Transcriptomic Analyses of Exercise Training in Alzheimer’s Disease Cerebral Cortex

Abstract

Background:

Research reported exercise could reduce Alzheimer’s disease (AD) symptoms in human and animals. However, the molecular mechanism of exercise training via transcriptomic analysis was unclear especially in AD in the cortex area.

Objective:

Investigate potential significant pathways in the cortex area that were affected by exercise during AD.

Methods:

RNA-seq analysis, differential expressed genes, functional enrichment analysis, and GSOAP clustering analysis were performed in the isolated cerebral cortex from eight 3xTg AD mice (12 weeks old) randomly and equally divided into control (AD) and exercise training (AD-EX) group. Swimming exercise training in AD-EX group was conducted 30 min/day for 1 month.

Results:

There were 412 genes significant differentially expressed in AD-EX group compared to AD group. Top 10 upregulated genes in AD-EX group against AD group mostly correlated with neuroinflammation, while top 10 downregulated genes mostly had connection with vascularization, membrane transport, learning memory, and chemokine signal. Pathway analysis revealed the upregulated interferon alpha beta signaling in AD-EX had association with cytokines delivery in microglia cells compared to AD and top 10 upregulated genes involved in interferon alpha beta were Usp18, Isg15, Mx1, Mx2, Stat1, Oas1a, and Irf9; The downregulated extracellular matrix organization in AD-EX had correlation with Aβ and neuron cells interaction and Vtn was one of the top 10 downregulated genes involved in this pathway.

Conclusion:

Exercise training influenced 3xTg mice cortex through interferon alpha beta signaling upregulation and extracellular matrix organization downregulation based on transcriptomics analysis.

Keywords

Alzheimer’s disease bioinformatics cortex exercise

INTRODUCTION

Alzheimer’s disease (AD) is an age-related neurodegenerative disease caused by accumulation of tau and amyloid-β (Aβ) protein which will affect neural cells, brain volume, and cognitive ability [1 –3]. Furthermore, AD affects multiple parts of the brain, but the cortex area receives the biggest impact because the cortex area has the highest amyloid deposition [4, 5]. The impact from amyloid deposition is accelerated atrophy in the cortex area which causes neuronal and vascular disruption [4, 6]. As a consequence from those disturbances, the functional connectivity between the cortex area and other parts of the brain is declined when compared to healthy patients [7]. In the end, the cortex area is highly affected by AD, which causes physiology disturbance and disconnection functional connectivity to other parts of the brain.

3xTg mice is the AD mice strain used to study the interaction between Aβ plaques and tau protein accumulation in the cortex area because 3xTg mice have gene mutations in APP Swedish, MAPT P301L, and PSEN M146V [8]. Aβ plaques are expressed in the intraneural part of the cortex area which contributes to cognitive dysfunction in 4-month-old or older 3xTg mice [9]. Genetic profiling based on microarray study reveals that 3xTg mice have hundreds of genes that are differentially expressed and affect multiple pathways related to neurological, immunological, and inflammatory disease [10].

Several meta-analysis studies reported physical activity and exercise training prevent cognitive decline or enhance cognitive ability in patients with AD by improving their quality of life, decreasing depressed symptoms, and improving cortex functional networks [11 –13]. Furthermore, in the cortex area, a systematic review reported that exercise training reduced amyloid protein concentration to support the positive impact of exercise training on the cortex [14]. Swimming exercise trainings improve cognitive and behavioral disorders in mice with Alzheimer-like disease particularly through oxidative markers and inflammatory cytokines in the cortex area [15, 16]. Exercise training affected synapse density through the BDNF pathway and mitochondrial integrity which caused energy metabolism to become more stable in mice with AD [15]. However, some pathways that were regulated by exercise training are still poorly investigated like ion regulation pathway and immune-related pathway in the mouse cortex of AD. Additionally, the molecular mechanism related to exercise training affected by AD is still not clear [17, 18].

Transcriptomics analysis was performed to detect thousand or hundreds candidate genes in AD research [19]. One of the transcriptomics analysis, RNA-seq analysis, produced high throughput result by detecting differential expressed genes (DEGs) which uncovered a dysregulation pattern in AD [20]. DEGs determine the differences of gene transcripts quantity between two conditions that can be visualized by volcano plot to show significant gene score and mean average (MA) plot to show mean expression gene score [21]. The DEG quality of the RNA-seq sample is also checked with principal component analysis (PCA) and heatmap analysis [22]. Principal component analysis is a DEG quality control technique through unsupervised method to explore genes expression variability in each sample, while heatmap analysis is a DEG quality control technique by compared normalized Z score of genes expression [23, 24]. Usually, after significant DEGs are determined and the quality is appropriate, the functional enrichment analysis is conducted to detect genes role in various pathways within multiple databases [25]. Three well-known databases, including the gene ontology biological process (GOBP) terms, Kyoto encyclopedia of genes and genomes (KEGG) pathways, and Reactome pathways, are often used in various research [26 –31]. For the visualization of functional enrichment analysis result, an enrichment plot was used to check the functional enrichment quality based on statistic score ranked genes [32]. However, recently, there was a novel method to visualize functional enrichment pathways through t-distributed stochastic neighbor embedding (tSNE) score based on overrepresentation analysis called gene set over-representation analysis plotter (GSOAP) [33]. Thus, all analyses from significant DEG determination, quality control, functional enrichment pathway analysis, and visualization were integrated to find meaningful information related to hundreds or thousands of genes.

From the information above, we conducted transcriptomics analysis to investigate which pathways were influenced by exercise training in the 3xTg AD mice cortex area because transcriptomics analysis could detect multiple pathways that were affected by exercise training and most AD-exercise cortex studies just focused on BDNF or AD pathway although exercise had potential to affect other pathways in cortex area. 3xTg AD mice were chosen because they represent the real AD condition which produces high Aβ and tau protein accumulation. Exercise training research using transcriptomics analysis has never been done in AD research, especially in the cortex area, and motivated us to conduct this experiment. Furthermore, we applied GSOAP clustering method in the functional enrichment pathway analysis which was another novelty in this study to extract massive functional enrichment information result. Therefore, we expected that the top 3 pathways in the AD cortex area were significantly affected by exercise using high-throughput analysis with GSOAP clustering.

