Alzheimer’s disease (AD) is the most common and complex neurological disease in the world. Neurological conditions such as AD are the leading cause of 70% of dementia worldwide (Mayeux and Stern, 2012). Global Alzheimer’s cases are increasing year by year and are expected to reach 78 million by 2030. While the pathogenesis of the disease is not yet known, many believe the accumulation of over-phosphorylated tau and Aβ plaques in neurofibrillary tangles are the main causes of AD (Andreasen et al., 1999; Shao et al., 2011; Arvanitakis and Bennett, 2019). When extracellular amyloid plaques build up in specific areas of the brain, they can lead to amyloid vascular disease or neurodegenerative diseases (Thanvi and Robinson, 2006; Jäkel et al., 2022). The neurofibrillary tangles (NFTs) are huge paired intracellular helical strands of hyperphosphorylated tau proteins, which induce neuronal and synaptic loss (Moloney et al., 2021). The predominant regions of the pathological process underlying AD in human brain are the association areas of the cerebral cortex and the hippocampus (Wang et al., 2010). Despite these extensive findings on both, there are few effective drugs to improve and treat AD. Worse, many patients have to wait a long time to be diagnosed with the disease. During this period, the Aβ burden is significant and memory loss has already occurred (Arvanitakis et al., 2019; Long and Holtzman, 2019). Typical AD has gone through a gradual and hidden development process, and there is no specific detection method, which cannot be accurately diagnosed, difficult to cure, difficult to control, and poor prognosis. Therefore, it is urgent to explore new potential biomarkers for early diagnosis and effective treatment of AD.

Alzheimer’s disease is the most common neurodegenerative disorder with limited therapeutics, and AD is characterized by the formation of plaques made by protein aggregates. Mounting studies have suggested that targeting transcription factors holds promise for treating neurodegenerative disorders including AD (Ping Yang et al., 2022). CMV promotor-driven transcription factor EB (TFEB), for example, is injected with adeno-associated viral particles in targeted mice, which are mainly localized in neuronal nuclei and upregulated lysosomes. This resulted in decreased steady state levels of Aβ in APP proteins and cerebral stromal liquid (Xiao et al., 2015; Song et al., 2020). A growing body of evidence also suggests that the accumulation of misfolding proteins in AD is caused by damage to macromolecular autophagy and autophagy lysosomal pathways, making TFEB, which regulates autophagic lysosomal pathway, a promising target for AD treatment (Zheng et al., 2021). We also found decreased expression of the transcription factor Nrf2 and its NQO1, HO-1, and GCLC driver genes, and changes in related Nrf2 pathways in AD brains. Nrf2 activation may provide cellular protection and prevent an increasing number of diseases including neurodegenerative diseases. These avenues of evidence point to the activation of the Nrf2 transcription factor as a potential new therapeutic scheme for AD (Osama et al., 2020). Additional research shows that ATF6 transcription factor activation can attenuate amyloidosis through BACE1 downregulation, and concomitantly ATF6 overexpression significantly reduces Aβ1-42. The results of this study suggest ATF6 may become a potential focus for targeted therapy of AD. Interestingly, TBP and CDk7 also belong to the basal transcription factor family like ATF6 among the 10 key genes selected in our study (Geng et al., 2016; Du et al., 2020).

Recently, progress has begun to clarify the physiological and pathological roles of non-coding RNAs (ncRNAs) in a variety of diseases, including cancer. ncRNAs do not have protein coding functions, but are important regulators of many cellular processes (Esteller, 2011). Of these, miRNAs are the most studied and have become key players in the pathogenesis of AD involved in regulating key growth regulatory pathways (Swarbrick et al., 2019; Toden et al., 2021). miRNAs are small endogenous non-coding RNA molecules that can inhibit or silence the expression of post-transcription genes. Many miRNAs play an important role in post-transcriptional regulation and are highly conserved (Sand et al., 2009). Previous research has found that numerous miRNAs, including miRNA-126a-3p, MicroRNA-455-3p, miR-501-3p, and miRNA-101a-3p, have a role in the pathogenesis of AD, showing that miRNAs are strongly linked to the development of AD (Hara et al., 2017; Kumar et al., 2017; Kumar and Reddy, 2019; Lin et al., 2022; Xue et al., 2022). However, the specific role and potential mechanisms of miRNA in AD pathogenesis are unclear.

