Comprehensive analysis identified a reduction in ATP1A2 mediated by ARID3A in abdominal aortic aneurysm

Abstract Abdominal aortic aneurysm (AAA) is characterized by abdominal aorta dilatation and progressive structural impairment and is usually an asymptomatic and potentially lethal disease with a risk of rupture. To investigate the underlying mechanisms of AAA initiation and progression, seven AAA datasets related to human and mice were downloaded from the GEO database and reanalysed in the present study. After comprehensive bioinformatics analysis, we identified the enriched pathways associated with inflammation responses, vascular smooth muscle cell (VSMC) phenotype switching and cytokine secretion in AAA. Most importantly, we identified ATPase Na+/K+ transporting subunit alpha 2 (ATP1A2) as a key gene that was significantly decreased in AAA samples of both human and mice; meanwhile, its reduction mainly occurred in VSMCs of the aorta; this finding was validated by immunostaining and Western blot in human and mouse AAA samples. Furthermore, we explored the potential upstream transcription factors (TFs) that regulate ATP1A2 expression. We found that the TF AT‐rich interaction domain 3A (ARID3A) bound the promoter of ATP1A2 to suppress its expression. Our present study identified the ARID3A‐ATP1A2 axis as a novel pathway in the pathological processes of AAA, further elucidating the molecular mechanism of AAA and providing potential therapeutic targets for AAA.

(approximately 80%) if not treated immediately. 6,7 However, at present, surgery is the only way to prevent AAA rupture with certain limitations (high cost, high requests for the surgeon and mortality 1-5%), especially for patients with large AAAs. 8 Therefore, it is necessary to fully understand the molecular and cellular mechanisms underlying AAA initiation and progression to establish an effective therapeutic strategy for AAA patients.
Further research has elucidated the underlying pathophysiological processes responsible for AAA, such as endothelial dysfunction, vascular smooth muscle cell (VSMC) apoptosis and depletion, elastic media destruction, and proteolytic degradation of the extracellular matrix (ECM) proteins elastin and collagen. 5 The molecular mechanisms that induce these pathological processes have been more or less clarified. 9,10 For example, infiltration of various immune cells has been regarded as a key feature and driver of AAA, especially macrophages which contribute to the destruction of the aortic normal lamellar architecture. 11,12 Although immune and inflammation have been proved to be the main culprits in initiation and progression of AAA, there are still lots of questions about the research on the mechanism of inflammatory responses. The most concern is the failure of clinical transformation application based on these findings about immune and inflammation, 8 which suggests that our understanding of the pathogenesis of AAA is still incomplete. Thus, except immune-related pathways, exploring the non-immune molecules involved in AAA formation could provide new insights into the therapeutic strategies of AAA.
In the present study, we reanalysed seven AAA datasets downloaded from the Gene Expression Omnibus (GEO) database ( Figure 1). We identified ATP1A2 as the key gene; its expression level was decreased in AAA samples compared with control samples, which was validated by immunostaining and Western blot in human AAA samples and an angiotensin II (Ang II)-induced mouse model of AAA. The ATP1A2 gene encodes α2-subunit of Na + /K + -ATPase, which is a transmembrane protein responsible for intracellular ion homeostasis critical for membrane potential and numerous cellular processes. 13,14 It has been reported that a 50% reduction in the α2subunit significantly enhances cardiac and vasculature contraction, resulting in an abnormal increase in blood pressure. 13 Although hypertension is one of the most important risk factors for AAA and ATP1A2 is involved in the regulation of blood pressure, whether ATP1A2 participates in the pathological processes of AAA remains unknown. Therefore, we next performed the single-cell RNA sequencing (scRNA-seq) data analysis and screened the transcription factors (TFs) that potentially regulate the expression of ATP1A2, to reveal the underlying mechanisms of ATP1A2 in AAA. Among the four isoforms of the α-subunit, ATP1A2 was significantly decreased in AAA, and its reduction mainly occurred in VSMCs of the aorta.
Furthermore, we found that the transcription factor ARID3A bound the promoter of ATP1A2 to suppress its expression. Thus, our present study identified the ARID3A-ATP1A2 axis as a novel pathway involved in AAA regulation, which not only provides new insights into the pathogenesis of AAA but also promotes the development of AAA diagnostic and treatment approaches.

