The following antibodies were used in the current study: mouse monoclonal antibodies: CD31 (Abcam), Mac-2 (Santacruz, Germany), α-SMA (Boster, Wuhan, China); rabbit polyclonal antibodies: FKBP11 (Bioss, Beijing, China), p65, p-p65 (Cusabio, Wuhan, China), and MMP-9 (Santacruz, Heidelberg, Germany); VCAM-1, ICAM-1, MCP-1 (Wanleibio, Shenyang, China); β-actin (New England Biolabs, Germany); IL1-β (Sigma-Aldrich, Taufkirchen, Germany). The following HRP conjugated secondary antibodies were used for immunoblotting: rabbit IgG (New England Biolabs, Germany), mouse IgG (Abcam). The following secondary antibodies for immunofluorescence were used: Texas red conjugated anti-rabbit and FITC conjugated anti-mouse (Vector Laboratories, CA, USA).

The following reagents used in the current study were: Ang II (Sigma-Aldrich, St. Louis, MO), Trypsin-EDTA and HEPES (Gibco, Germany), fetal bovine serum (Sigma-Aldrich, St. Louis, MO), Protein A-Agarose Immunoprecipitation Reagent (Santacruz, Germany); phosphatase inhibitor cocktail (Roche Applied Science, Penzberg, Germany) and protease inhibitor cocktail (Roche diagnostics GmbH, Mannheim, Germany); BCA reagent (Perbio Science, Bonn, Germany); Vectashield mounting medium with DAPI (Vector Laboratories, CA, USA); PVDF membrane and immobilion enhanced chemiluminescence reagent (Millipore GmbH, Germany).

All studies were conducted under protocols approved by the Ethics Committee of Tongji medical college, Huazhong University of Science and Technology (Wuhan, China). Written informed consent was obtained from the patients or their relatives in accordance with the Declaration of Helsinki. Six consecutive patients (3 males and 3 females, mean age 48.0 ± 9.8 years) underwent emergency surgery for Stanford type A AAD were recruited. Acute aortic dissection is defined as dissection detected within 2 weeks of the onset of symptoms (Steuer et al., 2013). Patients with Marfan syndrome, Ehlers-Danlos syndrome, and other known connective tissue disorders were excluded. None of the patients had bicuspid aortic valve, aortic coarctation, Takayasu's arteritis, coronary artery disease, or cocaine use. Dissected ascending aorta specimens were obtained through operation, during which the tear site was identified and a strip of aortic wall including the tear site was carefully excised. Any adherent mural thrombus was removed from the aortic wall. Each sample was divided into three parts. One part was used for histological assessment, one part for RNA isolation and the rest for protein extraction. As control samples, normal ascending aorta specimens were collected from 6 organ donors (mean age 42.6 ± 8.3 years). Normal control samples were treated in the exactly same manner as the AAD samples (Pan et al., 2014). All subjects were of Asian origin. Detailed clinical information of the individuals enrolled in the study is shown in Table 1. There are no statistically significant differences of the demographics between the two groups.

According to the microarray results, the top 10 hub genes were chosen for further validation by qRT-PCR in AAD patients vs. healthy controls. The qRT-PCR was performed as described elsewhere (Liu J. et al., 2017). Aorta tissue samples were collected and pulverized under liquid nitrogen, total RNA was extracted with Trizol reagent (TaKaRa, Japan). Then, 1 μg RNA was added to a 20 μL reaction volume for cDNA reverse transcription using the Prime Script™ RT reagent Kit with gDNA Eraser (TaKaRa, Japan), and quantitative real-time polymerase chain reaction (qRT-PCR) was performed using the SYBR® Premix Ex Taq™ (TaKaRa, Japan) in a Step One Plus real-time PCR system (Applied Biosystems, USA). Gene expression was quantified according to the 2^−ΔΔCt method. Primer sets for selected genes were designed by TianYi Huiyuan (Wuhan, China) and their sequences are available in Table 2. The expression data were normalized to the reference 18s rRNA.

The microarray data sets GSE 52093 was obtained from the NCBI Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo) database, which was performed in seven human type A AAD samples and five control samples, and the platforms was Illumina HumanHT-12 V4.0 expression beadchip (Illumina Inc., San Diego, CA, USA) (Pan et al., 2014). For multiple probes corresponding to one gene, there average expression value was taken as the gene expression value. After that, gene expression values were normalized using preprocessCore package (version 1.28.0, http://www.bioconductor.org/package/release/bioc/html/preprocessCore.html) (Bolstad et al., 2003), and were performed with log2 transformation. Only probes that were annotated to an Entrez gene ID were kept for further analysis. Expression profiles of probes mapping to the same Entrez gene ID were averaged.

