Exploring the pathogenesis and key genes associated of acute myocardial infarction complicated with Alzheimer’s disease

Despite mounting evidence linking Acute Myocardial Infarction (AMI) to Alzheimer’s disease (AD), the shared mechanism of these two conditions’ occurrence remains unclear. This research aims to delve deeper into the molecular process of the occurrence of the two diseases. We retrieved the gene expression profiles of AD (GSE5281) and AMI (GSE66360) from the Gene Expression Omnibus database. Then, a total of 22 common differentially expressed genes (DEGs) including one downregulated gene and 21 upregulated genes were chosen for further analysis. Following the discovery of the common DEGs between AMI and AD, we performed protein–protein interaction analysis and hub gene identification analysis. Next, ten important hub genes were identified. Additionally, the key genes were identified by the least absolute shrinkage and selection operator and support vector machine‐recursive feature elimination and multivariable logistic regression analysis. The BCL6 was identified to be the most connected with AMI and AD. Finally, the BCL6 gene was validated in the GSE40680 (AMI) and GSE122063 (AD) datasets. Our research indicates that AMI and AD share a comparable pathophysiology. The Hub genes, especially BCL6, were essential in developing AMI and AD. In addition, these hub genes and shared pathways can offer fresh perspectives for additional mechanism investigation.


Identification of differentially expressed genes (DEGs) in AMI and AD
After the data downloaded from the GEO database, the primary data was annotated to form an expression matrix, each probe was matched to their homologous gene symbols, and the repeated gene symbols in the matrix were excluded.Next, the matrix was then normalized using the RMA method 10 .Then, the R software's "limma" package was used to analyze the differential expression of the AMI and AD samples.DEGs were defined as genes with p < 0.05 and (|log fold change (FC)|≥ 1.As the demographics of age, sex, and weight of the individuals were not available from the primary data, we were unable to adjust the data for subgroup analysis based on these factors.DEG heat maps and volcano plots were produced using the "pheatmap" and "ggplot2" packages.Following our use of these screening parameters, two sets of DEGs were found.With the "venn" package, these DEGs were analyzed to determine the intersecting genes between the AMI and AD 11 .Further analysis was conducted using these common genes that intersected.

Functional enrichment analysis for common DEGs
The functional enrichment analysis was carried out in three GO domains: biological process (BP), cellular component (CC), and molecular function (MF) 12 .Databases of pathways involving illnesses, medications, chemicals, and biological processes can be found in the KEGG database [13][14][15][16] .In order to ascertain the biological roles of the genes and related pathways, the enrichment analysis was performed using the R package "clusterProfiler." A p value of less than 0.05 was deemed statistically significant.

Construction of protein-protein interaction (PPI) network and identification of hub genes
We created a PPI network 17 using the Search Tool for the Retrieval of Interacting Genes (STRING) (http:// stringdb.org/) in order to investigate the interactions between the common genes that were found above in more detail.Cytoscape 3.9.1 was used to display the findings.The proteins were represented by nodes in the network result, while the interactions between proteins were shown by lines.We installed the MCODE and cytoHubba plugin, which can be downloaded for free via Cytoscape software, after finding the hub genes among these common genes in order to analyze the correlation among these genes 18,19 .Additionally, genes for functional tests were prioritized by GeneMANIA (http:// genem ania.org/) 20 .

Establishment and evaluation of the prognosis model
In biomedical research, it is important to select the variables most associated with the studied outcome and to determine the strength of this association.The "glmnet" package was used to obtain the LASSO algorithm to minimize the dimensionality of the data 21 .Besides, using an "SVM" package, a support vector machine-recursive feature elimination (SVM-RFE) model was created, and the average misjudgment rates of their tenfold crossvalidations were compared 22 .Additionally, the impact of confounding factors was lessened by applying the multivariable logistic regression analysis 23 .A nomogram is primarily used to summarise the specific and statistically important parameters obtained from multivariable logistic regression analysis as a simple, two-dimensional graphic 24 .The "rms" package in R was used to estimate AMI and AD risk.A diagnostic nomogram model for AMI and AD was created by summarising logistic regression's independent components.

