Bioinformatics analysis of the potential mechanisms of Alzheimer’s disease induced by exposure to combined triazine herbicides

Abstract Background The development of Alzheimer’s disease (AD) is promoted by a combination of genetic and environmental factors. Notably, combined exposure to triazine herbicides atrazine (ATR), simazine (SIM), and propazine (PRO) may promote the development of AD, but the mechanism is unknown. Aim To study the molecular mechanism of AD induced by triazine herbicides. Methods Differentially expressed genes (DEGs) of AD patients and controls were identified. The intersectional targets of ATR, SIM, and PRO for possible associations with AD were screened through network pharmacology and used for gene ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) enrichment analysis. The binding potentials between the core targets and herbicides were validated by molecular docking and molecular dynamics. Results A total of 1,062 DEGs were screened between the AD patients and controls, which identified 148 intersectional targets of herbicides causing AD that were screened by network pharmacology analysis. GO and KEGG enrichment analysis revealed that cell cycling and cellular senescence were important signalling pathways. Finally, the core targets EGFR, FN1, and TYMS were screened and validated by molecular docking and molecular dynamics. Conclusion Our results suggest that combined exposure to triazine herbicides might promote the development of AD, thereby providing new insights for the prevention of AD.


Introduction
Alzheimer's disease (AD) is a devastating neurodegenerative disorder characterised by progressive memory loss, behavioural changes, and disability, ultimately resulting in death (Neth et al. 2020).At present, AD reportedly affects 46.8 million individuals worldwide, which is projected to increase to 131.5 million by the year 2050 (Khoury et al. 2017).Hence, the burden of care for AD patients has become a growing concern for both families and society as a whole.It is well recognised that both environmental factors and personal habits play significant roles in the onset of AD (Kikuchi et al. 2020).
Triazines are a class of herbicides that are widely applied as weed-control agents (Rodríguez-González et al. 2013).The most commonly used triazine herbicides include atrazine (ATR), simazine (SIM), and propazine (PRO) (Brandhonneur et al. 2020).Triazine herbicides specifically target broadleaf and grassy weeds and are often combined with other herbicides to manage an extensive variety of weed types (Freitas et al. 2014).However, the use of triazine herbicides has become a topic of discussion and debate due to detrimental environmental effects and potential risks to human health.
Various epidemiological studies have confirmed a correlation between pesticide exposure and the occurrence of sporadic AD (Hayden et al. 2010;Yegambaram et al. 2015;Killin et al. 2016).However, relatively few studies have investigated the relationship between exposure to triazine herbicides and the incidence of AD.A previous study reported that several widely used triazine herbicides, including ATR, SIM, and PRO, enhanced the production and oligomerisation of amyloid-beta 42, which have been associated with the early onset of AD (Portelius et al. 2016).ATR, SIM, and PRO are frequently detected in natural environments (Brandhonneur et al. 2020;Wang and Liu 2020).Due to the similar chemical structures and metabolic characteristics (Wang and Liu 2020), combined exposure to ATR, SIM, and PRO, especially occupationally, might be a key factor in the onset of AD.
In the present study, microarray data of AD patients from the Gene Expression Omnibus (GEO) database, in addition to network pharmacology, molecular docking, and molecular dynamics studies, were integrated to explore the mechanisms underlying the onset and development of AD induced by exposure to triazine herbicides.The results of this study provide evidence that combined exposure to ATR, SIM, and PRO is linked to the onset and development of AD.Hence, further research is warranted to elucidate the molecular mechanisms underlying this association.

Genetic microarray data sources
The dataset GSE28146 was obtained from the GEO database (http://www.ncbi.nlm.nih.gov/).The experiment was based on the GPL570 platform and conducted using the Affymetrix Human Genome U133 Plus 2.0 Array (Affymetrix, Inc., Santa Clara, CA, USA), which included 30 gray matter samples from the CA1 region of the human hippocampus.The gray matter samples were collected from eight healthy control patients and 22 AD patients.For analysis, subjects were assigned to four groups reflecting different levels of AD severity based on the MiniMental State Examination (MMSE) criteria: eight as "Control" (MMSE > 25), seven as "Incipient AD" (MMSE 20-26), eight as "Moderate A" (MMSE 14-19), and seven as "Severe AD" (MMSE < 14).

Screening of differentially expressed genes (DEGs)
Gene expression analysis was conducted using the interactive web tool GEO2R (https://www.ncbi.nlm.nih.gov/geo/geo2r/).DEGs between patients with severe AD and healthy controls were identified.The criteria for DEGs were p < 0.05 and |log2 fold change| ≥ 1.5.

