Abstract

Objective

   Atherosclerosis(AS) is a vascular disease characterized by the
   development of plaque in the arteries, and atrial fibrillation (AF) is
   a common heart arrhythmia. These two conditions share several risk
   factors in common, such as aging, diabetes, obesity, and hypertension.
   Ferroptosis is a new mode of non-apoptotic cell death that plays a key
   role in cardiomyocyte death and has been associated with a variety of
   cardiac diseases. This study aimed to investigate the ferroptosis
   biomarkers and underlying biological mechanisms associated with AF and
   AS.

Materials and Methods

   The gene expression dataset was obtained from GEO database,
   differentially expressed genes (DEGs) and ferroptosis expressed genes
   (FDGs) were obtained by data processing and screening, and then
   functional enrichment, network construction, transcription factor
   prediction, identification of biomarkers by LASSO and SVM - RFE
   algorithms, and also immune infiltration analyses and cellular
   experiments were performed.

Results

   In AF and AS, 1627 and 571 DEGs were identified respectively, and 128
   were intersected, and 47 common FDGs were also identified. Gene
   Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG)
   analysis of DEGs revealed that they were associated with biological
   processes and pathways such as leukocyte immunity, and FDGs were also
   involved in specific functions and pathways. Fifteen key genes were
   identified, CSF1R and ITGAM expression differences were verified, and
   seven transcription factors were predicted to be differentially
   expressed. Characterized genes were screened to construct models with
   good diagnostic efficacy, and immune infiltration showed that NUPR1 was
   associated with altered immune environments, and WB indicated that
   NUPR1 was highly expressed in the disease model.

Conclusion

   Our study demonstrates that the ferroptosis gene NUPR1 plays a role in
   the pathogenesis of atrial fibrillation and atherosclerosis, and also
   provides valuable insights into their molecular mechanisms, which may
   contribute to the development of new targets and strategies for the
   treatment of these diseases.

   Keywords: atrial fibrillation, AF, atherosclerosis, AS, ferroptosis,
   CSF1R, ITGAM, NUPR1

Introduction

   Atrial fibrillation (AF), the most prevalent cardiac arrhythmia, has
   become more widespread globally as the population ages.[38]^1^,[39]^2
   Despite advances in management, a considerable risk of morbidity and
   death still exists. The prevention of AF and its associated
   complications remains a challenge, and this multifactorial arrhythmia
   is intertwined with common concurrent cardiovascular diseases and is a
   classic cardiovascular risk factor.[40]^3–6 One such cardiovascular
   condition commonly associated with AF is atherosclerosis (AS). AS is
   characterized by arterial wall damage leading to plaque formation,
   vessel narrowing, and reduced blood flow.[41]^7–9 Both AF and AS share
   several risk factors, including aging, obesity, diabetes, and
   hypertension, suggesting potential common pathophysiological
   mechanisms.[42]^10–12 The pathogenesis of AS involves multiple cellular
   processes, including endothelial dysfunction, oxidative stress,
   inflammation, and vascular smooth muscle cell
   alterations.[43]^8^,[44]^9^,[45]^13^,[46]^14

   Recent research has highlighted the role of ferroptosis, a novel form
   of regulated cell death, in cardiovascular diseases. Unlike apoptosis,
   necrosis, or autophagy, ferroptosis is characterized by iron
   accumulation, mitochondrial changes, and lipid peroxidation.[47]^15–17
   While the mechanisms linking AF and AS are not fully understood,
   emerging evidence suggests that inflammatory processes and oxidative
   stress may represent common pathways in both conditions.[48]^18

   The objective of this research is to find similar genetic signatures
   between AF and AS by thorough transcriptional profiling, potentially
   revealing common molecular pathways that could explain their frequent
   coexistence and provide new insights for both conditions.

