Transcriptome alterations of vascular smooth muscle cells in aortic wall of myocardial infarction patients

This article contains further data and information from our published manuscript [1]. We aim to identify significant transcriptome alterations of vascular smooth muscle cells (VSMCs) in the aortic wall of myocardial infarction (MI) patients. Microarray gene analysis was applied to evaluate VSMCs of MI and non-MI patients. Prediction Analysis of Microarray (PAM) identified genes that significantly discriminated the two groups of samples. Incorporation of gene ontology (GO) identified a VSMCs-associated classifier that discriminated between the two groups of samples. Mass spectrometry-based iTRAQ analysis revealed proteins significantly differentiating these two groups of samples. Ingenuity Pathway Analysis (IPA) revealed top pathways associated with hypoxia signaling in cardiovascular system. Enrichment analysis of these proteins suggested an activated pathway, and an integrated transcriptome-proteome pathway analysis revealed that it is the most implicated pathway. The intersection of the top candidate molecules from the transcriptome and proteome highlighted overexpression.

discriminated between the two groups of samples. Mass spectrometry-based iTRAQ analysis revealed proteins significantly differentiating these two groups of samples. Ingenuity Pathway Analysis (IPA) revealed top pathways associated with hypoxia signaling in cardiovascular system. Enrichment analysis of these proteins suggested an activated pathway, and an integrated transcriptomeproteome pathway analysis revealed that it is the most implicated pathway. The intersection of the top candidate molecules from the transcriptome and proteome highlighted overexpression.
& 2018 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

Value of the data
Combination of multiple technologies and bioinformatics analysis performed in this study reveals the molecular changes induced by myocardial infarction on aortic smooth cells in humans.
The alterations of the VSMCs transcriptome are congruent with alterations at the protein levels.
Both levels show notably the up-regulation of the superoxide dismutase (SOD) with the activation of superoxide radical degradation pathway.
Differentially expressed genes and pathways identified in these comparisons may be used in future experiments investigating response in myocardial infarction.

Clinical analysis
The characteristics of the myocardial infarction (MI) and non-MI samples undergoing transcriptomics and proteomics studies are presented in Tables 1(A) and 1(B) respectively. The baseline demographic and clinical characteristics of samples undergoing transcriptomics study were  compared with those of the samples from the proteomics study (Table 2). In addition, the characteristics of the transcriptomic MI and non-MI samples with those of the independent cohorts comprising additional MI and non-MI patients undergoing RT-qPCR were compared (Tables 3 and 4).

Gene expression data analysis and class prediction by Prediction Analysis of Microarray (PAM)
The samples were preprocessed through several steps, including quality assessment and outlier identification, normalization, batch effect correction and evaluation ( Fig. 1). To interrogate differentially expressed genes between MI and non-MI we conducted gene-expression profiling using the Affymetrix U219 microarray platform. The R 'limma' package (https://www.bioconductor.org/help/ workflows/arrays/) identified 4,357 probe sets, selected at a 'limma'-defined p-value o 0.05. Based on this set of differentially expressed genes (DEGs), we performed principal component analysis (PCA) (Fig. 1).
To determine subgroup of genes distinguishing MI from non-MI subjects, we performed supervised PAM [2] and identified a set of differentially expressed genes (DEGs) that discriminated between the two subtypes at Wilcox FDR o 0.1 (Table 5).
Gene Ontology (GO) analysis of the DEGs was performed using DAVID Bioinformatics tools [3] (http://david.abcc.ncifcrf.gov/). The GO results for the down-regulated transcripts were not enriched for any GO terms. The GO analysis revealed biological processes ( Table 6).
Clustering of genes were done by two methods, hierarchical and k-mean clustering. Hierarchical clustering with multiscale bootstrap resampling was done by Pvclust, an R statistical software package [4]. The Pvclust is an R package for assessing the uncertainty in hierarchical cluster analysis. For each cluster in hierarchical clustering, quantities called p-values are calculated via multiscale bootstrap resampling. The parameters (https://cran.r-project.org/web/packages/pvclust/pvclust.pdf) used here were 10000 bootstrap replications, cluster method: Ward algorithm and distance method: Euclidean. For the heat maps plot, we used log2 scale. The k-mean clustering was performed by R (https://stat.ethz.ch/R-manual/R-devel/library/stats/ html/kmeans.html), showing that the selection of features gave a higher accuracy than PAM alone. The genes were discriminated between the MI and non-MI vascular smooth muscle cells (VSMCs) samples (Table 7). A clustered result is shown in Fig. 2 of Ref. [1].
1.3. Protein processing, electrostatic repulsion-hydrophilic interaction chromatography (ERLIC) and LC-MS/MS analysis using Q-Exactive mass spectrometer Differential expressed proteins identified are shown in Table 8. Only peptides identified with strict spectral false discovery rate of less than 1% (q-value r 0.01) were considered.

Hierarchical cluster analysis of RT-qPCR-based detected genes
Using six RT-qPCR-supported genes as a representative gene classifier characterizing the differences between MI and non-MI aortic samples, hierarchical clustering was performed with multiscale bootstrap resampling by Pvclust. The result is shown in Fig. 2.

Transcriptomic and proteomic pathways analysis
Systemic evaluation was performed using IPA (www.ingenuity.com) to identify transcriptomic and proteomic pathways, and significantly enriched canonical pathways are shown in Table 9. An integrated transcriptome-proteome correlation is performed to identify common enriched pathway and molecule (Table 10).

Sample collection
Aortic tissue samples were obtained from patients who presented with coronary artery disease undergoing coronary artery bypass graft (CABG) surgery at the National University Hospital of Singapore from 2009 to 2013. Patients underwent CABG either after a recent myocardial infarction (MI group) or as stable angina patients (non-MI group). An aortic punch tissue was collected at the time of proximal anastomosis between the aorta and saphenous vein grafts. The tissues from the aortic punch were immediately preserved on dry ice, and stored in liquid nitrogen tank. The study was approved by the National Healthcare Group Domain Specific Review Board (Tissue Bank registration: NUH/ 2009-0073), and written informed consent was obtained from all patients. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki.

Sample grouping
17 MI and 19 non-MI samples were recruited for laser capture microdissection (LCM) and microarray profiling. The proteomic study included 25 MI and 25 non-MI samples. Four MI and six non-MI samples overlapped between the microarray and proteomic studies. RT-qPCR was done on an independent cohort of samples, including an additional 20 MI and 20 non-MI samples. A schematic of the design and workflow is presented in Fig. 3. Table 4 Demographic characteristics of non-MI study group and non-MI validation group.

Characteristics
Non

Sample processing
The protocols for (1) cryosectioning and staining of aortic tissue, (2) LCM of VSMCs, total RNA isolation and complementary DNA (cDNA) synthesis, and (3) protein processing, ERLIC and LC-MS/MS analysis using Q-Exactive mass spectrometer are described in our manuscript [1].