| Human aortic samples acquisition
Full-thickness aortic samples were obtained from patients (

| AAA mouse model
The AAA mouse model was established with Ang II infusion as previously reported. 15,16 In brief, Alzet osmotic pumps (model 2004; 0000298, Durect Corporation) containing either Ang II (AAA group, 1 μg/kg/min, Sigma-Aldrich) or saline were implanted in 10-week-old ApoE −/− male mice (C57BL/6J background) for 28 days. The salineinjected mice were assigned to the Sham group. After sacrifice, aneurysmal (proximal to the renal arteries) or control segments of the aorta were harvested for further processing. All experimental protocols for mice in this study were approved by the Animal Experimental Ethics Committee of the Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology.

| Microarray datasets
Datasets search in GEO database (http://www.ncbi.nlm.nih.gov/geo/) using the following terms: 'Abdominal Aortic Aneurysm or AAA' retrieved 121 datasets, including 36 human sample datasets and 39 mouse sample datasets in Dec 2020. Then, we screened the datasets according to the organism of datasets ('human and mouse'), type of sequencing ('transcription profile'), type of sampling (aortic tissue) and time point of sampling (7 days after AAA in the mouse model). Finally, we downloaded six datasets from the GEO database, which were divided into four groups according to the platforms and species in the present study. The more detailed information about the datasets is shown in Table 1. [16][17][18][19][20][21]

| Data analysis
Data preprocessing, including background correction and quartile normalization, was performed by Limma package 22 for Groups 1&2 and marray package for Group 3&4 in R version 4.1.1. In particular, Group 1 contained three datasets with the same platform, which were merged after removing batch effects from nonbiological technical biases by applying ComBat algorithm of the SVA package. 23 Limma package was used to screen the differentially expressed genes (DEGs) between the AAA and control groups by setting significance cutoff criteria to adjusted p-value <0.05 corrected by Benjamini-Hochberg multiple test and fold changes (FC) >1.5. Volcano plots and heatmap plots were utilized to display the DEGs between AAA samples and controls by using some of the packages (ggplot2, pheatmap and circlize) provided by the bioconductor project.

| Functional enrichment
The clusterProfiler package was applied to carry out Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Set Enrichment Analysis (GSEA) analysis. 24

| Single-cell RNA sequencing data analysis
To locate and verify the ATP1A2 expression, we downloaded a scRNA-seq dataset (GSE16 4678) of CaCl 2 -induced AAA mouse model from GEO database. 25 The scRNA-seq data filtered with the

| Identification of transcription factors (TFs)
To explore whether transcription factors are involved in the dysregulation of ATP1A2, we used web databases to predict TFs for ATP1A2. We selected 'human' for the biosample category with other default options in The Signaling Pathways Project

| Correlation analysis
In order to identify the potential TFs that could regulate the ATP1A2 expression, we first screened for the common differentially expressed TFs targeting ATP1A2 in Groups 1&2, and then adopted

| Generation of reporter and expression vectors
The promoter region between −1002 and 0 of the human ATP1A2 gene was amplified by PCR and then inserted into the PGL3 reporter vector between the XhoI and HindIII sites. The full-length human ARID3A coding sequence was amplified and cloned into the pHAGE-CMV expression vector between the MluI and XhoI restriction sites. The

| Luciferase reporter assay
The ability of AR and ARID3A to regulate ATP1A2 expression was measured by dual-luciferase assay. Briefly, 3 × 10 4 HEK293 cells were seeded in 48-well plates, and after 12 hours, the empty vector or AR/ARID3A expression vectors (0.15 μg per well) were transiently cotransfected into cells with ATP1A2-PGL3 (0.05 μg per well) and TK (0.02 μg per well). ATP1A2-PGL3 was labelled with firefly luciferase, while TK was labelled with Renilla luciferase. Thirty-six hours later, cell lysates were harvested, and the firefly and Renilla luciferase activities were detected using a dual-luciferase reporter assay kit (Promega, E1910). The activity of ATP1A2 luciferase was normalized to that of the empty vector as a control group.

