Inflammatory cytokines promote mesenchymal transformation in embryonic and adult valve endothelial cells

Abstract

Objective

Inflammatory activation of valve endothelium is an early phase of aortic valve disease pathogenesis, but subsequent mechanisms are poorly understood. Adult valve endothelial cells retain the developmental ability to undergo endothelial to mesenchymal transformation (EndMT), but a biological role has not been established. Here we test whether and how inflammatory cytokines (TNF-α and IL-6) regulate EndMT in embryonic and adult valve endothelium.

Methods and Results

Using in vitro 3D collagen gel culture assays with primary cells, we determined that IL-6 and TNF-α induce EndMT and cell invasion in dose dependent manners. Inflammatory-EndMT occurred through an Akt/NFκB-dependent pathway in both adult and embryonic stages. In embryonic valves, inflammatory-EndMT required canonical TGFβ signaling through Alk2/5 to drive EndMT. In adult valve endothelium, however, inflammatory-induced EndMT still occurred when Alk2/5 signaling was blocked. Inflammatory receptor gene expression was significantly upregulated in vivo during embryonic valve maturation. Endothelial-derived mesenchymal cells expressing activated NFκB were found distal to calcific lesions in diseased human aortic valves.

Conclusions

Inflammatory cytokine induced EndMT in valve endothelium is present in both embryonic and adult stages, acting through Akt/NFκB but differently utilizing TGFβ signaling. Molecular signatures of valve EndMT may be important diagnostic and therapeutic targets in early valve disease.

Introduction

Degenerative heart valve disease (HVD) is a significant contributor to cardiovascular morbidity in the United States. Aortic valve dysfunction affects 2.5% of all Americans and 30% of the elderly¹. HVD directly claims 20,000 lives annually in the US and sufferers have twice the risk of heart attacks, strokes, and heart failure¹. In addition, congenital valve abnormalities affect 1%-2% of live births and lead to early structural deterioration². Historically, aortic valve disease (AVD) was thought to be pathogenically similar to vascular atherosclerosis. Both diseases are characterized by loss of endothelial integrity, infiltration of inflammatory cells, accumulation of plasma lipoproteins, release of inflammatory cytokines, mesenchymal proliferation, extracellular matrix remodeling, and the growth of plaque lesions³. Unlike atherosclerosis, however, AVD leads to large, obstructive lesions containing significant matrix mineralization⁴. Randomized clinical trials using lipid lowering agents to halt AVD severity or progression have been disappointing^5,6. These findings underscore the conclusion that mechanisms of aortic valve disease are distinctly different than vascular atherosclerosis. As a consequence, there are currently no established molecular biomarkers specific to AVD progression and no molecular targets for AVD therapy⁷.

We previously determined in mice that the degree of inflammation in aortic valves directly correlates with the degree of calcification⁸. Diseased human valve cusps become progressively thickened, with increases in the presence of macrophages and cytokines such as interleukin-6 (IL-6) and tumor necrosis factor-alpha (TNF-α)^9-11. Both IL-6 and TNF-α can activate arterial endothelial cells, which is a phenotypic change characterized by upregulation of adhesion molecules such as ICAM-1 and VCAM-1, recruitment of leukocytes, and increased monolayer permeability^10,12,13. In atherosclerotic plaques, both IL-6 and TNF-α drive transcription through nuclear translocation of nuclear factor kappa b (NFκB), but their role in valve endothelium is unknown^14,15.

Fetal human aortic valve endothelial cells also express ICAM-1 and VCAM-1, which suggests that endothelial activation also occurs during valve development¹⁶. TNF-α, IL-6, and NFκB expression levels are elevated in the myocardium of children with congenital heart defects affecting valve development^17,18. NFκB inhibition during avian heart development leads to valvuloseptal defects and impaired outlet formation¹⁹. Leptin, a member of the IL-6 superfamily, was shown to induce EndMT in embryonic endocardium, in part through activating TGFβ²⁰. Akt has also been shown to drive embryonic and adult EndMT^21,22. Adult ovine aortic valve endothelium has been shown to undergo EndMT in vitro and in vivo^23-25. Collectively, these findings motivate the hypothesis that inflammatory signaling toward EndMT may be a developmentally conserved signaling pathway in valve endothelium, but a biological rationale is unclear.

