Discovery of shear- and side-specific mRNAs and miRNAs in human aortic valvular endothelial cells

Abstract

The role of endothelial cells (ECs) in aortic valve (AV) disease remains relatively unknown; however, disease preferentially occurs in the fibrosa. We hypothesized oscillatory shear (OS) present on the fibrosa stimulates ECs to modify mRNAs and microRNAs (miRNAs) inducing disease. Our goal was to identify mRNAs and miRNAs differentially regulated by OS and laminar shear (LS) in human AVECs (HAVECs) from the fibrosa (fHAVECs) and ventricularis (vHAVECs). HAVECs expressed EC markers as well as some smooth muscle cell markers and functionally aligned with the flow. HAVECs were exposed to OS and LS for 24 h, and total RNA was analyzed by mRNA and miRNA microarrays. We found over 700 and 300 mRNAs down- and upregulated, respectively, by OS; however, there was no side dependency. mRNA microarray results were validated for 26 of 28 tested genes. Ingenuity Pathway Analysis revealed thrombospondin 1 (Thbs1) and NF-κB inhibitor-α (Nfkbia) as highly connected, shear-sensitive genes. miRNA array analysis yielded 30 shear-sensitive miRNAs and 3 side-specific miRNAs. miRNA validation confirmed 4 of 17 shear-sensitive miRNAs and 1 of 3 side-dependent miRNAs. Using miRWalk and several filtering steps, we identified shear-sensitive mRNAs potentially targeted by shear-sensitive miRNAs. These genes and signaling pathways could act as therapeutic targets of AV disease.

the aortic valve (AV) is composed of three layers: the fibrosa, facing the aorta; the ventricularis, facing the left ventricle; and the spongiosa, the layer between the fibrosa and ventricularis. The AV is composed of two cell types: interstitial cells, which make up the majority of the valve, and endothelial cells (ECs), which line the AV (5). AV disease is characterized by inflammation, sclerotic and calcific lesions, thrombus formations, and fractured matrix fibers (48). Furthermore, AV disease can be classified as either regurgitant, where the valve does not close completely after diastole, or stenotic, where the valve does not open fully during systole (1). In the developed world, 25% of patients 65 yr old or older have AV sclerosis, a characteristic of AV stenosis. AV calcification is the most common cause of AV stenosis (21, 41).

AV disease was thought to be a “wear and tear” process (3). Now it is thought to be histologically similar to atherosclerosis, in terms of lipid deposition and macrophage and T lymphocyte infiltration, with some differences [mineralization and small numbers of smooth muscle cells (SMCs)] (14, 33–35). Once beneath the endothelium, activated T lymphocytes then release cytokines (3, 16). The released cytokines facilitate extracellular remodeling by increasing metalloprotease production (3, 17). Furthermore, bone morphogenic protein (BMP) signaling and other osteogenic pathways have been shown to be upregulated, inducing differentiation of interstitial cells into an osteogenic phenotype and calcific nodule formation (40). However, it is important to note that antiatherosclerosis therapies, such as lipid-lowering statins, are not effective at slowing AV stenosis progression according to prognostic, randomized trials (7, 32, 42). This further highlights the need to better understand AV disease, a disease that is currently only treatable by replacement or repair surgeries. Recently, a microarray study completed by Bosse et al. (4) elucidated differences in gene expression between stenotic and nonstenotic valves and provided further understanding about the unique mechanism underlying AV disease.

AV disease is side dependent (24), coinciding with different hemodynamic conditions on either side. AV calcification occurs preferentially in the fibrosa, whereas the ventricularis is relatively unaffected. This side dependency may be due to the different local hemodynamic conditions. In the ventricularis, ECs experience pulsatile, unidirectional shear, whereas in the fibrosa, ECs experience disturbed flow (low and oscillatory) and become diseased (52); however, why differential hemodynamic conditions correlate with side-dependent AV calcification remains unknown. Shear stress regulates endothelial function both acutely and chronically via mechanotransduction events (37). Both vascular and valvular ECs respond to shear stress by inducing changes in gene expression profiles in vitro and in vivo (9, 29, 39, 45, 51). Previously, we (5) identified mRNA expression profiles in response to laminar shear stress using cultured porcine AVECs pooled from both fibrosa and ventricularis sides. In addition, Simmons et al. (45) showed differential mRNA expression patterns in the fibrosa and ventricularis endothelium in vivo using fresh porcine AVs with a frozen coverslip method. However, mRNA and microRNA (miRNA) expression profiles under both laminar (LS) and oscillatory shear stress (OS) in human AVECs (HAVECs), in a side-dependent manner, have not been determined, and this was the focus of the present study.

Furthermore, there is an emerging role of miRNAs in various cardiovascular-related diseases (50). miRNAs are short nucleotide sequences that bind to the 3′-untranslated region of mRNA, thereby regulating protein expression mainly by degradation of mRNA targets or by suppression of translation (13). Recently, miRNA-23b and miRNA-19a have been found to be shear responsive in human umbilical vein ECs (HUVECs) in vitro and to regulate cell growth and cyclin D₁ expression, respectively (39, 51). Fang et al. (11) showed that miRNA-10a regulates a proinflammatory state in the endothelium susceptible to atherosclerosis both in vivo in porcine aortic ECs and in vitro in human aortic ECs (HAECs). Moreover, Nigam et al.(30) recently identified miRNAs differentially expressed between aortic stenosis and aortic insufficiency (miRNA-26a, miRNA-30b, and miRNA-195) in vivo using whole bicuspid valves and linked them to calcification-related genes, such as Smad1/3, runt-related transcription factor 2 (Runx2), and Bmp2 in AV interstitial cells in vitro. Since the previous study used total RNA obtained from the entire valve or interstitial cells, the role of endothelial miRNA in AV disease remains unknown.