METHODS

Animals and exercise treatment

The animal model for AD research design used was the 3xTg mice genotype which has gene mutation in APP Swedish, MAPT P301L, and PSEN M146V [8]. 3xTg mice were fed regular mice food and water freely. Room temperature in the mice cage was kept at 25°C with controlled 12-h day and night cycle. 3xTg mice (n = 8) were separated equally into two groups, the Alzheimer disease (AD) group and exercise training (AD-EX) group. The mice in both groups started swimming at 12 weeks. The swimming protocol was adapted and modified from previous research that used a swimming exercise in disease-cortex related research [34]. The swimming pool was 60×90×50 cm water tub. The AD group did not undergo swimming exercise training as a control, while the AD-EX group underwent swimming exercise as a treatment. The AD-EX group was swimming 30 min/day, 5 days/week for 4 weeks. Water temperature for swimming was maintained at 25–28°C. The exercise mice group was dried gently with a hair dryer and a towel to maintain body temperature. After exercise training, the mice rested two days to avoid acute exercise effect. Then, the mice were sacrificed at 17 weeks old.

Mouse brain tissue preparation

Mice were put into CO₂ chamber for anesthetic. Initially, the CO₂ concentration chamber was around 0.04% but increased gradually until it reached 70% volume per minute. CO₂ source was from gas cylinder with appropriate pressure regulator in order to precisely control the gas flow. During the anesthetic condition after 30–60 s, the mice fainted and their heads were removed. Then, mice skulls were opened, and right side areas of the cortex were isolated. Each isolated cortex brain was put into the Eppendorf tube contain Trizol to retain the RNA and cells condition. After that, the tube which would be put in liquid nitrogen was sent to Genomics, BioSci & Tech Co. located in New Taipei City, Taiwan for RNA extraction and RNA-seq analysis at the same day the brain tissue isolation was conducted.

Next generation sequencing (NGS) preparation and sequencing

The NGS data from TruSeq Stranded mRNA Library Prep Kit (Illumina, USA) was prepared using purified RNA. The sequencing library procedure was based on the manufacturer’s procedures. First, 1μg total RNA concentration was purified by oligo(dT)-coupled magnetic beads and dissociated into small pieces under increasing temperature. Then, reverse transcriptase enzyme with fortuitous primers were used to synthesize the first-strand cDNA. Double-strand cDNA fragments were synthesized into several generations along with adenylation on 3’ ends of DNA. AMPure XP system (Beckman Coulter, USA) ligated and purified the adaptors. To check the quality of the data, Agilent Bioanalyzer 2100 system was used. Real-Time PCR system was also used to confirm the data quality. The qualified data were analyzed on an Illumina NovaSeq 6000 platform which used 150 bp paired-end reads to sequencing the data. That analysis was generated by Genomics, BioSci & Tech Co. located in New Taipei City, Taiwan.

Fig. 1

Work scheme in this experiment. NGS, next generation sequencing; MA, mean average; GSEA, gene set enrichment analysis; GSOAP, gene set over-representation analysis plotter.

Differential expression gene analysis

Trimmomatic (version 0.39) was applied to remove bases with low quality and sequences from adapters in raw data [35]. Bowtie2 (version 2.3.4.1) aligned filtered reads to the reference genomes [36]. Quantification of the transcript abundance was executed by RSEM software (version 1.2.28) [37]. DeSeq2 was used to identify DEGs between the AD-EX group and AD group that resulted in log2foldchange, log2mean score, and p-value [38]. A volcano plot was used to represent the relationship between p-value and log2foldchange, while an MA plot was applied to show log2mean score and log2foldchange. For PCA analysis and heatmap analysis to check DEGs quality, only significant genes were selected to process analysis using R language [39]. The result from PCA analysis was PC1 and PC2 that represented first and second largest variability in all samples, while heatmap analysis would show normalized Z score and clustered samples.

Functional enrichment analysis

Gene enrichment analysis (GSEA) was done using fgsea package with GOBP terms, KEGG pathways, and Reactome pathways as the database [25 –28]. GSEA plot was visualized using clusterprofiler and enrichplot [32, 40]. For the enrichment plot, the gene list was ranked based on DEGs statistic score. The another functional enrichment visualization, GSOAP package, was used for pathway selection and clustering based on tSNE score between all the genes in selected pathways [33].

Statistical analysis

The differential expression (DE) analysis statistic was based on Wald T-test [38]. The significant value for DEG and functional enrichment analysis was p-value<0.05. Gene expression was transformed into Z score calculation to check DE quality control. Functional enrichment analysis used functional class scoring (FCS) method which used statistic score from DEG result to rank the genes. For GSOAP analysis, the result was based on tSNE which was a technique to reduce high dimensionality data into two dimension through modified stochastic neighbor embedding which allowed the data easier to be understood [41]. The significant in GSOAP analysis table represented -log10 of p-value from FCS functional enrichment result. The statistical analysis used R programming language to analyze all the result from differential expression analysis to functional enrichment analysis.

Ethics approval

The animal research study procedure was approved by IACUC China Medical University with ID number 2018-158. IACUC was Institutional Animal Care and Use Committee which was the standard guidelines applied in Taiwan.

Procedure

The procedure started with exercise for AD 3xTg mice as a treatment group (AD-EX group) and non-exercise AD 3xTg mice as a control group (AD group). Mouse brain tissue in cortex area was isolated after swimming training or sedentary condition. Then, the cortex sample was analyzed with NGS. The result from NGS was analyzed to identify significant DEGs which would be visualized with volcano plot and MA plot. After significant genes were detected, functional enrichment analysis was conducted to check which significant pathway related to AD with pathway analysis, GSOAP, and GSEA plot as visualization.

RESULTS

The DE analysis comparing the AD-EX group against the AD group from RNA-seq data was conducted to identify the significant genes that was affected by exercise training effect in AD (Fig. 2, Supplementary Table 1). The result displayed 412 genes that were significant differentially expressed in the AD-EX group compared to the AD group (Fig. 2A). The top 10 upregulated genes were Usp18, Isg15, Mx1, Mx2, Cxcl10, Gm6451, Stat1, Oas1a, Oasl2, and Irf9 and the top 10 downregulated genes were Clec14a, Slc16a2, Vtn, Cxcl12, Cldn5, Cytl1, Tek, Tagln, Gkn3, and Itm2a in AD-EX related to AD group (Fig. 2B). The highlight in this data was Usp18 and Clec14a because they were the most significant genes in upregulation and downregulation respectively in exercise training group based on p-value. Meanwhile, mean expression in top 10 upregulated and downregulated genes were quite high compared to other genes was another confirmation that the genes were significant differentially expressed (Fig. 2B).