| Statistical analysis
Bioinformatics analysis was described above. The data from all experiments are described as the mean ±standard deviation (SD). Two-tailed Student's t-test was used to identify differences between two groups and one-way ANOVA followed by Tukey's post hoc analysis was applied to compare multiple groups. Statistical analyses were performed with R (version 4.1.1) software and GraphPad Prism (version 8.00) software. p-value < 0.05 was considered statistically significant.

| Comprehensive analysis of human datasets
After data preprocessing, the normalized data were presented in boxplots ( Figure S1A  common DEGs in AAA ( Figure 2F).

| Identification of ATP1A2 as the key gene involved in AAA
To further validate the bioinformatics analysis results of the human datasets, we explored the expression profile of mouse AAA datasets (GSE51227 & GSE17901). After data preprocessing (Figure S1C and  Figure 3A).
In the GSE17901 dataset, 68 DEGs (36 upregulated and 32 downregulated genes) were identified ( Figure 3B). Furthermore, we found 15 overlapping DEGs in these two datasets, which were shown in the Venn diagram and heatmap plot ( Figure 3C, D). In addition, we validated the expression of 16 common DEGs involved in five nonimmune pathways (shown in Figure 2F) in mouse AAA datasets, and found a decreased expression of ATP1A2 in AAA group ( Figure 3E).
Notably, ATP1A2 was the only common DEGs that was significantly downregulated in both human and mouse AAA samples after intersecting the DEGs of Groups 1-4 ( Figure 3F), which suggested the ATP1A2 as a conservative regulator involved in human and mouse aneurysms. As we known, ATP1A2 belongs to the subfamily of Na + /K + -ATPases which is composed of two subunits, alpha and beta. 31 Four well-known isoforms of the catalytic α-subunit (α1, α2, α3 and α4, which are encoded by ATP1A1, ATP1A2, ATP1A3 and ATP1A4) have been identified. 32 Given that the expression level of that α1 and α2 subunits encoded by ATP1A1 and ATP1A2 genes are mainly α-subunits of Na + /K + -ATPases presented in the aorta ( Figure 3G), which was consistent with the previous studies. 14,32,33 Although ATP1A4 was significantly elevated in human AAA samples, there was no significant difference between AAA and sham samples in mice, while significant expression level of ATP1A2 rather than ATP1A1 was observed between the control and AAA groups in both human and mice ( Figure 3G). Above results identified ATP1A2 as a key gene involved in the pathological condition of AAA, which reminded us whether ATP1A2 expression is cell-specifically decreased in AAA.

| Cellular localization of the ATP1A2 in AAA
In order to reveal the underlying mechanisms of ATP1A2 in AAA,  Figure 4B), which were consistent with previous study. 25 Next, we assessed the expression levels of ATP1A1, ATP1A2, ATP1A3 and ATP1A4 in each cell population after identifying the cell type of each cell cluster ( Figure 4C). Similar to the results above described, α-subunits of Na + /K + -ATPase presented in the aorta are mainly encoded by ATP1A1 and ATP1A2, in contrast, ATP1A3 and ATP1A4 are seldom expressed in aortic cells. In addition, unlike ATP1A1 which is ubiquitously expressed in a variety of cells, for example VSMCs, fibroblasts, macrophages, neutrophils,

T cells and ECs, ATP1A2 is specifically expressed in VSMCs and
fibroblasts ( Figure 4C). Furthermore, when we focused on the RNA expression levels of ATP1A1 and ATP1A2 in VSMCs and fibroblasts, we found that there was no significant difference between AAA and sham samples in fibroblasts, while ATP1A2 showed a decreased expression in VSMCs ( Figure 4D, E). In brief, these results suggested that among the four isoforms of the α-subunit, ATP1A2 was significantly decreased in AAA, and its reduction mainly occurred in VSMCs of the aorta, which required further experimental verification.

