negative infections that may contribute to an increased susceptibility to allergic conjunctivitis [15, 16]. When peripheral blood mononuclear cells (PBMCs) from patients were stimulated with allergens such as Der p, a significant increase in both TLR4 expression and activation markers (CD69) was observed in CD⁴⁺ T cells. This suggests that allergen-specific stimulation can enhance the TLR4 activity, which may contribute to the Th2 inflammatory microenvironment commonly seen in allergic responses [15]. Bioinformatics tools play a crucial role in understanding the genetic foundations of infectious diseases, which is essential for early detection of eye infections and effective epidemiological responses. By utilizing a range of prediction tools, we categorized the nsSNPs in the TLR4 gene and identified those most likely to disrupt receptor function, potentially leading to diseases such as conjunctivitis. Future research will focus on exploring the in-silico structural and functional effects of missense mutations in human TLR4, specifically with pink eye infections. This approach will aim to identify potential treatment targets and enhance drug precision. Such strategies hold promises for improving treatment efficacy and developing preventative measures to reduce infection rates.

1. MATERIALS AND METHODS

2.1 Screening datasets

The National Center for Biotechnology Information (NCBI) (https://www.ncbi.nlm.nih.gov) has made available the hTLR4 gene data, including unique SNP findings, through their dbSNP database (http://www.ncbi.nlm.nih.gov/snp). Additionally, UniProt (https://www.uniprot.org) is used to obtain protein sequence information, allowing for detailed insights into its structure and function. The Overall methodology of the present study is given in Figure 1.

       
           
       
Figure 1. Workflow of the current study

2.2 Prediction of nsSNPs

SNPNexus tool (https://www.snp-nexus.org/v4) simplifies the selection and prioritization of known and novel genomic variants [17]. Sorting Intolerant From Tolerant (SIFT) uses sequence homology to identify deleterious SNPs, analyzing amino acid variations that affect protein phenotypic and functional alterations [18]. PolyPhen analyzes protein composition and functionality by analyzing amino acid substitutions using protein sequence, and substitution information. The score ranges from 0.0 to 1.0 indicating deleterious or tolerated variants [19]. Protein Analysis through Evolutionary Relationship (PANTHER) (https://www.pantherdb.org/tools) uses evolutionary relationships and the Hidden Markov Model (HMM) for analysis [20]. The TLR4 sequence, mutant, and position are submitted to the query [21]. Amino acid substitutions on functional proteins are categorized as benign or harmful by the web server MutPred2 (http://mutpred2.mutdb.org), with "pathogenic mutation" values ranging from 0.5 to 1.0 [22]. Polyphen-2-Polymorphism Phenotyping v2 (http://genetics.bwh.harvard.edu/pph2/) predicts the consequence of an amino acid substitution on a protein's structure and, hence, function [23]. The PredictSNP (https://loschmidt.chemi.muni.cz/predictsnp1) uses input data from various tools to predict the impact of changing one amino acid, offering increased effectiveness and accuracy [24].

2.3 Impact of disease-associated nsSNPs on protein stability.

SuSpect (http://www.sbg.bio.ic.ac.uk/suspect) analyzes disease prediction using sequence-based and annotation methods, reducing annotation bias and ranking genes based on multiple evidence lines [25]. SNPs & GO (https://snps.biofold.org/snps-and-go/snps-and-go.html) uses functional annotation of protein sequences to determine variation associated with a disease [22]. Meta SNP (https://snps.biofold.org/meta-snp) is used to predict a single nucleotide variation in protein sequence. The output value>0.5 is considered as a disease, and <0>[26].  iStable (http://predictor.nchu.edu.tw/iStable) is a server that predicts protein stability changes using sequence or structure information, combining five forecasting instruments and a variety of feature coding and prediction techniques [27].

2.4 Homology Modeling Prediction

The SWISS-MODEL algorithm (https://swissmodel.expasy.org) is used for protein structural homology modeling that constructs accurate protein structures by utilizing closely related templates. The process begins by clustering known protein structures and selecting the template with the highest sequence identity to the target protein [28]. Following model construction, SWISS-MODEL evaluates the model's quality using several validation metrics, including the Ramachandran plot, QMEAN score, and MolProbity score. These assessments provide insights into the geometric quality and reliability of the predicted structure [29].

2.5 Structural Verification Analysis

SAVESv6.1 (https://saves.mbi.ucla.edu) was used to select and validate a structural model, incorporating PROCHECK and ERRAT for overall quality. A RAMACHANDRAN plot assessed model quality, determining the amino acid preferred area and dihedral angle. This plot of the torsional angles phi (?) and psi (?) of the residues (amino acids) contained in a peptide. By Ramachandran plot helps to determine which torsional angles are permitted and can obtain insight into the structure of peptides. The model with satisfying stereo-chemical quality was further assessed by QMEAN and ProSA was used to compute Z score values for the models [30].

