Abstract Alzheimer’s disease (AD) is a degenerative illness that accounts for the common type of dementia among adults over the age of 65. Despite extensive studies on the pathogenesis of the disease, early diagnosis of AD is still debatable. In this research, we performed bioinformatics approaches on the AD-related E-MTAB 6094 dataset to uncover new potential biomarkers for AD diagnosis. To achieve this, we performed in-depth in silico assays, including differentially expressed genes analysis, weighted gene co-expression network analyses, module-trait association analyses, gene ontology and pathway enrichment analyses, and hub genes network analyses. Finally, the expression of the identified candidate genes was evaluated in AD patients PBMC samples by qRT-PCR. Through computational analyses, we found that RN7SK LncRNA and its co-expressed genes of TNF, TNFAIP3, CCLT3, and FLT3 are from key genes in AD development that are associated with inflammatory responses. Our experimental validation revealed that RN7SK LncRNA and TNF were substantially up-regulated in AD samples (P = 0.006 and P = 0.023, respectively). Whereas, TNFAIP3 expression was significantly decreased (P = 0.016). However, the expression of CCL3 and FLT3 did not differ significantly between two groups (P = 0.396 and P = 0.521, respectively). In conclusion, in this study a novel LncRNA associated with AD pathogenesis were identified, which may provide new diagnostic biomarker for AD. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-024-82490-9. Keywords: Alzheimer’s disease, Bioinformatics, Biomarker, LncRNA, RN7SK, PBMC Subject terms: Computational biology and bioinformatics, Genetics, Neuroscience Introduction Alzheimer’s disease (AD), the most frequent type of dementia, is defined by impairment in cognition, function, and behavior, which often begins with memory loss of recent events^[34]1. Dementia affects around 55 million people in the world^[35]2. This statistic will virtually double every 20 years, extending to 82 million by 2030. As a result, this illness has far-reaching social and economic consequences^[36]3. AD is characterized by amyloid plaques, neurofibrillary tangles, soluble amyloid-β (Aβ) oligomers, synaptic loss, neuritic dystrophy, and neurodegenerative processes^[37]4. In clinical settings, the diagnosis of AD is generally based on cerebral spinal fluid (CSF) biomarkers or positron emission tomography (PET) imaging, and it is usually established only after cognitive symptoms have started to appear. The main drawbacks of these diagnostic methods are their invasiveness, high expense, poor patient compliance, and above all their inability to identify AD in its early stages^[38]5,[39]6. The intrinsic function of blood cells in immunological responses to physiological and pathological alterations has made blood a useful resource for investigating disease related molecular markers, enabling earlier and more efficient therapeutic intervention^[40]7. Additionally, research has shown that there is a substantial degree of gene expression correlation between peripheral blood cells and brain tissue^[41]8, and the expression profiling of whole blood or peripheral blood mononuclear cells (PBMCs) may be used to diagnose brain diseases^[42]9. To this end, and in view of the recent studies that have emphasized the link of early alterations in peripheral blood markers with the early development of AD, blood-based biomarkers of AD will be promising due to being accessible, cost-effective, and less invasive^[43]10. Recent availability of bioinformatics tools has facilitated identification of disease markers^[44]11,[45]12. For instance, the NCBI GEO database ([46]https://www.ncbi.nlm.nih.gov/geo/) and ArrayExpress database ([47]https://www.ebi.ac.uk/biostudies/arrayexpress) include a huge number of publically available gene expression data to discover the informative genes related with AD. Furthermore, correlation network analysis is extensively used in bioinformatics research to screen genes and uncover potential biomarkers or therapeutic targets in a variety of disorders. In this way, a well-liked systems biology technique, namely Weighted Gene Co-expression Network Analysis (WGCNA) can be utilized to characterize the correlation of gene expression obtained from microarray databases^[48]13. WGCNA can be utilized to define highly correlated expression gene clusters (called modules), correlate the modules to phenotypes in order to determine which module is most closely related to phenotypic traits, and compute module membership values. Additional correlation network analysis can be used to pinpoint the major participants within each module, which will help find potential therapeutic targets or biomarkers of disease^[49]14. In this regard, several researches in the context of neurodegenerative illnesses, including AD and Parkinson’s, have used WGCNA to study gene expression profiles and search for new molecular biomarkers for diagnosis and treatment^[50]14,[51]15. In the present study, we aimed to use WGCNA and other bioinformatics tools to analyze publicly accessible microarray data and candidate differentially co-expressed genes for experimental validation in AD patients by qRT-PCR. These candidate genes were included RN7SK, TNF, CCL3, TNFAIP3, and FLT3. 