Linear models for microarray data (Limma) package (version1.22.0, http://www.bioconductor.org/packages/release/bioc/html/limma.html) in R was utilized to identify the Differentially-Expressed Genes (DEGs) between type A AD and control samples. Genes with fold change >2 and false discovery rate (FDR) < 0.05 were selected for subsequent analysis (Smyth, 2004).

Functional enrichment of GO and KEGG pathways analyses were performed with the DAVID Bioinformatics Tool (https://david.ncifcrf.gov/, version 6.8) (Huang da et al., 2009a,b). P-value < 0.05 was considered to be significant enrichment. Functional enrichment analysis of the network module genes was also conducted according to the above methods.

The hub genes of modules, described as the most closely associated with disease, often have more biological significance compared with the hub genes of global networks (Goh et al., 2007). A gene can be considered as a hub gene if it has a unique character, such as high gene significance (GS), high module membership (MM), and high intramodular connectivity (IC) in the network (Zhang and Horvath, 2005). GS showed the different IC and MM described the significance of genes in various networks. We therefore identified the hub genes in modules through the GS and MM.

PPI assessment was employed to infer the interactions among proteins in the selected WGCNA modules with STRING database (https://string-db.org/) (Szklarczyk et al., 2017).

THP-1 cells, a pro-monocytic cell line (ATCC® TIB-202™), were cultured in RPMI 1640 (Gibco, CA,USA) supplemented with 10% fetal bovine serum (Sigma-Aldrich, St. Louis, MO), 10 mM HEPEs, 0.1 mM MEM non-essential amino acids, 1 mM sodium pyruvate, and 100 nM penicillin/streptomycin (Life Technologies), were maintained at 5% CO₂, 37°C. EA.hy926 cells were obtained from ATCC (Manassas, VA) and cultured in RPMI 1640 medium (Gibco, CA, USA) with 10% fetal bovine serum (Sigma-Aldrich, St. Louis, MO) in a humidified atmosphere of 5% CO₂. Primary human vascular endothelial cells (HUVECs, Cat No.: CP-H082, Procell Life Science&Technology Co,.Ltd., Wuhan, China)were cultured in endothelial culture medium (Cat No.: CM-H084, Procell Life Science&Technology Co,.Ltd., Wuhan, China). Specific siRNAs targeting FKBP11 purchased from Ruibo (Guangzhou, China), and transfected into cells with TurboFect Transfection Reagent (Thermo Fisher Scientific, Waltham, MA USA) according to the manufacturer's protocol. The sequence of FKBP11-siRNA was listed as following: FKBP11-Si1, GGGCAATCATTCCTTCTCA; FKBP11-Si2, GAGAAGCGAAGGGCAATCA; FKBP11-Si3, CTCACTTGGCCTATGGAAA. The scrambled siRNA controls were used as a negative control.

Immunofluorescence analyses of immortalized human endothelial cell line EA.hy926 cells were done on collagen-coated coverslips. EA.hy926 cells were fixed in 3.7% paraformaldehyde (PFA) and washed three times with PBS containing 0.2% Trition X-100. Coverslips were blocked with blocking reagent for 1 h and incubated overnight at 4°C with a primary antibody against p-p65. Coverslips were washed three times with PBS and corresponding fluorescently labeled secondary antibodies were added for 60 min. Coverslips were rinsed thrice in PBS and placed on vectashield mounting medium containing nuclear stain DAPI. Specimens were analyzed on a Leica SP5 confocal microscope. Nuclear localization of p-p65 was analyzed using an ImageJ-Fiji (http://imagej.nih.gov) macro including automated detection of nuclei in the DAPI channel and comparison of p-p65 levels within the nucleus vs. within a band mask surrounding the nucleus. DAPI stained and fluorescently labeled images were acquired individually (Madhusudhan et al., 2015).