We hypothesized that disturbed flow present on the fibrosa side of the AV stimulates ECs to regulate miRNAs and mRNAs to induce AV disease progression. Identification of miRNAs and mRNAs that respond to shear stress (shear sensitive) in HAVECs can uncover the potential molecular mechanisms underlying AV disease. Furthermore, circulating genes may also provide potential biomarkers for AV disease (10). Here, we report the isolation and characterization of side-specific HAVECs, and, using these cells, we carried out microarrays to identify shear- and side-dependent mRNAs and miRNAs.

METHODS

Cells and cell culture.

Side-specific HAVECs [from the fibrosa endothelium (fHAVECs) and ventricularis endothelium (vHAVECs)] were isolated from noncalcified AVs obtained from heart transplant surgeries (n = 6) (according to an Insititutional Review Board-approved protocol at Emory University and Georgia Institute of Technology) using a brief collagenase digestion and gentle scraping method as previously described (5) and detailed in the Supplemental Material.¹

Confluent cells were sorted for endothelial purity in the following manner: fHAVECs and vHAVECs were incubated in 5 μl of DiI-acetylated LDL (acLDL; BTI) per 1 ml of complete media for 4 h before cell sorting using FACS Aria I (BD Biosciences). HAECs and human umbilical vein ECs (HUVECs) were used as positive controls. Human aortic SMCs (HASMCs) were used as a negative control. Before being sorted, fluorescent images of cells incubated with acLDL were taken using an Axiocam MRm camera (Zeiss) and an Axiovert 200M inverted microscope (Zeiss) with a ×5 (Plan-Neofluar, numerical aperture: 0.15) objective lens. Axiovision 3.1 software (Zeiss) was used for image acquisition and processing.

Shear conditions.

Upon confluency, HAVECs were exposed to steady LS using either the parallel plate flow chamber or the cone-and-plate viscometer, as we have previously reported (18, 26) and described in further detail in the Supplemental Material. OS was applied using the cone-and-plate viscometer (26). For LS, we used a unidirectional shear stress of 20 dyn/cm²; for OS, we used a bidirectional shear stress of ±5 dyn/cm² at 1 Hz to approximate the complex shear stress conditions surrounding AVECs in vivo (47).

HAVEC alignment under laminar shear.

fHAVECs and vHAVECs were sheared under LS for 48 h using the parallel plate flow chamber and shear medium 1 (n = 4). (All media formulations are in the Supplemental Material.) After being sheared, HAVECs were washed with PBS and fixed with 4% formaldehyde. HAVECs were then stained for F-actin using rhodamine phalloidin (Invitrogen). Slides were mounted with Fluoro-Gel (Electron Microscopy Sciences). Images were taken at ×40 magnification (numerical aperture: 1.3) using a Zeiss LSM 510 UV confocal microscope. Images were acquired at room temperature using Zeiss LSM 510 software. LSM Image Browser was used for processing. The shape index and angle of alignment were assessed as previously described (18). The shape index ranges from 0 to 1, where a line has a shape index of 0 and a circle has a shape index of 1.

Characterization of side-specific HAVECs.

HAVECs were characterized at the gene and protein levels. Expression of three EC-specific genes [von Willebrand Factor (vWF), platelet/EC adhesion molecule (PECAM)-1, and vascular endothelial (VE-)cadherin] and two smooth muscle markers [α-smooth muscle actin (α-SMA) and basic calponin] was assessed by quantitative PCR using StepOne Plus and SYBR green reagents (ABI). The total number of copies per marker was determined using the standard curve method. ANOVA with a post hoc Tukey's test (GraphPad Prism 5) was used to determine differences among cell types. Immunocytochemical staining was performed using the same markers and is described in the Supplemental Material. For flow cytometry analysis, PECAM-1 antibody (1:100) was used to calculate the percentage of positive ECs. Fluorescence intensity was detected from 10,000 events using the BD LSR II flow cytometer, and FlowJo 7.6 was used for data analysis. Gates were set so that 2% of the negative control was positive.

Isolation of total RNA and mRNA and miRNA microarrays.

HAVECs were sheared under either LS or OS for 24 h in shear medium 2. Total RNAs were generated from the following conditions: 1) fHAVECs exposed to OS (FO), 2) fHAVECs exposed to LS (FL), 3) vHAVECs exposed to OS (VO), and 4) vHAVECs exposed to LS (VL). After the exposure to shear, cell alignment was verified by light microscopy. Cells were washed three times with ice-cold PBS and scraped in Qiazol (Qiagen) to isolate total RNA following the microRNeasy kit (Qiagen). Kruppel-like factor 2 (Klf2) and endothelial nitric oxide synthase (eNOS), well-known shear-responsive genes, were then analyzed by quantitative PCR to determine whether HAVECs responded as anticipated. Next, samples from six different HAVEC isolations were selected based on quality control analysis using an Agilent Bioanalyzer and a Taqman assay for RNU24. Full coverage miRNA (Illumina Human miRNA BeadChips) and mRNA arrays (Illumina Human HT-12 Expression BeadChips) were completed (n = 6 each). Twenty-four mRNA microarrays and twenty-four miRNA microarrays (4 groups, n = 6 each) were run by the Emory Biomarker Center from total RNA samples. After hybridization, BeadChips were scanned on the Illumina BeadArray Reader to determine the probe fluorescence intensity of 47,231 probes on the mRNA array and 1,145 probes on the miRNA array. Non-normalized, raw probe intensities were used for analysis. Microarray data have been uploaded to the Gene Expression Omnibus (GSE26953).