7SK small nuclear RNA (RN7SK) is a highly conserved small nuclear non-coding RNA with 331 nucleotides in animals. It is an essential regulatory molecule that activates the Positive Transcriptional Elongation Factor b (P-TEFb) to control RNA polymerase II transcription elongation^[52]16. RN7SK is essential for the strong bidirectional transcription of gene pairs with high expression, implying that RN7SK has an organizational role at promoters with complicated and high Polymerase II turnover. Bidirectional gene transcription frequently occurs in cell cycle, DNA repair, and RNA metabolism, providing an effective mechanism to coordinate the expression of genes involved in similar cellular response pathways^[53]17. Although its function in neurogenesis and nervous system development has been discussed, there is no information regarding this LncRNA’s involvement in the pathophysiology of AD^[54]18,[55]19. Moreover, RN7SK can influence the development of monocytes into macrophages, control macrophage polarization, and regulate innate immunological responses, such as phagocytosis and antigen processing^[56]20. Tumor necrosis factor (TNF) is a key pro-inflammatory cytokine that initiates and regulates the cytokine cascade in response to an inflammatory event^[57]21. TNF levels have been found to be elevated in both the brains and plasma of AD patients^[58]22. Aβ activates the transcription factor NFkB, leading to increased TNF production by microglia^[59]23. In addition, TNF can enhance Aβ burden by up-regulating β-secretase synthesis and increasing ɣ secretase activity^[60]24,[61]25. TNF-associated protein 3 (TNFAIP3), also known as A20, is thought to be a key regulator of inflammation and the peripheral immune system by inhibiting NF-kB. TNFAIP3 expression rises quickly after NFkB activation in response to stimulation with various stimuli, such as inflammatory cytokines and microbial-derived metabolites. TNFAIP3 regulates and suppresses NFkB activity through a negative feedback loop mechanism^[62]26. Although there have been no direct investigations on the association between TNFAIP3 and AD in human, there is evidence that deregulation of this enzyme can contribute to the pathophysiology of neurodegenerative illnesses such as Parkinson’s disease^[63]27. CCL3 (also known as Macrophage Inflammatory Protein-1 α) belongs to the CC chemokine subfamily (CC subfamily Ligand 3). The basic function of chemokines is to recruit leukocytes to the site of inflammation. CCL3 is produced continuously by microglial cells^[64]28, and it has been linked to Alzheimer’s disease^[65]29. Receptor Protein Tyrosine Kinase FLT is associated with the class III receptor tyrosine kinase family that regulates multiple important physiological processes, including signaling cascades that affect cell proliferation and survival^[66]30. While the role of this factor in the pathophysiology of AD is unclear, it has been reported that FLT3 has a role in microglial activation, and inhibiting FLT3 can decrease the pro-inflammatory cytokines levels in microglia treated with lipopolysaccharides^[67]31. On this basis, we studied for the first time the biomarker potency of RN7SK LncRNA and its co-expressed genes. As a result, our findings might offer a new blood-based biomarker for gene-based early diagnosis of AD. Materials and methods Data acquisition and preprocessing Figure [68]1 displays a flowchart of the study plan, data processing, and analyses. After an extensive data search in the GEO and ArrayExpress databases, the E-MTAB-6094 dataset was selected from the ArrayExpress database, whose expression profile is based on Microarray. This data set is based on the Agilent Human Gene Expression 4 × 44 K v2 Microarray 026652 G4845A platform and contains expression information related to 22 samples of AD patients and 13 samples of healthy controls^[69]32. We selected this dataset because it included both male and female samples in sufficient numbers, provided raw data, and had a reliable data matrix. Secondly, the platform used in this dataset included both LncRNAs and mRNA genes. Fig. 1. [70]Fig. 1 [71]Open in a new tab A flow chart outlining the study plan, data preparation, and analysis. After downloading the information related to the E-MTAB-6094 dataset, R software (Version 4.2.1) and Limma package were used to perform preprocessing of the data. The normexp method was used to correct the background, and the quantile algorithm was used to normalize the data. Identification of DEGs (DEmRNAs and DELncRNAs) In this step, after obtaining the mean expression value related to the types of probes which map to the same gene symbol in the platform, the HGNC database ([72]https://www.genenames.org) was used to extract the information of the genes, whose gene symbols are available in the HGNC database. Next, Limma package and FDR method were used to classify differentially expressed genes (DEGs). Selection criteria were determined as |log2FC| ≥ 1 and adjusted P-value < 0.001 for DEmRNAs, and |log2FC| ≥ 0.585 and adjusted P-value < 0.001 for DELncRNAs. WGCNA and detection of expression modules In parallel with the identification of DEGs, the pre-processed raw data matrix, which included genes with specific HGNC gene symbols and rank in the top 25% of the expression value distribution, was imported into the WGCNA package in the R environment. First, data cleaning and preprocessing steps were performed on the data to remove missing values and outliers. In the following, by using the Pearson correlation coefficient test, the matrix of the gene expression profiles was transformed into the matrix of pairwise gene similarity. Afterwards, a correlation adjacency matrix was created by setting the soft-thresholding value in accordance with the scale-free topology condition^[73]33. Next, for every gene pair, the adjacency matrix was converted to a topology overlap matrix (TOM), and a dynamic tree-cut technique was employed to find distinct modules that each contained at least 30 interconnected genes. Then, the adjacency matrix was altered to a TOM for all pairs of genes, and a dynamic tree-cut algorithm was used to determine diverse modules that contained at least 30 interconnected genes. Finally, the dissimilarity between the modules was checked again, and if the difference between them was less than 25%, they were merged with each other. Construction of module-trait relationships and significant module selection After identifying the expression