Fig. 2

Differential expression (DE) analysis & sample quality control analysis based on AD and AD exercise training group comparison. A) Volcano plot. B) MA plot. C) PCA analysis. D) Heatmap analysis. Volcano plot showed relationship between fold change and significant value. In this graph, red color meant it was significant while the green plot was not significant. MA plot showed relationship between fold change and mean expression. Red color in MA plot showed gene expression was expressed higher in AD-EX group compared to AD group while blue color showed gene expression was expressed lower in AD-EX group compared to AD group. Heatmap analysis depicted sample in the x axis and gene in the y axis. Usp18, ubiquitin specific peptidase; Isg15, interferon-stimulated gene 15; Mx1, myxovirus resistance protein 1; Mx2, myxovirus resistance protein 2; Cxcl10, C-X-C motif chemokine ligand 10; Gm6451, predicted gene 6451; Stat1, signal transducer and activator of transcription 1; Oas1a, 2^′5^′ oligoadenylate synthetase 1A; Oasl2, 2^′–5^′ oligoadenylate synthetase 2; Irf9, interferon regulatory factor 9; Clec14a, C-type lectin domain containing 14A; Slc16a2, solute carrier family 16 member 2; Vtn, vitronectin; Cxcl12, C-X-C motif chemokine ligand 12; Cldn5, claudin 5; Cytl1, cytokine like 1; Tek, endothelial tyrosine kinase; Tagln, transgelin; Gkn3, gastrokine 3; Itm2a, integral membrane protein 2A; AD ⟶ AD exercise, AD group was compared to AD exercise group; PC1, principal component 1; PC2, principal component 2.

After the DE result was determined, the DE gene list quality was checked to confirm whether it was differentially expressed or not. The methods to conduct the quality control process were PCA analysis and heatmap analysis (Fig. 2). From significant DE genes only, PCA analysis showed there was different variation between the AD and AD-EX groups (Fig. 2C). Each sample in each group was clustered together which meant there was no outlier according to the PCA result. To confirm this result, heatmap analysis based on normalized Z score was performed to understand the cluster between each sample. It revealed that each AD group was assembled into different dendrogram compared to AD-EX group (Fig. 2D).

Next, we conducted functional enrichment analysis to examine the molecular mechanism of exercise training in AD. Three databases (GO, KEGG, Reactome) were used to confirm the molecular pathway. The top 10 significant terms of up- and downregulation are shown in Fig. 3 and Supplementary Table 2. In the GO biological process, there were sixteen pathways that were known to be associated with the top 10 upregulation and downregulation genes. Response to type I interferon, defense response to virus, response to virus, regulation of ribonuclease activity, regulation of nuclease activity, negative regulation of viral process, interferon gamma mediated signaling pathway, and response to interferon gamma were more upregulated in the AD-EX group compared to the AD group. For downregulation pathways, there were skeletal system development, regulation of neuron projection development, divalent inorganic cation homeostasis, regulation of body fluid levels, metal ion transport, cell junction assembly, axon development, and wound healing which expressed lower in the AD-EX group than the AD group (Fig. 3A). For KEGG pathway, there were seven pathways related to the top 10 upregulation and downregulation genes. Those pathways were cytosolic DNA sensing pathway, RIG I receptor pathway, TOLL-like receptor pathway, focal adhesion, tight junction, leukocyte transendothelial migration, and cell adhesion CAMS that were expressed higher in the AD-EX group compared to the AD group in the cortex area (Fig. 3B). RIG I and TOLL-like receptor pathway were correlated to the interferon pathway which confirmed the GO biological process result was also upregulated in the AD-EX group versus the AD group. On the other hand, the KEGG downregulated pathways which related to cell adhesion, junction, and vascular were more downregulated in the AD-EX group compared to the AD group. The last database, Reactome, showed nine pathways correlated to the top 10 upregulated and downregulated genes in the AD-EX group compared to AD group (Fig. 3C). Interferon alpha beta signaling, antiviral mechanism by IFN stimulated genes, DOX58 IFIH1 mediated induction of interferon alpha beta, interferon signaling, and interferon gamma signaling were more upregulated in the AD-EX group compared to the AD group, while SLC mediated transmembrane transport, extracellular matrix, cell surface interaction at the vascular wall, and integrin cell surface interactions were more downregulated in the AD-EX group than the AD group. Overall, the enrichment result showed connection with the top 10 upregulated and downregulated DE genes and there were consistency functional pathways between those three databases.

Fig. 3

Top 20 Enrichment analysis result from various database. Upregulated, genes-related pathway in AD-EX group expression was higher than AD group. Downregulated, genes-related pathway in AD-EX group expression was lower than AD group. A) Gene ontology biological process (GOBP), B) KEGG pathway, C) Reactome pathway. Bold text was the pathway that involved top 10 upregulation or downregulation genes. NES, normalized enrichment score. Red color showed the pathway had higher expression in the AD exercise training group while green color showed the pathway had lower expression in the AD exercise training group.