Microarray analysis and heat map generation.

Significance of microarray analyses (SAMs; SAM 3.0) were completed to determine shear- and side-dependent differences in HAVEC miRNA and mRNA arrays. Paired comparisons were made among all four groups. Significance was assessed with a false discovery rate of <6% for mRNA arrays and <25% for miRNA arrays. Heat maps of shear-responsive miRNAs and mRNAs were created using Cluster 3.0 for clustering and Java Treeview for visualizing, as previously described (29).

Validation by quantitative PCR analysis.

Quantitative PCR with the same RNA preparations used in the microarray experiments was used to validate selected shear- and side-dependent miRNAs and mRNAs. Significance testing was performed using Student's t-tests with P < 0.05. mRNAs were validated for those changed in the most physiologically relevant condition, mimicking in vivo shear conditions: the fibrosa endothelium exposed to OS and the ventricularis endothelium exposed to LS (FO compared with VL). From this group, we selected genes that 1) showed a large fold change, 2) were transcription regulators, 3) were well-known, shear -sensitive genes, or 4) were novel targets that may be involved in AV disease. All miRNAs changed in FO versus VL, side-dependent miRNAs, and those miRNAs previously found to be shear sensitive in HUVECs (miRNA-192) were validated (28). Quantitative PCR using Taqman assays with predesigned primers and probes from Applied Biosystems were used for validation. All values were normalized to internal controls: 18S for mRNA arrays or RNU6B for miRNA arrays. Fold changes were calculated using the ΔΔC_t method (where C_t is threshold cycle).

RESULTS

Isolation and purification of fHAVECs and vHAVECs.

fHAVECs and vHAVECs were sorted using acLDL for purity. After being sorted, the percentage of acLDL-positive cells was compared among fHAVECs, vHAVECs, and HUVECs. Supplemental Figure S1, A–D, shows the cobblestone morphology of HAVECs as well as acLDL uptake before sorting. Flow cytometry analysis after sorting revealed that 95% and 94% of vHAVEC and fHAVEC populations, respectively, were acLDL positive compared with HUVECs, which had a 96% acLDL-positive population.

Characterization of side-specific HAVECs.

Quantitative PCR analysis was performed using the following endothelial markers: VE-cadherin (Fig. 1A), PECAM-1 (Fig. 1B), and vWF (Fig. 1C), and the following SMC markers: α-SMA (Fig. 1D) and basic calponin (Fig. 1E). vHAVECs and fHAVECs were compared with well-known cells: HUVECs and HAECs as positive controls and HASMCs as a negative control. Expression of VE-cadherin and PECAM-1 in fHAVECs and vHAVECs was not significantly different from HUVECs and HAECs, whereas they were not detected in HASMCs (Fig. 1, A and B). Expression of vWF in HAVECs tended to be lower than HUVECs and HAECs; however, there was only a significant difference between vHAVECs and HUVECs. α-SMA expression was also significantly different between vHAVECs and HUVECs. As expected, HASMCs expressed α-SMA and basic calponin (Fig. 1), but, surprisingly, we observed α-SMA and basic calponin expression in HAVECs. HUVECs and HAECs also had low but measurable mRNA levels of α-SMA and basic calponin. Furthermore, expression levels of α-SMA in fHAVECs and vHAVECs were ∼24% and 41% of HASMCs, respectively, and expression levels of basic calponin in fHAVECs and vHAVECs were 34% and 69% of HASMCs, respectively.

Fig. 1.

Human aortic valve (AV) endothelial cell (EC) (HAVEC) characterization with quantitative PCR. mRNA expression profiles of HAVECs from the fibrosa (fHAVECs) and ventricularis (vHAVECs) were compared against other ECs [human umbilical vein ECs (HUVECs) and human aortic ECs (HAECs)] and against human aortic smooth muscle cells (HASMCs). Three positive EC markers [vascular endothelial (VE)-cadherin (A), platelet/EC adhesion molecule-1 (PECAM-1; B), and von Willebrand factor (vWF; C)] and two smooth muscle markers [α-smooth muscle actin (α-SMA; D) and basic calponin (E)] were assessed. Values are means ± SE; n = 3 HUVEC donors, 3 HAEC donors, 3 HASMC donors, 5 fHAVEC donors, and 5 vHAVEC donors. *P < 0.05.

Next, we examined the above cell markers at the protein level using immunocytochemistry in HAVECs. Similar to mRNA results, we observed expression and expected localization of PECAM-1, vWF, and VE-cadherin, as shown in Fig. 2, A–F. As expected, α-SMA staining was intense and well organized into filaments in HASMCs (Fig. 2G); however, fHAVECs and vHAVECs also showed α-SMA staining, although it was not well organized (Fig. 2, E, F, I, and J). To determine whether α-SMA staining in our cultured HAVECs was due to a contaminating subpopulation of interstitial cells or whether, in fact, HAVECs expressed both α-SMA and EC markers, we costained HAVECs with both α-SMA and VE-cadherin. We found that most α-SMA-positive cells also expressed VE-cadherin, suggesting that HAVECs indeed coexpressed both endothelial markers and a smooth muscle marker. To determine if this in vitro finding was also observed in vivo, we costained human AV frozen sections with vWF and α-SMA. As shown in Fig. 2, K and L, some HAVECs expressed both vWF and α-SMA. This result is consistent with a previous report by Paranya et al. (38) in the ovine AV endothelium.