modules, in order to detect the appropriate phenotype associated module for further analysis, the associations of the eigengene variable of each module to the clinical traits data of the E-MTAB-6094 dataset were calculated by Pearson’s correlation analysis. At last, the correlation coefficient between the module memberships with gene significance was calculated using the labeleHeatmap function, and the significant module was selected based on the obtained P-values. Identification of shared genes between DEGs and the selected gene module and construction of co-expression LncRNA-mRNA network for shared genes After identifying DEGs and significant gene modules, common genes between these two lists were extracted for additional analyses using the Venny tool software (version 2.1). Next, a co-expression network for all shared genes, which include both mRNAs and LncRNAs, was created. GO and pathway enrichment analyses of mRNAs related to the mRNA-LncRNA co-expression network and selection of genes involved in immune processes In order to perform functional and pathway enrichment analyses, shared mRNAs list of mRNA-LncRNA co-expression network were submitted into the Enrichr database ([74]https://maaya.nlab.cloud/enrichr/), and functional gene ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis were executed. Considering that the Enrichr submitted mRNA genes were significantly correlated (P-value < 0.05) to the immune processes, the mRNAs list involved in the immune system-related functions was extracted using the ImmPort database ([75]https://www.immport.org). Then, co-expression network analysis was accomplished for the new extracted mRNAs, and a co-expression mRNA-LncRNA sub-network was constructed. Selection of desired LncRNA and its co-expressed mRNA genes for experimental validation To select target genes, interactions in the co-expression mRNA-LncRNA sub-network were envisaged using Cytoscape software ([76]https://cytoscape.org/, version 3.9.0), and the CytoHubba plugin was used to screen the vital nodes based on the Degree algorithm. Also, in addition to reviewing previous studies, the LncRNADisease v3.0 database ([77]http://www.rnanut.net/lncrnadisease/) was used to ensure that the LncRNA selected in this study had not been experimentally investigated in AD. Patients Fifty people with late-onset AD, diagnosed based on the psychiatric interview, clinical examination, laboratory investigation, and magnetic resonance brain imaging according to DSM-5 diagnostic criteria, were incorporated in this research. The control group encompassed fifty volunteers who were matched in terms of age and sex with the patient group (Table [78]1). The patient group’s inclusion criteria included age above 65, no defined neurologic or psychiatric illnesses, and no cardiovascular or cerebrovascular disease. Inclusion criteria for the healthy control group included age over 65, no immunosuppressive use in the previous two weeks, no malignancy, no inflammatory or autoimmune disease, and no history of AD in first-degree relatives. This study was approved by the ethical committee of Shahid Beheshti University of Medical Science (ethical code: IR.SBMU.MSP.REC.1401.428). All subjects provided informed consent before participating in this study. Table 1. Characteristics of AD and healthy control groups. Variables AD patients (n = 50) Controls (n = 50) P-value Gender ratio (male: female) 29:21 26:24 0.54 Age (mean ± SD) 74.1 ± 7.619 72.46 ± 6.0817 0.23 Stage of disease (number) Stage 1: 22 Stage 2: 19 Stage 3: 9 Not applicable Not applicable [79]Open in a new tab Total RNA extraction and qRT-PCR Finally, to confirm the results obtained from the bioinformatics study, 3 cc of blood samples were taken from 100 participants and collected in EDTA tubes. Considering that the storage time of the collected blood sample at room temperature has the greatest effect on the degradation rate of RNA molecules and the half-life of mRNA and LncRNA molecules in the blood sample is reported as 16.4 h and 17.46 ± 3.0 h, respectively^[80]34, the blood fresh samples collected in tubes containing anticoagulant were immediately transferred to the laboratory and RNA extraction procedures were performed on them. To extract total RNA, first PBMC cells were isolated using the Ficoll method (Lymphodex, Germany)^[81]35 and total RNA was isolated from the PBMC cell sediment following the manufacturer’s instructions (RNXplus, Iran). After checking concentration and purity of extracted RNA (OD A260/A280 and A260/230), reverse transcription to cDNA was achieved using the AddScript cDNA synthesis kit (AddBio, Korea). qRT-PCR was performed using the SinaSYBR Green qPCR Master Mix, (SinaClon, Iran) in a Rotor-Gene Q real-time PCR system (QIAGEN, Germany). Melting peak analysis was used to check that the primers were specific. The efficiency of qRT-PCRs, were determined using the Linreg software (version 2021.2). The transcription levels of the target genes were normalized to those of the reference gene (GAPDH). Data were presented as the relative fold change between the cDNA of the study and the calibrator sample using pfaffl method^[82]36. Primer sets utilized in this work were provided by metabion international AG (metabion, Germany), and their sequences are available in Table [83]2. Table 2. PCR primers used for qRT-PCR. Genes Primer sequences Location Product size (bp) GAPDH F: TTGACCTCAACTACATGGTTTACA R: GCTCCTGGAAGATGGTGATG F: 311 R: 436 126 RN7SK F: CAAACAAGCTCTCAAGGTCCA R: GCCTCATTTGGATGTGTCTG F: 225 R: 304 80 TNF F: AGCCTCTTCTCCTTCCTGAT R: AAGATGATCTGACTGCCTGG F: 269 R: 418 142 TNFAIP3 F: GCTATGATACTCGGAACTGGA R: ATTGGCCTTCTGAGGATGTT F: 724 R: 871 148 CCL3 F: AACCAGTTCTCTGCATCACT R: CTTGGTTAGGAAGATGACACC F: 140 R: 286 147 FLT3 F: GAAGGCATCTACACCATTA R: GTAAGGATTCACACCAAGT F: 2650 R: 2733 84 [84]Open in a new tab Statistical analysis R software (version 