Materials and Methods

Data Collection

   We obtained gene expression datasets from GEO
   ([49]http://www.ncbi.nlm.nih.gov/geo), a public database. Using the
   search terms atherosclerosis(AS) and atrial fibrillation (AF).
   Following was decided upon as the inclusion criteria: 1. The included
   test products should be from human beings 2. Two separate expression
   profiles, and comprise a high number of samples (≥30). Finally,
   download four microarray datasets ([50]GSE41177, [51]GSE2240,
   [52]GSE28829, [53]GSE100927). The [54]GSE41177 dataset comprises 32
   pairs of patients with AF and 6 pairs of patients with sinus rhythm.
   The [55]GSE2240 dataset has 20 pairs of patients with AF and 40 pairs
   of sinus rhythm, we utilize the [56]GPL96 platform, which has 10 pairs
   of patients with AF and 20 pairs of sinus rhythm, which is used for
   validation. The [57]GSE28829 dataset comprises of 13 early AS Composed
   of sclerotic plaque samples and 16 advanced atherosclerotic plaque
   samples. [58]GSE100927 comprised of 35 non-atherosclerotic plaque
   samples and 69 atherosclerotic plaque samples, which were utilized for
   validation. Data analysis and collation were done out using R software
   (version 4.2.3) and Perl software (v5.30.0).

Screening Differentially Expressed Genes(DEGs) in AF and AS

   The expression profiles of the two sets of GEOs were transformed into
   gene expression matrices using Perl software, and R was used to
   batch-correct the data sets to get normalized expression levels. Log
   transformation and DEGs screening were done on the normalized data
   using the “Limma” and “Pheatmap” packages in R software. |logFC|>0.585
   and adjusted P-values <0.05 were selected as threshold thresholds of
   statistical significance for DEGs samples. The “VennDiagram” tool in R
   software was used to collect their public DEGs, and 128 genes were
   obtained.

Screening Ferroptosis Differentially Expressed Genes(FDGs) in AF and AS

   Then we downloaded the latest Ferroptosis-related genes from the
   website ([59]http://www.zhounan.org/ferrdb/current/), and analyzed the
   difference of iron death-related genes in [60]GSE41177 and [61]GSE28829
   datasets.172 Ferroptosis differential genes (FDGs) were obtained in the
   AF group and 109 in the AS group. Taking the intersection of the two
   finally obtained 47 FDGs.

Enrichment Analysis of DEGs

   The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes
   (KEGG) enrichment analysis of these 128 genes was done using the
   “clusterProfiler” package in R. The GO enrichment analysis split the
   gene biological function into three parts: cellular component (CC),
   molecular function (MF), and biological process (BP); KEGG is a
   database connected to important metabolic and signal transduction
   pathways. Where p-values and adjusted p-values are both < 0.05.

Protein-Protein Interaction (PPI) Network Construction and Hub Gene Analysis

   Use the Search Tool for the Retrieval of Interacting Genes/Proteins
   (STRING, [62]https://string-db.org/) to identify PPIs in DEGs. STRING
   is a database that looks for connections between proteins, including
   direct or indirect interactions. Interactions with a cumulative score
   of higher than 0.4 were judged statistically significant, and 0.7 was
   eventually picked. Use Cytoscape (version 3.9.1) to visualize this PPI
   network. The essential functional modules were examined utilizing
   Cytoscape plug-in molecular complex identification technology (MCODE).
   Set the selection criteria to be: node density cutoff = 0.1, node score
   cutoff = 0.2, K-core = 2, max depth = 100, and ultimately get 4
   sub-networks. Obtain the hub genes using the CytoHubba plug-in in
   Cytoscape program.

GeneMANIA Network Construction and Functional Enrichment

   GeneMANIA ([63]http://www.genemania.org/), which is frequently used to
   generate hypotheses about gene function, evaluate gene lists, and
   choose genes for functional inquiry, created a co-expression network
   comprising these hub genes. The involved module hub genes were then
   subjected to GO and KEGG analyses using R software.

Verification of Hub Genes Expression

   Expression of the identified hub gene was validated in [64]GSE2240 and
   [65]GSE100927. The data of the two groups were compared using t-test,
   and the P value was < 0.05.

Prediction of Transcription Factors (TFs) and Validation

   Transcriptional Regulatory Relationships Unraveled by Sentence-based
   Text mining (TRRUST, [66]https://www.grnpedia.org/trrust/) is a
   regulatory network used to predict transcription factors(TFs) and the
   regulatory link between the target gene of transcription factors and
   transcription factors. Eight hundred human TFs and eighty-two mouse TFs
   are connected to 6552 and 8444 regulatory targets, respectively, in the
   current version of TRRUST. The TRRUST database provided the TFs for the
   regulatory hub genes, with P value < 0.05. We utilize these TFs in
   [67]GSE41177 and [68]GSE28829 to determine if there are differences.