Fig. 2.

HAVEC protein characterization. To assess EC markers and a smooth muscle cell marker, we performed immunocytochemical staining on fHAVECs and vHAVECs. A, C, and E: PECAM-1 (A), vWF (C), and costaining of VE-cadherin and α-SMA (E) in fHAVECs taken with a confocal microscope. n = 4. Scale bar = 20 μm. B, D, and F: analogous staining in vHAVECs (n = 4). G–J: images of α-SMA staining taken with traditional fluorescence. n = 4. Scale bar = 50 μm. G: α-SMA staining in HASMCs. H: α-SMA staining in HUVECs. I and J: representative images of α-SMA staining in fHAVECs (I) and vHAVECs (J). K and L: images of frozen sections from human AVs costained with vWF and α-SMA. Original magnifications: ×10 (K) and ×40 (L). n = 4. Green signifies EC marker staining; red signifies α-SMA; blue is a Hoechst counterstain. M–P: PECAM-1 expression via flow cytometry (n = 4). M and N: representative histograms of fHAVEC (M) and vHAVEC (N) staining of PECAM-1 compared with the isotype control. O and P: respresentative histograms of fHAVEC (O) and vHAVEC (P) PECAM-1 staining compared with HASMCs, respectively.

To further confirm the purity of isolated ECs, we investigated the percentage of cells expressing an EC marker (PECAM-1) by flow cytometry. Compared with its isotype or cell type controls, 91–93% of all isolated HAVECs expressed PECAM-1, similar to a positive control, HAECs (Fig. 2, M–P, and Supplemental Table S1). In contrast, HASMCs did not express the endothelial marker PECAM-1 at either the mRNA (quantitative PCR; Fig. 1B) or protein (flow cytometry; Fig. 2, O and P) level.

Shear stress regulates HAVECs.

To examine whether HAVECs isolated separately from both the fibrosa and ventricularis respond to shear stress, we studied shear-induced cell alignment and shape index as well as two well-known mechanosensitive genes, Klf2 and eNOS. After being sheared for 48 h, both fHAVECs and vHAVECs aligned in the direction of the flow, as shown in Fig. 3, B and D. This shear-induced shape change was further confirmed by the quantitative analysis of shape index and angle of alignment (Fig. 3, E and F). fHAVECs showed small, but statistically significant, increased alignment over vHAVECs (P < 0.05).

Fig. 3.

HAVECs respond to shear stress. fHAVECs (B) and vHAVECs (D) were sheared under laminar shear (LS; 20 dyn/cm²) for 48 h and compared with static controls [ fHAVECs (A) and vHAVECs (C)]. n = 4. *Differences in shear condition (P < 0.05); †difference in side (P < 0.05). Shape index (E) and angle of alignment (F) were compared between sides and conditions. G: LS increased the phosphorylation of endothelial nitric oxide synthase (eNOS) over static and oscillatory shear (OS) samples. n = 4. The following groups were analyzed: fHAVECs exposed to OS (FO), fHAVECs exposed to LS (FL), vHAVECs exposed to OS (VO), and vHAVECs exposed to LS (VL). *P < 0.05 compared with FS and VO; #P = 0.06 compared with FS and P = 0.12 compared with FO.

Next, we found that LS exposure increased total eNOS expression in both fHAVECs and vHAVECs over static and OS conditions (see Fig. 5C). Also, LS significantly increased phosphorylated eNOS in vHAVECs, and there was a trend for increased phosphorylated eNOS in fHAVECs after LS, as shown in Fig. 3G. Furthermore, we found that eNOS and Klf2 were increased by LS in HAVECs at the mRNA level by ∼3-fold and 12-fold, respectively (see Fig. 5). These shear-dependent responses are consistent with other vascular ECs (29).

Fig. 5.

mRNA validation. A: of the 28 shear-sensitive mRNAs changed in the most physiological conditions (FO and VL) according to microarray analysis, 26 mRNAs were confirmed with quantitative PCR. n = 6. *P < 0.05. For validation at the protein level, the protein lysate was used for Western blot analysis using antibodies against bone morphogenetic protein-4 (BMP-4; B), total (t-)eNOS (C), and IκB (D). FS, fHAVECs under static conditions; VS, vHAVECs under static conditions. Representative blots for each antibody, including an internal control (β-actin), are shown. The loading control shown in C is the same and appropriate loading control shown in Fig. 3G. Bar graphs quantify the blots. For BMP-4, n = 5; *P < 0.05 compared with FS, VS, and VO and **P < 0.05 compared with FS and FO. For IκB, n = 3; *P < 0.05 compared with FS, VS, and VO and **P < 0.05 compared with FS, VS, and FO. For total eNOS, n = 5; *P < 0.05 compared with FS, FO, VS, and VO and **P < 0.05 compared with FS, FO, VS, and VO.

mRNA microarray of sheared, side-specific HAVECs.