Many pathways were affected by doing exercise training (Fig. 3). In order to know those pathways relationship statistically, GSOAP analysis was conducted to know whether there were any clusters or not. The database that was used for this analysis was the Reactome database. The reason this database was selected because in this database, molecular mechanism pathway could be tracked and shows more complete genes compared to the KEGG pathway database. Furthermore, GO was also not selected because the database just showed annotation function only. The GSOAP analysis showed there were three different clusters that contributed to three different mechanisms (Fig. 4A, Supplementary Table 3). The first cluster contributed to innate immune pathways, the second cluster contributed to cell structure, and third cluster contributed to molecule exchange membrane protein in AD [42 –45]. The highlight in those clusters were interferon alpha beta signaling, extracellular matrix organization, and ion channel transport because all of them were correlated to Aβ accumulation in different ways [42 , 46]. From that result, an enrichment plot was conducted to know the enrichment pattern. Interferon alpha beta signaling pathway which contained upregulated significant gene expression was enriched positively in the AD-EX group in contrast to the AD group according to GSEA plot analysis (Fig. 4B). The extracellular matrix organization pathway showed there was negative enrichment in AD-EX group compared to AD group (Fig. 4C). The last pathway, ion channel transport was also enriched negatively in the AD-EX group compared to the AD group (Fig. 4D). Several top 10 upregulation and downregulation genes contributed to top 20 Reactome pathways (Fig. 4E, Supplementary Table 4). Seven out of the top ten upregulated genes (Usp18, Isg15, Mx1, Mx2, Stat1, Oas1a, Irf9) contributed interferon alpha beta signaling and another pathway like interferon signaling. Three out of the top ten downregulated genes (Vtn, Tek, Slc16a2) contributed to various downregulated pathways like extracellular matrix organization, cell surface interactions at the vascular wall, integrin cell surface interactions, and SLC mediated transmembrane transport. Ion channel transport which was the major highlight in GSOAP clustering result did not have any top 10 upregulated or downregulated genes. From those GSOAP and GSEA plots, exercise training affected AD through innate immune system, cell structure, and ion channel transport.

Fig. 4

GSOAP clustering and Gene Set Enrichment Analysis (GSEA) plot analysis. A) GSOAP clustering analysis. B) GSEA interferon alpha beta signaling. C) GSEA extracellular matrix organization. D) GSEA ion channel transport. E) Top 10 upregulated genes and downregulated genes that contributed to the top 20 Reactome pathways. The top part in the plot showed the running enrichment score in the gene list. GSOAP analysis (A) showed how many genes contributed to pathway by circle diameter and significant score by how dark the color was. For the enrichment plot (B–D), the peak plot which was further from 0.0 meant the enrichment score for the pathway. The middle part was the members of the gene contributed to the pathway. The bottom part, ranking metric described a gene’s correlation with the exercise training treatment. The positive value described positive correlation with exercise while the negative value exhibited negative correlation with exercise training treatment. For the top 10 upregulation and downregulation genes contribution in top 20 Reactome (E), red text represented upregulation pathways while the black text represented downregulation pathways. proj. 1, projection 1 based on tSNE method; proj. 2 = projection 2 based on tSNE method. Rank in ordered dataset, gene rank based on statistic value.

DISCUSSION

The transcriptomic analysis research in the current study suggested 10 significantly upregulated genes after exercise training in AD cortex, which include Usp18, Isg15, Mx1, Mx2, Cxcl10, Gm6451, Stat1, Oas1a, Oasl2, and Irf9, while 10 significantly downregulated genes included Clec14a, Slc16a2, Vtn, Cxcl12, Cldn5, Cytl1, Tek, Tagln, Gkn3, and Itm2a. Functional enrichment analysis showed that multiple pathways involved in the top 10 upregulated and downregulated genes were related to neuroinflammation and cell maintenance after exercise training. To specify which pathway received the most impactful of exercise training in the cortex, GSOAP analysis and GSEA plots were conducted using the Reactome database to elicit three significant pathways: interferon alpha beta signaling, extracellular matrix organization, and ion channel transport.

Some of the top 10 upregulated and downregulated genes after exercise training in the cortex had correlation with AD studies related to neuroinflammation, immune system, vascularization, membrane transport, learning memory, and chemokine signal. Usp18 is known as a negative regulator in the inflammation system and when the gene was knocked out in mice, the mice became sensitive to AD-related proinflammatory compound like lipopolysaccharide (LPS) which caused reduced survival rate [47 –49]. Isg15 was detected in AD patients and relates to microglia activation [50]. Mx1, which contributes to the interferon alpha pathway, had high expression in white matter in AD brain, especially near the senile plaques and also affects patients cognitive decline as well [51, 52]. Cxcl10 expression in cerebrospinal fluid is high in mild AD patients and usually the gene is co-localized with Aβ in APP/PS1 mice [53]. Stat1 contributes to tau phosphorylation and Aβ₄₂ during the aging process by upregulating BACE1 when δ-secretase is activated [54]. Oas1a contributes to AD progression by regulating the interferon molecular pathway in microglia [55]. Irf9 correlates with AD by activating the interferon molecular pathway with Stat1 and Stat2, which was initiated by toll-like receptor interacting with Aβ or tau protein [56]. Mx2, Gm6451, and Oasl2 have not been reported in any AD studies. On the other hand, related to the top 10 significant downregulated genes, Vtn protein reportedly interacts directly with APP in APP mice genotype based on proteomics study [57]. Cxcl12 affects impaired learning and memory in the Tg2576 mouse model [58]. Cldn5 contributes to memory impairment in AD mice by regulating GABAergic neurotransmission [59]. Cytl1 correlates with cerebrospinal fluid Aβ₄₂, tau, and phosphorylated tau in impaired cognition patients based on a proteomics study [60]. The Tek gene polymorphism is highly related to AD pathology based on a genome wide association study analysis using AD human sample datasets [61]. Tagln, which regulates the cytoskeletal system, is dysregulated significantly in peripheral blood of AD patients according to microarray analysis [62]. Slc16a2, Clec14a, Gkn3, and Itm2a were significant downregulated genes in this AD mouse model after exercise and have not been reported in any other AD studies. Overall, the effect of exercise on the cortex of AD mice revealed 13 genes from the top 10 upregulated and top 10 downregulated genes were associated with AD mechanism, while 7 genes had not been reported in any previous AD studies.

Interferon alpha beta signaling was reported to have an impact in AD by regulating neuroinflammation in the cortex area [50]. According to our transcriptomic findings, the interferon alpha beta signaling pathway was upregulated after exercise training in the cortex of AD mice. Upregulation in interferon alpha reported could induce neuroinflammation in the brain but upregulation in interferon beta could increase memory and alleviated neuroinflammation [42, 63]. Although interferon alpha beta contribution to AD was ambiguous, several genes related to interferon alpha beta were also upregulated especially during exercise using human and rodent datasets similar to our transcriptomics study [64]. Since several of the top 10 upregulated and downregulated genes from the differential expression analysis (Usp18, Isg15, Mx1, Mx2, Stat1, Oas1a, and Irf9) were also involved in interferon alpha beta pathway, this indicates that this pathway is significantly affected during exercise on the cortex area. Interestingly, only top 10 upregulated genes contributed to interferon alpha beta pathway while top 10 downregulated genes did not have any contribution in interferon alpha beta pathway. Therefore, exercise significantly affected the interferon alpha beta system in the cortex of AD mice but it needs to be explored further whether interferon genes alleviates the pathology symptoms or not during exercise.