| Validation of the ATP1A2 expression in human and mouse aortic aneurysm
Since ATP1A1 and ATP1A2 are the two main isoforms expressed in the aorta, 14 we verified the expression and distribution of ATP1A1 and ATP1A2 in human and mouse specimens by using immunohistochemistry and immunofluorescence. HE and EVG staining showed that AAA was successfully induced by Ang II infusion in ApoE −/− mice ( Figure 5A). Compared with saline treatment, Ang II remarkably reduced the ATP1A2 protein level in the abdominal aorta of the mice, while the expression level of ATP1A1 was similar in these two groups ( Figure 5B, C). In addition, we observed extensive co-localization of ATP1A1/2 and α-SMA (a biomarker for VSMC) in aortic samples, demonstrating the predominant expression of ATP1A1 and ATP1A2 in VSMCs ( Figure 5D, E). Furthermore, we detected the protein levels of ATP1A1 and ATP1A2 in human samples. The results showed that ATP1A1 and ATP1A2 were highly expressed in VSMCs of the normal aorta ( Figure 5F, G), and the protein level of ATP1A2 was dramatically reduced in the aorta of patients with AAA ( Figure 5G-I), which confirmed the above analysis results.
Next, we were curious whether the expression patterns of ATP1A1 and ATP1A2 were limited to AAA or conserved in TAA and

AD. Thus, IHC was used to detect the expression levels of ATP1A1
and ATP1A2 in the aortas of patients with or without TAA/AD. The results demonstrated that ATP1A2 was also downregulated in the aortas of patients with TAA, while there was no significant difference in ATP1A1 between the two groups ( Figure S3A and S3B). The same expression pattern (i.e. significant reduction in ATP1A2 expression) was also observed in AD samples ( Figure S3C and S3D). Moreover, the Western blot results further confirmed the decrease in ATP1A2 expression in AAA and TAA samples ( Figure S3E). Thus, these results indicated that ATP1A2 may be a key regulator of both AAA and TAA.

| Identification of potential TFs regulating ATP1A2 expression
The GEO data and our own verification data both show that the expression level of ATP1A2 is significantly reduced in aneurysms.
Therefore, it is very interesting to reveal the molecular mechanism that regulates the reduction in ATP1A2 expression. TFs are one of the most important molecules that regulate gene expression by directly binding to the promoters of target genes. 34 Thus, to identify the TFs that regulate the expression of ATP1A2, the Signaling Pathways Project database is used to predict the TFs that bind to the promoter of ATP1A2. 28

