Alzheimer's disease

Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized primarily by cognitive impairment and behavioral disturbances, commonly occurring in the elderly or late-middle-aged individuals[1, 2].

Clinical symptoms of AD include memory loss, aphasia, apraxia, agnosia, and visuospatial deficits, often accompanied by changes in personality and behavior[3-5]. Previous studies suggest that the development of AD results from a combination of genetic, epigenetic, and environmental factors[6,7].

Given the unclear mechanisms underlying AD and the lack of effective treatments to cure or substantially reverse the disease, the search for effective therapies continues[8]. Pathologically, Alzheimer's disease patients exhibit the accumulation of large amyloid-beta (Aβ) plaques and neurofibrillary tangles (NFTs) in their brains, accompanied by neuroinflammation, synaptic dysfunction, mitochondrial and bioenergy disturbances, and vascular abnormalities[9].

Major depressive disorder

Major depressive disorder (MDD) is a mental health condition characterized by significant physical changes, such as fatigue, weight loss, and decreased appetite, alongside widespread episodes, low treatment adherence, and high relapse rates. Approximately 300 million people worldwide are affected by MDD, making it one of the leading causes of disability [10].

The clinical symptoms of MDD include persistent low mood, loss of interest, changes in weight or appetite, and an increased risk of suicide [11]. However, due to the lack of in-depth understanding of pathogenesis and objective diagnostic evidence for MDD, it is difficult to identify and prevent early in clinic[12].

Due to the complexity of the pathophysiological mechanisms underlying MDD, accurate diagnostic methods and pharmacological treatments are relatively limited. Genetic factors, stress, comorbidities, and endocrine imbalances are considered to be related to the pathogenesis of MDD[11, 13].

Relationship between gut microbiota and AD, MDD

In recent years, the gut-brain axis has gained significant attention for its role in maintaining homeostasis and contributing to disease development. The gut microbiota, often referred to as the "second brain," interacts with the central nervous system through various pathways, including the immune system, tryptophan metabolism, the vagus nerve, and the enteric nervous system. These interactions involve microbial metabolites such as short-chain fatty acids, branched-chain amino acids, and peptidoglycans[14]. It is worth noting that emerging studies have suggested that intestinal microbial metabolites play an important role in mediating the development of AD [15-17] and MDD [18-20] by influencing central nervous system receptors and cellular microenvironment. Based on literature research, nNOS and lilrB3 were selected as potential key targets for MDD and AD(Project. Background).

Therefore, in this study, we plan to screen and identify key gut microbiome metabolite molecules that interact with AD and MDD disease receptors nNOS and LilrB3, using computer-aided drug design (CADD) methods and computational chemistry based on the gut microbiome metabolic molecule library. In addition, we evaluated the upstream mechanisms by which key molecules of gut metabolites influence disease development through metagenomic sequencing of gut microbiota, combined with multi-omics databases for case-control studies of patients with AD and MDD.

1. Small-molecule metabolite libraries of the intestinal microbiota

A total of 1723 Human gut microbiota metabolite molecules were obtained from the Virtual Metabolic Human database (https://www.vmh.life/#microbes/metabolites)[21]. The metabolite molecular structure formula was obtained from PubChem, and manually drawn using Chemdraw (version 20.0). The format was converted through Openbable software (version 3.1.1).

2. Virtual screening

PyRx software (version 0.8) was used for virtual screening[22]. The predicted 3D structure files of nNOS (AF-P29475-F1-model_v4) was obtained from AlphaFold Protein Structure Database[23, 24], and lilrB3 (8GRX) protein was obtained from PDB database[25, 26]. The ligands and receptors were pretreated with Pymol for dehydrating, hydrogenation and charge balance. Batch virtual filtering through the Vina Wizard in PyRx software, with getbox set to automatically maximize. Binding Affinity, rmsd/ub, rmsd/lb and other results were exported for subsequent analysis.

3. Molecular docking

Autodock software (version 4.2.3) was used for molecular docking[27]. The specific operation is as follows: After preprocessing receptors and ligands, Autogrid is carried out according to the software instructions, with getbox set to automatically maximize. Autodock operation was performed based on the gpf file generated by Autogrid. Set the Number of GA Runs to 200 and the Maximum Number of evals to 2,500,000, the Maximum Number of generations to 27,000. The original file of the conformation of the docking result and Binding energy data are exported and visualized using PyMOL software (version 2.5.8).