Identification of AS and AF Ferroptosis Gene Biomarkers

   The glmnet package was used to use the least absolute shrinkage and
   selection operator (LASSO) method to minimize the dimensionality of the
   data. The FDGs between AS and AF patients and normal samples were kept
   for feature selection, and LASSO algorithms were used to find the gene
   biomarkers for AS and AF. The average misjudgement rates of their
   10-fold cross-validations were used to compare the support vector
   machine-recursive feature elimination (SVM-RFE) model, which was
   created with an SVM. Additionally, overlapping biomarkers produced from
   the two algorithms were used to find the best gene biomarkers for AS
   and AF. By computing the receiver operating characteristic (ROC) curve
   and calculating the area under the curve (AUC), accuracy, sensitivity,
   and specificity, the best gene biomarkers’ diagnostic potential was
   evaluated.

Immunoinfiltration Analysis

   CIBERSORT, an algorithm for analyzing the gene expression profile to
   determine the cell composition of complex tissues. In this work, we
   used the [69]GSE41177 and [70]GSE28829 datasets from CIBESORT to
   estimate the proportion of 22 kinds of invading immune cell types in
   each tissue. The total analyzed immune cell type fractions for each
   sample were equal to 1.

Cell Culture

   The American Typical Culture Collection (Manassas, VA, USA) is where
   HL-1 and THP-1 were obtained. AF modeling: HL-1 cells were cultured in
   DMEM/F12 medium (Sparkjade, China) containing 10% fetal bovine serum
   (FBS) (BI, Israel), and 1% penicillin-streptomycin (MCE, USA) at 37 °C
   with 5% CO2. As previously reported in investigations,[71]^19^,[72]^20
   HL-1 cells were tachypaced using a YC-3 stimulator (Nanjing, China).
   The cells were stimulated at 10 hz using 1.5 V/cm pulse voltage and
   square pulses lasting 5 ms. AS modeling: As we have done before, THP-1
   cells were utilized to promote differentiation into the AS
   model.[73]^21 The cells were cultivated at 37 °C with 5% CO2 in 1640
   medium (BI, Israel), which included 10% fetal bovine serum (FBS) (BI,
   Israel), 5% β-mercaptoethanol (Sigma, USA), and 1%
   penicillin-streptomycin (MCE, USA). To create atherosclerotic
   macrophages for more research, phorbol 12-myristate-13-acetate (PMA)
   (Sigma, USA) was treated for 24 hours, and then oxidized-LDL (80 μg/mL)
   (Yiyuan Bio, China) was added for 48 hours.

Western Blot Analysis

   Proteins were extracted from treated cells, The total protein
   concentrations in cells were measured using the BCA Kit. Western
   blotting was performed according to standard protocols. The following
   targets were incubated on membranes: NUPR1 (YT8077, Immunoway, China);
   and GAPDH (D110016, BBI, China). The ECL substrate chemiluminescence
   solution (E422-01, Vazyme China) was applied uniformly to the
   membranes. The gray value of each band was quantified using the ImageJ
   software.

Result

Identification of Differentially Expressed Genes in AF and AS

   The [74]GSE41177 dataset comprises 32 pairs of patients with AF and 6
   pairs of patients with sinus rhythm. The [75]GSE28829 dataset comprises
   13 samples of early AS plaques and 16 samples of late atherosclerotic
   plaques. The expression files of GEO were evaluated using |log2 FC| >
   0.585 and corrected p-value < 0.05, with batch correction. The former
   tested a total of 1627 DEGs, comprising 780 up-regulated genes and 847
   down-regulated genes, whereas the latter screened a total of 571 DEGs,
   including 161 up-regulated genes and 410 down-regulated genes. The heat
   map displayed the 50 most significantly up- and down-regulated genes,
   and the differentially expressed genes were then taken as an
   intersection and represented as a Venn diagram ([76]Figure 1A and
   [77]B), of which 104 were co-up-regulated genes and 24 were
   co-down-regulated genes. These visualizations were done using heat maps
   and volcano maps ([78]supplementary Figure 1).

Figure 1.

   [79]Figure 1
   [80]Open in a new tab

   (A) These two data sets demonstrate the upregulation of 104 DEGs that
   overlap. (B) These two data sets demonstrate the downregulation of 24
   overlapping DEGs.(C) The GO enrichment analysis histogram (D) Bubble
   plot of GO analysis. (E) Circle diagram showing the analysis of GO
   enrichment. (F) The KEGG pathway analysis histogram. (G) A KEGG pathway
   analysis bubble diagram. The threshold for significance was an adjusted
   P value <0.05.

