{"paper_id":"6894cf84-e401-403a-aef6-2cd2c9a37e0d","body_text":"Alzheimer’s disease (AD) is the most common form of dementia in the world and the fifth leading cause of death among older Americans [ 1 ]. AD impairs the central nervous system, including the cerebral cortex, hippocampus, dorsal raphe, and ventral tegmental area [ 2 , 3 ], and causes a progressive decline in cognitive functions, psychological changes, and sleep disorders [ 4 ]. Melatonin, secreted by the pineal gland, normally regulates the sleep–wake cycle but is greatly inhibited in patients with AD [ 5 , 6 , 7 ]. Consequently, melatonin has been prescribed to treat sleep disorders in AD patients. Some studies report that melatonin is also able to inhibit the progression of AD neuropathology and reverse cognitive impairment [ 8 , 9 ], and further investigations in systematic reviews and meta-analyses of randomized controlled trials support its potential as a promising therapeutic approach for improving cognitive decline in mild AD and mild cognitive impairment [ 9 , 10 , 11 , 12 ].\nAlthough the precise causes of AD are still unknown, clinical studies suggest the involvement of neuronal degeneration, the accumulation of abnormal protein turnover, disturbed neurotransmission systems/cell–cell communications, and energy metabolism/mitochondrial malfunction [ 6 ]. The characteristic hallmarks of AD are the presence of the abnormal accumulation of amyloid-beta (Aβ) plaques and neurofibrillary tangles (NFTs) [ 13 ]. In preclinical studies of AD, melatonin was shown to enhance Aβ lymphatic clearance in a transgenic mouse model of amyloidosis [ 14 ] and restore neurotransmission functions by regulating acetylcholine [ 15 , 16 ] and glutamate levels [ 17 , 18 ]. However, systematic analyses and explorations of gene–gene or protein–protein interaction networks of melatonin treatment in AD remain elusive.\nAdvancements in microarray and next-generation sequencing (NGS) technologies have allowed large genomic repositories for data mining, including the UniProt [ 19 ], Gene Expression Omnibus (GEO) [ 20 ], PubChem [ 21 ], and EMBOSS [ 22 ] databases, for analyzing differentially expressed genes (DEGs) [ 23 ] in various pathogenic conditions [ 24 ]. There are also many bioinformatics software tools and algorithms available to identify the gene targets and signaling pathways of melatonin in AD. In this study, we used different databases to mine the genes of interest and then performed a cluster analysis, Gene Ontology (GO) enrichment analysis, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis to identify the gene targets and associated pathways of melatonin treatment in AD.\n\nWe identified 3397 genes that were related to AD pathology from DisGeNet and 329 genes that were predicted to be targets of melatonin from ChEMBL. A total of 223 genes were found to be overlapped across both sets of genes. These genes were examined in the subsequent analyses. All genes are listed in  Supplementary Table S1 .\nThe general PPI network of AD protein-encoding genes was constructed ( Figure 1 ). This PPI network had 143 nodes and 823 edges, with an average node degree of 11.4 and PPI enrichment  p -value of <1.0 × 10 −16 . The clustered genes in the center with surrounding dense edges were considered to be promising therapeutic targets. These candidate genes included  CREB ,  MET ,  ERBB2 ,  AR ,  MME ,  MMP2 ,  IKBKB ,  HDAC1 ,  PIL3R1 ,  GSK3B ,  MAPK4 ,  ESR1 ,  PGR ,  ITGB1 ,  CAP3 ,  ATM , and  NR3C1 . On the bottom right of the network, we also observed a second cluster of genes with dense edges, which included  SLC6A4 ,  SLC6A3 ,  CHRNA3 ,  DRD2 , etc.\nThe melatonin–AD overlapping genes formed two clusters with K-core set to 6. Cluster 1 contained 16 genes, and Cluster 2 contained 18 genes ( Figure 2 a,b). Detailed information on each cluster and the corresponding gene lists are shown in  Table 1 .\nCluster 1, with 16 nodes and 68 edges, had a score of 9.067, indicating its robustness and significance, with  MMP2  as the key seed node ( Figure 2 a). The  MMP2  gene is located on chromosome 16 at position 12.2 and encodes for matrix metallopeptidase 2 (MMP2), a type IV collagenase that is involved in the breakdown of extracellular matrix (ECM) in normal physiological processes and acts as a novel H3NT protease possibly related to epigenetic modifications in AD [ 25 , 26 ]. Cluster 2 had 18 nodes and 60 edges, with a score of 7.059 ( Figure 2 b). The seed node in Cluster 2 was  NR3C1 , which is a gene located on chromosome 5. The  NR3C1  gene encodes for glucocorticoid receptor, which is involved in upregulating the expression of anti-inflammatory proteins in the nucleus and downregulating proinflammatory proteins in the cytosol [ 27 , 28 ].