| ARID3A negatively regulates ATP1A2 expression in VSMCs
Based on the correlation analysis of ATP1A2 and candidate upstream TFs, AR and ARID3A were regarded as potential TFs that regulate ATP1A2. Then, we cotransfected HEK293T cells overexpressing AR or ARID3A plus an ATP1A2 luciferase reporter vector. Given that AR requires activation by androgen to play a role in transcriptional regulation, 35 we adopted different concentrations of androgen analogs (testosterone propionate, TP) to activate AR. The results showed that regardless of TP stimulation, overexpression of AR did not significantly increase the expression of ATP1A2 ( Figure 7A), while overexpression of ARID3A resulted in an approximately 50% reduction in ATP1A2 ( Figure 7B). Furthermore, ARID3A was overexpressed in primary cultured human aorta smooth muscle cells (HASMCs) through lentivirus, and ATP1A2, but not ATP1A1 was obviously suppressed by ARID3A overexpression (Figure 7C, D). These results indicate that ARID3A is the TF that binds to the promoter of ATP1A2 and inhibits its expression. Furthermore, scRNA-seq data analysis implies that the increased ARID3A in AAA might be derived from macrophages ( Figure S4). integrity due to its roles in the regulation of blood pressure and ion homeostasis.
As we currently known, ATP1A2 gene encodes an α2-subunit of Na + /K + -ATPase which is firstly discovered as the molecular machine for pumping Na + and K + across cell membrane. 36 Na + / K + -ATPase is composed of two essential subunits, alpha (the catalytic subunit of the enzyme) and beta. There are four isoforms of the α-subunit of Na + /K + -ATPase such as α1, α2, α3 and α4. 13,31 Among them, the α1 and α2 subunits encoded by ATP1A1 and ATP1A2 genes are present in the aorta (70% of all Na + / K + -ATPases in mice contain the α1-subunit, and the remaining 30% contain the α2-subunit), 14,32,33 which is similar to our results.
Although α2-subunit exhibits low expression in the cardiovascular system compared with the α1-subunit, previous studies have confirmed its regulation for cardiovascular function and blood pressure by applying a global genetic knockout of α2-subunit mice. 37,38 Currently, no study has revealed the relationship between Na + / K + -ATPase and AAA, but some studies have found its roles in regulating the contraction of VSMCs, thereby leading to dynamic changes in blood pressure. 14 Currently, the regulatory effects of Na + /K + -ATPase on VSMCs are mainly manifested in the following three aspects: 1. Na + /K + -ATPase maintains intracellular ion homeostasis.
It is well known that Na + /K + -ATPase maintains high cytosolic K + and low cytosolic Na+concentrations in animal cells by transporting two K + ions into the cell and three Na + ions out of the cell with the hydrolysis of an ATP molecule, 40  implies that the α2-isoform is less active at resting conditions but may be activated during abnormal agonist stimulation. 32,43 2. Na+/K+-ATPase is involved in VSMC apoptosis and proliferation.
It has been reported that inhibition of Na + /K + -ATPase suppresses apoptosis of VSMCs in various unfavourable conditions. For example, Na + /K + -ATPase inhibition by ouabain (a special type of Na + /K + -ATPase inhibitor) sharply suppressed apoptosis in serum-deprived VSMCs. 44 Knocking out ATP1A2 suppressed the rate of apoptosis following PM2.5 exposure. 45 The apoptosis of VSMCs and cardiomyocytes transfected with ATP1A2 siRNA was alleviated in anoxia/reoxygenation injury. 46 Notably, in addition to the apoptosis, some studies also demonstrated that Na + /K + -ATPase was involved in VSMC proliferation. A novel butyrolactone derivative could inhibit migration and proliferation of VSMCs through the inhibition effects on the activity of Na + /K + -ATPase. 47 Furthermore, low concentrations of ouabain reflecting the affinity to the α1-isoform can activate proliferation via Src/EGFR and ERK1/2 activation in the synthetic phenotype of VSMCs. 48 The Ang II, a molecule used to induce AAA mouse model, could stimulate Na + /K + -ATPase activation by upregulating ATP1A1/2 gene transcription via PI3K-p42/44 signal pathways, which was responsible for Ang II-induced VSMCs proliferation. 49 The above findings confirmed the crucial roles of Na + /K + -ATPase in determining the fate of VSMCs during the pathological conditions. Data are presented as the mean ± SD. **p < 0.01; ns indicates no significance vs. controls, Student's t-test.
The Na + /K + -ATPase (specifically the α2-isoform) plays an important role in modulating VSMC tone and blood pressure. 50 It has been elucidated that Na + /K + -ATPase activation in VSMCs leads to membrane hyperpolarization which consequently reduces the levels of intracellular Ca 2+ , thereby contributing to vasodilation. 14

| CON CLUS IONS
In conclusion, in the present study, we reanalysed seven AAA datasets to examine the global gene expression profile of AAA and identified ATP1A2 as a key gene involved in the progression of AAA. Furthermore, we screened TFs that potentially regulate ATP1A2, which showed that ARID3A negatively regulated ATP1A2. Considering our findings, we believe that ATP1A2 and/ or ARID3A play important roles in AAA initiation and progression, which broadens our understanding of the pathogenesis of AAA, and provides valuable insights for the future AAA therapeutic research.

ACK N OWLED G EM ENT
This work was supported by grants from the National Natural Science Foundation of China (NO. 82070488, NO. 82000440).

CO N FLI C T S O F I NTE R E S T S
The authors declare that they have no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated and analysed during the present study are included in the article/supplementary material or GEO database (http://www.ncbi.nlm.nih.gov/geo/).