Identification FDGs in AF and AS

   The 172 and 109 FDGs identified from the [81]GSE41177 and [82]GSE28829
   datasets, respectively, were differentially expressed compared to
   normal samples. Clustered heatmaps showed the expression patterns of
   FDGs in the samples ([83]supplementary Figure 2). The intersection of
   these two groups of FDGs was later taken and 47 common genes were
   obtained ([84]supplementary Figure 2). The correlation between these
   genes in is shown in [85]Figure 2A and [86]B.

Figure 2.

   [87]Figure 2
   [88]Open in a new tab

   (A and B) The correlation of FDGs between AF and AS. (C and D) 11
   AF-related characteristics were chosen using the LASSO logistic
   regression approach with penalty parameter adjustment carried out using
   10-fold cross-validation. Similar to (E and F), 14 AS-related
   characteristics were obtained.(G and H) SVM-RFE algorithm to filter the
   4 FDGs to find the ideal mashup of feature genes. The ideal AF feature
   genes were ultimately found to be 4 genes. (I and J) Four AS optimum
   features were also obtained. *p < 0.05; **p < 0.01; ***p < 0.001.

Functional Enrichment Analysis of DEGs

   To investigate the biological processes and pathways involved, 94
   frequently occurring DEGs were submitted to GO and KEGG pathway
   enrichment analyses. These genes were primarily enriched in
   leukocyte-mediated immunity, leukocyte movement, phagocytosis,
   processing of antigens, MHC class II presentation of peptide antigens,
   and positive control of cytokine production, according to the results
   of the GO analysis ([89]Figure 1C–E), all p<0.05. In terms of KEGG
   pathways, Staphylococcus aureus infection, Phagosome, and Tuberculosis
   were three highly enriched pathways ([90]Figure 1F and [91]G), all
   p0.05. These findings imply that the emergence and progression of both
   illnesses are concurrently influenced by leukocyte-mediated immunity
   and positive control of response to external stimulation ([92]Figure
   1E).

Functional Enrichment Analysis of FDGs

   GO enrichment and KEGG pathway analysis were carried out to clarify the
   biological processes and pathways connected to the FDGs. Therefore, GO
   enrichment studies revealed that the function of “cellular response to
   chemical and oxidative stress” was substantially connected to FDGs of
   AF and AS ([93]Figure 3A-F). Lipid and atherosclerosis, autophagy, and
   the FoxO signaling pathway were shown to be abundant in KEGG pathway
   studies ([94]Figure 3G-J). These pieces of data suggested that FDGs may
   be involved in the control of autophagy and oxidative stress, which may
   contribute to their potential role in the pathogenesis of AF and AS.

Figure 3.

   [95]Figure 3
   [96]Open in a new tab

   Functional analyses for the FDGs. (A-F) GO enrichment studies revealed
   that the function of “cellular response to chemical and oxidative
   stress” was substantially connected to FDGs of AF and AS. (G-J) KEGG
   pathway analyses indicate Lipid and atherosclerosis, autophagy, and the
   FoxO signaling pathway were enriched.

DEGs PPI Network Construction and Selection of Hub Genes

   Using Cytoscape, a PPI network with 80 nodes and 248 interaction
   pairings was created for frequent DEGs with a composite score higher
   than 0.7 ([97]Figure 4A). The top 15 hub genes were determined by
   utilizing the Cytoscape plug-in’s cytoHubba feature and using degree as
   the scoring type, including LY86, ITGAM, CSF1R, TYROBP, FCER1G, TLR2,
   HLA-DRA, ITGB2, CD14, LYN, C1QB, CD86, LILRB2, HCK, and VAV1
   ([98]Figure 4B and [99]Table 1). With the help of the GeneMANIA
   database, we investigated the co-expression network and related
   functions of these genes. With 69.88% coexpression, 20.59% physical
   interaction, 3.98% pathway, 3.12% colocalization, and 2.44% prediction,
   these genes displayed a sophisticated PPI network. ([100]Figure 4C).
   The MCODE component of Cytoscape was used to derive four closely
   related gene modules, comprising 31 shared DEGs and 63 interaction
   pairings ([101]Figure 4D-G).