In this study, we determine that both TNF-α and IL-6 induce EndMT in embryonic and adult valve endothelium via an Akt/NFκB dependent pathway. In embryonic valve endocardium, NFκB acted upstream of TGFβ to induce EndMT through Alk2/5. We found increasing levels of TNF-α and IL-6 receptor gene expression in embryonic valve primordia during valve morphogenesis. Interestingly, while TGFβ could drive EndMT in adult valve endothelium through Alk2/5, TGFβ was not required for NFκB induced EndMT. Finally, we identify evidence of EndMT in calcified human aortic valves. Taken together, these findings support a developmentally conserved mechanism of valve remodeling via NFκB induced EndMT that may be a previously unrecognized molecular signature of early AVD.

Methods

We employed two well-characterized valve endothelial cell culture systems to test inflammatory signaling on EndMT in vitro: Porcine aortic valve endothelial cells (PAVEC) and quail embryonic endocardial explants (QEE)^26-28. Additional details for each source are given in the supplement. PAVEC were isolated through collagenase digestion as previously described²⁹, with purity confirmed via non-detection of the mesenchymal ACTA2 gene expression (> 37 cycles via RT-PCR). PAVEC (< passage 5) were cultured as confluent surface monolayers on the surface of type 1 collagen hydrogels (1.5 mg/ml, 50,000 cells/cm²). Pre-transformed (HH14⁻ staged) quail endocardial explants (QEE) from the atrioventricular (AV) and outflow tract (OFT) valvulogenic regions were cultured as monolayer patches (without myocardium) on 1.5 mg/ml collagen gels as previously described²⁷. EndMT was quantified via real-time PCR, immunohistochemistry, and Western blotting; while cell invasion was quantified at 60 μm depth with brightfield microscopy (Supplemental Figure I). The protein extraction and Western blot protocols are described in detail in the online supplement. NFκB protein nuclear co-localization was quantified for 100 cells for each treatment using Metamorph 7.1 software (Molecular Devices, Sunnyvale, CA). RNA extraction, purification, and RT-PCR were performed according to the manufacturer’s protocols. Primers used for RT-PCR are listed in Supplemental Table I. For NFκB transfection, PAVEC cells were trypsinized and electroporated with plasmid encoding the RelA subunit of NFκB (Addgene plasmid 23255) using Neon transfection system (Invitrogen, Carlsbad, CA) and further cultured in 5% serum, antibiotic-free DMEM for 24 hours³⁰. The cells were then trypsinized, re-suspended in normal media (DMEM with 10% fetal bovine serum and 1% penicillin-streptomycin), and seeded onto gels. Invasion assay was performed as described above. Human diseased aortic valves were obtained from adults undergoing planned, non-elective valve replacement surgery at Robert Packer Hospital in Sayer, PA. All procedures were approved by Institutional Review Boards at Cornell and Robert Packer Hospital. Pediatric valves were obtained from Dr. Jonathan Chen, Cornell-Weill Medical School and NY Presbyterian Hospital. Results are expressed as mean ± S.E>M., n ≥ 3 independent cultures per treatment condition, with 5-6 explants pooled for each QEE sample. Data was analyzed with GraphPad Prism version 4 for Windows (GraphPad Software, San Diego, CA). Treatment effects were compared using Analysis of Variance (ANOVA) with Tukey’s post hoc paired tests, and data was transformed when necessary to obtain equal sample variances. Differences between means were considered significant at p ≤ 0.05. Expanded methods are provided in the Supplemental Information.