\nTo  further analyze the two clusters, we performed a GO enrichment analysis. In Cluster 1, we found 140 biological process (BP) GO terms, 1 GO term related to the cellular component (CC), and 41 GO terms related to molecular function (MF). The top 10 GO terms were visualized in Cluster 1 ( Figure 3 a) and Cluster 2 ( Figure 4 a). The connectivity of each gene to the top 10 BPs in Cluster 1 and Cluster 2 is shown in  Figure 2 c,d. The top 40 BP GO terms in Cluster 1 ( Table 2 ) and Cluster 2 ( Table 3 ) were also examined for a more comprehensive analysis.\nAmong the top 10 BP GO terms in Cluster 1, the most significantly enriched processes were related to steroid hormone signaling and apoptotic signaling pathways ( Figure 3 a). The most frequent genes in these pathways included  AR ,  XIAP ,  ES1 ,  CASP8 ,  PARP1 ,  PGR , and  HDAC1 . However, the seed node of Cluster 1,  MMP2 , was not observed in the top 10 BP GO terms.  MMP 2 was found in the mid-20 to mid–late 30 pathways related to cellular responses to UV, abiotic stimuli, environmental stimuli, and chemical stress, and female pregnancy. Among the top 10 BP GO terms in Cluster 2, the most significantly enriched processes were related to monoamine neurotransmission, such as dopamine and catecholamine. The most frequent genes in these pathways included  SLC6A4 ,  DRD4 ,  DRD2 ,  DRD1 ,  SLC6A3 ,  HTR2A ,  CNR1 ,  TPH1 , etc. The detailed gene connectivity to each biological process is shown in  Figure 2 d. However, the seed node of Cluster 2,  NR3C1 , was not observed even in the top 40 BP GO terms and was not found in any glucocorticoid pathways in Cluster 2.  NR3C1 , a glucocorticoid receptor, regulates stress response genes, while  MMP2 , an enzyme involved in extracellular matrix remodeling, plays a key role in tissue repair, highlighting their interplay in cellular stress and injury responses.\nThe KEGG pathway analysis identified 13 pathways in Cluster 1 ( Figure 3 b) and 8 pathways in Cluster 2 ( Figure 4 b). An analysis of the results showed that Cluster 1 genes were associated with apoptosis, whereas Cluster 2 genes were associated with monoamine neurotransmission, particularly serotonin and dopamine ( Figure 5 ).\n\nBy integrating the cluster analysis with GO and KEGG analyses, we identified that Cluster 1, with its  MMP2  seed node, mainly involved genes related to ECM breakdown and epigenetic modifications in AD, whereas Cluster 2, with its  NR3C1  seed node, mainly involved genes related to the regulation of glucocorticoid receptors. However, both  MMP2  and  NR3C1  were not in the top 10 BP GO terms of their respective clusters, and  NR3C1  was not even in the top 40 BP GO terms. This result suggests that these two gene targets have never been studied in association with melatonin treatment in AD, despite several studies reporting the potential role of  MMP2  and  NR3C1  in AD [ 29 , 30 ].\nMMP2  plays a vital role in tissue remodeling and turnover by specifically targeting and degrading proteins such as collagen and fibronectin [ 31 ].  MMP2  also plays a crucial role in epigenetic modifications related to AD by facilitating myogenic gene activation and Aβ degradation [ 25 ], as demonstrated by the effects of novel HDAC inhibitors that target  MMP2  activity [ 32 ]. Melatonin is primarily synthesized and secreted by the pineal gland in the brain and regulates circadian rhythms [ 33 ], sleep–wake cycles [ 34 ], and various physiological processes [ 35 ]. Melatonin has also been shown to regulate the expression of  MMP2 , mediated by SIRT1, a key enzyme involved in various cellular processes [ 36 , 37 ]. Melatonin can inhibit SIRT1 to modulate  MMP2  activity, which was shown to impact the degradation of the ECM and exert anti-proliferative effects on prostate tumor cells [ 36 , 38 ]. Moreover,  MMP2  has been linked to the regulation of neurotransmission [ 39 ] and is also involved in the remodeling of synapses via the degradation and turnover of ECM components within the synaptic environment [ 40 , 41 ]. Consequently,  MMP2  is involved in regulating the structural plasticity and functionality of synapses, and it plays a crucial role in shaping the efficiency and effectiveness of neurotransmission processes.\nNR3C1  functions as a transcription factor by modulating the expression of target genes upon binding with glucocorticoids [ 27 ]. Notably,  NR3C1  interacts with melatonin receptors [ 42 , 43 ], suggesting potential cross-regulation between glucocorticoid and melatonin signaling pathways [ 44 ]. Although the precise mechanisms and implications of this interaction are still under investigation, this underscores  NR3C1 ′s potential role in mediating the effects of melatonin in various cellular processes [ 43 , 45 ]. Additionally,  NR3C1  enhances  DRD2  expression in the miR-124-1+/− prefrontal cortex in mice, indicating its critical role in modulating neurotransmitter pathways and PFC function [ 46 ].  