Network pharmacological analysis
The SwissTargetPrediction database (http://www.swisstargetprediction.ch),Comparative Toxicogenomics Database (CTD, http://ctdbase.org/),STITCH (Search Tool for InTeractions of CHemicals) database (http://stitch.embl.de/),and GeneCards database (http://www.genecards.org/)were accessed to predict potential targets of ATR, SIM, and PRO.The three-dimensional (3D) structures of ATR, SIM, and PRO were obtained from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) and searched against these four databases.Analyses of the SwissTargetPrediction and CTD databases were conducted with default parameters with the species specified as Homo sapiens.Analyses with the STITCH database were also conducted with default parameters with the species specified as H. sapiens and a minimum required interaction score of 0.4.
The Draw Venn Diagram tool (http://bioinformatics.psb.ugent.be/webtools/Venn/)was used to identify intersectional targets of triazine herbicides that may contribute to AD.Following identification of the intersectional targets, a protein-protein interaction (PPI) network was generated using the STRING (Search Tool for the Retrieval of Interacting Genes) database (http://string-db.org/) with the specified species as H. sapiens and a minimum required interaction score of 0.4.The PPI network was visualised using Cytoscape 3.6.0software (https://cytoscape.org/).

Gene ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) enrichment analysis
The Metascape database (https://metascape.org/)was used for enrichment analysis of GO terms and KEGG pathways (species H. sapiens), with p < 0.05 considered as significantly enriched.The Top 10 GO terms for the categories biological process (BP), cellular component (CC), and molecular functions (MF), as well as the top 10 KEGG pathways, were used for further study.

Core targets screening
The cytoHubba (http://hub.iis.sinica.edu.tw/cytohubba)topological analysis methods betweenness, closeness, degree, edge percolated component (EPC), maximal clique centrality (MCC), and maximum neighbourhood component (MNC) were selected to identify the top 10 hub targets.Then, the online Draw Venn Diagram tool was used to intersect the top 10 identified targets to obtain the core targets.

Molecular docking analysis
The 3D structures of ATR, SIM, and PRO obtained from the PubChem database were converted into the Mol2 format using Chimaera software 1.16 (https://www.cgl.ucsf.edu/chimaera/).The 3D structures of the core targets were retrieved from the Protein Data Bank (https://www.rcsb.org/)and the hydrogenation, dehydration, ligand removal, and other preparatory processes were assessed using Chimaera software.Molecular docking of the triazine herbicides and core targets to analyse binding activity was performed using the SwissDock online platform (http://www.swissdock.ch/).The results of molecular docking analysis were visualised with Chimaera software.

Molecular dynamics analysis
Molecular dynamics analysis was carried out using the Gromacs2022 program at constant temperature and pressure as well as periodic boundary conditions.The Amber99sb-ildn force field and the TIP3P water model were used.During the analysis, all hydrogen bonds involved were constrained using the LINCS algorithm with an integration step of 2 fs.Electrostatic interactions were calculated using the Particle-mesh Ewald method with a cut-off value set at 1.2 nm.The non-bonded interaction cut-off value was set at 10 Å and updated every 10 steps.The analysis temperature was controlled to 295 K using the V-scale temperature coupling method and the pressure was controlled to 1 bar using the Berendsen method.Then 100 ns of simulations were performed for the complex system and the conformation was saved every 10 ps.After the simulations were completed, plots were made using Python and Origin, and the Molecular Mechanics-Poisson Boltzmann Surface Area binding free energy analysis between the proteins and the triazine herbicides was performed using the g_mmpbsa program.

Construction of a herbicide-target-disease regulatory network
The herbicide-target-disease network was generated with Cytoscape 3.6.0software.The Network Analyser tool was used to score and rank nodes based on network features.

Screening of DEGs and network pharmacological analysis
In total, 20,788 mRNAs were identified between patients with severe AD and healthy controls by analysis of the GSE28146 dataset with the interactive web tool GEO2R.Of the 20,788 mRNAs, 1,062 were differentially expressed (567 upregulated and 495 downregulated, Figure 1A).Screening with the SwissTargetPrediction, CTD, STITCH, and GeneCards databases identified 3,063 targets of ATR (Figure 1B), 31 of SIM (Figure 1C), and 39 of PRO (Figure 1D).In addition, 3,076 unique targets of triazine herbicides were obtained after removing duplicates.Venn analysis identified 148 intersectional targets (66 upregulated and 82 downregulated, Figure 1E).The intersectional targets were used to construct a PPI network using the STRING database, which included 132 nodes and 473 edges with an average node degree of 7.17 (Figure 1F and Supplementary Table S1).