Thus far, our results confirmed that our HAVECs responded to shear as expected at least for some of the well-known mechanoresponses. Therefore, we carried out genome-wide microarray experiments using these HAVECs to determine the expression profiles of both mRNAs and miRNAs in response to shear stress. Both fHAVECs and vHAVECs were exposed to LS (20 dyn/cm²) and OS (±5 dyn/cm²) for 24 h. Total RNA was prepared for mRNA and miRNA microarray analyses of the following four groups: FO, FL, VO, and VL.

Results from the mRNA microarray using Illumina Human HT-12 BeadChips (47,231 human probes) were analyzed using SAM. This analysis revealed that >1,000 genes were differentially regulated by OS compared with LS (>700 downregulated genes in OS and >300 upregulated genes in OS, false discovery rate: <6%), as shown in Supplemental Table S2. Significantly altered gene probes were analyzed by hierarchical clustering to investigate intragroup and intergroup variations as represented in the heat map (Fig. 4, A–C). We focused on three comparisons: 1) FO versus VL, the most physiologically relevant condition in vivo; 2) FO versus FL; and 3) VO versus VL, because of their physiological relevance. These three heat maps showed low variability within groups, showing the reproducibility of the data. Venn diagrams (Fig. 4D) were also used to investigate differences and overlaps in shear-responsive genes among these three groups. The majority of shear-sensitive genes identified were common to all three groups (53% for FO vs. VL, 67% for VO vs. VL, and 69% for FO vs. FL). The Venn diagram suggested minimal side-specific differences in gene expression. Interestingly, SAM also did not detect any side-dependent differences between fHAVECs and vHAVECs at the mRNA level, confirming the Venn diagram analysis.

Fig. 4.

Differential expression profiles of HAVECs in response to OS versus LS. The heat map shows all shear-sensitive genes changed in the following three groups: FO versus VL (A), FO versus FL (B), and VO versus VL (C). False discovery rate: <6%. n = 6. Red shows upregulated genes and green shows downregulated genes. D: number of common genes among groups.

Validation of shear-sensitive mRNAs in HAVECs by quantitative PCR.

To validate the mRNA microarray data, we carried out quantitative PCR for 28 genes selected based on the following criteria: 1) top changers in both directions [solute carrier organic anion transporter family, member 2A1 (Slco2a1), IGF-binding protein 5 (Igfbp5), synuclein, α-interacting protein (Sncaip), and adrenomedullin (Adm)]; 2) known shear-sensitive genes (Klf2, eNOS, and Bmp4); 3) well-known targets in cardiovascular disease [cathepsin K (Ctsk)]; 4) novel targets that may be related to AV disease [histone deacetylase 1 (Hdac1), inhibitor of DNA binding 1 (Id1), and desert hedgehog (Dhh)]; 5) transcription factors that have the potential to regulate many genes [activating transcription factor 3 (Atf3), early growth response 1 (Egr1), and FosB]; and 6) those that showed many relationships with other genes in Ingenuity Pathway Analysis (IPA; Thbs1 and Nfkbia). We found that the mRNA array data were highly accurate, with 26 of 28 genes (93%) confirmed (Fig. 5A and Supplemental Table S3). We further confirmed the mRNA microarray results using cell lysates obtained from fHAVECs and vHAVECs exposed to LS, OS, or static conditions for 24 h. Western blot (Fig. 5, B–D) results showed that OS increased BMP4 protein expression while decreasing IκB-α and eNOS protein expression compared with LS (P < 0.05).

Functional annotation and categorization of mechanosensitive genes.

To better understand the mRNA microarray data set as well as demonstrate possible areas of research stemming from this work, we used IPA to investigate the top pathways and functions changed in FO compared with VL, the most physiologically relevant condition in vivo, as shown in Table 1. The top molecular and cellular functions identified were cell movement, cell death, and cellular growth and proliferation. Cardiovascular system development and function was the top physiological system development and function identified.

Table 1.

Overrepresented gene ontology categories regulated 24 h after shear stress in human aortic valve endothelial cells by Ingenuity Pathway Analysis

	Number of Genes
Gene Ontology Category
Molecular and cellular functions
Cell movement	175
Cell death	264
Cellular growth and proliferation	257
Cellular development	221
Cellular compromise	28
Physiological system development and function
Cardiovascular system development and function	118
Organismal development	136
Organismal survival	124
Tissue development	157
Skeletal and muscular system development and function	92
Top Canonical Pathways
LPS/IL-1-mediated inhibition of retinoid X receptor function
Sulfur metabolism
Hepatic fibrosis/hepatic stellate cell activation
Keratan sulfate biosynthesis
Neuregulin signaling

Moreover, we used the pathway builder to uncover known relationships between shear-sensitive genes to further our understanding of AV disease, as shown in Supplemental Fig. S2. This analysis highlights the following highly connected nodes: Nfkbia, Thbs1, peroxisome proliferator-activated receptor-γ (Pparg), Klf2, and endothelin-1 (Edn1). Next, we used the comparison function of IPA to compare our mRNA microarray data set with other shear- and valve-related data sets. About 45% of the shear-sensitive genes (466 of 1,041 genes) identified in our HAVEC mRNA microarray (FO vs. VL) were also reported in HUVECs (28). We then compared our HAVEC mRNA microarray result (FO vs. VL) to those of the in vivo porcine AV endothelium (fibrosa vs. ventricularis) reported by Simmons et al. (45). This comparison revealed 24 genes that changed in the same direction (FO vs. VL compared with the fibrosa vs. ventricularis porcine endothelium). Of those 24 genes, 8 genes had known relationships, as shown in Supplemental Fig. S3. This analysis showed that some mechanosensitive genes are conserved in vitro compared with in vivo across species.