Figure 4.

   [102]Figure 4
   [103]Open in a new tab

   (A) PPI network diagram. Champagne yellow denotes down-regulated genes,
   whereas lavender blue denotes up-regulated genes. (B) Cytoscape was
   used to screen 15 overlapping hub genes; the deeper the hue, the more
   likely the gene is a network core gene.(C) GeneMANIA was used to
   evaluate the hub genes and the genes that co-expressed with them. Four
   important gene clustering modules are shown in (D–G).

Table 1.

   Top 15 hub Gene Scores
   Node_name MCC DMNC MNC Degree EPC Bottleneck Eccentricity Closeness
   Radiality Betweenness Stress Clustering Coefficient
   LY86 1 0.47886 10 1 14.869 3 0.24375 38.83333 3.6974 130.67155 660
   0.43636
   ITGAM 1 0.44065 15 1 17.169 7 0.325 43.83333 3.95065 800.06075 2792
   0.32353
   CSF1R 1 0.42605 18 1 17.831 2 0.325 43.5 3.9 317.39795 1470 0.37908
   TYROBP 1 0.31982 24 1 19.556 13 0.325 48.83333 4.1026 1104.31658 3630
   0.21846
   FCER1G 1 0.41061 15 1 16.695 6 0.325 43 3.91266 311.46895 1596 0.34167
   TLR2 1 0.27524 17 1 15.917 6 0.325 43.66667 3.93799 549.88683 1996 0.25
   HLA-DRA 8 0.37893 4 4 7.425 1 0.24375 30.75 3.21623 3.13799 26 0.66667
   ITGB2 6 0.64826 5 3 5.008 1 0.195 28.15 2.86169 0 0 1
   CD14 201 0 1 11 3.242 1 0.195 25.58333 2.7224 0 0 0
   LYN 4 0.66569 6 3 11.406 2 0.24375 34.33333 3.43149 153.15404 544
   0.66667
   C1QB 393 0 1 14 2.206 1 0.195 23.5 2.50714 0 0 0
   CD86 135 0.30779 2 9 4.942 1 0.24375 28.83333 3.10227 0 0 1
   LILRB2 4 0.28529 6 3 8.256 1 0.24375 31.66667 3.21623 23.52383 96 0.4
   HCK 17 0.30779 2 5 5.747 1 0.24375 29.25 3.06429 7.81746 32 0.33333
   VAV1 1 0.3063 24 1 19.389 13 0.325 49.33333 4.19123 1281.57541 4634
   0.24638
   [104]Open in a new tab

Functional Analysis of Hub Genes

   We conducted GO and KEGG pathway enrichment analysis to examine the
   biological pathways and functions of hub genes. According to the
   results of the GO analysis, these genes were primarily involved in
   phagocytosis, Cellular reactions to compounds of bacterial origin and
   lipopolysaccharides, the effective controlling of immune effector
   mechanisms, cellular responses to biotic stimuli, positive regulation
   of leukocyte activation, leukocyte-mediated immunity, and positive
   regulation of response to external stimuli ([105]Figure 5A–C). The
   results show that leukocyte immune responses and cellular immunological
   effects have an influence on both illnesses. They are implicated in
   Staphylococcus aureus infection, Hematopoietic cell lineage, Phagosome,
   Pertussis, Leishmaniasis, Tuberculosis, and Legionellosis, according to
   KEGG pathway analysis ([106]Figure 5D–F). GO and KEGG enrichment data
   show a substantial correlation between hub genes and the immune system.

Figure 5.

   [107]Figure 5
   [108]Open in a new tab

   GO and KEGG pathway enrichment analysis. (A-C) Analysis of GO
   enrichment (D-F) analysis of KEGG pathway enrichment.

Validation of Hub Genes

   To validate these hub genes’ reliable expression levels. We chose two
   more datasets that had atherosclerotic plaques and AF, and we examined
   how much these hub genes were expressed in those datasets. Only CSF1R
   and ITGAM genes were shown to be downregulated in AF as compared to
   normal tissue, according to the findings ([109]Figure 6A). In
   comparison to healthy vascular tissue, atherosclerotic plaques had
   greater expression levels of the genes CSF1R, ITGAM, TLR2, FCER1,
   TYROBP, and LY86 ([110]Figure 6B).

Figure 6.