Screening of core targets
The top 10 targets were identified by the betweenness, closeness, degree, EPC, MCC, and MNC methods (Figure 3A-F and Supplementary Tables S4-S9).Venn analysis revealed that the top three core targets were thymidylate synthase (TYMS), fibronectin 1 (FN1), and epidermal growth factor receptor (EGFR) (Figure 3G).A PPI network of TYMS, FN1, and EGFR was constructed using the STRING database (Figure 3H).

Molecular dynamics analysis
Molecular dynamics analysis is used to validate the results of molecular docking.The ATR-EGFR, ATR-FN1, and ATR-TYMS complexes showed decreasing fluctuations in RMSD throughout the simulation, gradually reaching stability (Figure 5A-C).The SIM-EGFR complex reached stability after 60 ns with a plateaued RMSD (Figure 5D).There were fluctuations during the RMSD simulations of the SIM-FN1 complex, but overall it was smooth (Figure 5E).The SIM-TYMS complex showed smaller fluctuations before 40 ns and between 80-100 ns, indicating stability (Figure 5F).The PRO-EGFR complex remained relatively stable with small fluctuations after 70 ns (Figure 5G).The PRO-FN1 complex showed significant fluctuations, but 90 ns the fluctuations decreased and stability was achieved (Figure 5H).The PRO-TYMS complex showed small fluctuations up to 40 ns and then large fluctuations, suggesting that it was gradually becoming unstable (Figure 5I).

Construction of the herbicide-target-disease regulatory network
A herbicide-target-disease regulatory network based on the above results was generated to clarify the relationships among the triazine herbicides, targets, GO and KEGG pathways, and AD.The final network included 177 nodes and 906 edges, with an average number of neighbours of 10.24 (Figure 6).

Discussion
Although epidemiological studies have confirmed a strong correlation between environmental pesticide exposure and the onset of AD (Gauthier et al. 2001;Parrón et al. 2011), the underlying mechanisms remain unknown.The objective of this study was to elucidate the molecular mechanism underlying the potential contributions of environmental triazine herbicides (i.e.ATR, SIM, and PRO) to the development of AD.The results of the present study strengthen the argument that environmental exposure to triazine herbicides is a potential risk factor for AD.
Bioinformatics analysis of the gene microarray data file GSE28146 from the GEO database identified 1,062 DEGs (567 up-regulated and 495 down-regulated) between patients with severe AD and healthy controls.Then, network pharmacology analysis identified the action targets of the three triazine herbicides and further identified 148 intersectional with potential activities.In addition, GO and KEGG enrichment analysis of the intersectional targets was conducted, and three core targets were validated by molecular docking analysis.
KEGG analysis of the targets revealed that the signalling pathways of cell cycling and cellular senescence were significantly enriched.The cell cycle, which classifies the stages of growth and division, is composed of two principal stages: interphase, consisting of the G1, S, and G2 phases, and the mitotic phase (Fisher 2012;Matthews et al. 2022).AD is characterised by alterations to the normal cell cycle and increased neurogenesis before the onset of significant deterioration (García-Osta et al. 2022).Regrettably, the majority of the new neurons fail to fully develop and instead undergo cell death (García-Osta et al. 2022).A previous study reported that exposure to ATR at 100-500 μm for 24 h arrested the development of neural stem cells at the G1 phase and inhibited transition from the G1 to S phase via down-regulation of cyclin D1 and cyclin dependent kinases 2 and 4 (Shan et al. 2021).Although no prior study has linked exposure to SIM and PRO with alterations to the neuronal cell cycle, tome sequencing of male Xenopus laevis revealed that SIM at 100.9 µg/L for 100 days caused significant down-regulation of the cell cycle and expression levels of related genes in addition to reproductive toxicity (Sai et al. 2015).These findings suggest that exposure to combinations of triazine herbicides could cause AD via alterations to the cell cycle of neurons.
Cellular senescence, which is characterised by permanent cessation of cell growth, is thought to play a significant role in AD (Guerrero et al. 2021).Studies have reported the presence of senescent astrocytes, microglia, endothelial cells, and neurons in the brains of both AD patients and animal models of AD (Liu 2022).In senescent neuronal cells, mitochondrial malfunction is a common manifestation that leads to decreased energy production and oxidative stress (Martini and Passos 2023).Analysis of BLTK1 murine Leydig cells suggested that ATR, SIM, PRO, and related chlorometabolites may lead to abnormal cellular steroidogenesis via altered mitochondrial function (Forgacs et al. 2013).In addition, exposure of mice to SIM was reported to cause mitochondrial dysfunction, resulting in decreased maturation of oocytes and liver steatosis (Vancova et al. 2000;Shang et al. 2021).In addition, numerous in vivo and in vitro studies have demonstrated that ATR exposure can cause mitochondrial dysfunction in different cell types (Zhang et al. 2018;Saalfeld et al. 2019;Karadayian et al. 2023).A previous study by our group confirmed that ATR contributes to hippocampal neurotoxicity in rats via structural and functional changes to the mitochondria (Li et al. 2018(Li et al. , 2019)).Collectively, these findings suggest that mitochondrial dysfunction following exposure to triazine herbicides could be a major contributing factor to the pathogenesis of AD.
Finally, three possible key targets (i.e.TYMS, FN1, and EGFR) of herbicide-induced AD were screened by multiple analytical approaches.TYMS is an enzyme that plays an important role in the production of thymidine monophosphate, a crucial component of DNA synthesis (Zhang et al. 2022), and folate metabolism (Rahimi et al. 2021).Abnormalities in the above biological processes are closely associated with AD.FN1 plays central roles in various biological processes, such cell adhesion and migration, in addition to embryonic development (Cai et al. 2018).Previous studies have found that FN1 is also associated with lipopolysaccharide-induced oxidative stress of hippocampal neurons (Xie et al. 2020) and presents a novel serum biomarker for AD (Long et al. 2016).However, relatively few studies have investigated the association of TYMS and FN1 with the pathogenesis of AD.In the present study, molecular docking and molecular dynamics analysis showed that ATR, SIM, and PRO all have high potential for binding to TYMS and FN1, thereby providing clues for subsequent mechanistic studies.
EGFR is essential for early development of the brain.As the nervous system matures, expression of EGFR decreases.However, in response to decreased neural functioning and brain shrinkage, EGFR expression reappears in brain cells in order to maintain the number of neurons (Jayaswamy et al. 2023).Recent pharmacological studies for treatment of AD have focused on various drugs targeting EGFR to reduce the activities of reactive astrocytes, improve autophagy, alleviate the toxicity of amyloid beta, reduce neuroinflammation, and reverse axonal degradation (Mansour et al. 2021).Triazine herbicides are well-known environmental endocrine disruptors.Notably, ATR was reported to inhibit tyrosine kinase activities and is predicted to bind to the adenosine triphosphate pocket within the TK domain of EGFR (Hardesty et al. 2018).Moreover, ATR was reported to stimulate the EGFR signalling pathway and increase the rate of androgenesis induced by human chorionic gonadotropin in rat Leydig cells (Pogrmic-Majkic et al. 2016).However, no study has yet investigated the effects of SIM and PRO on EGFR activities.In the present study, molecular docking and molecular dynamics analysis confirmed that both SIM and PRO can bind to EGFR, suggesting that EGFR is a key target of AD induced by triazine herbicides.