4.2.1) ([85]http://www.r-project.org/index.html) and a variety of R/Bioconductor packages^[86]37 including WGCNA, Limma, annotate, AgiMicroRna, Affy, edgeR, ggplot2, and pheatmap were applied for statistical analysis. For comparisons of gene expression between two groups Mann-Whitney U test was used. Chi-square test was used to investigate the association between clinicopathological factors and gene expression levels. To evaluate the sensitivity and the specificity for each value of the measure, Receiver operating characteristic curve (ROC curve) analysis were performed, and the area under the curve (AUC) was calculated with a 95% confidence interval, confirming the biomarker’s capacity to discriminate between the two groups^[87]38. All statistical analyses were performed using the SPSS 26.0 statistical program (SPSS, Chicago, IL). P-values < 0.05 were considered significant. Results Data filtering and preprocessing The E-MTAB-6094 dataset was selected for expression analysis to identify potential biomarker genes implicated in late-onset AD. The features of this dataset are listed in Table [88]3. This data set is based on the Agilent Human Gene Expression 4 × 44 K v2 Microarray 026652 G4845A platform^[89]2 and contains expression information related to 22 samples of AD patients and 13 samples of healthy controls. E-MTAB-6094 was found to contain 43,271 gene probes in total. These genes were screened and preprocessed using the criteria stated in the Method section, and 17,275 genes were identified. Data normalization is depicted as a diagram in Fig. [90]2. Table 3. Properties of the E-MTAB-6094 dataset. Group Source name Number of samples Mean age Gender (Male: Female) clinical dementia rate (1:3) Array design Protocol description Healthy control samples Peripheral blood mononuclear cell 13 77.3 3:10 Not applicable Agilent Human Gene Expression 4 × 44 K v2 Microarray 026652 G4845A Labeling using QuickAmp kit (monocolor system-Cy3) (Agilent technologies, Santa Clara, CA) AD patient samples Peripheral blood mononuclear cell 22 79.36 8:14 15:7 Agilent Human Gene Expression 4 × 44 K v2 Microarray 026652 G4845A Labeling using QuickAmp kit (monocolor system-Cy3) (Agilent technologies, Santa Clara, CA) [91]Open in a new tab Fig. 2. [92]Fig. 2 [93]Open in a new tab The boxplot displays the distributions of the selected sample’s values prior to (A) and after (B) normalization, which were appropriate for differential expression analyses. Differentially expressed genes (DEmRNAs and DELncRNAs) A volcano plot was used to identify DEGs in both patient and healthy samples. Totally, 267 mRNA genes were found to be differentially expressed (DEmRNAs) using thresholds of adjusted P-value < 0.001 and |logFC| ≥ 1, including 162 up-regulated and 105 down-regulated genes in AD samples compared to the controls. In addition, 11 LncRNA genes were identified as DELncRNAs using the |log2FC| ≥ 0.585 and adjusted P-value < 0.001, which included 3 up-regulated and 8 down-regulated lncRNAs in AD samples compared with normal samples. The volcano plot in Fig. [94]3A depicts the up-regulated and down-regulated mRNAs in patients and healthy controls. Heatmap illustration of the DELncRNA is alsow shown in Fig. [95]3B. Fig. 3. [96]Fig. 3 [97]Open in a new tab (A) Volcano plot displaying differentially expressed mRNAs between AD and normal samples. The cut-off criteria for DEmRNAs analysis were set as |logFC| ≥ 1, adj. P-value < 0.001. Blue dots indicate down-regulated DEmRNAs and red dots indicate up-regulated DEmRNAs. (B) Heatmap illustrating the DELncRNA between AD and healthy control samples. Gene co-expression network creation and module recognition After preprocessing step, 4,266 normalized genes reported from the E-MTAB-6094 dataset were loaded into the R environment’s WGCNA tool. The scale independence and mean connectivity analysis revealed that when the weighted value equals 14, the mean degree of connectivity was close to 0, and scale independence was 0.9, so the weighted value was set to 14 (Fig. [98]4A). Then, the modules were built applying a dynamic tree-cut technique, and 18 modules with size range of 32 to 1,481 genes were found, including 9 modules with a positive connection and 9 modules with a negative correlation (Fig. [99]4B and C). Fig. 4. [100]Fig. 4 [101]Open in a new tab Analysis of the E-MTAB-6094 dataset using weighted gene co-expression networks (WGCNA) method. (A) The soft-thresholding power and the scale-free fit index. (B) Dynamic dendrogram showing clustered genes according to a dissimilarity measure. (C) Heatmap illustrating the correlation between module eigengenes and clinical phenotype of Alzheimer’s disease. (D) The number of genes in each module. To determine which module was most strongly correlated with clinical characteristics, the Pearson’s correlation coefficient between the module and the AD phenotype was assessed. The blue module and clinical feature had the strongest correlation in the module trait connection. So, the blue module was selected as the module of interest for subsequent analyses (Fig. [102]4D, r= -0.97, P = 2e-22). Common genes between DEGs and the interested module and their co-expression LncRNA-mRNA network Following the identification of DEGs (DEmRNAs and DELncRNAs) and the interested module, common genes between these two lists were retrieved for additional analyses using the Venny tool program (version 2.1). A Venn diagram revealed that three potential genes were common between DELncRNAs and the blue module, while 139 genes were common between DEmRNAs and the blue module (Fig. [103]5). Then, a co-expression network for all 142 common genes, which included both mRNA and LncRNA, was constructed. Fig. 5. [104]Fig. 5 [105]Open in a new tab Shared genes between DEGs and interested blue module. GO and KEGG analysis of constructed co-expression LncRNA-mRNA network The common 142 genes were