Proteins were collected from the aortic tissues or cultured endothelial cells using RIPA lysis buffer. Protein concentration was quantified using BCA reagent. Equal amounts of protein were electrophoretically separated on 10% SDS polyacrylamide gel, transferred to PVDF membranes and probed with desired primary antibodies overnight at 4°C. Membranes were then washed with TBST and incubated with anti-mouse IgG (1:5,000) or anti-rabbit IgG (1:2,000) horseradish peroxidase conjugated antibodies as indicated. Blots were developed with the enhanced chemiluminescence system. To compare and quantify levels of proteins, the density of each band was measured using Image J software. Equal loading for total cell or tissue lysates was determined by β-actin western blot (Dong et al., 2015; Madhusudhan et al., 2015).

Immunoprecipitation of total cellular proteins from endothelial cells was done as described elsewhere (Shahzad et al., 2016) (Dong et al., 2015). In brief, total cellular proteins were extracted with protein lysate containing phosphatase inhibitor and complete protease inhibitor cocktail. Leave part of the centrifuged protein supernatant for input. Lysates were combined with 3 μg of specific antibody and incubated for 2 h at 4°C. Immunoprecipitates were purified with protein A/G agarose beads and washed with PBS containing phosphatase inhibitor and protease inhibitor cocktail. Immunoprecipitates were fractionated by 10% SDS-PAGE, transferred to membranes, and subjected to immunoblotting with appropriate primary and secondary antibodies as described above.

Transmigration assay was performed as previously described (Seehaus et al., 2009). In brief, Corning® Transwell® polycarbonate membrane inserts (6.5 mm diameter, 8 μm pore size) were used. Endothelial monolayer was incubated with Ang II 24 h after transfected with FKBP11-Si RNA or NC-Si RNA for 24 h. Then 5 × 10⁵ THP-1 cells were seeded into each upper compartment containing 100 μL serum-free RPMI medium. In preincubation experiments cells were washed twice before THP-1 cells were added to the endothelial cell monolayer. Following incubation at 37°C in a humidified 5% CO₂ incubator for 12 h the upper compartments were removed and THP-1 cells transmigrated into the lower compartment were counted after stained with crystal violet. Each assay was performed in triplicates and repeated at least three times. Cells were counted in five randomly selected squares per well and presented as the number of migrated cells/field.

One-way ANOVA and the Student's t-test were used to evaluate any variations in outcomes between different groups by SPSS 19.0 statistical software. Most bioinformatics analyses including WGCNA were performed in R (version 3.3.1) with default test statistics and cutoff values as specified in individual method sections. P-value < 0.05 was considered as statistically significant. Results are expressed as mean ± SEM. A P-value < 0.05 was considered as statistically significant.

We used R software and corrected batch effect by Bayesian methods, then removed the probes without corresponding annotation information. To determine the expression values of each gene, we used multiple GSE52093 probes corresponding to the median expression value of that gene. After excluding Aorta Dissected No.3 through clustering analysis (Supplementary Figure 1), six human type A AAD samples and five control samples were included in the subsequent analysis (Figure 1).

By using algorithms provided in the Limma package, we calculated the data and obtained lists of differentially expressed genes. Hierarchical clustering analysis was obtained for the DEGs from 11 samples of the AAD patients and control groups. The general gene expression patterns were evidently different in the two groups by Tree View (Figure 2A).

To reveal AAD–related biological processes, we conducted a gene ontology (GO) functional enrichment analysis of the DEGs was performed. The majority of GO terms were associated with the biological processes (BP) such as DNA replication, mitotic nuclear division, extracellular matrix organization, G1/S transition of mitotic cell cycle, and sister chromatid cohesion, with cellular component (CC) like cytosol, cytoplasm, Z disc, spindle pole and actin cytoskeleton, and with molecular function (MF) like protein binding and actin binding (Figure 2B and Table 3).

To analyze the significant enrichment of the above mentioned DEGs in pathway terms, we used Fisher's exact test to calculate the significance level of the pathway (P < 0.05). We performed pathway annotation of DEGs and obtained DEGs involved in all pathway terms with the help of KEGG databases. The DEGs of AAD were enriched mainly in pathways such as cell cycle, DNA replication, ECM-receptor interaction, regulation of actin cytoskeleton, vascular smooth cell contraction, focal adhesion, and Wnt signaling (Figure 2C and Table 4).