Results

TNF-α and IL-6 induce EndMT in adult valve endothelial cells

We first exposed 3D cultured adult PAVEC monolayers to doses of TNF-α or IL-6 and quantified changes in EndMT related gene expression, protein expression, and collagen matrix invasion. Both IL-6 and TNF-α in 3D culture induced loss of endothelial cell-cell contacts including PECAM-1 and VE-cadherin, acquisition of mesenchymal α-smooth muscle actin (α-SMA) and invasion into collagen matrix in a dose-dependent manner (Figure 1A,D,E,G; Supplemental Figures I-II). Genetic overexpression of NFκB alone was sufficient to cause 3D matrix invasion in PAVEC, supporting an NFκB-dependent EndMT mechanism (Figure 1A). Pro-EndMT related gene expression (ACTA2, Snail, TGFβ1) was significantly upregulated in response to NFκB overexpression (238±20.6, 15.0±1.1, and 4.2±0.5 fold respectively), TNF-α (14.3±4.8, 3.8±0.7, and 3.8±0.6 fold respectively, p<0.05), but not IL-6 (Figure 1B). EndMT-related protein expression (α-SMA, Snai1 increase and VE-cadherin decrease) was shown in PAVEC, but not porcine aortic endothelial cells (PAEC) in response to TNF-α (Figure 1G). TGFβ-1, however, induced EndMT protein expression in both PAVEC and PAEC (Figure 1G). Nuclear translocation of NFκB was significantly upregulated in response to either IL-6 or TNF-α (1.7±0.1 and 2.0±0.1 fold increase, p<0.05, Figure 1C,F; Supplemental Figure III). To control for cell number, we quantified PAVEC proliferation (via BrdU incorporation) and apoptosis (via TUNEL) in response to doses of TNF-α. PAVEC proliferation increased with TNF-α dose (up to 2.7±0.1 at 100 ng/mL, p<0.05, Supplemental Figure ID), while apoptosis remained unchanged (Supplemental Figure IE). Collectively, these results suggest that IL-6 and TNF-α induce EndMT and matrix invasion in adult PAVEC in vitro at least in part through NFκB, with TNF-α being a more potent EndMT inducer.

Figure 1.

Exposure to inflammatory cytokines induces mesenchymal transformation in adult endothelial cells. A.) Porcine aortic valve endothelial cell (PAVEC) mesenchymal transformation and invasion of the collagen matrix after 48 hour exposure to 100 ng/mL IL-6, 100 ng/mL TNF-α, or genetic overexpression of NFκB. B.) PAVEC EndMT-related gene expression after 48 hour exposure to 100 ng/mL IL-6 or TNF-α, or genetic overexpression of NFκB C.) PAVEC NFκB nuclear localization quantification. D.) Confocal images of control, +100 ng/mL IL-6, and +100 ng/mL TNF-α PAVEC at a 48 hour time point stained for CD31 (red), α-SMA (green), and DNA (blue). E.) Confocal images of control, +100 ng/mL IL-6, and +100 ng/mL TNF-α PAVEC at a 48 hour time point stained for VE-cadherin (red), α-SMA (green), and DNA (blue). F.) Confocal images of control, +100 ng/mL IL-6, and +100 ng/mL TNF-α PAVEC at a 48 hour time point stained for NFκB (green) and DNA (blue). G.) Western blots for PAVEC and porcine aortic valve endothelial cells (PAEC) exposed to 100 ng/mL TNF-α or TGFβ-1 for 48 hours. H.) Diseased human aortic valve endothelial cells co-expressing α-SMA and CD31 (on the fibrosa or outflow side and invaded, calcified nodule shown in inset, I.) Expanded frame of Figure 1H. J.) Nuclear translocation of NFκB in diseased human valve endothelial cells. Error bars show ±SEM, n ≥ 3 culture wells. Bars that do not share any letters are significantly different according to a one-way ANOVA with Tukey’s post test (p ≤ 0.05). For EndMT gene expression (Figure 1B), differences that are significantly different than the control according to an unpaired Student’s t-test are indicated with a *. Differences were considered significant at p ≤ 0.05. Scale bars = 25 μm (D, E) or 10 μm (F).

EndMT is present in calcified human aortic valves

In explanted human aortic valves with calcific lesions (Figure 1G-I) we identified a population of sub-endothelial cells co-expressing α-SMA and PECAM1, but not CD45 (Supplemental Figure IV). This suggests the cells were EndMT-derived, but not from an immune origin³¹. The number of invaded fibrosal endothelial cells was 11.13 ± 1.88 cells/40X field (SEM, n = 3 calcified adult valves). Endothelial cells located near the calcified nodule and on the fibrosa side showed decreased VE-cadherin expression (Supplemental Figure IVD). EndMT and potential EndMT-derived cells were only found in the fibrosa layer, distal to sites of calcified lesions. Furthermore, we found that many of these transformed and invaded cells also co-expressed nuclear NFκB (Figure 1H). Healthy pediatric human valves did not co-express α-SMA and VE-cadherin or nuclear NFκB (Supplemental Figures V and VI). These findings support that NFκB-mediated EndMT occurs in human calcific aortic valve disease, but is likely not involved in the mineralization process.