   [111]Figure 6
   [112]Open in a new tab

   (A) The expression level of the hub gene in [113]GSE2240. A comparison
   of data between the two groups was performed using the mean t-test.
   p-value <0.05 was considered statistically significant. *P<0.05;
   **P<0.01; ***P<0.001. (B) The expression level of the hub gene in
   [114]GSE100927. A comparison of data between the two groups was
   performed using the mean t-test. p values < 0.05 were considered
   statistically significant. ns, not significant,*P<0.05; ***P<0.001.

Prediction of TFs and Validation

   Seven transcription factors (TFs) that may regulate the expression of
   these genes were found using the TRRUST database ([115]Figure 7A and
   [116]Table 2). We discovered that the transcription factors HIF1A,
   SPI1, and TRERF1 were substantially expressed in AF ([117]Figure 7B) to
   provide additional support. While transcription factor TRERF1 was
   reduced in atherosclerotic plaques, transcription factors RELA and SPI1
   were upregulated ([118]Figure 7C).

Figure 7.

   [119]Figure 7
   [120]Open in a new tab

   (A) TFs regulate the network. TFs are in the periphery, marked in light
   green, and hub genes are in the middle, marked in light purple. (B and
   C) Expression levels of TFs in [121]GSE41177 and [122]GSE28829.
   Comparison of data between the two groups was performed using the mean
   t-test. p values < 0.05 were considered statistically significant. ns,
   not significant, *p < 0.05; **p < 0.01; ***p < 0.001.

Table 2.

   Key Transcriptional Factors (TFs) of Hub Genes
   Key TF Description Of Overlapped Genes P value Q value List of
   Overlapped Genes
   TRERF1 Transcriptional regulating factor 1 2 1.24E-05 5.23E-05 CD86,
   LILRB2
   SPI1 Spleen focus forming virus (SFFV) proviral integration oncogene
   spi1 3 1.49E-05 5.23E-05 ITGAM, HCK, ITGB2
   RUNX1 Runt-related transcription factor 1 2 0.000452 0.00105 ITGB2,
   CSF1R
   HIF1A Hypoxia inducible factor 1, alpha subunit (basic helix-loop-helix
   transcription factor) 2 0.00193 0.00338 TLR2, ITGB2
   SP1 Sp1 transcription factor 3 0.00565 0.00792 ITGB2, TLR2, CD14
   RELA v-rel reticuloendotheliosis viral oncogene homolog A (avian) 2
   0.0232 0.0235 CD86, TLR2
   NFKB1 Nuclear factor of kappa light polypeptide gene enhancer in
   B-cells 1 2 0.0235 0.0235 CD86, TLR2
   [123]Open in a new tab

2 FDGs Identified as Co-Diagnostic Genes for AS and AF

   Considering the differences between AF and AS patients and healthy
   individuals, we aimed to estimate the diagnostic potential of FDGs
   ([124]Figure 2A and [125]B). Next, we used two different machine
   learning algorithms, LASSO and SVM-RFE, in the [126]GSE41177 and
   [127]GSE28829 datasets to screen for important FDGs that can
   differentiate between AF, AS, and normal individuals. We used the LASSO
   logistic regression algorithm to screen for 11 features associated with
   AF, adjusting the penalty parameter through 10-fold cross-validation.
   We used the LASSO logistic regression algorithm to filter out 11
   features associated with AF ([128]Figure 2C and [129]D) and 14 features
   associated with AS ([130]Figure 2E and [131]F). Then, we used the
   SVM-RFE algorithm to screen the FDGs of AF and AS to determine the best
   combination of feature genes. Finally, four genes (maximum accuracy =
   1, minimum RMSE = 0) were identified as the best feature genes for AF
   ([132]Figure 2G and [133]H), and the best feature genes for AS were
   also four ([134]Figure 2I and [135]J) (maximum accuracy = 0.9, minimum
   RMSE = 0.1). Crossover of marker genes obtained from the LASSO model of
   AF and the SVM-RFE model identified 2 marker genes (NOX5, ADAMTS13) for
   subsequent analyses ([136]Figure 8A), and using the same approach, 4
   genes (CTSB, CP, NUPR1, PARP12) were identified in the model crossover
   for AS. Regrettably, we did not find AF intersecting with AS in several
   of the characterized genes, but we used these characterized genes to
   intersect with the previous 47 FDGs, and we obtained two
   Ferroptosis-related genes, CTSB, and NUPR1, and were associated with
   both AF, AS. Based on the above four marker genes, we built a logistic
   regression model using the R package glm, and the subsequent ROC curves
   showed that the logistic regression model based on the four marker
   genes could discriminate between normal samples and AFsamples, with
   AUCs of at least 0.781 ([137]Figure 8B). In addition, ROC curves for
   the marker genes were generated to elucidate the ability of individual
   genes to distinguish AF and normal samples. As shown in [138]Figure 8C,
   the AUC of all genes was 1. Meanwhile, normal and AS samples had an AUC
   of at least 0.846 ([139]Figure 8D) for the former and 1 ([140]Figure
   8E) for the latter. The above evidence suggests that the logistic
   regression model is more accurate and specific than individual marker
   genes in distinguishing AF, AS and normal samples.