then subjected to GO functional and KEGG pathway enrichment analyses. The GO analysis found that the 142 AD-related genes were primarily implicated in the biological process of cellular response to lipopolysaccharides, cytokine-mediated signaling pathway, and the regulation of IL-1β production. The molecular function indicated cytokine activity, chemokine activity, and chemokine receptor binding genes. The cellular component was enriched for Clathrin-Coated Endocytic Vesicle, Clathrin-Coated Vesicle Membrane, and Clathrin-Coated Endocytic Vesicle Membrane genes (Fig. [106]6). The KEGG pathway analysis identified genes related to Rheumatoid Arthritis, viral protein interactions with cytokines and receptors, and the NFκB signaling pathway (Fig. [107]7). Fig. 6. [108]Fig. 6 [109]Open in a new tab Gene ontology analysis of shared mRNA genes in terms of: (A) Biological Process, (B) Cellular Component, and (C) Molecular Function. Fig. 7. [110]Fig. 7 [111]Open in a new tab KEGG pathway analysis of the mRNA genes of LncRNA-mRNA co-expression network. Permission has been obtained from Kanehisa laboratories for using KEGG pathway database^[112]39. Immune genes extraction and reconstruction of LncRNA-mRNA co-expression sub-network Considering that the most of the 142 mRNAs in the LncRNA-mRNA co-expression network were involved in inflammation and immune process, in the following step, we used the ImmPort database ([113]https://www.immport.org) to extract the mRNA genes related to the immune system from the network. As a result, a sub-network of LncRNA-mRNA co-expression was obtained for this series. Target LncRNA and its co-expressed mRNAs for experimental analysis We visualized mRNA-LncRNA co-expression sub-network interactions using Cytoscape software (version 3.8.0), and screened the hub genes of the network using the cytoHubba plug-in (Fig. [114]8). Finally, RN7SK and its co-expressed mRNAs, namely TNF, TNFAIP3, CCL3, and FLT3 were selected for experimental analysis. Fig. 8. [115]Fig. 8 [116]Open in a new tab Vital nodes of mRNA-LncRNA co-expression sub-network visualized by Cytoscape. Expression assay and validation of the selected genes in AD patients To attain more evidence for the importance of the key genes in the selected LncRNA-mRNA co-expression sub-network, we calculated the expression level of the selected genes in PBMC of the two groups of AD patients and controls through qRT-PCR. We showed that in PBMC of late-onset AD patients, expression levels of RN7SK and TNF were in line with the microarray findings and were significantly increased. CCL3 and FLT3 also showed a trend of increase in their expression, which was not significant. On the other hand, and on the contrary of the microarray result, TNFAIP3 expression level was significantly decreased in AD patients (Fig. [117]9A; Table [118]4). Figure [119]9B depicts the agarose gel electrophoresis of the transcripts to evaluate the specificity of primers. Fig. 9. [120]Fig. 9 [121]Open in a new tab qRT-PCR results of candidated transcripts. (A) Relative expression level of the target genes in AD patients and healthy controls measured by qRT-PCR analysis. The results are expressed as mean ± SEM. (B) 2% Agarose gel electrophoresis of the PCR products showing, the specificity of the amplified transcripts. Table 4. The values of relative expression level (fold change) and standard error of mean (SEM) in AD patients and healthy controls. RN7SK TNF TNFAIP3 CCL3 FLT3 Fold Change P-value Fold Change P-value Fold Change P-value Fold Change P-value Fold Change P-value Healthy controls (n = 50) Mean 2.079 0.006 1.618 0.023 2.873 0.016 2.016 0.396 2.734 0.521 SEM 0.38 0.27 0.599 0.42 0.59 AD patients (n = 50) Mean 6.398 6.829 0.988 2.307 3.001 SEM 1.44 1.38 0.189 0.55 0.67 [122]Open in a new tab Association between gene expression level and clinicopathological data Furthermore, we performed association analysis between the expression levels of genes and clinicopathological data (Table [123]5). There were significant positive associations between stage 3 of the disease with the relative expression level of all of the transcript except CCL3. However, the expression levels of mentioned genes were not significantly associated with any of assessed demographic factors of age and gender. Table 5. Comparison of the association between the expression levels of the transcripts with different clinicopathologic factors. Parameters Subclasses RN7SK relative expression (mean ± SEM) P-value TNF relative expression (mean ± SEM) P-value TNFAIP3 relative expression (mean ± SEM) P-value FLT3 relative expression (mean ± SEM) P-value CCL3 relative expression (mean ± SEM) P-value Age 65–80 (n = 82) 81–95 (n = 18) 3.78 ± 0.53 6.29 ± 3.57 0.597 4.14 ± 0.88 4.59 ± 1.02 0.206 1.95 ± 0.34 1.80 ± 0.96 0.560 2.69 ± 0.41 3.67 ± 1.67 0.44 2.45 ± 0.41 0.84 ± 0.23 0.034 Gender Female (n = 45) Male (n = 55) 4.86 ± 1.46 3.72 ± 0.73 0.366 2.75 ± 0.47 5.42 ± 1.28 0.737 1.56 ± 0.34 2.22 ± 0.52 0.355 3.02 ± 0.5 2.74 ± 0.7 0.12 2.03 ± 0.44 2.26 ± 0.51 0.833 Stage Stage 1 (n = 22) Stage 2 (n = 19) Stage 3 (n = 9) 4.58 ± 1.35 3.52 ± 0.77 16.9 ± 6.24 0.322 0.251 0.000 4.65 ± 1.45 4.43 ± 1.08 17.21 ± 5.42 0.557 0.567 0.001 0.66 ± 0.18 1.11 ± 0.36 1.5 ± 0.55 0.003 0.009 0.015 2.39 ± 0.87 3.79 ± 1.43 2.8 ± 0.76 0.797 0.971 0.073 1.45 ± 0.54 2.16 ± 0.77 4.69 ± 2.16 0.156 0.247 0.489 [124]Open in a new tab ROC curve analysis The potential diagnostic values of RN7SK, TNF, and TNFAIP3 expression in AD patients were evaluated by AUC value. The results of our research proposed that RN7SK expression may function as a biomarker with a sensitivity of 50% and specificity of 80% (AUC = 0.658, P = 0.005). Figure [125]10 shows the sensitivity and specificity of this LncRNA and the associated TNF and TNFAIP3 genes. Fig. 10. [126]Fig. 10 [127]Open in a new tab ROC curve analysis of RN7SK, TNF, and TNFAIP3 as biomarker for differentiating between AD patients and controls. The sensitivity and specificity of RN7SK, TNF, and TNFAIP3 transcription were 50% and 84% (AUC = 0.658, P = 0. 