Identification of shear-sensitive and side-dependent miRNAs in HAVECs.

To discover shear-sensitive and side-dependent miRNAs in HAVECs, we carried out miRNA microarrays using Illumina Human microRNA BeadChips (1,145 human probes) using total RNA as described above for the mRNA microarray. Analyses were conducted using the following four groups: FO, FL, VO, and VL. Comparisons were made among three groups: 1) FO versus VL, 2) FO versus FL, and 3) VO versus VL to investigate the differential responses of HAVECs to OS versus LS. We did not include the comparison between VO and FL because of its lack of physiological relevance in vivo. To investigate side-dependent miRNAs in HAVECs, we compared 1) FL versus VL and 2) FO versus VO.

Among all comparisons, signicance of microarray analysis of the miRNA microarray data showed 30 shear-responsive miRNAs and 3 side-dependent changes, as shown in Supplemental Table S4. Heat map analysis, shown in Fig. 6, showed more intragroup and intergroup variation compared with the mRNA microarray analysis (Fig. 4).

Fig. 6.

MicroRNA (miRNA) heat maps. The heat maps show miRNA expression profiles for the following shear response comparisons: FO versus VL (A), FO versus FL (B), and VO versus VL (C). D and E: side-dependent differences between the following groups: FO versus VO (D) and FL versus VL (E). False discovery rate: <25%. n = 6.

Validation of shear-sensitive and side-dependent miRNAs in HAVECs by quantitative PCR.

We validated shear-sensitive and side-dependent miRNAs using quantitative PCR (Supplemental Table S3). For validation, we chose all 16 miRNAs identified in the FO versus VL group (Fig. 7A) as well as all side-dependent miRNAs (FL vs. VL and FO vs. VO; Fig. 7, C and D). We also selected miRNA-192, identified in VO versus VL (Fig. 7B), since it was previously found as a shear-sensitive miRNA (28). Three of sixteen miRNAs (miRNA-139-3p, miRNA-187, and miRNA-486–5p) identified in FO versus VL were validated. Also, there was a trend showing increased expression of miRNA-217 in FO versus VL (P < 0.1). Moreover, miRNA-192 was also confirmed. Interestingly, we validated one side-dependent miRNA, miRNA-370 (FL vs. VL). We attempted in situ hybridization using human AV frozen sections and fluorescent probes to validate the in vivo levels of shear-sensitive miRNAs, but it was unsuccessful due to high background in human AV samples (data not shown).

Fig. 7.

miRNA validation. A: of the 16 miRNAs that were found to be shear responsive in the most physiological conditions (FO and VL) through microarray analysis, 3 miRNAs were confirmed with quantitative PCR. B: miRNA-192, found to be shear sensitive in vHAVECs, was confirmed. C: side-dependent miRNA-370 was confirmed. D: side-dependent miRNA-485-3p and miRNA-485-5p were not confirmed. n = 6. *P < 0.05 and +P < 0.1.

Identification of potential gene targets of shear-sensitive miRNAs.

Our goal was to discover potential shear-sensitive mRNAs that are regulated by shear-sensitive miRNAs. To achieve our goal, the miRWalk program, which combines 10 Web-based gene target prediction programs, was used to initially identify potential miRNA targets. Of those targets, we selected genes identified by more than three prediction programs. These selected genes were further filtered for those identified as shear sensitive in our mRNA microarray. Finally, we selected shear-sensitive mRNAs that showed an inverse relationship compared with miRNAs. An example of this filtering process is shown in Supplemental Fig. S4. From this filtering process, we identified two potential targets for miRNA-139-3p, 16 potential targets for miRNA-187, 22 potential targets for miRNA-192, and 8 potential targets for miRNA-486-5p, as shown in Table 2 and categorized by cellular function in Fig. 8. Interestingly, miRNA-139-3p identifies the FosB transcription factor as a putative target, whereas miRNA-192 potentially targets additional transcription factors: Atf3, Egr1, and FosB.

Table 2.

Predictive gene targets for shear-sensitive miRNAs

miRNA	Predictive mRNA Target
hsa-miRNA-139-3p	fosB, Rnf41
hsa-miRNA-187	Armc7, Cd276, Cyyr1, Dync1li2, Ets1, Flnc, Ifnar1, Lypd1, Mbnl2, Nfkbiz, Pgm2l1, Plod3, Sema3f, Snx27, Ssh2, Trib2
hsa-miRNA-192	Acp1, Apln, Asb1, Atf3, C10orf10, Ccnd2, Chrnb1, Ctnnbip1, Dynlt1, Egr1, Fam129a, FosB, Hoxb5, Mcm6, Myo1d, Nav1, Ndst1, Nmt2, Nrip3, Osbpl10, Phactr2, Phlda1
hsa-miRNA-486-5p	Ak2, Als2cr4, Camk2n1, Ctdspl, Efna1, Mylk2, Prnd, Rnf41

miRNA, microRNA; hsa, Homo sapiens.

Fig. 8.

Predicted targets in AV disease. This chart organizes predicted targets from Table 2 by cellular functions important in AV disease using Ingenuity Pathway Analysis and AmiGO.

DISCUSSION

The major findings of this study were 1) the isolation and culture of ECs in a side-specific manner (fibrosa and ventricularis) from human AVs, 2) responses of HAVECs to shear stress by aligning in the direction of the imposed flow, 3) the discovery of shear-dependent mRNAs and the most interconnected shear-sensitive genes as potential master mechanoregulators, and 4) the identification of shear- and side-dependent miRNAs and their potential shear-sensitive target genes.