TNFα and IL-6 induce EndMT in embryonic valve endocardium

Similarly, we quantified the EndMT response characteristics of 3D cultured HH14⁻ pre-transformed QEE monolayers from OFT or AV regions to doses of TNF-α and IL-6. As in adult valve endothelium, both IL-6 and TNF-α induced loss of QH1, acquisition of α-SMA, and collagen matrix invasion (Figure 2A, D). Embryonic endocardial EndMT-related genes (ACTA2, Slug, and TGFβ3) were significantly upregulated in valvulogenic endocardium by both IL-6 and TNF-α (Figure 2B). PECAM1 gene expression was significantly increased in response to inflammatory cytokine (Figure 2A). PECAM1 has been shown to be an early marker of cardiovascular development, and the increase in CD31 in culture is likely related to embryonic endocardial differentiation³². Downstream NFκB protein nuclear translocation also occurred in response to either IL-6 or TNF-α (Figure 2C, E; Supplemental Figure VII). There was no difference in explant area (a control for cell number) with IL-6 or TNF-α stimulation in comparison to control (Supplemental Figure VIIIC,D). Collectively, these results suggest that both TNF-α and IL-6 induce EndMT in embryonic valve endocardial cells in vitro with similar potency. We further quantified the expression of inflammatory signaling related genes in the AV and OFT valve forming regions of embryonic hearts across the period of valve morphogenesis (Day 2-Day 10). We determined that ICAM1, NFKB1, IL6RA (coding for IL-6Rα), and TNFRSF1A (coding for TNFR1) gene expression were significantly upregulated in both valve regions as EndMT and valve remodeling progressed (Supplemental Figure IX). These results suggest that inflammatory signaling is active in valves during morphogenesis in vivo and directly cause EndMT in valve endocardium in vitro.

Figure 2.

Exposure to inflammatory cytokines induces mesenchymal transformation in embryonic endocardial cells. A.) Quail endocardial explant (QEE) mesenchymal transformation and invasion of the collagen matrix after 48 hour exposure to 100 ng/mL IL-6 or TNF-α. B.) QEE EndMT-related gene expression after 48 hour exposure to 100 ng/mL IL-6 or TNF-α. C.) QEE NFκB nuclear localization quantification. D.) Confocal images of control, +100 ng/mL IL-6, and +100 ng/mL TNF-α QEE at a 48 hour time point stained for QH1 (red), α-SMA (green), and DNA (blue). E.) Confocal images of control, +100 ng/mL IL-6, and +100 ng/mL TNF-α QEE at a 48 hour time point stained for NFκB (green) and DNA (blue). Error bars show ±SEM, n ≥ 3 culture wells with pooled explants. Bars that do not share any letters are significantly different according to a one-way ANOVA with Tukey’s post test (p ≤ 0.05). Scale bars = 20 μm.

Inflammatory EndMT acts through an Akt/NFκB pathway present in both adult and embryonic valve endothelium

We next sought to identify the signaling pathway by which inflammatory cytokines induced EndMT through NFκB. Previous studies have shown that TNF-α activates NFκB-mediated transcription through Akt and/or MEK1^33-37. IL-6 activates STAT3 most potently, but there is also evidence of IL-6-induced NFκB activation through PI3K and Akt^38-41. Utilizing our 3D culture system, molecular inhibition of either Akt (via 5 μm Akt inhibitor XI) or MEK1 (via 25 μm PD98059) significantly reduced TNF-α-induced NFκB expression and nuclear translocation in PAVEC, but only Akt inhibition resulted in complete knockdown (Figures 3C,D). Akt inhibition also completely blocked IL-6-induced NFκB gene expression and protein nuclear translocation, while STAT3 inhibition (via 5 μm PpYLKTK STAT3 inhibitor) had no effect (Supplemental Figures XC,D). Inhibition of Akt completely blocked TNF-α induced inflammatory receptor activation (VCAM1, ICAM1), EndMT associated gene expression (ACTA2, Snail, TGFβ1), and subsequent matrix invasion by PAVEC (Figure 3A-C). MEK1 blockade also inhibited TNF-α-induced inflammatory receptor activation, EndMT gene expression, and subsequent invasion, but to a lesser degree than Akt inhibition. In a similar fashion, Akt inhibition completely blocked IL-6-induced upregulation of inflammatory receptors, EndMT gene expression, and subsequent matrix invasion (Supplemental Figures XA-C). STAT3 inhibition, on the other hand, had no effect on any of these responses from IL-6.