0063), 58% and 88% (AUC = 0.632, P = 0.0229), and 32% and 94% (AUC = 0.64, P = 0.0158), respectively. Discussion AD is a neurodegenerative condition and the most prevalent type of dementia in adults. Despite the fact that the first case of this disease was described more than a hundred years ago, the pathological mechanisms leading to the development of this disease are still debated. On the other hand, due to the problems of clinical diagnosis, including the invasiveness of CSF tests or cost of PET scan the existing methods, more and more studies have been conducted using blood samples in order to find possible markers that showing the diagnostic signature and prognosis of the disease^[128]6,[129]7. Changes in peripheral blood cell parameters, a common indication of inflammation and immunological dysfunction, have been described in patients with neurodegenetative illnesses. A meta-analysis research, for example, found that AD patients had significantly lower lymphocyte counts while having significantly higher leukocyte counts, neutrophil counts, and neutrophil-lymphocyte ratio (NLR) than healthy controls. Furthermore, AD patients had a significantly lower percentage of B lymphocytes and CD8^+ T cells and a much higher CD4/CD8 ratio than healthy controls. Furthermore, significant alterations in hemoglobin levels and platelet distribution width were detected in patients with AD or MCI compared to healthy controls^[130]40. Furthermore, due to the development of ultrasensitive technologies for assessing proteins in fluids such as blood, significant progress has been achieved in the study of biomarkers for neurodegenerative diseases such as AD. From these blood base biomarkers (BBBs) we can list the Neurofilament Light Chain (NfL)^[131]41, Amyloid-Beta (Aβ42/40 ratio)^[132]42, hyperphosphorylated tau like P-Tau181^[133]43, P-Tau217^[134]44, P-Tau231^[135]45, chitinase 3-like protein 1 (CHI3L1)^[136]46, and Glial fibrillary acidic protein (GFAP)^[137]47. Currently, plasma Aβ1–42/Aβ1–40, p-tau 217, NfL, and GFAP among these BBBs satisfy steps 1, 2, 3, and partially phase 4, but there is not enough data to support their clinical relevance (phase 5)^[138]48. Two reasons that limit their routine clinical use are robustness and confounding factor. For example, there is a theoretical issue that the plasma Aβ42/40 ratio may have a reduced diagnostic accuracy than the Aβ42/40 ratio in CSF. This could present a concern for clinical robustness due to the challenge of applying a cutoff at the individual level^[139]49. Furthermore, various comorbidities, including medical diseases like dyslipidemia, hypertension, diabetes, and chronic renal disease, stroke, and myocardial infarction, impacted plasma biomarkers of Ab40, Ab42, total tau, and NfL even in an age- and sex-adjusted model^[140]50. Advanced age (oldest old), race, and gender are further relevant demographic characteristics that may alter the interpretation of NfL, p-tau181, and p-tau231 data in the clinical setting^[141]48,[142]51,[143]52. Therefore, based on the provided information about the current limitations of blood biomarkers, more research is still needed to identify appropriate and accessible potential biomarkers, so recently molecular biomarkers, including the expression pattern of coding and non-coding genes in peripheral blood cells, have gained more attention. So, the use of these new specific types of blood biomarkers, such as RN7SK LncRNA, in combination with its co-expressed genes, in addition to the advantages of availability and non-invasiveness, may be useful in the field of early diagnosis and the management of AD. In recent years, WGCNA has been primarily used to investigate new biomarkers in various disorders. This allows for the discovery of possible biomarkers or therapeutic targets based on the interconnections of genes and the association between genes and patients’ clinical characteristics^[144]13. In the present study, we used bioinformatics analyses methods to find candidate genes whose expression changes can partake in the development of AD. To do this, first we evaluated the E-MTAB-6094 ArrayExpress database using WGCNA in order to identify the genes that were differentially expressed from among those that were involved in a co-expression network and had a strong association with the phenotype of AD. Next, for the collection of these genes which included both mRNAs and lncRNAs, an mRNA-LncRNA co-expression network was created. In the following, we performed pathway enrichment analysis and found that most of the obtained mRNAs were associated with immune system functions. So, immune-related mRNAs were extracted out of the co-expression network of mRNA genes and were used for reconstruction of the LncRNA-mRNA co-expression sub-network. Finally, from this sub-network, RN7SK LncRNA, TNF, CCL3, TNFAIP3, and FLT3 were chosen for experimental validation. Several LncRNAs have been linked with the development of AD^[145]53, which may impact the nervous system in a variety of biological ways, such as posttranscriptional regulation and epigenetic control^[146]54. For example, it was discovered that silencing the LncRNA X-inactive specific transcript (XIST) reduced AD-related beta-amyloid cleaving enzyme 1 (BACE1) changes via miR-124/BACE1 signaling pathways^[147]55. Furthermore, it has been observed that elevated plasma level of LncRNA BACE1-antisense (BACE1-AS) could act as a powerful blood-based biomarker for AD^[148]56. The BACE1-AS, transcribed from the reverse strand of the beta-amyloid cleaving enzyme 1 gene, interacts with the BACE1 mRNA, and augment its stability and translation in a positive feed-forward way^[149]57. RN7SK is a highly conserved small nuclear non-coding RNA, which, so