We here established a miRNA-mRNA, TF-mRNAs regulatory network by integrating relevant TFs (transcription factors), miRNAs and mRNAs to try to gain further insight into the potential functions of ncRNAs as well as transcription factors in AD. This study may provide a better understanding of the underlying pathogenesis of AD and potentially new biomarkers for diagnosis and treatment of AD.

Due to the heterogeneity of AD pathology, there is a lack of sufficient efficacy in treating AD (Lam et al., 2013; Byun et al., 2015). In the past decades, neurodegenerative therapy for AD has made more and more progress, and drug resistance is often allowed in traditional histology (Nandigam, 2008). Hence, identifying more appropriate molecular regulation of transcriptional clusters is critical to guiding personalized AD treatment. In this work, the top 7,200 genes based on MAD values were chosen for WGCNA analysis. The “Midnight” module of GSE1297 was chosen as it is significantly associated with clinical AD (Figure 2). Subsequently, according to the cytoHubba plug-in of Cytoscape, we made use of the algorithm Between Centrality to screen 10 key DEGs as hub genes in the PPI network, including CCNB1, GRIA1, TBP, MAPT, PPP2R2D, CDK7, USO1, PRPF19, GRM5, and MRPS15. It’s suggesting that these genes may play important role in the mechanism of AD and the specific flow chart is shown in Figure 9G.

Luckily, we discovered that in GO analysis, GRM5 and GRIA1 were predominantly enhanced in the synaptic membrane of CC (cellular component), Neuron to Neuron synapse, postsynaptic membrane, postsynaptic density, and postsynaptic specialization. In BP (biological process), GRM5 and GRIA1 were predominantly enriched in the modulation of chemical synaptic transmission, regulation of trans-synaptic signaling, and regulation of membrane potential. More and more studies have shown that neuron synapse and AD are closely related, synapse loss and Tau pathology are hallmarks of AD and other tauopathies (Yoshiyama et al., 2007; Skaper et al., 2017; Dejanovic et al., 2018; Kurucu et al., 2022). In KEGG pathway analysis, GRM5 and GRIA1 are co-enriched in glutamatergic synapse and retrograde endocannabinoid signaling pathways. TBP, GRM5, and GRIA1 were co-enriched in Huntington’s disease pathway. Both Huntington’s disease (HD) and Alzheimer’s disease (AD) are similar, albeit clinically distinct, neurodegenerative disorders that share pathologic features related to selective brain injury (Naia et al., 2017; Mukherjee, 2021). TBP and CDK7 are basal transcription factors among them. In addition, through FpClass analysis, we know that TBP and CDK7 have high reliability based on gene expression and topological network analysis, and GRM5 and GRIA1 also have high scores (Supplementary Data 7). In the verification in the GeneMANIA database, they still have the same relationship network (Supplementary Data 5, Figure A).

All three nuclear RNA polymerases have a subunit of the TATA binding protein (TBP) complex required for transcription (Davidson et al., 2004). In all fields of life, transcription regulation by DNA-dependent three-types RNA polymerase is largely achieved at the initial level. Among them, TBP is the most conservative initiation factor, which is crucial to the transcription initiation of all ancient eukaryotes, and is the only initiation factor required for the complete start of RNA polymerase in all eukaryotes (Kramm et al., 2019; Mishal and Luna-Arias, 2022). Back in 2004, they demonstrated significant differences in TBP between the AD group of normal subjects and patients. In addition, their number and distribution are not proportional to the number and the distribution of positive tau or β-amyloid structures, and the TBP accumulates in AD brains, suggesting that TBP might be a contributing factor to AD due to its own entanglements (Reid et al., 2004). Evidence implicating TBP in the molecular mechanism of several neurodegenerative diseases has emerged in the past few years. TBP may contribute to these diseases through a loss of normal function (likely to be catastrophic to a cell) which affects neurodegenerative diseases, such as AD and Huntington’s disease (van Roon-Mom et al., 2005; Bech et al., 2010).