Here, we were able to isolate, culture, and characterize side-specific HAVECs. As far as we are aware, this is the first report of HAVECs isolated in a side-dependent manner in human AVs. Given our ultimate goal of understanding and treating human AV disease, it is important to use valvular ECs from human sources. Our isolated HAVECs retain many molecular markers (vWF, VE-cadherin, PECAM-1, Klf2, and eNOS), functions, and phenotypes characteristic of vascular ECs.

Although HAVECs are ECs, they are a unique class of ECs. To our initial surprise, we found that HAVECs express α-SMA and basic calponin, SMC markers. One potential explanation was that our HAVECs were contaminated with human AV interstitial cells, which express SMC markers. Although we cannot completely rule out the possibility of a small contamination, our flow cytometry and immunocytochemistry (Fig. 2) data strongly support that most of the cells were ECs. Smooth muscle marker expression in ECs has also been reported by others previously both in vivo and in vitro. Porcine coronary microvascular ECs express α-SMA, whereas porcine aortic ECs do not (2). Also, vascular ECs express α-SMA and basic calponin as they undergo endothelial-to-mesenchymal transdifferentiation (EMT) (15, 22). Furthermore, the ovine AV endothelium expressing α-SMA has been shown to undergo EMT in vivo (38). Interestingly, we identified miRNA-370 as a side-dependent miRNA in HAVECs and increased in vHAVECs, and, recently, miRNA-370 has been reported to be upregulated during EMT (6). Further statistical analysis between fHAVECs and vHAVECs showed that there was an increase in α-SMA (P = 0.0243) and basic calponin (P = 0.0676) expression in vHAVECs compared with fHAVECs, coinciding with increased expression of EMT-implicated miRNA-370. Therefore, it is possible that some of our HAVECs, especially vHAVECs, that express α-SMA may be undergoing EMT. This EMT could have been exacerbated since we isolated HAVECs from patients undergoing heart transplantation. However, currently, we do not have access to healthy AVs, and future studies using healthy animal or human AVs are necessary.

Our functional experiments showed that HAVECs aligned in the direction of the flow using ∼25 HAVEC isolations. While this is consistent to that of vascular ECs, it is different from porcine AVECs, which aligned perpendicular to flow (5). Potential explanations include 1) a species difference (human vs. porcine), 2) side dependency (fHAVECs and vHAVECs vs. pooled porcine AVECs), 3) sorting and purity (sorted HAVECs and unsorted porcine AVECs), and 4) matrix difference (gelatin vs. collagen I). Future studies examining species differences and cell-matrix interactions in HAVECs will be critical for a thorough understanding of valvular biology in human disease and in animal models of disease.

Using fHAVECs and vHAVECs, we carried out gene microarray experiments to identify shear-sensitive and side-dependent mRNAs and miRNAs. The samples used for microarray analysis were selected based on the following critieria: 1) LS-induced alignment, 2) eNOS and Klf2 upregulation by LS, and 3) total RNA quality. We chose to focus on comparisons made between FO and VL because these conditions mimicked in vivo hemodynamic conditions for each side of the AV. Additionally, SAM did not detect any side-dependent differences. The absence of side-dependent differences in mRNA expression suggests that the microenvironment (i.e., shear stress patterns, etc.) could be responsible for the differences between fHAVECs and vHAVECs seen in vivo rather than them being inherently different cell types. Further investigation is required to address this question.

When FO versus VL was compared, >300 genes were upregulated and >700 genes were downregulated by OS (Supplemental Table S2 and Fig. 5). The mRNA microarray data are remarkably accurate, as we were able to validated 26 of 28 genes tested (93% validated). As described in the results, we validated some of the most well-known shear-sensitive genes, Klf2, eNOS, and Bmp4, providing further confidence in our microarray results. In addition, we identified expected targets such as Ctsk as well as interesting targets that may be related to AV disease, including Hdac1, Id1, and Dhh. Moreover, we identified shear-sensitive transcription factors such as Atf3, Egr1, and FosB that could serve as master regulators.

To better understand the shear-sensitive changes in gene expression profiles, IPA was completed and showed that Thbs1 and Nfkbia were the genes with the most connections to other flow-sensitive genes. Interestingly, Thbs1, which plays a role in blocking angiogenesis and has a proapoptotic role (23), was found to be increased under FO versus VL. Likewise, Nfkbi (IκK-α), which plays an antiapoptotic role by sequestering inactive NF-κB (19), was found to be decreased under FO versus VL. Strikingly, the data suggest that Thbs1 and Nfkbia are highly connected, shear-regulated genes that promote a proapoptotic environment under oscillatory shear in HAVECs; however, their role in AV disease has not yet been reported, and their potential as therapeutic targets of AV disease has not been fully realized. Further gene ontology analysis, using IPA, showed that the top canonical pathway (Table 1) regulated by shear (OS/LS) was LPS/IL-1-mediated inhibition of retinoid X receptor function. Retinoid X receptor inhibition has recently been hypothesized to induce a proatherogenic environment, which could be important in AV disease (36). Another top pathway related to shear in HAVECs was sulfur metabolism. Recent research has highlighted the importance of sulfur metabolism, particularly hydrogen sulfide, in cardiac protection (43), but the role of sulfur metabolism in the AV has yet to be well described. Investigation of the role of these mechanosensitive genes and pathways in AV disease is highly warranted in the future.