Figure 3.

Inflammatory cytokines induce EndMT through Akt/MAPK/NFκB in PAVEC. A.) Cell invasion after a 48 hour to exposure to 100 ng/mL TNF-α or 100 ng/mL TNF-α with 5 μm Akt inhibitor XI or 25 μm PD98059 MEK1 inhibitor. B.) EndMT-related gene expression after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. C.) Inflammatory activation-related genes after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. D.) PAVEC NFκB nuclear localization quantification after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. Error bars show ±SEM, n ≥ 3 culture wells. Bars that do not share any letters are significantly different according to a one-way ANOVA with Tukey’s post test (p ≤ 0.05).

Identical experiments were conducted on avian pre-EndMT valvulogenic endocardial monolayers. As in adult valve endothelium, molecular inhibition of Akt or MEK1 both blocked TNF-α-induced NFκB gene expression and nuclear translocation. Akt inhibition completely blocked NFκB while MEK1 inhibition only partially blocked NFκB (Figure 4C,D). IL-6-induced NFκB gene expression and protein nuclear translocation was also blocked with Akt inhibition, but not STAT3 inhibition (Supplemental Figures XIC,D). Akt or MEK1 inhibition also blocked inflammatory receptor (VCAM1, ICAM1) upregulation, EndMT-related gene expression (ACTA2, Slug, TGFβ3), and collagen invasion by QEE cells (Supplemental Figures XIA-C). As a positive control, Akt1 gene expression was blocked by Akt inhibition in both PAVEC and QEE subjected to TNF-α (Supplemental Figures XII and XIII). Collectively, these results suggest that inflammatory cytokines initiate EndMT at the gene, protein, and function levels through an Akt/NFκB pathway that is conserved in both adult valve endothelial and embryonic valvulogenic endocardium. TNF-α was a more potent inducer of EndMT and resulted in more consistent in pathway responses in both cell types when compared with IL-6. IL-6 may also induce EndMT genes more strongly at an earlier or later time point.

Figure 4.

Inflammatory cytokines induce EndMT through Akt/MAPK/NFκB in QEE. A.) Cell invasion after a 48 hour exposure to 100 ng/mL TNF-α or 100 ng/mL TNF-α with 5 μm Akt inhibitor XI or 25 μm PD98059 MEK1 inhibitor. B.) EndMT-related gene expression after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. C.) Inflammatory activation-related genes after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. D.) QEE NFκB nuclear localization quantification after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. Error bars show ±SEM, n ≥ 3 culture wells with pooled explants. Bars that do not share any letters are significantly different according to a one-way ANOVA with Tukey’s post test (p ≤ 0.05).

TGFβ signaling through Alk2/5 is required for inflammation-induced embryonic EndMT, but not for adult EndMT

Previous studies have identified a canonical TGFβ signaling pathway that induces EndMT and is isoform specific: the TGFβ1 isoform in adult VEC, while EndMT invasion is mediated by TGF–3 in avian embryonic valve endocardium^23,42,43. We therefore tested whether TGFβ signaling was involved in inflammatory-induced EndMT. As a positive control, exogenous TGF–3 (100 ng/ml) induced EndMT matrix invasion in embryonic valve endocardium (Figure 5A), while exogenous TGF–1 induced EndMT in PAVEC (Figure 6A, Supplemental Figure XIV). TNF-α induced TGFB1 gene expression in PAVEC and TGFB3 in QEE (3.79±0.57 and 12.03±1.64, Figures 1B and 2B). IL-6 induced TGFB3 only in QEE (7.61±0.53, Figure 2B). Previous studies demonstrate that TGFβ type I activin receptor-like kinases 2 and 5 (Alk2/5) mediate TGFβ-induced EndMT^44,45. Molecular inhibition of Alk2/5 signaling (via 10 μm SB 431542) in QEE reduced EndMT-related gene expression (Figure 5B, Supplemental Figure XVB), cell invasion induced by either TNF-α or IL-6 (Figure 5A, Supplemental Figure XVA), and Smad2/3 physphorylation and nuclear localization (Figure 5E), but did not affect inflammatory activation-related gene expression (Figure 5C, Supplemental Figure XVC) or NFκB protein nuclear localization (Figure 5D, Supplemental Figure XVD). Together, these results support that inflammatory cytokine-induced EndMT in embryonic valves acts at least in part by canonical TGFβ-Alk-Smad signaling that is downstream of NFκB.