Gene set variation analysis (GSVA) analysis
Gene set enrichment analysis (GSEA) is a popular method for condensing data from gene expression profiles into pathways or signature summaries.More accurately than GSEA, gene set variation analysis (GSVA) may identify minute changes in pathway activity in sample populations.GSVA is the basis for developing biological models that are centered around pathways 25 .The GO set served as the study's background set, and each marker gene was subjected to GSVA analysis using the R software's "GSVA" package.In the event that t > 0 signifies the activation of a function in the high-expression group, t < 0 implies that this activation is present in the low-expression group.

Differential expression analysis of DEGs in AMI and AD
Figure 1 displays the flowchart for this article.In the GSE5281 dataset, we found 414 upregulated genes and 295 downregulated genes for AD patients after excluding the batch effect (Fig. 2A) (Supplementary Table 1).The GSE66360 dataset revealed 350 genes with substantial changes for AMI patients, including 288 upregulated and 62 downregulated genes (Fig. 2B) (Supplementary Table 2).Based on this, we compared the differentially expressed genes in AMI and AD patients, ultimately obtaining a total of 22 differentially expressed genes, consisting of 21 upregulated genes and 1 downregulated gene (Fig. 2C,D).

Functional enrichment analysis
In order to obtain a more comprehensive understanding of the role of DEGs, enrichment studies of KEGG and GO pathways were made.Changes in BP mainly included inflammatory reactions (e.g., regulation of inflammatory response) and signal transduction (e.g., maintenance of protein location and positive regulation of cold-induced thermogenesis).Besides, the changes in CC were notably focused on cell secretion, such as secretory granule lumen, cytoplasmic vesicle lumen, and vesicle lumen.Moreover, in the MF section, changes were significant in DNA-binding transcription activator activity, RNA polymerase II-specific (Fig. 3A-C) (Supplementary Table 3).In particular, changes in the KEGG pathway mainly were enriched in several immune-related pathways (IL-17 signaling pathway, TNF signaling pathway, and Chemokine signaling pathway) (Fig. 3D-F) (Supplementary Table 4).

PPI network analysis and hub gene selection
To identify hub genes and show potential interactions between proteins encoded by common DEGs, STRING screened the DEGs' PPI network, which had 19 nodes and 64 edges (Fig. 4A).To find important clustering modules, module analysis was carried out using the Cytoscape plug-in MCODE.Two modules were retrieved from the PPI network constructed using common DEGs.Module 1 included ten nodes and 18 edges, and module 2 with three nodes and three edges (Fig. 4B).Besides, through the seven plug-in cytoHubba algorithms, we calculated the top ten hub genes for the following analysis (Supplementary Table 5), and the relationship among ten hub genes was displayed in Fig. 4C.GeneMANIA conducted PPI analysis of the ten hub genes and their twenty interacting genes to predict correlations between colocalization, pathways, shared protein domains, coexpression, and prediction (Fig. 4D).The inner circle contains the hub genes, and the outside circle contains the predicted genes.As shown in Fig. 4D, the network illustrates that these genes were mainly enriched in positive regulation of transcription by RNA polymerase II, positive regulation of transcription from RNA polymerase II promoter in response to stress, and myeloid leukocyte differentiation.

Hub gene functional enrichment analysis
GO analysis showed that these hub genes are mainly involved in regulation of inflammatory response, chromatin DNA binding, and ossification (Fig. 5A) (Supplementary Table 6).In addition, KEGG Pathway analysis showed that they are mainly involved in the IL-17 signaling pathway, TNF signaling pathway, and Chemokine signaling pathway (Fig. 5B) (Supplementary Table 7).The above hub genes functional analysis shows that the AMI and the AD have a strong relationship with inflammation.