The transcription cycle of RNA polymerase II is regulated by a group of cyclin dependent kinases (CDK). CDK7 related to TFIIH, a transcription initiation factor, is not only an effector of RNA polymerase II phosphorylation and other targets in the transcription mechanism, but also a CDK activated kinase involved in transcription, and also plays a key role in regulating eukaryotic cell division (Sansó and Fisher, 2013; Fisher, 2019). There is growing evidence that CDK regulates the transcriptional cycle of RNA polymerase II. Specific CDKs are considered as important molecular mechanisms in the transcription cycle, and describe that recently emerged transcriptional CDK can serve as promising drug targets in cancer (Fisher, 2017; Parua and Fisher, 2020). Moreover, some studies show that neuronal activity of CDK7 in hippocampus is related to aging and AD (Zhu et al., 2000). CDK7 is also required for activity-Dependent expression of Neuronal genes, synaptic plasticity at long-term, and memory at long-term (He et al., 2017). It can be hypothesized that CDK7 is a promising drug target for AD.

Although GRM5 (coding for metabotropic glutamate receptor 5, mGluR5) is a promising target for treating cognitive deficits in schizophrenia and AD, its association with cognitive and brain phenotypes within this disorder has received little attention. Early researchers found that the common genetic variation of GRM5 in schizophrenic patients would affect cognitive function, hippocampal volume and hippocampal mGluR5 protein level compared with the healthy control group, among them, the metabotropic glutamate receptor subtype 5 (mGluR5) is encoded by GRM5 gene, which is an attractive new drug target for the treatment of schizophrenia (Matosin et al., 2018). MGluR5 is a postsynaptic G-protein coupled glutamate receptor, which is closely related to several key cellular processes destroyed in schizophrenia. In preclinical models of schizophrenia, positive mGluR5 modulators have shown encouraging therapeutic potential, especially in the treatment of cognitive dysfunctions (Matosin et al., 2017). Early in the course of AD inhibition, some have attempted to do so by downregulating cholinergic receptors and glutamate receptors, which, as the disease progresses, play key roles in inflammation and oxidative stress, thereby participating in the neurodegenerative process. Accumulating evidence suggests that perturbation of the excitatory amino acid L-glutamate (L-Glu) in systems may underlie the pathogenesis of hypoxia ischemia, epilepsy, and chronic neurodegenerative diseases (e.g., Huntington’s disease and AD) (Hynd et al., 2004; Schaeffer and Gattaz, 2008). It is reported that glutamate excitatory neurotransmission is an important process in learning and memory, which is seriously damaged in AD, possibly due to β Amyloid peptides (1–42) increase in relation to oxidative stress. The researchers also found that some glutamate receptors are overactivated in AD. This sustained mild activation may lead to neuronal damage and impaired synaptic plasticity (learning). More and more evidence shows that glutamate mediated neurotoxicity is involved in the pathogenesis of AD, and this metabolic glutamate receptor (GRM5), whose signal transduction will activate the second messenger system of calcium phosphatidyl inositol, will also participate in the regulation of neural network activity and synaptic plasticity, which makes us speculate that GRM5 may be a very promising target for exploring treatment and improving AD disease (Danysz and Parsons, 2003; Doraiswamy, 2003; Tanović and Alfaro, 2006; Świetlik et al., 2022).