Figure 8.

   [141]Figure 8
   [142]Open in a new tab

   (A) The marker genes obtained from the AF, AS LASSO and SVM-RFE models.
   (B and D) Logistic regression model to identify the AUC of AF, AS
   samples. (C and E) ROC curves for the 4 marker genes of AF, AS.

CTSB and NUPR1 are Closely Related to AF, AS-Related Pathways

   To further explore the potential function of marker genes to
   distinguish disease samples from normal samples, we performed
   single-gene GSEA-KEGG pathway analysis. The first 6 pathways enriched
   for each marker gene are shown in [143]Figure 9. After comprehensive
   analysis, we found that in AF, these CTSBs were enriched in “natural
   killer cell-mediated cytotoxicity”, “nitrogen metabolism”, “PPAR
   signaling pathway”, “starch and fructose metabolism”, “steroid hormone
   biosynthesis”, and “alanine-leucine and isoleucine degradation”;
   whereas, in AS, these CTSBs were enriched in “cytokine signaling
   pathway”, cytokine- cytokine receptor interactions”, this has the same
   result as NUPR1 enrichment in AF.

Figure 9.

   [144]Figure 9
   [145]Open in a new tab

   AF, AS Single-gene GSEA-KEGG pathway analysis in CTSB (A and C), NUPR1
   (B and D).

Analysis of Immune Cell Infiltration Landscape in AF, AS Patients

   The earlier findings suggested that AF, AS and the immunological
   microenvironment are inextricably linked.[146]^22–24 The earlier
   findings suggested that AF, AS and the immunological microenvironment
   are inextricably linked. In order to investigate the variations in the
   immune microenvironment between AF, AS patients and normal samples, we
   developed the CIBERSORT technique. According to [147]Figure 10A and
   [148]B, whereas T-cell follicular helper, Macrophages M2 were more
   expressed in normal samples than in AF samples, which had a larger
   number of plasma cells, monocytes, and neutrophils. B cells memory,
   macrophages M0, and macrophages M2 were more prevalent in AS samples
   than in normal samples, although T cells regulatory (Tregs), mast cells
   resting, and macrophages M0 were more prevalent in normal samples. As
   compared to the normal sample, AS, AF had a higher percentage of T
   cells gamma delta, but less dendritic cells that were activated. In
   addition, Pearson correlation analysis revealed that Mast cells resting
   and activated had strong positive and negative correlations with NUPR1
   in AF. CTSB was positively correlated with macrophages M0 while NUPR1
   was negatively correlated with NK cells resting in AS ([149]Figure 10C
   and [150]D). These pieces of information suggested that NUPR1 may be
   related to modifications in the immune milieu of AF, AS patients.

Figure 10.

   [151]Figure 10
   [152]Open in a new tab

   Analysis of immune cell infiltration landscape in AF, AS patients.(A
   and B) Implemented the CIBERSORT algorithm to explore the differences
   in the immune microenvironment between AF, AS patients and normal
   samples. (C and D) Mast cells resting and activated had strong positive
   and negative correlations with NUPR1 in AF. CTSB was positively
   correlated with macrophages M0 while NUPR1 was negatively correlated
   with NK cells resting in AS (*p < 0.05, **p < 0.01).

Expression of NUPR1 in AF and AS Models

   The expression of NUPR1 in AF and AS model was detected by Western
   blotting. The results showed that NUPR1 were weakly expressed in the
   normal model and strongly expressed in AF and AS models ([153]Figure
   11A-D).

Figure 11.