Extracellular matrix organization was affected by exercise training in AD cortex and several studies reported this pathway was affected by AD pathology. In AD, the upregulated matrix adhesion system increased the interaction between Aβ₄₂ and nerve cells that led to neuroinflammation [65]. This condition was confirmed by proteomics analysis that reported there was increased protein expression in matrisome and cell-extracellular matrix (ECM) interaction in AD patients compared to healthy patients [66]. Surprisingly, our data showed the matrisome and cell-ECM related pathways genes expression were reduced after exercise in cortex area which indicated exercise affected AD pathology and reverse extracellular matrix genes expression. The top 10 upregulated genes did not contribute to extracellular matrix organization but one of the top 10 downregulated genes, Vtn, involved in this pathway. However, whether extracellular matrix organization downregulation during exercise in the cortex area reduced AD symptoms needs to be investigated further because transcriptomics study did not uncover that topic and there is no study exploring this idea.

Ion channel transport was a unique pathway that was a major highlight in the third cluster due to large gene size and high significant score, but the pathway did not involve the top 10 upregulated or downregulated genes (Supplementary Table 3). Our exercise transcriptomics data exhibited downregulated enrichment related to ion channel transport in AD cortex with Atp2a3, Sln, Atp1a1, Atp6ap1, Atp2b3, Atp10d, and Atp2b2 were significant differentially expressed. Based on the significant gene list in this pathway, our exercise data affected ion channel in AD cortex through reticulum endoplasmic Ca^2 + transportation by Atp2a3 and Atp2b2 gene expression and extracellular vesicle by Atp1a1 gene expression based on bioinformatics studies but the other genes had never been mentioned in previous AD research [67 –69]. In the AD situation, there was a dysregulation of Ca^2 + exchange in reticulum endoplasmic membrane by increased Ca^2 + transportation into the cytoplasm [70]. The dysregulation itself occurred because of presenilin mutation which is common in AD stimulating Aβ accumulation and tau hyperphosphorylation [71, 72]. On the other hand, extracellular vesicle affected AD by disrupted synapse formation by delivering the tau oligomers to neuron cells [73]. However, the contribution of Atp2a3, Atp2b2, and Atp1a1 was questionable because there are zero studies investigating those genes directly to the effect of AD. Overall, the ion channel transport pathway related to the reticulum endoplasmic Ca^2 + transportation and extracellular vesicle transportation in AD cortex was affected by exercise activity but further research is needed to elaborate Atp2a3, Atp2b2, and Atp1a1 direct contribution to ion channel pathway.

Our findings were different than any AD-exercise cortex research articles because those articles usually only analyzed the animal samples using western blotting, behavioral test, and only focused on neuron development or AD pathology-related proteins. As a consequence, it was difficult to find a comparison or reference that mentioned inflammation, extracellular matrix, or ion channel-related proteins in AD-exercise cortex research articles because they did not measure those proteins in detail. The closest evidence that exercise affected the immune system in AD mice was that exercise affected the chemokines in 3xTg mice blood significantly after exercise but the research article did not mention the top 10 upregulated genes [74]. On the other hand, no AD-exercise research articles used cortex to address the extracellular matrix pathway or Vtn gene that was part of the top 10 downregulated genes contributing to extracellular matrix pathway. Furthermore, ion channel pathway was not addressed in AD-exercise research in cortex articles as well indicated molecular pathway exploration in cortex during AD-exercise condition needed more future investigation.

Ultimately, we found more detail explanation how exercise training affected AD especially in cortex area. Nevertheless, further experiments were needed to be done to confirm our transcriptomic result because some genes have not been investigated directly to AD in cortex condition. Furthermore, our experiment also did not specify cell isolation for transcriptomics analysis condition because the gene expression might be different in neuron cell, microglia, or astrocytes during exercise. However, this study showed exercise affected multiples pathways related to interferon alpha beta signaling, extracellular matrix organization, and ion channel transport in the cortex area which have never been addressed in AD-exercise studies although those pathways were related to AD pathology mechanism. RNA-seq analysis in this AD-exercise research was replicated four times which made the differential expression genes and functional enrichment result pathways were legitimate. On the other hand, top 10 upregulated and downregulated genes that involved interferon alpha beta signaling, extracellular matrix organization during exercise in cortex area could be explored further for AD drug development like Usp18 that contributed to negative regulation of innate immune system. Hopefully, our data will help enlighten the AD mechanism and treatment development because it has not been progressed well until now.

Conclusion

Exercise training affected hundreds of significant differentially expressed genes and resulted in three significant major pathways based on GSOAP analysis which were interferon alpha beta signaling, extracellular matrix organization, and ion channel transport in AD cortex area. The interferon alpha beta signaling pathway which associated with Aβ clearance or neuroinflammation involved several top 10 upregulated genes during exercise in cortex area and they were Usp18, Isg15, Mx1, Mx2, Stat1, Oas1a, and Irf9. The extracellular matrix organization pathway in the cortex area which associated with the interaction between Aβ and neuron cells also involved Vtn, one of the top 10 downregulated genes during exercise. The ion channel transport pathway which correlated with Ca^2 + transportation in the reticulum endoplasmic and extracellular vesicle was a unique significant pathway affected by exercise in the AD cortex area because it did not involve the top 10 upregulated or downregulated genes but it reached the top 10 downregulated pathways in functional enrichment analysis result. Furthermore, the large gene size and high significant score in GSOAP analysis was another reason the ion channel transport pathway should be paid attention to. However, this idea needs wet lab experiments to confirm and uncover more detailed molecular mechanisms. Furthermore, the specific cell types impacting AD animals during exercise training needed to be explored to observe the genes expression levels in different cell type.

Footnotes

ACKNOWLEDGMENTS

from this study is created by using BioRender web application.