far as we know, has not been studied in relation to AD. Our in silico and experimental expression analysis in AD samples revealed that RN7SK was considerably over-expressed in AD patients’ blood samples compared to controls. We showed that the expression of RNSK in AD patients was almost threefold higher than that of healthy subjects, which was consistent with the studied microarray results. It is interesting to note that the mouse brain showed higher Rn7SK small nuclear RNA expression. Additionally, it is shown that Rn7SK has a dynamic pattern of expression throughout brain development from embryo to adult and is significantly up-regulated during neuronal differentiation in vivo^[150]18. It has been suggested that RN7SK regulates a number of significant RNA-binding proteins and chromatin effectors to play an essential mediator function in transcriptional and post-transcriptional regulatory processes in neurons^[151]58. As a result, disrupting such RN7SK signaling pathways by mutation or aggregation of RN7SK-interacting proteins may lead to neuronal dysfunction in neurological disorders^[152]59,[153]60. RN7SK may also play a role in cell senescence. Overexpression of RN7SK decreased cell viability^[154]61, while knockdown of RN7SK and siRNA transfection in adipose tissue-derived mesenchymal stem cells (AD-MSCs) resulted in enhanced proliferation, viability, and differentiation potential. These findings could imply that a decrease in RN7SK levels in MSCs may cause senescence to be delayed or stopped and improve their engineering and dedifferentiation for successful patient treatment^[155]62. Apart from its involvement in brain development and neuron function, RN7SK is also associated with inflammatory reactions, which may explain its possible link to neurodegenerative disease like AD. For example, a previous research has shown that RN7SK up-regulation in the serum of Multiple Sclerosis (MS) patients plays a significant role in the inflammation and/or neurodegeneration. This study hypothesized that up-regulation of RN7SK in 7SK snRNP complex may disregulate P-TEFb complex, affecting CD4^+ T cell regulation by altering CD4^+ T cell differentiation. This emphasizes the importance of RN7SK in regulating the progression and modulating the inflammatory responses^[156]63. Additionally, RN7SK is also known for its role in macrophage differentiation, polarization, and innate immune response regulation, including antigen processing and phagocytosis. Knockdown of RN7SK can reduce the levels of M2 markers (CD206, CD163, or Dectin) while increasing the levels of M1 markers (MHC II or CD23), indicating that RN7SK can induce M2 phenotypic polarization. Since M1 macrophages are more potent in antigen uptake and phagocytosis compared with M2, RN7SK negatively regulates these innate functions^[157]20. Our qRT-PCR data also showed that TNF expression levels in AD samples were increased approximately 4-fold compared to healthy samples, which corresponds to a 2.6-fold change in the microarray results. TNF is a key pro-inflammatory cytokine in regulating and initiating the cytokine cascades in response to inflammatory stimuli in AD^[158]64 and its level has been increased in both the brain and the plasma of AD patients^[159]22. TNF can enhance Aβ burden by up-regulating β-secretase synthesis and increasing ɣ-secretase activity^[160]25,[161]65. Meanwhile, Aβ activates the transcription factor NFkB, causing microglia to produce more TNF^[162]66. Moreover, an increased level of TNF in the brain can prevent microglial clearance of Aβ, leading to synaptic disruption, disease progression, and cognitive loss^[163]67. Regarding the changes in the expression level of TNFAIP3, our qRT-PCR results were in the opposite of the data obtained from the microarray analysis. Our data showed that TNFAIP3 gene was significantly down-regulated by 2.9 folds in the AD samples compared to the controls. Whereas according to the microarray dataset of E-MTAB-6094, TNFAIP3 had 2.5 fold higher expression in AD patients. TNFα-induced protein 3 (TNFAIP3) encodes the A20 protein, a powerful anti-inflammatory molecule^[164]68 that inhibits TNF-NFκB activation. Genetic ablation of Tnfaip3 in mice model causes multiorgan inflammation, cachexia, and perinatal mortality^[165]69. The protective function of TNFAIP3 has also been demonstrated in the nervous system and microglia, where it has been observed that microglia lacking in TNFAIP3 up-regulate pro-inflammatory factors, develop an inflammatory response, and exhibit increased neuronal damage. This is followed by an increase in microglial counts and changed morphology as CD8^+ T lymphocytes infiltrate the CNS^[166]70,[167]71. Recently, it was also shown that TNFAIP3 inhibits Myc-independent late-phase microglial proliferation which occurs in a variety of neurological illnesses, such as AD. Inhibiting microglial proliferation protects synaptic degeneration and behavioral impairments in an AD mice model^[168]72. Consistent with our results, Antonio Cuadrado et al. found that Tnfaip3 was highly down-regulated in AD mouse models and responded to basic pathogenic processes in AD brain, such as apoptosis^[169]73. Also, we can refer to a previous study that showed tnfaip3 was down-regulated in Parkinson’s disease. Based on this study, tnfaip3 performs its neuroprotective role by targeting autophagy and inflammatory responses through restricting NFκB and mTOR signaling pathways^[170]27. In the following, regarding the changes of FLT3 and CCL3 expression levels, we did not perceive any significant changes between the two groups of AD patients and healthy peoples in terms of the expression levels of these two factors. Previously, it was reported that inhibition of receptor tyrosine kinase Flt3 (FMS-like tyrosine-3) modifies microglia cell activity by decreasing activation-induced IL-6 production and expression of