Biomarkers of Blood from Patients with Atherosclerosis Based on Bioinformatics Analysis

Abstract

Atherosclerosis is a multifaceted disease characterized by the formation and accumulation of plaques that attach to arteries and cause cardiovascular disease and vascular embolism. A range of diagnostic techniques, including selective coronary angiography, stress tests, computerized tomography, and nuclear scans, assess cardiovascular disease risk and treatment targets. However, there is currently no simple blood biochemical index or biological target for the diagnosis of atherosclerosis. Therefore, it is of interest to find a biochemical blood marker for atherosclerosis. Three datasets from the Gene Expression Omnibus (GEO) database were analyzed to obtain differentially expressed genes (DEG) and the results were integrated using the Robustrankaggreg algorithm. The genes considered more critical by the Robustrankaggreg algorithm were put into their own data set and the data set system with cell classification information for verification. Twenty-one possible genes were screened out. Interestingly, we found a good correlation between RPS4Y1, EIF1AY, and XIST. In addition, we know the general expression of these genes in different cell types and whole blood cells. In this study, we identified BTNL8 and BLNK as having good clinical significance. These results will contribute to the analysis of the underlying genes involved in the progression of atherosclerosis and provide insights for the discovery of new diagnostic and evaluation methods.

Background

Atherosclerosis (AS) is the leading cause of peripheral vascular disease, coronary heart disease, and cerebral infarction.¹ The development of atherosclerotic lesions may be caused by low-density lipoprotein, a lipoprotein that carries cholesterol into peripheral tissue cells and can be oxidized to become oxidized low-density lipoprotein. Other risk factors contribute to atherosclerosis and its thrombotic complications, including diabetes, smoking, and high blood pressure.² Growing evidence also indicates a role for emerging risk factors, including clonal hematopoiesis and inflammation. A range of auxiliary examination methods, both invasive (such as selective coronary angiography) and non-invasive (such as nuclear scans, CT, stress tests, and blood biomarkers), allow assessment of cardiovascular disease risk and treatment targets. However, there is no simple blood biochemical index or biological target for the diagnosis of atherosclerosis at present; instead, more ultrasonographic screening or angiography are used.² Therefore, it would be valuable to identify a biochemical blood marker for atherosclerosis.

With the development of omics and the availability of clinical blood samples, many studies have focused on the blood transcriptome of patients with atherosclerosis. Transcriptome analysis of blood cells, divided into those of atherosclerotic patients and matched controls, will potentially supply biomarkers for diagnostic purposes and provide insights into the mechanism of atherosclerosis.^3-6 One study focused on differences in various cells in the blood of patients with AS to explore the biological functions of macrophages and CD34 cells⁷; other studies have examined the transcriptome of peripheral blood and the transcriptional expression of circulating cells in patients with acute myocardial infarction or artery plaque.⁸ Meanwhile, with the development of high-throughput sequencing and bioinformatics analysis techniques, a bioinformatics gene analysis related to the increased risk of atherosclerosis due to familial hypercholesterolemia provides a basis for the development of therapies for atherosclerosis.⁹ In addition, bioinformatics analysis of oncology,^10,11 endocrine diseases,¹² and respiratory diseases¹³ drives basic research and provides directions for treating patients.

In this study, after the detection of differential expression genes in multiple data sets, the Robust rank aggregation algorithm was used for integration evaluation, and 21 possible genes were screened out as potential biomarkers for biological diagnostic screening. We looked at the expression of these genes in different circulating cells. Interestingly, we found a good correlation between RPS4Y1, EIF1AY, and XIST.

Methods

Retrieve

Keywords “atherosclerosis” and “blood” were searched in the GEO database and the species was limited to “Homo sapiens.” Fifty-nine data sets were retrieved, and then we manually excluded the mRNA chip data sets unrelated to the blood of atherosclerosis patients and not clearly grouped, and finally 3 data sets were screened out (Table 1).

Table 1.

Information of data sets.

GEO	Platform	Normal	Patient	Reference number
GSE27034	GPL570	18	19	Masud et al5
GSE90074	GPL6480	50	93	Ravi et al6
GSE12288	GPL96	112	110	Sinnaeve et al7

Differentially expressed genes (DEGs) analysis, network analysis, and functional enrichment analysis. We assessed the primary data using R language assessment and quality control, all the expression of matrix through log2 processing, and use ggpubr package for draw violin plots (Supplemental Figure S1). The samples were divided into case and control groups according to the information on GEO. The LIMMA package was used to analyze the differential genes¹⁴ (Supplemental Tables S1-S3 and Figure S2). We screened genes with P-value less than .01 and |LogFc| > 0.5. We drew the network diagram based on the strings database and analyzed the path of the network diagram (Supplemental Figure S3).