Conclusion
In this study, multiple bioinformatics approaches were performed to investigate the potential molecular mechanisms of AD induced by exposure to combinations of triazine herbicides.The key signalling pathways included cell cycling and cellular senescence, while TYMS, FN1, and EGFR were identified as key targets of action.This study is only a preliminary exploration of triazine herbicide-induced AD.These results provide ideas for future vivo and in vitro studies.

Figure 1 .
Figure 1.screening of DEgs and network pharmacological analysis.(A) Volcano plot of DEms between AD patients and control.(B) The 3D structure of ATR.(C) The 3D structure of sim.(D) The 3D structure of PRo.(E) The Venn map of triazine herbicides related targets and AD related targets.(f) The PPi network.The width of each edge is proportional to the combined score, while the colour (from red to purple) and size of each node are proportional to interaction strength.

Figure 2 .
Figure 2. go and KEgg enrichment analysis.(A) KEgg pathway enrichment analysis.(B) BP enrichment analysis.(C) CC enrichment analysis.(D) mf enrichment analysis.Circos panels from outside to inside represent the following: go or KEgg terms, total gene number, enriched gene number, and enrichment factor.

Figure 3 .
Figure 3. screening of core targets.(A) Top 10 targets calculated through betweenness method.(B) Top 10 targets calculated through closeness method.(C) Top 10 targets calculated through degree method.(D) Top 10 targets calculated through EPC method.(E) Top 10 targets calculated through mCC method.(f) Top 10 targets calculated through mnC method.The width of each edge is proportional to the combined score, while the colour (yellow red to purple).(g) The Venn map of top three core targets.(H) The PPi network.

Figure 6 .
Figure 6.Construction of the herbicides-targets-disease regulatory network.The green rectangle represents triazine herbicides, the pink rectangle represents AD, the light purple circles represent KEgg pathways, the yellow circles represent BP terms, the light green circles represent CC terms, the light green circles represent BP terms, the orange circles represent hub targets, and the blue circles represent the other targets.