FUNDING

The study is supported by Ministry of Science and Technology, Taiwan (MOST 110-2314-B-468-004; MOST 107-2314-B-468-002-MY3) as well as partially supported by China Medical University and Weifang Medical University.

CONFLICT OF INTEREST

The authors have no conflict of interest to report.

DATA AVAILABILITY

The data supporting the findings in this study are available within the article and its . More detail information about the data can be received on request from the corresponding author.

The supplementary material is available in the electronic version of this article: .

References

Scheltens

, De Strooper

, Kivipelto

, Holstege

, Chételat

, Teunissen

, Cummings

, van der Flier

(2021) Alzheimer’s disease. Lancet 397, 1577–1590.

Sharma

, Singh

, Garg

, Dhiman

(2021) Apoptotic pathways and Alzheimer’s disease: Probing therapeutic potential. Neurochem Res 46, 3103–3122.

Sampath

, Sathyanesan

, Newton

(2017) Cognitive dysfunction in major depression and Alzheimer’s disease is associated with hippocampal-prefrontal cortex dysconnectivity. Neuropsychiatr Dis Treat 13, 1509–1519.

Fjell

, McEvoy

, Holland

, Dale

, Walhovd

(2014) What is normal in normal aging? Effects of aging, amyloid and Alzheimer’s disease on the cerebral cortex and the hippocampus. Prog Neurobiol 117, 20–40.

Salat

, Kaye

, Janowsky

(2001) Selective preservation and degeneration within the prefrontal cortex in aging and Alzheimer disease. Arch Neurol 58, 1403–1408.

Guo

, Zhang

, Zeng

, Huang

, Xu

, Zhao

(2020) Molecular and cellular mechanisms underlying the pathogenesis of Alzheimer’s disease. Mol Neurodegener 15, 40.

Jobson

, Hase

, Clarkson

, Kalaria

(2021) The role of the medial prefrontal cortex in cognition, ageing and dementia. Brain Commun 3, fcab125.

Oddo

, Caccamo

, Shepherd

, Murphy

, Golde

, Kayed

, Metherate

, Mattson

, Akbari

, LaFerla

(2003) Triple-transgenic model of Alzheimer’s disease with plaques and tangles: Intracellular Abeta and synaptic dysfunction. Neuron 39, 409–421.

Billings

, Oddo

, Green

, McGaugh

, LaFerla

(2005) Intraneuronal Abeta causes the onset of early Alzheimer’s disease-related cognitive deficits in transgenic mice. Neuron 45, 675–688.

10.

Castillo

, Leon

, Mazzei

, Abolhassani

, Haruyama

, Saito

, Saido

, Hokama

, Iwaki

, Ohara

, Ninomiya

, Kiyohara

, Sakumi

, LaFerla

, Nakabeppu

(2017) Comparative profiling of cortical gene expression in Alzheimer’s disease patients and mouse models demonstrates a link between amyloidosis and neuroinflammation. Sci Rep 7, 17762.

11.

Jia

, Liang

, Xu

, Wang

(2019) Effects of physical activity and exercise on the cognitive function of patients with Alzheimer disease: A meta-analysis. BMC Geriatr 19, 181.

12.

López-Ortiz

, Valenzuela

, Seisdedos

, Morales

, Vega

, Castillo-García

, Nisticò

, Mercuri

, Lista

, Lucia

, Santos-Lozano

(2021) Exercise interventions in Alzheimer’s disease: A systematic review and meta-analysis of randomized controlled trials. Ageing Res Rev 72, 101479.

13.

Dauwan

, Begemann

MJH

, Slot

MIE

, Lee

EHM

, Scheltens

, Sommer

IEC

(2021) Physical exercise improves quality of life, depressive symptoms, and cognition across chronic brain disorders: A transdiagnostic systematic review and meta-analysis of randomized controlled trials. J Neurol 268, 1222–1246.

14.

Vasconcelos-Filho

FSL

, da Rocha Oliveira

, de Freitas

TBC

, de Pontes

PADS

, da Rocha-e-Silva

, Chaves

EMC

, da Silva

CGL

, Soares

, Ceccatto

(2021) Neuroprotective mechanisms of chronic physical exercise via reduction of β-amyloid protein in experimental models of Alzheimer’s disease: A systematic review. Life Sci 275, 119372.

15.

Pang

, Wang

, Pei

, Zhang

, Shen

, Gao

, Chang

(2019) Regular exercise enhances cognitive function and intracephalic GLUT expression in Alzheimer’s disease model mice. J Alzheimers Dis 72, 83–96.

16.

Bashiri

, Enayati

, Bashiri

, Salari

(2020) Swimming exercise improves cognitive and behavioral disorders in male NMRI mice with sporadic Alzheimer-like disease. Physiol Behav 223, 113003.

17.

Belaya

, Kucháriková

, Górová

, Kysenius

, Hare

, Crouch

, Malm

, Atalay

, White

, Liddell

, Kanninen

(2021) Regular physical exercise modulates iron homeostasis in the 5xFAD mouse model of Alzheimer’s disease. Int J Mol Sci 22, 8715.

18.

Kim

, Cho

, Kang

(2019) Protective effect of exercise training against the progression of Alzheimer’s disease in 3xTg-AD mice. Behav Brain Res 374, 112105.

19.

Yang

, Diao

, Wang

, Sun

, Zhou

, Xie

(2020) Identification of key regulatory genes and pathways in prefrontal cortex of Alzheimer’s disease. Interdiscip Sci 12, 90–98.

20.

Chen

, Mills

, Janitz

(2016) RNA-sequencing to elucidate early patterns of dysregulation underlying the onset of Alzheimer’s disease. In Systems Biology of Alzheimer’s Disease, Castrillo JI, Oliver SG, eds. Springer New York, New York, NY, pp. 327–347.

21.

Chen

, Wong

(2019) Transcriptome informatics. In Encyclopedia of Bioinformatics and Computational Biology, Ranganathan S, Gribskov M, Nakai K, Schonbach C, eds. Academic Press, Oxford, pp. 324–340.

22.

Yin

, Zeng

, Hu

, Gao

, Liu

, Chen

(2020) Parental uveitis influences offspring with an increased susceptibility to the experimental autoimmune uveitis. Front Immunol 11, 1053.

23.