Robustrankaggreg

Robustrankaggreg R package was used to integrate the up-down-regulated genes,¹⁵ respectively. RRA is a rank aggregation method based on sequential statistics, which can achieve the purpose of removing the noise of individual experimental results while increasing the signal and reducing the proportion of false-positive results in high-throughput data integration. There were n rank vectors and normalized the rank vectors were sorted from small to large. When the $l$ rank vector was greater than or equal to the mean of all rank vectors, the row ${\displaystyle \sum }_{k=l}^n\left(\begin{array}{c}n\\ k\end{array}\right)x^k{(1-x)}^{n-k}$ of all rank vectors was calculated, and the minimum value was taken as the score. Genes with a score less than 0.05 were screened out as the marker genes we considered, and a heat map of logFC in different datasets was drawn (Figure 1).

Figure 1.

LogFC of Genes were identified in 3 datasets, red represents high values, and blue represents low values.

Genetic alignment and correlation analysis

The expression matrices of the identified genes were selected from the original data set and GSE9820,³ and the unclustered and clustered heat maps were constructed with pheatmap function (Figures 2–6). Correlation analysis was performed for all identified markers (Supplemental Tables S4-S6), and regression analysis was performed for the most interesting genes, and scatter plots and residual plots were plotted (Figures 3–5 and Supplemental Figure S4). The genes of interest were plotted in a scatter plot. P value<.05.

Figure 3.

Clustered heat map of gene expression in GSE90074 (A) scatter diagram and regression line of XIST and EIF1AY, regression equation: y = −1.32619x − 0.61110, residual standard error: 1.075 on 141 degree of freedom, n:143, multiple R-squared: 0.9411, adjusted R-squared: 0.9407, F-statistic: 2253 on 1 and 141 DF, P-value: <2.2e⁻¹⁶ (B) scatter diagram and regression line of XIST and RPS4Y1, regression equation: y = −1.19600x − 2.91454, residual standard error: 1.009 on 141° of freedom, n:143, multiple R-squared: 0.9482, adjusted R-squared: 0.9478, F-statistic: 2579 on 1 and 141 degree of freedom, P-value: <2.2e⁻¹⁶ (C).

Figure 4.

Clustered heat map of gene expression in GSE12288 (A) scatter diagram and regression line of XIST and RPS4Y1, regression equation: y = −0.67326x + 12.23007, residual standard error: 0.9911 on 220 degree of freedom, N:222, multiple R-squared: 0.7398, adjusted R-squared: 0.7386, F-statistic: 625.5 on 1 and 220 DF, P-value: <2.2e⁻¹⁶ (B) scatter diagram and regression line of XIST and EIF1AY, regression equation: y = −1.01055x + 12.24370, residual standard error: 1.239 on 220 degree of freedom, N:222, multiple R-squared: 0.5933, adjusted R-squared: 0.5914, F-statistic: 320.9 on 1 and 220 DF, P-value: <2.2e⁻¹⁶ (C).

Figure 5.

Clustered heat map of gene expression in GSE27037 (A) scatter diagram and regression line of XIST and EIF1AY, regression equation: y = −0.61712x + 0.14227, residual standard error: 0.5456 on 35 degree of freedom n:37, multiple R-squared: 0.9219, adjusted R-squared: 0.9197, F-statistic: 413.2 on 1 and 35 DF, P-value: <2.2e⁻¹⁶ (B) scatter diagram and regression line of XIST and RPS4Y1, regression equation: y = −0.51754x − 0.02932, residual standard error: 0.4404 on 35 degree of freedom, n:37, multiple R-squared: 0.9491, adjusted R-squared: 0.9477, F-statistic: 652.8 on 1 and 35 DF, P-value: <2.2e⁻¹⁶ (C).

Figure 6.

Gene expression in GSE9820. Unclustered heat map of gene expression in GSE9820 (A) clustered heat map of gene expression in GSE9820 (B).

Figure 2.

Gene expression in 3 datasets. Red represents high expression, blue represents low expression, each column represents a sample, and each row represents a gene. Unclustered heat map of gene expression in GSE27034 (A) unclustered heat map of gene expression in GSE90074 (B) unclustered heat map of gene expression in GSE12288 (C).

Results

Genes detected according to the integrated DEGs

Deg analysis was performed on all data sets, and specific DEG results can be seen in the Supplemental Data. Only 1 pathway “Cytokine Signaling in Immune system” was enriched after network analysis and pathway analysis of the genes considered significant. We still obtained 21 genes based on RRA algorithm integration with good scores, including up-regulated genes: BTNL8, GPR15, STX11, DDX3Y(DBY), TMEM158. G0S2, PS4Y1 (RPS4Y), ZNF80, PTGS2, EIF1AY (IF1AY), and FFAR2. Among them, BTNL8, GPR15, STX11, and TMEM158 have relatively high logFC in multiple data sets, while DDX3Y(DBY), G0S2, PS4Y1(RPS4Y), PTGS2, EIF1AY(IF1AY), and FFAR2 have relatively high logFC in a single data set. The down-regulated genes included BLNK, XIST, PSPH, LOC10272435, SCGB3A1, AKR1C3, KLRC1, EFHB, KIZ, and FCRL2, among them BLNK showed significant differences in multiple data sets, while XIST showed a considerable difference in GSE90074. These genes may be used for screening and evaluation of AS or vascular plaques.