WGCNA, which defines transcriptional modules based on Pearson correlation and determines relationship between these modules and different clinical traits, was performed to identify gene co-expression networks associated with AAD clinical-pathological factors. The microarray data Set GSE52093 was used for this analysis. We identified 9 distinct co-expression modules (Figure 3A). The expression data from different genes within each calculated module were used to determine the module eigengenes. Eigengenes were then correlated with external traits to identify modules that are significantly associated with the clinical traits and to find the most significant associations. The calculating module-trait correlation was presented as a heatmap (Figure 3B). We observed that the disease state (AAD vs. Control) showed the most significant correlation with several modules. According to the correlation coefficients, genes clustered in black, magenta, dodgerblue, red modules are positive correlation, while genes in purple, yellow, orange, turquoise, darkgreen, greenyellow modules are negative correlation in AAD tissues.

As shown in Figure 3B, genes clustered in dodgerblue module have the strongest positive correlation with AAD, while genes in darkgreen module have negative correlation with the disease status. To further explore the key regulatory notes, we applied GO and pathway enrichment analysis of the genes in the dodgerblue module (Figure 3B). Similarly, as showed in overall correlation, the most GO terms in dodgerblue module were associated with BP like mitotic nuclear division, cell division, DNA replication, G1/S transition of mitotic cell, and sister chromatid cohesion, CC like cytosol, cytoplasm and spindle pole, and MF like protein binding (Figure 3C). The DEGs of AAD were enriched mainly in cell cycle, DNA replication, ECM-receptor interaction, biosynthesis of amino acid, focal adhesion (Figure 3D). The data indicate that the genes highly significantly associated with the patients' disease state (AAD) are also the most important elements of this module.

A major goal of our research was to analyze the degree of correlation between genes and disease, and to ascertain the significance of genes in the corresponding modules. Then a PPI (protein protein interaction) map was constructed based on the DEGs in the dodgerblue module (Supplementary Figure 2). Network of DEGs in dodgerblue module was comprised of 706 genes, including the top 10 hub genes SLC20A1, GINS2, CNN1, FAM198B, MAD2L2, UBE2T, FKBP11, SLMAP, CCDC34, and GALK1 (Table 5) with the interactived genes of 1, 66, 3, 0, 6, 53, 3, 7, 0, and 11, respectively (Figure 3E).

To obtain further evidence for the significance of the expression changes seen in the dodgerblue module, we evaluated the expression of the top 10 hub genes, such as SLC20A1, GINS2, CNN1, FAM198B, MAD2L2, UBE2T, FKBP11, SLMAP, CCDC34, and GALK1 by qRT-PCR in an independent sample set of 6 AAD cases and 6 controls originated from our study center (Table 1). Validating the hub genes module, we observed that many members of this module also showed significant differential expression changes between AAD and controls with the same directionality as in our discovery set. Overall, the qRT-PCR results were consistent with the results of the microarray analysis (Table 5). However, the qRT-PCR analysis tended to give higher up-regulation levels than those calculated from the microarray data. The highest change was found for FKBP11 (Figure 4). To investigate the pathophysiological relevance of FKBP11 gene as biomarker or interventional target for AAD, we performed additional in vitro and ex vivo analysis.

To ascertain the results of the above-described in silico analyses, we selected FKBP11 as representative gene from the top 10 hub genes to determine its expression ex vivo in the aorta samples of the patients who underwent surgical procedure after aortic dissection. We determined the relative FKBP11 gene expression with qRT-PCR and protein expression with Western Blot in the aorta samples. Consistent with the in silico analysis, the FKBP11 expression level was significantly higher at both mRNA and protein level as compared with to control samples (Figures 4, 5A). EVG staining of the aortic wall showed significantly distorted and deficient elastic fibers in AAD group (Figure 5B). Immunohistochemical staining of FKBP11 on the paraffin embedded aorta sections further proved the expression pattern and intriguingly revealed an endothelial prone staining of FKBP11 (Figure 5C). In order to determine the cell type specific expression of FKBP11, double immunofluorescent staining was performed. As we assumed, the expression of FKBP11 was mostly colocalized with an endothelial marker CD31 (Figure 5D). These data indicate that within aorta of AAD patients, the expression of FKBP11 is predominantly induced in the endothelium.