Ringnér

(2008) What is principal component analysis? Nat Biotechnol 26, 303–304.

24.

Hubbard

, Zhai

, Peng

(2020) Exploring mechanisms of quantitative resistance to leptosphaeria maculans (blackleg) in the cotyledons of canola (brassica napus) based on transcriptomic and microscopic analyses. Plants 9, 864.

25.

Sergushichev

(2016) An algorithm for fast preranked gene set enrichment analysis using cumulative statistic calculation. bioRxiv, 060012.

26.

Ashburner

, Ball

, Blake

, Botstein

, Butler

, Cherry

, Davis

, Dolinski

, Dwight

, Eppig

, Harris

, Hill

, Issel-Tarver

, Kasarskis

, Lewis

, Matese

, Richardson

, Ringwald

, Rubin

, Sherlock

(2000) Gene ontology: Tool for the unification of biology. The Gene Ontology Consortium. Nat Gen 25, 25–29.

27.

Kanehisa

, Goto

(2000) KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res 28, 27–30.

28.

Gillespie

, Jassal

, Stephan

, Milacic

, Rothfels

, Senff-Ribeiro

, Griss

, Sevilla

, Matthews

, Gong

, Deng

, Varusai

, Ragueneau

, Haider

, May

, Shamovsky

, Weiser

, Brunson

, Sanati

, Beckman

, Shao

, Fabregat

, Sidiropoulos

, Murillo

, Viteri

, Cook

, Shorser

, Bader

, Demir

, Sander

, Haw

, Wu

, Stein

, Hermjakob

, D’Eustachio

(2022) The reactome pathway knowledgebase 2022. Nucleic Acids Res 50, D687–D692.

29.

Zhang

, Tong

, Chen

, Wu

, Wang

(2018) Analysis of NFKB2-mediated regulation of mechanisms underlying the development of Hodgkin’s lymphoma. Mol Med Rep 17, 8129–8136.

30.

Chen

, Chu

, Lu

, Kong

, Huang

, Cai

(2015) Gene ontology and KEGG pathway enrichment analysis of a drug target-based classification system. PloS One 10, e0126492.

31.

Paczkowska

, Barenboim

, Sintupisut

, Fox

, Zhu

, Abd-Rabbo

, Mee

, Boutros

; PCAWG Drivers and Functional Interpretation Working Group; Reimand

; PCAWG Consortium (2020) Integrative pathway enrichment analysis of multivariate omics data. Nat Commun 11, 735.

32.

(2022) enrichplot: Visualization of Functional Enrichment Result, R package version 1.14.2.

33.

Tokar

, Pastrello

, Jurisica

(2020) GSOAP: A tool for visualization of gene set over-representation analysis. Bioinformatics 36, 2923–2925.

34.

Cheng

, Lee

(2022) Exercise training enhances BDNF/TrkB signaling pathway and inhibits apoptosis in diabetic cerebral cortex. Int J Mol Sci 23, 6740.

35.

Bolger

, Lohse

, Usadel

(2014) Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120.

36.

Langmead

, Salzberg

(2012) Fast gapped-read alignment with Bowtie 2. Nat Methods 9, 357–359.

37.

, Dewey

(2011) RSEM: Accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinformatics 12, 323.

38.

Love

, Huber

, Anders

(2014) Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15, 550.

39.

Team R (2020) RStudio, PBC, Boston, MA.

40.

, Wang

, Han

, He

(2012) clusterProfiler: An R package for comparing biological themes among gene clusters. OMICS 16, 284–287.

41.

van der Maaten

, Hinton

(2008) Viualizing data using t-SNE. J Mach Learn Res 9, 2579–2605.

42.

Taylor

, Minter

, Newman

, Zhang

, Adlard

, Crack

(2014) Type-1 interferon signaling mediates neuro-inflammatory events in models of Alzheimer’s disease. Neurobiol Aging 35, 1012–1023.

43.

Sun

, Xu

, Jiang

, Liu

, Yang

, Bai

, Yang

(2021) Role of the extracellular matrix in Alzheimer’s disease. Front Aging Neurosci 13, 707466.

44.

Thei

, Imm

, Kaisis

, Dallas

, Kerrigan

(2018) Microglia in Alzheimer’s disease: A role for ion channels. Front Neurosci 12, 676.

45.

Iacobas

, Iacobas

, Stout

, Spray

(2020) Cellular environment remodels the genomic fabrics of functional pathways in astrocytes. Genes 11, 520.

46.

Wang

, Zheng

(2019) Ca2+ homeostasis dysregulation in Alzheimer’s disease: A focus on plasma membrane and cell organelles. FASEB J 33, 6697–6712.

47.

Knobeloch

, Utermöhlen

, Kisser

, Prinz

, Horak

(2005) Reexamination of the role of ubiquitin-like modifier ISG15 in the phenotype of UBP43-deficient mice. Mol Cell Biol 25, 11030–11034.

48.

Harland

, Torres

, Liu

, Wang

(2020) Neuronal mitochondria modulation of LPS-induced neuroinflammation. J Neurosci 40, 1756.

49.

Yang

, Xian

, Hu

, Tian

, Qin

, Wang

, Cui

(2015) USP18 negatively regulates NF-κB signaling by targeting TAK1 and NEMO for deubiquitination through distinct mechanisms. Sci Rep 5, 12738–12738.

50.

Roy

, Wang

, Wan

, Chiu

, Cole

, Yin

, Propson

, Xu

, Jankowsky

, Liu

, Lee

VMY

, Trojanowski

, Ginsberg

, Butovsky

, Zheng

, Cao

(2020) Type I interferon response drives neuroinflammation and synapse loss in Alzheimer disease. J Clin Invest 130, 1912–1930.

51.

Yamada

, Horisberger

, Kawaguchi

, Moroo

, Toyoda

(1994) Immunohistochemistry using antibodies to alpha-interferon and its induced protein, MxA, in Alzheimers and Parkinson’s disease brain-tissues. Neurosci Lett 181, 61–64.

52.

, Huang

, Tang

NLS

, Lam

LCW

(2012) MxA polymorphisms are associated with risk and age-at-onset in Alzheimer disease and accelerated cognitive decline in Chinese elders. Rejuvenation Res 15, 516–522.

53.