4. GWAS data acquisition and analysis

Genome-wide Association Study (GWAS) sequencing data were obtained from the GWAS catalog . GWAS sequencing data of AD were obtained from GCST002245[28], GCST90079819[29], GCST90083805[29], GCST013196[30], GCST013197[30], GCST90276158[31], and data of MDD were obtained from GCST90078468[29], GCST90038650[32], GCST90078469[29], GCST90078637[29], GCST90043709[33] comprehensively. The R studio (version 4.3.1) was used for data quality control, and MAGMA (version 1.07), a multivariate regression model-based gene and pathway analysis tool, was used to analyze key single nucleotide polymorphism (SNPs) in the GWAS summary data[34]. SNPS are annotated based on snp151_hg38 (GRCh38).

5. Metagenomic data acquisition and analysis

Metagenomic sequencing data of Gut Microbiota in AD and MDD population cohorts were obtained from ENA Browser and CNGBdb[35] (China National GeneBank DataBase) database.

The metagenomic sequencing data of 75 amyloid-positive AD patients and 100 cognitively healthy controls participating in the AlzBiom study[36], and data of 156 MDD patients and 155 healthy controls were obtained[37] and quality controlled.

Metagenomic data was processed using inStrain (version 1.9.0) [38]. inStrain is an open-source tool designed for analyzing and comparing genomic data, particularly aimed at tracking genomic variation in bacteria and other microorganisms. The analysis was done by introducing a publicly available reference genome, the UHGG genome run inStrain (version 1.9.0), a collection of all microbial species known to exist in the human gut[39]. The profile and compare functions of inStrain can obtain the expected output results, mainly generated in the form of tsv tables.

1. Identification of key drug molecules binding to target receptors of AD and MDD based on computer-assisted drug design

1.1 Virtual screening of small-molecule metabolites of the gut microbiota potentially bound to the target receptor

A molecular database of human gut microbiota metabolites was constructed and virtual screening was performed as described above (Methods 1-2). Using an absolute value of predicted binding affinity > 7.6 as the criterion, filter out 34 small molecules with potential interactions with nNOS and 33 small molecules with potential interactions with lilrB3. Considering indicators such as toxicity, off-target effects, and the feasibility of engineered bacterial metabolic pathways, a comprehensive assessment was performed (Table 1-2).

1.2 Molecular docking of potential metabolite drug molecules to the key receptors of AD and MDD

Based on the virtual screening results, some promising small molecules were selected for molecular docking to further explore their ability to bind to disease receptors. Autodock was used for molecular docking as described above (Methods 3).

We found that inosine showed the largest clustering center of conformation binding energy among the molecular docking results of nNOS, and tetrahydrofolic acid showed the largest among the molecular docking results of lilrb3 (Figure 1-2). By further analysis of binding conformation, we found that the chemical bonds of inosine and nNOS were mainly distributed in amino acids such as PHE-709, ARG-419, ASN-702, MET-337, GLU-637, etc. (Figure 3). In the binding of tetrahydrofolic acid and lilrb3, the chemical bonds are mainly distributed in ALA-93, GLN-205, ARG-95, TYR-96, SER-65 and other amino acids (Figure 4).

Fig 1.Clustering binding energy center of candidate intestinal microbiota metabolite molecules docking with nNOS molecules.

Fig 2.Clustering binding energy center of candidate intestinal microbiota metabolite molecules docking with lilrB3 molecules.

Fig 3.Conformation diagram of molecular docking between inosine ligand and nNOS receptor.

Fig 4.Conformation diagram of molecular docking between tetrahydrofolic acid ligand and lilrB3 receptor.

Further, the Surface Plasmon Resonance (SPR) technique was used to evaluate the affinity, affinity constant and binding kinetics of ligand-receptor binding (Wet lab. Results. SPR Biacore molecular interaction). Ultimately, tetrahydrofolic acid and inosine were selected as potential drug molecules with specific binding to receptor targets associated with Alzheimer's disease (AD) and major depressive disorder (MDD), respectively (Figure 5).