NR3C1  is expressed in various brain regions and is essential for regulating neurotransmission and neuronal circuit activity [ 47 , 48 ]; its dysregulation is linked to altered neurotransmitter systems, impaired synaptic plasticity, and increased susceptibility to neuroinflammation and oxidative stress [ 49 , 50 ]. Such impairments can significantly impact the overall efficiency and integrity of neurotransmission processes, potentially contributing to the development of neurological disorders such as AD. Increased  NR3C1  activation has been shown to cause neuronal damage and cognitive decline [ 51 ], modulate Aβ metabolism, and influence neuroinflammation [ 52 ], further linking it to AD pathogenesis. Epigenetic modifications, such as DNA methylation in the  NR3C1 , have also been associated with altered hypothalamic–pituitary–adrenal axis function [ 53 , 54 ] and increased stress vulnerability [ 55 ], both of which may heighten the risk of AD. These findings highlight  NR3C1  as a critical factor in understanding AD and underscore its potential as a therapeutic target.\nMMP2  and  NR3C1  have been identified to have significant associations with melatonin, and they were shown to be differentially expressed in AD in a preclinical study. Increased levels of  MMP2  have been observed in postmortem AD brains [ 56 ] and in the cerebrospinal fluid of AD patients [ 57 ], indicating its potential as a biomarker for disease progression. Animal models of AD frequently exhibit altered  MMP2  expression, highlighting its connection to Aβ pathology and blood–brain barrier disruption [ 58 ]. Furthermore,  MMP2  plays a crucial role in Aβ clearance, neuroinflammation, and synaptic plasticity [ 59 ], making it a vital factor in maintaining neuronal health. In AD,  MMP2  was shown to degrade Aβ via cleaving soluble Aβ peptides, was upregulated around astrocytes in AD brain [ 60 ], and was shown to accumulate near NFTs to eliminate the toxic form of tau [ 61 ]. Melatonin supplementation has been reported to be effective in increasing  MMP2  activity in gastric ulceration [ 23 ] and decreasing  MMP2  expression in cancer stem cells and SKOV3 cells [ 62 ]. However, the effect of melatonin treatment on  MMP2  expression in AD has never been studied. The GO enrichment analysis suggested that Cluster 1 was enriched in steroid hormone signaling and apoptotic signaling pathways.  MMP2  has been found to be associated with changes in steroid hormones in endometriosis sera and peritoneal fluid [ 63 ].\nAccording to the results of the top 40 BP GO terms,  MMP2  is also associated with female pregnancy. This might indicate the possible mechanism of melatonin treatment in AD via manipulating estrogen and progesterone levels. Existing studies have found that melatonin can inhibit estrogen receptor transactivation in breast cancer stem cells [ 64 ] and induce progesterone production in human granulosa lutein cells [ 65 ]. Estrogen and selective estrogen receptor modulators have both beneficial and harmful roles in AD [ 66 ]. On the other hand, progesterone has been found to have neuroprotective effects [ 67 ]. These lines of evidence highlight the need to further investigate the target genes identified in the cluster analysis.\nIn Cluster 2, the GO enrichment analysis also indicated that melatonin can interfere with dopaminergic and catecholaminergic neurotransmission. Dopamine was found to be downregulated in the hippocampus and VTA and upregulated in the frontal cortex and SNR networks in a mouse model of AD [ 6 ]. Melatonin was also shown to reduce the dopamine content in the neuro-intermediate lobe of male hamsters [ 68 ]. Melatonin was also able to suppress catecholamine synthesis by inhibiting the MT1-cAMP pathway in adrenomedullary cells [ 44 ] and promoting Smad signaling [ 44 ]. However, the molecular mechanism of how melatonin regulates these two monoamine neurotransmission pathways remains unclear. Notably, the seed node  NR3C1  in Cluster 2 was not observed in any pathway of the top 40 BP GO terms.  NR3C1  encodes glucocorticoid receptors, which are widely expressed in the brain. Glucocorticoids and corticotropin-related hormones can induce AD-associated pathologies [ 69 ]. Although melatonin has been found to regulate glucocorticoid receptors [ 70 ], whether there is melatonin and glucocorticoid receptor crosstalk in AD is currently not known. Based on the GO results, the involvement of glucocorticoid receptors in monoamine neurotransmission signaling pathways in AD should be further investigated.