Establishment and evaluation of the prognosis model
We then applied the SVM-RFE algorithm to filter the ten hub genes to identify the optimal combination of feature genes.Nine genes (maximal accuracy = 0.919, minimal errors = 0.0811) were identified as the optimal feature genes for the AMI patients (Fig. 6A a,b).Meanwhile, we used the LASSO method to construct a model for the target gene set, and based on the ten-fold cross-validation, we obtained a six gene model graph (Fig. 6A c,d).In addition, we took the intersection genes between the SVM-RFE and LASSO results to obtain six model genes (Fig. 6B).Next, we analyzed these six genes with logistic regression and found that only three hub genes with a p-value less than 0.05 were obtained: GAJ1, BCL6 and RGS1.Further, we constructed the predictive model using the two hub genes.To calculate the risk score of each one, the following formula was applied: Risk score    www.nature.com/scientificreports/Meanwhile, six genes (maximal accuracy = 0.888, minimal errors = 0.112) were identified as the optimal feature genes for AD patients through the SVM-RFE algorithm (Fig. 7A a,b).Meanwhile, we used the LASSO method to construct a model for the target gene set, and based on the ten-fold cross-validation, we obtained an eight gene model graph (Fig. 7A c,d).In addition, we took the intersection genes between the SVM-RFE and LASSO results to obtain five model genes (Fig. 7B).Next, we analyzed these six genes with logistic regression and found that three hub genes with a p-value less than 0.05 were obtained: CEBPB, BCL6, and NFKBIA.Further, we constructed the predictive model using the three hub genes.To calculate the risk score of each one,  www.nature.com/scientificreports/

GSVA analysis
We intersected the model genes based on the AMI and AD prediction models and discovered that the BCL6 gene differed between the two disorders.Subsequently, we ran GSVA analysis on a BCL6-based training set.Based on GO functional analysis, we found that AMI individuals had a high expression about cell proliferation, such as regulation of endodermal cell-differentiation, while had a lower expression about immune response, such as lipopolysaccharide immune receptor activity (Fig. 8A).Besides, for AD patients, we found that AD individuals had a high expression about cell proliferation and migration, such as sphingosine-I-phosphate phosphatase activity, while had a lower expression about immune response, such as tumor necrosis factor binding (Fig. 8B).

Key gene validation
Moreover, we corroborated that BCL6 was expressed in the validation dataset.The expression trends of BCL6 were found to be in agreement with the test dataset.We created ROC curves and determined the associated area under the curve (AUC) of these gene expression levels to validate the diagnostic value of BCL6 found in the investigation.The BCL6 gene was discovered to be expressed at a higher level in AMI patients than in normal patients, and the single gene prediction ability for AMI patients is strong (AUC = 0.734) (Fig. 10A,B).Additionally, AD patients' BCL6 expression levels were still rising, and the single gene prediction performance is strong (AUC = 0.817) (Fig. 10C,D).