Guedes

, Lao

, Cardoso

, El Khoury

(2018) Roles of microglial and monocyte chemokines and their receptors in regulating Alzheimer’s disease-associated amyloid-beta and tau pathologies. Front Neurol 9, 549.

54.

Zhang

, Li

, Wang

, Song

, Yu

, Kang

, Liu

, Zhang

, Xie

, Liu

, Wang

, Ye

(2021) delta-Secretase-cleaved tau stimulates A beta production via upregulating STAT1-BACE1 signaling in Alzheimer’s disease. Mol Psychiatry 26, 586–603.

55.

Magusali

, Graham

, Piers

, Panichnantakul

, Yaman

, Shoai

, Reynolds

, Botia

, Brookes

, Guetta-Baranes

, Bellou

, Bayram

, Sokolova

, Ryten

, Sala Frigerio

, Escott-Price

, Morgan

, Pocock

, Hardy

, Salih

(2021) A genetic link between risk for Alzheimer’s disease and severe COVID-19 outcomes via the OAS1 gene. Brain 144, 3727–3741.

56.

Sanford

SAI

, McEwan

(2022) Type-I interferons in Alzheimer’s disease and other tauopathies. Front Cell Neurosci 16, 949340.

57.

Badhwar

, Brown

, Stanimirovic

, Haqqani

, Hamel

(2017) Proteomic differences in brain vessels of Alzheimer’s disease mice: Normalization by PPARγ agonist pioglitazone. J Cereb Blood Flow Metab 37, 1120–1136.

58.

Parachikova

, Cotman

(2007) Reduced CXCL12/CXCR4 results in impaired downregulated in a mouse model of Alzheimer learning and is disease. Neurobiol Dis 28, 143–153.

59.

Zhu

, Wei

, Yuan

, He

, Chen

, Ji

, Zhang

(2022) Claudin-5 relieves cognitive decline in Alzheimer’s disease mice through suppression of inhibitory GABAergic neurotransmission. Aging (Albany NY) 14, 3554–3568.

60.

Dayon

, Núñez Galindo

, Wojcik

, Cominetti

, Corthésy

, Oikonomidi

, Henry

, Kussmann

, Migliavacca

, Severin

, Bowman

, Popp

(2018) Alzheimer disease pathology and the cerebrospinal fluid proteome. Alzheimers Res Ther 10, 66.

61.

Sun

, Song

, Wang

, Han

, Bai

, Xie

, Feng

, Jia

, Duan

, Lei

(2014) Hidden risk genes with high-order intragenic epistasis in Alzheimer’s disease. J Alzheimers Dis 41, 1039–1056.

62.

Naughton

, Duncan

, Murrey

, Meadows

, Newsom

, Stoicea

, White

, Scharre

, McCarty

, Fu

(2015) Blood genome-wide transcriptional profiles reflect broad molecular impairments and strong blood-brain links in Alzheimer’s disease. J Alzheimers Dis 43, 93–108.

63.

Mudò

, Frinchi

, Nuzzo

, Scaduto

, Plescia

, Massenti

, Di Carlo

, Cannizzaro

, Cassata

, Cicero

, Ruscica

, Belluardo

, Grimaldi

(2019) Anti-inflammatory and cognitive effects of interferon-β1a (IFNβ1a) in a rat model of Alzheimer’s disease. J Neuroinflammation 16, 44.

64.

Hill

, Gammie

(2022) Alzheimer’s disease large-scale gene expression portrait identifies exercise as the top theoretical treatment. Sci Rep 12, 17189.

65.

Leshchyns’ka

, Sytnyk

(2016) Synaptic cell adhesion molecules in Alzheimer’s disease. Neural Plast 2016, 6427537–6427537.

66.

Johnson

ECB

, Carter

, Dammer

, Duong

, Gerasimov

, Liu

, Betarbet

, Ping

, Yin

, Serrano

, Beach

, Peng

, De Jager

, Haroutunian

, Zhang

, Gaiteri

, Bennett

, Gearing

, Wingo

, Lah

, Levey

, Seyfried

(2022) Large-scale deep multi-layer analysis of Alzheimer’s disease brain reveals strong proteomic disease-related changes not observed at the RNA level. Nat Neurosci 25, 213–225.

67.

, Li

, Tong

, Yang

, Huang

, Xu

, Dai

(2020) Systematic metabolic analysis of potential target, therapeutic drug, diagnostic method and animal model applicability in three neurodegenerative diseases. Aging (Albany NY) 12, 9882–9914.

68.

Kawalia

, Raschka

, Naz

, de Matos Simoes

, Senger

, Hofmann-Apitius

(2017) Analytical strategy to prioritize Alzheimer’s disease candidate genes in gene regulatory networks using public expression data. J Alzheimers Dis 59, 1237–1254.

69.

Ceylan

(2022) Integrated bioinformatics analysis to identify alternative therapeutic targets for Alzheimer’s disease: Insights from a synaptic machinery perspective. J Mol Neurosci 72, 273–286.

70.

Popugaeva

, Bezprozvanny

(2013) Role of endoplasmic reticulum Ca2+ signaling in the pathogenesis of Alzheimer disease. Front Mol Neurosci 6, 29.

71.

Cao

, Guan

, Liang

, Huang

, Wang

(2019) Calcium ions stimulate the hyperphosphorylation of tau by activating microsomal prostaglandin E synthase 1. Front Aging Neurosci 11, 108.

72.

Chami

, Checler

(2020) Alterations of the endoplasmic reticulum (ER) calcium signaling molecular components in Alzheimer’s disease. Cells 9, 2577.

73.

Ruan

, Pathak

, Kalavai

, Yoshii-Kitahara

, Muraoka

, Bhatt

, Takamatsu-Yukawa

, Hu

, Wang

, Hersh

, Ericsson

, Gorantla

, Gendelman

, Kayed

, Ikezu

, Luebke

, Ikezu

(2021) Alzheimer’s disease brain-derived extracellular vesicles spread tau pathology in interneurons. Brain 144, 288–309.

74.

Haskins

, Jones

, Lu

, Bareiss

(2016) Early alterations in blood and brain RANTES and MCP-1 expression and the effect of exercise frequency in the 3xTg-AD mouse model of Alzheimer’s disease. Neurosci Lett 610, 165–170.