Fig 5.Molecular structure diagram of inosine and tetrahydrofolic acid.

2. Discovery of mechanistic evidence for key drug molecules in treating patients with AD and MDD based on multi-omics high-throughput sequencing data

2.1 Evaluating the single nucleotide polymorphisms of key molecular metabolites in intestinal metabolites of AD and MDD patients based on GWAS analysis

Based on the method described previously (Methods 4), MAGMA was used to analyze AD and MDD patient-control GWAS sequencing data to identify mutations in genes encoding metabolic enzymes associated with key metabolic molecules inosine and tetrahydrofolate in the gut microbiota. In the analysis of AD patients, the genes related to tetrahydrofolate metabolism DHFR (rs80650414, P = 2.98*10^-6), ALDH1L2 (rs374255885, P = 1.47*10-5), MTHFR (rs542060620, P = 3.96*10-^-5) were found to have significant SNPs (Figure 6). In addition, a strong effect mutation of APOE4 (rs1490374271, P = 1.95*10⁷) was verified. For the analysis of MDD patients, the gene related to inosine metabolism, PPAT (rs1491299790, P = 1.82*10^-6), NAMPT (rs1491400110, P = 9.3*10^-6), ADSS1 (rs1490898184, P = 1.08*10^-5) were found to have significant SNPs (Figure 7). These results suggest that the potential metabolic enzyme dysfunction caused by genetic variation of host metabolic enzymes is the cause of inosine and tetrahydrofolate deficiency.

Fig 6. Manhattan map for GWAS analysis of MDD.

Fig 7. Manhattan map for GWAS analysis of AD.

2.2 Evaluating the functional variation of Gut Microbiota metabolic enzymes in patients with AD and MDD based on metagenomic sequencing analysis

Based on the profile and compare functions in the inStrain analysis, the obtained table and image data can compare the differences between the sequences of the disease group and the control group. Specifically, it is whether there is a significant difference in the ANOVA analysis of variance of nucleotide diversity between the disease group and the control group and the proportion of variant sites in the disease group and the control group. The calculation formula for nucleotide diversity is: The calculation formula for nucleotide diversity is the sum of the square frequencies of each base: 1 - [(frequency of A)^{2 +} (frequency of C)^{2 +} (frequency of G)^{2 +} (frequency of T)²] Among them, the presentation of nucleotide mutation ratio is mainly divided into four parameters: con-SNV, pop-SNV, SNS and SNV. Pop-SNV takes into account minor alleles compared to con-SNV. The variant sites can be further verified through the differences of these parameters between the disease group and the control group and the ANOVA p-value.

For the AD population, the significance comparison and variation site distribution analysis of the ANOVA p-value of the nucleotide diversity of clpX, folA, folC, folE, folK and folP genes in the disease group and the control group were mainly conducted (Figure 8). For the MDD population, add, deoD, gsk and yfiH were mainly analyzed (Figure 9). A significant difference in nucleotide diversity, along with corresponding differences in the composition of variant types were observed when the P value is less than 0.05, suggesting that there are potential genetic mutations in key metabolic genes of gut microbiota.

Fig 8. Comparison of the ANOVA p-value for nucleotide diversity between the AD and control groups, along with the proportion of variant types in both groups. a-b. gene folE. c-d. gene folk. e-f. gene folP. g-h. gene folC. i-j. gene folA. k-l. gene clpX.

Fig 9. Comparison of the ANOVA p-value for nucleotide diversity between the MDD and control groups, along with the proportion of variant types in both groups. a-b. gene gsk. c-d. gene deoD. e-f. gene add. g-h. gene yfiH.

In this study, we focused on disease targets lilrB3 and nNOS for cognitively related diseases AD and MDD, identified key gut microbiota metabolite molecules tetrahydrofolic acid and inosine, and evaluated their potential as potential therapeutic agents. In addtion, we explain the upstream mechanisms of potential genetic variation leading to tetrahydrofolate and inosine deficiencies at the host and gut microbiota genomics levels. Based on the modeling and analysis work, combined with the wet experiment, we constructed the engineering bacteria that stably expressed tetrahydrofolic acid and inosine.