Discussion
In modern Western countries, cardiovascular diseases were the leading causes of morbidity and mortality [27][28][29] .
It is presently believed that pathogenetic pathways resulting in atherogenesis and atherosclerosis complications are abnormal lipid metabolism and inflammatory processes.For example, a slight increase in blood C reactive protein (CRP) levels, a measure of inflammation, is linked to an increased risk of cardiovascular events 30 .Promoter polymorphisms having functional significance in the expression of the cognate inflammatory gene are frequently observed in patients with AMI 31 .AD was the most frequent type of dementia in the elderly 32 .Pathologically, it was distinguished by the loss of neuronal synapses, extracellular amyloid protein deposits known as neuritic plaques, and intracellular production of defective neurofilaments that form neurofibrillary tangles 33 .Previous studies had revealed that inflammatory processes have been detected in the brains of people with AD 34 .Besides, gene variations that increase inflammation or change cholesterol transport had been commonly discovered in AD patients 35,36 .
Despite these similarities, little data explicitly connected AD and AMI.Epidemiological studies that demonstrated the reduction of inflammation by the use of statins or non-steroidal anti-inflammatory medicines (NSAIDs) decreased the incidence of both AD and AMI might provide the most substantial evidence yet that AMI and AD are related 28,37 .Nevertheless, no effort had been undertaken to assess the possibility that particular immunological genetic risk factors represent a significant pathogenetic and etiologic connection between AMI and AD.Therefore, it was essential to investigate the molecular mechanisms behind these two diseases and identify targets to stop the early disease's progression.
We discovered 22 shared DEGs between AD and AMI in this study by exploring the GEO databases for these two disorders.In addition, we carried out analyses of the KEGG pathway and GO enrichment, and we built a PPI network to determine which ten hub genes were among the DEGs.Next, we used the SVM-RFE and multivariable logistic regression analysis technique to build models for patients with AMI and AD.The BCL6 gene exhibited differential expression in both illnesses and demonstrated a strong predictive capacity (p < 0.05, AUC > 0.8).This means we must accept the genetic resemblance between AMI and AD.
While vigorous acute inflammatory responses were necessary for survival, uncontrolled inflammation was at the root of many diseases, including atherosclerosis.Chronic inflammation was a hallmark of atherosclerosis to the extent that acute transcriptional induction is a driver of inflammation.From early fatty streaks to mature lesions, MØ played a critical role in atherogenesis while the BCL6 played a pivotal role in MØ quiescence and the regulation of inflammation 38 .Previous research had found that BCL6 reduces oxidized LDL and TLR-induced inflammation and atherogenesis, indicating that BCL6 plays an important role in reducing early vascular lesion development 39 .Surprisingly, nevertheless, our research revealed that BCL6 gene expression was up-regulated, which may be connected to negative feedback regulation in individuals with high levels of atherosclerosis.Interestingly, prior research had also discovered increased BCL6 gene expression in individuals suffering from coronary heart disease 40 .Hence, additional mechanisms must be proven through additional tests.
Previous data unraveled that bone marrow mesenchymal stem cells-derived extracellular vesicles carrying miR-302d-3p repressed cardiomyocyte injury and inflammatory response after AMI by disrupting the NF-κB pathway via the BCL6/MD2 axis, which means the BCL6 played an essential role in early inflammation response 41 .Concurrent with this, a prior work discovered that BCL6 upregulation participated in the promotion of oxygen-glucose deprivation-induced increase of apoptosis and decreased cell proliferation after ischemic stroke 42 .Also, BCL6 upregulation was able to aggravate the pulmonary inflammation following influenza virus infection 43 .These data indirectly supported the effect of BCL6 on apoptosis, inflammation, and viability.
Although prior studies individually investigated the hub genes related to AMI and AD, few studies have used bioinformatics approaches to investigate their common molecular mechanism 44,45 .Due to the significant

Conclusion
In conclusion, we carried out enrichment and PPI network analysis, as well as determined the shared DEGs of AMI and AD.Our research revealed that AMI and AD shared numerous pathogenic pathways, maybe mediated by certain hub genes.In the development of AMI and AD, the hub gene, specifically the BCL6, is crucial.This study adds to our understanding of the molecular mechanism of AMI in combination with AD.

Figure 1 .
Figure 1.The flow chart of this study.

Figure 2 . 4 =
Figure 2. Differentially expressed genes.(A) The volcano plot for AD.(B) The volcano plot for AMI.(C) Upregulated genes between AMI and AD.(D) Downregulated genes between AMI and AD.

Figure 3 .
Figure 3. Functional enrichment analysis.(A-C) The circle plot (A), bar plot (B), and bubble plot (C) via GO analysis.(D-F) The circle plot (D), bar plot (E), and bubble plot (F) via KEGG analysis.

Figure 4 .
Figure 4. PPI analysis.(A) PPI network of the DEGs.(B) Two significant gene clustering modules via MCODE.(C) Ten hub genes via cytoHubba.(D) The gene-gene interaction network for hub genes via the GeneMANIA database.

Figure 5 .
Figure 5. Hub gene functional enrichment analysis.(A) The circle plot via GO analysis.(B) The circle plot via GO analysis.

Figure 6 .
Figure 6.Establishment and Evaluation of the AMI Prognosis Model.(A) The SVM-RFE and LASSO algorithm, (B) The six marker genes obtained from the LASSO and SVM-RFE models.(C) The nomogram of the Prognosis Model.(D) ROC analysis of this model.(E) The calibration curve.

Figure 7 .
Figure 7. Establishment and Evaluation of the AD Prognosis Model.(A) The SVM-RFE and LASSO algorithm, (B) The five marker genes obtained from the LASSO and SVM-RFE models.(C) The nomogram of the Prognosis Model.(D) ROC analysis of this model.(E) The calibration curve.

Figure 8 .
Figure 8. Gene set variation analysis.(A) GSVA between AMI and normal samples.(b) GSVA between AD and normal samples.