   [154]Figure 11
   [155]Open in a new tab

   Western blot: (A-D) NUPR1 are lowly expressed in normal group and
   highly expressed in AS. All data were represented as mean ± SEM, n=3.
   *P <0.05, **P <0.01.

Discussion

   In this study, we identified 128 shared DEGs between AF and AS
   conditions, along with 47 ferroptosis-related genes (FDGs) common to
   both diseases. Notably, CSF1R and ITGAM were identified as potential
   co-diagnostic markers, while NUPR1 emerged as a key ferroptosis-related
   gene involved in both conditions. Functional enrichment analysis
   revealed that these genes are mostly engaged in immune system
   functions, oxidative stress pathways, and inflammatory responses. The
   identification of FDGs represents a novel finding, suggesting that
   ferroptosis, an iron-dependent programmed cell death process, may serve
   as a crucial mechanistic link between AF and AS. Particularly, NUPR1,
   as a ferroptosis-related gene may be crucial in controlling the
   mechanisms that lead to cell death in both conditions, potentially
   providing novel targets for treatment.

   Our results have important clinical practice implications. The
   identification of CSF1R and ITGAM as co-diagnostic markers could
   potentially improve early detection strategies for both AF and AS. A
   diagnostic panel for risk stratification might be created using these
   indicators and the ferroptosis-related gene NUPR1. Previous studies
   have primarily focused on individual disease markers, whereas our
   integrated approach provides a more comprehensive understanding of the
   shared pathogenic mechanisms. This could lead to more effective
   targeted therapies addressing both conditions simultaneously.

   Our study provides some new insights compared to the studies by Wei et
   al and Wu et al[156]^25^,[157]^26 We found both overlap and new
   contributions. While previous studies explored the association between
   AF and AS, our study delves into the specific genetic mechanisms
   between the two and highlights the role of FDGs in both diseases.
   Unlike previous studies that have focused primarily on clinical
   outcomes and risk factors, our approach emphasises the genetic basis
   and common pathways of AF and AS. This distinction is critical as it
   allows us to propose potential therapeutic targets based on gene
   expression profiles.

   The relationship between AF and AS has been well proven in previous
   studies. Risk factors such as age, hypertension, diabetes, gender and
   obesity contribute to the development of both
   diseases.[158]^1^,[159]^10^,[160]^11^,[161]^27 The underlying
   mechanisms may involve inflammation, endothelial dysfunction and
   platelet-mediated thrombosis.[162]^1^,[163]^10–12 Local hemodynamic
   changes appear to be the primary mechanism in AF development. AS,
   stenosis, and carotid plaque formation can result from carotid
   endothelial injury due to abnormal blood flow.[164]^28

   Inflammation biomarkers play a crucial role in predicting
   cardiovascular events in AF patients.[165]^18^,[166]^29 Studies have
   shown the importance of various biomarkers, including IL-6, GDF-15,
   CRP, and cTnI in AF patients.[167]^30–33 Our findings regarding CSF1R
   and ITGAM add to this growing list of potential biomarkers,
   particularly highlighting their role in both AF and AS pathogenesis.

   The enrichment analysis of the DEGs provided further insights into the
   biological functions and pathways that may be dysregulated in AF and
   AS. The gene ontology (GO) analysis identified cellular components,
   molecular functions, and biological processes associated with the
   DEGs.[168]^5 KEGG analysis revealed key pathways potentially involved
   in disease development, particularly highlighting the role of
   inflammatory and immune response processes.

   However, our study still has some limitations. First, our study was
   based on the analysis of public databases, and further validation of
   these results in clinical samples is needed. Second, while we performed
   functional enrichment analysis, additional experimental validation of
   specific DEG mechanisms and transcription factors is needed. Third, the
   role of identified FDGs in disease progression requires further
   investigation through in vitro and in vivo studies.

Conclusions

   In conclusion, we identified ferroptosis as a potential mechanistic
   link between AF and AS through the discovery of 47 shared
   ferroptosis-related genes, particularly NUPR1, and two novel
   co-diagnostic markers (CSF1R and ITGAM). These findings reveal
   previously unrecognized molecular connections between the two
   conditions, extending beyond traditional shared risk factors. While
   experimental validation is needed, our results suggest new
   possibilities for developing integrated diagnostic approaches and
   therapeutic strategies targeting both AF and AS through
   ferroptosis-related pathways.

Acknowledgments