The correlation between RPS4Y1, XIST, and EIF1AY

Because the logFC value is low, the difference between the case and control groups is not visible to the naked eye. However, after clustering the heat maps, we found an interesting phenomenon for the first time: XIST is negatively correlated with RPS4Y1 in all 3 data sets, and XIST is negatively correlated with EIF1AY. The sample expressing XIST, RPS4Y1, and EIF1AY are basically not expressed, and vice versa. This mechanism may also be involved in atherosclerosis.

Validation in different cell types

We picked up the expression of these selected genes in the data set of GSE9820,³ which is a sequencing data of Mononuclear Cell Transcriptomes, and identified 5 kinds of cells, including CD34⁺ stem cells, CD4⁺ T-cells, resting CD14⁺ monocytes, stimulated monocytes, and macrophages. It can be seen that the expression level of BTNL8 is relatively low in these 5 kinds of cells, while it is still relatively high in other data sets, so it should be highly expressed in a cell that does not belong to these 5 kinds of cells. RPS4Y1 and EIF1AY were not tissued specific, but individual specific. GPR15 and ZNF80 were highly expressed in T cells, G0S2, PTGS2, and FFAR2 were highly expressed in stimulated monocytes, and stem cells mainly highly expressed BLNK, AkR1C3, and FCRL2. Good consistency between RPS4Y1 and EIF1AY can also be seen in the cluster diagram of GSE9820.

Discussions

This study combines 3 coronary atherosclerosis in patients with blood samples mRNA array dataset to filter possible coronary atherosclerosis possible genetic detection objects in the blood. We found there are 21 genes that may have specific significance and also discussed these gene expressions between different cells in the blood. This study first reported RPS4Y1, EIF1AY own the correlation between XIST.

Many of these genes are associated with inflammation and immunity. BTNL8, which has the best score, may stimulate the primary immune response acts on T-cell stimulated sub-optimally through the TCR/CD3 complex stimulating their proliferation and cytokine production.¹⁶ G0S2, G0/G1 switch protein 2, promotes apoptosis by binding to BCL2, resulting in preventing the formation of protective Bcl2-Bax heterodimers.¹⁷ GPR15L is a chemotactic factor that mediates recruitment of lymphocytes to epithelia through binding and activation of the G-protein coupled receptor GPR15 seems to be epithelia related.¹⁸ BLNK, B-cell linker protein, functions as a central linker protein downstream of the B-cell receptor (BCR), bridging the SYK kinase to a multitude of signaling pathways, and regulating biological outcomes of B-cell function and development.¹⁹ What is more, BLNK plays a role in the activation of ERK/EPHB2, MAP kinase p38, and JNK. Modulates AP1, BCR-mediated PLCG1, Ca²⁺ mobilization, PLCG2, NF-kappa-B, and NFAT. It plays a critical role in orchestrating the pro-B cell to pre-B cell transition²⁰ and may play an essential role in BCR-induced B-cell apoptosis. These differentially expressed genes between patients and normal controls can explain, to some extent, the genetic susceptibility of patients and the body’s response to AS.

XIST is a key initiator of X chromosome inactivation in Eutherian mammals, which may also be part of the inflammatory response.²¹ EIF1AY, Eukaryotic translation initiation factor 1A, seems to be required for the maximal rate of protein biosynthesis. Enhances ribosome dissociation into subunits and stabilizes the binding of the initiator Met-tRNA(I) to 40S ribosomal subunits.²² RPS4Y1, the ribosomal protein S4 40S ribosomal protein S4, Y isoform 1, is was extensively involved in RNA binding, multicellular organism development, nuclear-transcribed mRNA catabolic process, nonsense-mediated decay, SRP-dependent cotranslational, protein targeting to membrane, translation, translational initiation, and viral transcription. These genes are involved in the more basic biological functions of replication, translation, transcription, and they are identified by the DEG algorithm.²³ The basic blood metabolism of AS patients has certain differences, which may be correlated with risk factor clonal hematopoiesis.

Conclusion

These mRNA molecules are still lacking clinical cohort verification, and their use as a marker of screening is still to be debated. However, the differences between normal population and AS patients to some extent can explain their correlation with AS, indicating that repeated activation of inflammation is involved in the formation and development of AS. The specific roles of XIST, RPS4Y1, and EIF1AY in transcription and translation and how they are related need to be verified by molecular biology, which will be of great help for us to understand the central principle further. In general, we have only scratched the surface, which provides some targets for subsequent cohort studies. Through bioinformatics analysis, our results may be beneficial for the clinical molecular diagnosis,^24,25 treatment,²⁶ and prognosis.²⁷ The associations we have found may also be helpful for more fundamental studies of biological function.²⁸