\nMelatonin can help mitigate AD symptoms by enhancing neurotransmitter balance and reducing neuroinflammation [ 6 , 35 ], both of which are crucial for maintaining cognitive function. Specifically, melatonin could influence dopaminergic and catecholaminergic neurotransmission, potentially addressing behavioral and mood-related aspects of AD [ 71 , 72 ]. Moreover, the modulation of stress-related hormones by  NR3C1  could further protect neuronal health and support cognitive resilience [ 73 , 74 ]. Also, we should acknowledge that the STRING network likely treats interactions as binary (present/absent), which limits its ability to capture the variability in interaction strength, context, and biological relevance, highlighting the need for future analyses to incorporate models that reflect these complexities. Further experimental and clinical research is needed to validate these interactions to ultimately clarify the therapeutic potential of melatonin in AD.\nIn summary, we identified a total of 3397 genes related to AD from DisGeNet and 329 melatonin target genes from ChEMBL, revealing an overlap of 223 genes. A cluster analysis highlighted two key clusters centered on the  MMP2  and  NR3C1  genes, which both play crucial roles in steroid hormone signaling, apoptosis, and monoamine neurotransmission. Our findings provide important gene targets for future research on melatonin treatment in AD, paving the way for further investigations into their roles in AD pathophysiology.\n\nGenes related to AD were retrieved from DisGeNET ( https://www.disgenet.org ; accessed on 15 June 2021), a publicly available database of human disease-related genes curated from Genome-Wide Association Studies (GWASs) and the scientific literature [ 75 ]. Gene targets of melatonin were obtained from the ChEMBL database, a database that includes drug, chemical, bioactivity, and genomic data. ChEMBL also has a unique target prediction function that uses quantitative structure–activity relationship (QSAR) models and conformal predictors to predict the gene target of a drug of interest [ 76 ]. DisGeNET was chosen for its comprehensive gene–disease associations, while ChEMBL provides detailed data on bioactive compounds, offering insights that are not as effectively covered by UniProt and GeneCards.\nWe retrieved all genes related to AD and then used STRING to study the functional associations between proteins. STRING is a resource that can predict protein–protein or functional protein-encoding gene interactions [ 77 ]. The network was generated based on “homo sapiens” using a minimum required interaction score cutoff of 0.4 to balance the inclusion of relevant interactions while minimizing false positives, ensuring moderate to high confidence in the results. After network generation, the genes in the center with dense surrounding edges represented the gene hallmarks of AD and potential therapeutic targets. These genes are integral to critical pathways regulating cellular functions. The interconnections imply that targeting a central gene could yield broader therapeutic effects by influencing multiple related pathways in AD. Additionally, central genes often play pivotal roles in disease mechanisms, enhancing their potential as therapeutic targets. Their evolutionary conservation further underscores their importance in understanding and addressing AD.\nWe performed a simple manual comparison and filtering of the AD gene list and melatonin gene targets, identifying overlapping genes. Cytoscape 3.8.2 was used to segregate the overlapping genes into clusters with the K-core cutoff set at 6; this was chosen to ensure that only the most interconnected genes were included, thereby highlighting the core components of the network that are likely to play critical roles in the biological processes under investigation. Each cluster represents a sub-set with the highest connectivity in terms of their involvement in the same biological processes or the same specific functions [ 78 , 79 ]. The results were further processed by the MCODE module in Cytoscape to analyze clusters in the network, utilizing the default parameters to ensure a standardized approach in identifying the most densely connected regions of the network.\nA Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were performed separately on each cluster in R 4.1.1 version. The GO analyses utilized the clusterProfiler package 4.14.0 version. The  p -value thresholds of ≥0.01 for the top 15 results of the GO term components and the top 20 KEGG pathways were optimized based on preliminary analyses. We also selected the top 40 results for each cluster to avoid missing any important findings. Finally, we compared the GO and KEGG results with the cluster analysis results and AD PPI results to reveal new gene targets and pathways for future study.","source_license":"CC-BY-4.0","license_restricted":false}