GluA1 (also known as GluRA or GluR1) is a glutamate receptor subunit of AMPA encoded by the GRIA1 gene, there is genome-wide association between GRIA1 and schizophrenia (Barkus et al., 2014; Ang et al., 2018). Hippocampal proteomics study has pointed out that proteins involved in neuronal excitability and synaptic plasticity (e.g., GRIA1, GRM3, and SYN1) were altered in both “normal” aging and AD (Neuner et al., 2017). Similarly, a bioinformatics study identified 464 differentially expressed genes that were modulated by silent transcription factor (REST) between AD patients and controls. REST is strongly associated with glutamatergic synapses and long-term potentiation, and among them, GRIA1 shows a significant difference in its tendency to change with REST (Xu et al., 2021). From this, we speculate GRIA1 also may be regulated by other transcription factors, thereby affecting the progression of AD disease. From RNA-Seq Expression Data from GTEx (53 Tissues, 570 Donors), the highest median expression of GRM5 and GRIA1 all in Brain—Cerebellum (Supplementary Data 5, Figure B).

A microRNA (miRNA) is a short non-coding RNA with regulatory functions in a variety of biological processes, which has been implicated in many cellular processes including cell proliferation, apoptosis, gene expression, cellular differentiation, and development (Krol et al., 2010). Many studies have shown that miRNA often binds to target mRNA through complementary sequences and induces translational inhibition or target degradation as a negative regulation of mRNA expression (Towler et al., 2015). We tried to use the ENCORI platform to predict the miRNAs that regulate GRM5 and GRIA1, respectively, and then intersected with 82 up-regulated DEmiRNAs in the blood samples of the previous experiment to determine the hub miRNAs with significant differences. Respectively, hsa-miR-186-5p and hsa-miR-425-5p were identified to be potentially involved in the regulation of GRM5 and GRIA1. Among them, As shown in Figures 8A,B,D, from the AD transformation state diagram and ROC curve, we identified two significantly up-regulated miRNAs, hsa-miR-186-5p and hsa-miR-425-5p may with good prognostic values. Recent studies have found that hsa-miR-425-5p and hsa-miR-186-5p can be used as therapeutic targets for other diseases. For example, hsa-miR-425-5p may promote tumor occurrence and metastasis by activating CTNND1 related pathway (Liu et al., 2020), hsa-miR-186-5p regulates TGFβ signaling pathway through expression suppression of SMAD7 and SMAD6 genes in colorectal cancer (Bayat et al., 2021). In our study, GRM5 and GRIA1 were found to be significantly downregulated at mRNA level and regulated by the up-regulated hsa-miR-186-5p and hsa-miR-425-5p, respectively (Figures 4D,E). It has been reported that negatively regulated miRNA-mRNA pairs contribute to the improvement of AD and provide new reliable targets for the treatment of AD (Fu et al., 2019; Qian et al., 2019; Chen et al., 2021).