Matrix metalloproteinases (MMPs), especially MMP2 and MMP9, have been shown to be induced during the onset of aortic disease including AAD, promoting degradation of extracellular matrix components directly, such as elastin (Wu Z. et al., 2016). Western Blot analysis and immunohistochemical staining of MMP9 on clinical aorta samples showed that its expression was significant higher in AAD group (Figures 6A,B), which was consistent with the literature. In addition, MMP9 expression is positively correlated with the FKBP11 expression. But unlike the expression pattern of FKBP11 in endothelium, MMP9 localized dominantly in the middle layer, where mostly smooth muscle cells and infiltrated macrophages exist. In order to trace the source of MMP9, double immunofluorescent staining was performed, Immunofluorescence staining showed that MMP9 colocalized exclusively with Mac-2 (Figure 6C), a macrophage marker (Gough et al., 2006; Shahzad et al., 2011). The increased expression of MMP9, predominantly in macrophages has also been implicated in acute vascular crises such as atherosclerotic plaque rupture, abdominal aortic aneurysm, and aortic dissection (Wu Z. et al., 2016). But how the increased FKBP11 expression on endothelium could promote the expression of MMP9 in infiltrated macrophages in AD requires further elucidation.

P65 subunit of NF-kB has been shown to provoke a pro-inflammatory state and induce MMPs and cytokines in aortic aneurysm and aortic dissection (Saito et al., 2013). In accordance with the literature, the expression of phospho-p65 was induced in a endothelial cell line EA.hy926 (ATCC® CRL-2922™) when treated with Angiotension II (Ang II), and the expression of MMP9 and other pro-inflammatory factors such as MCP1, VCAM1, ICAM1, and IL1beta were simultaneously augmented (Tieu et al., 2009). We hypothesize that the overexpression of FKBP11 could activate the endothelium through a p65 associated pathway. To further determine the role of FKBP11 in aortic dissection, we applied FKBP11 specific siRNA to knock down the expression in EA.hy926. As shown in Figure 7A, FKBP11-Si2 and FKBP11-Si3 can effectively suppress the protein expression as compared to the scrambled control. Importantly, FKBP11 knockdown suppressed the phosphorylation and nuclear translocation of p65 and subsequently the expression of the pro-inflammatory cytokines (Figures 7B–D). Intriguingly, by searching the Bio-GRID database (The Biological General Repository for Interaction Datasets: https://thebiogrid.org), which is an open access database dedicated to the annotation and archival of protein, genetic and chemical interactions for all major model organism species and humans, we found an association between FKBP11 and p65 (Chatr-Aryamontri et al., 2017). This association has been indicated by a yeast two-hybrid technology in a study mapping the interactions of human liver proteins (Wang et al., 2011). However, a functional study to validate this interaction is hitherto unknown. And here using co-immunoprecipitation assays, we demonstrate the direct interaction of these two proteins in vitro; in addition, this interaction is further augmented by Ang II treatment (Figure 7E). Importantly, Pan et al. (2014) found JAK2 as key hub gene in AAD based on the same microarray data, however knockdown of FKBP11 in EA.hy926 resulted in unaltered expression of JAK2 regardless of Ang II treatment (Supplementary Figure 4). Furthermore, we could recapitulate the above findings in a primary endothelial cells HUVECs (Supplementary Figure 5). Hence, we could posit that in aortic dissection the increased expression of FKBP11 could elicit the pro-inflammatory state of the endothelium by interacting with NF-kB p65 subunit and promoting its nuclear translocation.

As we showed in Figure 7B, the increased FKBP11 expression on endothelium could promote the expression of MMP9 in infiltrated macrophages in AD. In addition, we found that increased expression of FKBP11 could elicit endothelial cell inflammation. Based on these data, we speculate that the inflamed endothelium could promote the transmigration of the circulating monocytes to the middle layer of the aorta and accelerate the formation and progression of aortic dissection. To ascertain this hypothesis, we performed a transwell experiment (Seehaus et al., 2009). Ang II treatment induced transmigration of THP1 cells, a monocytic cell line, more toward the endothelial monolayer when treated with Ang II. However, knockdown of the FKBP11 expression in endothelial cells abrogated the Ang II induced transmigration of THP1 cells (Figures 8A–D).

Taken together that the current study involving both WGCNA and complementary gene expression analysis identify FKBP11 is a key player in aortic dissection. FKBP11 expression predominantly induced within the endothelium of the dissected aorta could provoke endothelial cells inflammation. The pro-inflammatory effects of FKBP11 are mediated through its interaction with NF-kB p65 subunit, thereby promoting its nuclear translocation and pro-inflammatory cytokine production, further facilitating transendothelial migration of the circulating monocytes to the middle layer of the aorta. The infiltrated monocytes differentiate into active macrophages that can propagate sustained local inflammation by secreting MMPs and other ECM degrading proteins, thus contributing to the formation and progression of aortic dissection (Figure 9).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.