This is the first study comparing AVECs under LS and OS in vitro; however, Simmons et al. (45) showed side-dependent mRNA expression profiles of the porcine AV endothelium in vivo. When HAVECs (FO vs. VL) were compared with porcine AVECs in vivo (fibrosa vs. ventricularis endothelium), we found that 24 of 48 common genes changed in the same direction. Eight of the twenty-four common genes had known relationships according to IPA (Supplemental Fig. S3), including Bmp4, Ctsk, Hdac1, and Atf3. The following experimental differences between the two data sets may explain the relatively low number of common genes: 1) varied donor health history (diseased human AVs vs. healthy young pigs), 2) species difference (human vs. porcine), 3) in vitro versus in vivo, and 4) different array platforms with different probes (Illumina vs. Affymetrix). Interestingly, we found >450 shear-responsive genes conserved between our HAVEC shear data set and HUVECs sheared in the same cone-and-plate system using the same array platform (28), providing additional confidence in our data.

Our miRNA microarray data showed 30 shear-sensitive miRNAs and 3 side-dependent miRNAs. We validated 3 of 16 miRNAs tested in FO versus VL (miRNA-139-3p, miRNA-187, and miRNA-486-5p). Furthermore, miRNA-192 was validated in VO versus VL. Finally, side-dependent miRNA-370 was validated in FL versus VL. As demonstrated by heat map analysis (Fig. 6) and quantitative PCR validation (Fig. 7), the miRNA microarray data were less reproducible and accurate than the mRNA microarray. This may be due to a low-quality miRNA microarray platform, given its relative infancy.

The identified shear-responsive miRNAs are reportedly involved in several cell functions that may provide insight into the role of miRNAs in AV disease and stenosis (Fig. 8). For instance, it is known that proliferation, apoptosis, and migration all play a role in tissue remodeling in AV disease and that AV disease is characterized by increased monocyte binding and invasion into the interstitial space, hallmarks of inflammation (3). Here, we hypothesized that OS modulates these shear-responsive miRNAs to induce disease. Several studies have implicated the role of these miRNAs in AV disease, either directly or in a paracrine manner. Notably, OS-induced miRNA-187 was also found to be increased in thyroid tumors compared with hyperplastic nodules (31) and increased with LPS in the mouse lung (25). Moreover, a reduction of miRNA-187 caused a decrease in cell growth in HeLa cells (8). These studies suggest that a role of OS-induced miRNA-187 in promoting tissue remodeling and inflammation in AV disease. Additionally, we found miRNA-486-5p to be reduced under OS in HAVECs. Another study (27) found that miRNA-486-5p was decreased in eight types of cancer, suggesting that a reduction of miRNA-486-5p supports tissue remodeling as characterized by increased proliferation, an important component of AV disease. Moreover, OS-diminished miRNA-192 was also found to be decreased in mice whose mothers were fed a high-fat diet (53), suggesting a potential role in atherogenesis. Furthermore, OS-diminished miRNA-192 has been linked to cellular proliferation and is thought to be a tumor suppressor decreased in primary cancers (46). Finally, a study (12) has shown that a decrease in miRNA-192 correlates with a decrease in cell growth. These studies support that the OS-dependent loss of miRNA-192 promotes proliferation and an atherogenic environment. Research regarding OS-diminished miRNA-139-3p is limited and unclear; however, one study has shown miRNA-139-3p downregulation in malignant adrenocortical carcinomas compared with benign adrenocortical adenomas, suggesting a role for miRNA-139-3p in migration suppression (20, 44, 49). Although the specific role of these shear-sensitive miRNAs in AVECs must be established, these findings raise interesting hypotheses regarding their potential role in AV disease.

Interestingly, some shear-responsive miRNAs identified in this study were not shared by other shear-response studies in ECs. Although miRNA-192 is common to both our HAVEC data set and the HUVEC data set of Ni et al. (29), the identification of miRNA-139-3p, miRNA-187, and miRNA-486-5p as shear-sensitive in ECs is a first, to our knowledge. Conversely, shear-responsive miRNA-23b and miRNA-19a, previously identified as shear sensitive in HUVECs, were not present in our data set, further supporting the idea that shear regulation of miRNAs in HAVECs is unique from other ECs (39, 51). Additional contributing factors may include the cell source (young, relatively healthy vs. heart transplant patients) and shear conditions (LS/static vs. OS/LS).

Through the miRWalk analysis (as described in Supplemental Fig. S4), we have identified shear-sensitive potential mRNA targets regulated by shear-sensitive miRNAs. This analysis involved several filtering steps to reduce the number of predicted mRNA targets. Validation of these relationships is required in the future.

In summary, we found shear-sensitive miRNAs and mRNAs in side-specific human AVECs by microarray analysis. Using IPA, we also found shear-sensitive signaling nodes and pathways. The genes and signaling pathways identified in this study could be used as therapeutic targets of AV disease.

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Grants HL-75209, HL-87012, and HL-80711 (to H. Jo and R. M. Nerem), World Class University Project of Korea Grant R31-2008-000-10010-0 (to H. Jo), an Ada Lee and Pete Correll Professorship (to H. Jo), an American Heart Association Predoctoral Fellowship (to R. F. Ankeny), a National Science Foundation Graduate Research Fellowship (to C. J. Holliday), and a Georgia Tech/Emory Center Fellowship in Tissue Engineering (to C. J. Holliday).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).