Anomalous aggregation of microtubule associated tau protein (MAPT) is a prominent pathological feature in various neurodegenerative diseases including AD (Zhou and Wang, 2017; Strang et al., 2019). From in Figure 9, we can observe that the correlation coefficients of GRM5 and GRIA1 with MAPT were 0.69 and 0.18, respectively, with P < 0.05. A reliable index of AD-related cognitive state at a particular time point is the MiniMental State Examination (MMSE) (Clark et al., 1999). Furthermore, dementia-NFT of the senile type has been termed tangle-only dementia, NFT-predominant form of SD and limbic NFT dementia (Hyman, 1997). Therefore, MMSE and NFT were chosen as our main markers to quantify AD progression (Schmitt et al., 2000; Yamada, 2003). Since NFT scores increase and MMSE scores decrease with AD severity, genes upregulated with AD could only correlate positively with NFT or negatively with MMSE, whereas genes downregulated with AD could only positively correlate with MMES or negatively correlated with NTF scores. In our study (Figure 10), the significantly down-regulated gene GRM5 was positively correlated with MMSE with a correlation coefficient of 0.58 and negatively correlated with NFT with a correlation coefficient of −0.4, GRIA1 and MMSE were also positively correlated with a correlation coefficient of 0.52 and negatively correlated with NFT with a correlation coefficient of −0.5. At the same time, their p < 0.05. As shown in Figures 9A–F, we also can find that the transcription factors TBP and CDK7 are well correlated with their downstream regulated genes. For the transcription factors TBP and CDK7, and the differential expression changes of their regulated target genes GRM5 and GRIA1. It can be seen from Figures 5, 6 that the expression of CDK7 and GRIA1 in AD group is two times lower than that in normal group in two databases. The expression of TBP and GRM5 also decreased significantly compared with the normal group. Grouped according to gene expression profiles and phenotypes, related pathways and molecular mechanisms were evaluated by using GSEA software to validate the regulate of hsa-miR-425-5p, hsa-miR-186, and TBP to target genes GRIA1 and GRM5, respectively (Figure 11).

We also performed regression analysis using Lasso-Cox to construct and validate a prognostic signature associated with the four key genes, which had significant predictive value for AD patients and achieved high AUC values close to 90% in both databases (Figure 12). In addition, we screened 15 drugs targeting key genes from the DrugBank database, most of which are used to relieve pain, systemic anesthetics, inhibit signaling, inhibit cyclin-dependent kinases, inhibit multiple enzyme targets (including CDK7) and change the growth phase of treated cells, treat depression, it is used in potassium channel blockers, ameliorates multiple sclerosis (MS), and so on. We can try to study these drugs, and maybe they may help us to alleviate and treat AD (Supplementary Data 7). In conclusion, based on the bioinformatics analysis in this study, we propose GRM5 and GRIA1 as novel potential prognostic markers of AD and suggest miRNAs and transcription factors that may be related to AD and regulate GRM5 and GRIA1. We hope our findings will inform future research to improve outcomes for patients and try to mitigate and treat AD.

Our research identified a miRNA/TF-gene network that is potentially relevant for AD. The four hub genes (including TBP, CDK7 GRIA1, and GRM5) were markedly downregulated, which may have a critical influence on the pathophysiological mechanism of AD. Two potential target miRNAs (hsa-miR-425-5p and hsa-miR-186-5p) were furthermore forecast. The significantly down-regulated transcription factors TBP and CDK7 may be involved in the transcription of GRM5 and GRIA1, thereby affecting their expression and thus the progression of AD. TBP accumulating in the AD brain, localizing to neurofibrillary tangle structures, may be a contributing factor in AD. CDK7 is necessary for Activity-Dependent Neuronal Gene Expression, Long-Term Synaptic Plasticity, and Long-Term Memory, some studies have demonstrated that CDK7 neuronal activity in the hippocampus is linked to aging and AD. The downregulation of both TBP and CDK7 makes GRM5 and GRIA1 possible to cause neuronal damage and impaired synaptic plasticity in the pathogenesis of AD. As the metabotropic glutamate receptor encoded by GRM5 gene and the AMPA glutamate receptor subunit encoded by GRIA1 participate in the perturbed glutamatergic neurotransmission. Meanwhile, up-regulated hsa-miR-425-5p and hsa-miR-186-5p inhibited the transcription of GRIA1 and GRM5 by binding to them. These findings may contribute to the early diagnostic strategies, treatment targets and prognostic markers of AD disease.

The original contributions presented in this study are included in the article/Supplementary material, further inquiries can be directed to the corresponding authors.

QZ and PY designed the research study and wrote the original article. QZ, WG, and PY performed the data analysis. QZ, XP, YS, YW, and CP participated in the conception of the study and revision of the manuscript. YW was responsible for guiding the research framework, including concepts and theories. All authors approved the final version to be published.