antigen-presenting surface components^[171]31. Furthermore, it is found that Flt3 is crucial for the cell death stimulated by glutamate oxidative stress in primary cerebrocortical neurons and multiple neuronal cell lines. As a result, Flt3 inhibitors may be a promising pharmacological treatment for AD and Parkinson’s disease, which are caused by oxidative stress^[172]74. Similarly, increased expression of macrophage inflammatory protein 1α (CCL3) was found in neurons of the AD brains. Moreover, exposure to Aβ induced over-expression of CCL3 in AD model cell lines^[173]75,[174]76. But still, there is not enough information about the role of CCL3 in the pathogenesis of AD. In terms of the likely mechanism for involvement of our co-expressed genes in AD, we can consider the STAT3 signaling pathway. As a result of microglial activation and an increase in neuroinflammatory statue, pro-inflammatory TNF cytokine bind to the TNFR1 receptor and activate Janus tyrosine kinases (Jak). Activation of these kinases increases STAT3 phosphorylation and promotes its transport into the nucleus, where STAT3 interacts with NFkB to activate it and trigger target gene transcription^[175]77. Interestingly, there is evidence that STAT3 has a substantial role in increasing the expression of RNS7SK. This can help to explain the possible association between the increased expression of TNF and RN7SK in this work^[176]78. Furthermore, STAT3’s significance in microglial polarization and neuroinflammation is well understood. Knocking down STAT3 in rat microglia increases neuronal function and survival, probably by enhancing M2 polarization and starting the anti-inflammatory response after induction of injury^[177]79. Regarding the mechanism of the reduced TNFAIP3 expression in AD, it is possible to consider the regulatory role of miRNAs, for example, in the form of possible LncRNA-miRNA regulatory networks. In this way, we can refer to a study in which it was reported that microglial activation leads to an increased miR-27b-3p expression, which can in turn down regulate TNFAIP3 as its target. In the following, reduced activation of TNFAIP3 could increase the IL-6, IL-1β, and TNF-α expression, induce microglial activation, and worsen the neuroinflammation^[178]80. In this regard, there are studies which confirmed that miR-27b-3p has an increased expression in the AD samples^[179]81,[180]82. Also, there is a study about MS patients that indicated the up-regulation of miRNA-27b-3p, which may give a theory about the potential link between this miRNA and increased RN7SK expression which was mentioned earlier^[181]83. However, these are possible explanations, and more research is desired to illuminate the precise mechanisms. Overall, based on our findings, RN7SK LncRNA, TNF, and TNFAIP3 might be associated with AD, proposing a key set of genes that could be used as novel possible biomarkers of AD. Nevertheless, this research is not without limitations. For example, to reduce bias in our study, we excluded the datasets that failed to meet the selection criteria, and we only used one dataset for choosing deregulated genes. The number of subjects involved in the study could also be a possible shortcoming, and the diagnostic efficacy of these biomarkers should be validated in a larger and more diverse population in future works, especially with respect to ethnicity and potential co-morbidities like elderly patients who are prone to high incidence of cognitive decline. Moreover, evaluation of the relationship of these potential biomarkers in other neurological conditions is needed to evaluate their specificity. In conclusion, qRT-PCR verified the differential expression of diagnostic RN7SK; nevertheless, additional prospective studies are required to confirm its potential biomarker significance. Conclusion In summary, through WGCNA on microarray data from human PBMC samples, we discovered RN7SK LncRNA, TNF, TNFAIP3, CCL3, and FLT3 as key genes in AD pathogenesis. In the following, using the qRT-PCR, the biomarker characteristics of these factors were investigated at the experimental level, and it was observed that there is an important difference between the two groups of AD patients and healthy controls in terms of RN7SK LncRNA, TNF, and TNFAIP3 expression. SO, RN7SK LncRNA probably plays an associated role in the molecular pathogenesis of AD, and investigating its mRNA-LncRNA co-expression network may help identify new AD blood-based biomarker. Electronic supplementary material Below is the link to the electronic supplementary material. [182]Supplementary Material 1^ (35.5KB, docx) [183]Supplementary Material 2^ (23.8KB, txt) [184]Supplementary Material 3^ (111KB, xlsx) Author contributions M.K performed the in silico steps and experiments and wrote the draft. R.N.S evaluated the patients. M.S.K and M.R contributed in data analysis. S.M.D and S.G-F supervised the study. All the authors read and approved the submitted version. Funding The study was supported by a grant from Shahid Beheshti University of Medical Sciences. Data availability The datasets generated and/or analysed during the current study are available in the Agilent repository (https://www.agilent.com/en/human-gene-expression-microarrays-details-s pecifications). Declarations Ethics approval and consent to participate Informed consent has been obtained from all patients. Ethical approval for this study was obtained from the Ethical Committee of Shahid Beheshti University of Medical Sciences. All methods were performed in accordance with relevant guidelines and regulations. Competing interests The authors declare no competing interests. Consent to publish Informed consent has been obtained from all patients. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Sima Mansoori Derakhshan, Email: mansooris@tbzmed.ac.ir. Soudeh Ghafouri-Fard, Email: s.ghafourifard@sbmu.ac.ir. References