2.6 Structural Superimposition

Template Modeling alignment (TM-Align) tool (https://zhanggroup.org/TM-align) used for comparing the structural similarities between wild-type and mutant protein models. It calculates the TM-score and Root Mean Square Deviation (RMSD) to assess the degree of structural divergence quantitatively [31]. In contrast, RMSD quantifies the average distance between corresponding atoms in the aligned proteins; higher RMSD values signify greater structural divergence [32].

2.7 Structural analysis of variants

Mutation 3D (http://www.mutation3d.org/about.shtml) is a valuable computational tool designed to analyze the effects of amino acid substitutions on protein models and structures. This platform facilitates the identification of mutation clusters and functional hotspots, contributing to a deeper understanding of the implications of these alterations on protein function. The HOPE (Have (y) Our Protein Explained) program (https://www3.cmbi.umcn.nl/hope/input) collects details from multiple sources to offer insights into the effects of variants on TLR4 structure. After the FASTA format is submitted, it finds the mutant residue and goes on to the following step of analysis [33].

 2.8 Ligand Preparation

PubChem (https://pubchem.ncbi.nlm.nih.gov) is a comprehensive database that houses a wide range of chemical compounds, primarily small molecules, but also includes larger compounds such as lipids, peptides, carbohydrates, nucleotides, and chemically modified macromolecules. It provides detailed information on chemical structures, identifiers, chemical and physical properties, biological activities, patents, and health, safety, and toxicological data [34].

2.9 Molecular screening analysis

PyRx (https://pyrx.sourceforge.io) is a virtual screening tool used in computational drug discovery to evaluate libraries of compounds for potential therapeutic targets. It allows users to perform Virtual Screening on various platforms, guiding them through each step of the process from data preparation to job submission and result analysis. Discovery Studio Visualizer (https://discover.3ds.com/discovery-studio-visualizer-download) was then employed to examine the ligand-protein interactions, and both the ligand and protein structures were represented in 2D and 3D formats as PNG files for visualization. This workflow enables efficient drug discovery by predicting molecular interactions and identifying potential drug candidates [35].

RESULTS

1. 1. Dataset download

In this study, a total of 8,702 SNPs of the human TLR4 gene were downloaded from the dbNCBI database. These SNPs include 4,275 in UTR regions, with 3,824 located in the 3' UTR region and 451 in the 5' UTR region. There are also 15,843 SNPs in intronic regions, 799 synonymous variants, and 1,928 non-synonymous SNPs. Additionally, the dataset comprises 2,601 SNPs in the 3' downstream region, 2,902 SNPs in the 5' upstream region, 157 exonic mutations, and 208 SNPs in non-coding regions, as illustrated in Figure 2. For further analysis, the protein sequence of O00206 was retrieved from the UniProt database. The human TLR4 gene is located in the 9q33.1 region of chromosome 9. It has a molecular weight of 95 kDa and consists of 4 exons and 10 distinct introns.

       
           
       
Figure 2. Bar Plot presents the SNPs in different regions of the TLR4 Gene.

3.2 Functional Prediction of nsSNPs

A total of 8,702 rsIDs were submitted to SNPnexus, resulting in predictions for 1,724 variants, with 878 classified as deleterious and 972 as tolerated based on the SIFT algorithm (Figure 3a). Additionally, PolyPhen analysis revealed that 469 nsSNPs were likely to be damaging, 355 were categorized as possibly damaging, and 1,026 were identified as benign (Figure 3b). Among these, 132 SNPs were found to be common and were all predicted to be probably damaging (Table 1). These findings highlight the potential significance of these variants in terms of their impact on protein function and their possible associations with disease.

       
           
       
Figure 3. Detrimental missense nsSNPs are predicted by a) SIFT algorithm 0 and b) PolyPhen

Table 1. Prediction of functional consequences of coding variants identified by SNPnexus

*D=Deleterious; PD= Probably Damaging

3.3 Prediction of functional impact of mutation

The analysis of nsSNPs using PANTHER, Predict SNP, and PolyPhen2 revealed significant insights into their potential impact on protein function (Table 2). PANTHER classified 96 nsSNPs as Possibly Damaging, 28 as Probably Benign, and 8 as Probably Damaging, with specific mutations (R810L, C306W, and L401Q) identified as likely impairing protein function. Predict SNP found 104 SNPs to have a detrimental effect and 28 to be neutral. PolyPhen2 predicted all 128 SNPs to be Probably Damaging, with a few exceptions, such as S415T, I769T, and P168H, which were considered benign, and Y652C, which was marked as possibly damaging. These results suggest that many of the nsSNPs may have a detrimental effect on protein function, warranting further experimental investigation.

Table 2. List of disease linked variants predicted by computational tools.