Figure 5.

Embryonic endocardial monolayers co-opt TGFβ in inflammatory-EndMT signaling. A.) Cell invasion after a 48 hour exposure to 100 ng/mL TGFβ-3, 100 ng/mL TNF-α or 100 ng/mL TNF-α with 10 μm SB 431542 ALK5 inhibitor. B.) EndMT-related gene expression after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. C.) Inflammatory activation-related genes after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. D.) QEE NFκB nuclear localization quantification after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. E.) pSMAD2/3 nuclear localization following a 48 hour exposure to 100 ng/mL TGFβ-3, 100 ng/mL TNF-α, or 100 ng/mL TNF-α with 10 μm SB 431542 ALK5 inhibitor. Staining shows p-SMAD2/3 (green) or DNA (blue). Error bars show ±SEM, n ≥ 3 culture wells with pooled explants. Bars that do not share any letters are significantly different according to a one-way ANOVA with Tukey’s post test (p ≤ 0.05).

Figure 6.

Adult cells do not require TGFβ in inflammatory-EndMT signaling. A.) Cell invasion after a 48 hour exposure to 100 ng/mL TGFβ-1, 100 ng/mL TNF-α or 100 ng/mL TNF-α with 10 μm SB 431542 ALK5 inhibitor. B.) EndMT-related gene expression after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. C.) Inflammatory activation-related genes after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. D.) PAVEC NFκB nuclear localization quantification after a 48 hour exposure to 100 ng/mL TNF-α with inhibitors. E.) pSMAD2/3 nuclear localization following a 48 hour exposure to 1 ng/mL TGFβ-1, 100 ng/mL TNF-α, or 100 ng/mL TNF-α with 10 μm SB 431542 ALK5 inhibitor. Staining shows pSMAD2/3 (green) or DNA (blue). F.) Western blots for PAVEC or porcine aortic valve endothelial cells (PAEC) exposed to 100 ng/mL TNF-α or TGFβ-1 for 48 hours. Error bars show ±SEM, n ≥ 3 culture wells with monolayers. Bars that do not share any letters are significantly different according to a one-way ANOVA with Tukey’s post test (p ≤ 0.05).

In contrast to embryonic valve endocardium, Alk2/5 inhibition in adult valve endothelial cells (PAVEC) treated with TNF-α significantly decreased Smad2/3 phosphorylation and nuclear translocation (Figure 6E), but did not result in decreased cell invasion (Figure 6A), did not effect EndMT-related or inflammatory activation-related gene expression (Figures 6B,C), and did not inhibit NFκB nuclear localization (Figure 6D). Interestingly, endothelial cells from the porcine aorta show increased pSmad2 in response to TGFβ-1, but not to TNF-α treatment, suggesting a unique signaling behavior of PAVEC in response to TNF-α (Figure 6F). These results suggest that Alk2/5 dependent TGFβ signaling is not downstream of TNF-α-induced EndMT in adult valve endothelium. Therefore, while both embryonic and adult valve endothelium activate an EndMT program in response to TNF-α through Akt/NFκB, TGFβ signaling is required and downstream in embryonic, but not required in adult valve endothelium. Adult cells may bypass TGFβ signaling because activated NFκB has been shown to stabilize Snail, an EndMT-regulating transcription factor, in adult cells⁴⁶. In mouse cardiac endothelial cells, treatment with SB431542, but not PD98059 completely blocked TGFβ-2-induced EndMT⁴⁷.

Discussion

The molecular and cellular events that initiate and propagate aortic valve disease are poorly understood and understudied, particularly with the respect to endothelial dysfunction. While the occurrence of EndMT has been previously noted in adult valve endothelium, a biological role remains unclear. EndMT studies in embryonic valves have identified over 100 different regulatory genes, including members of the TGFβ, BMP, and VEGF signaling pathways and matrix proteins such as periostin and versican^48-51. Misexpression and mutation of many of these genes have been identified in diseased aortic valves³. Our results identify a novel, inflammatory cytokine-induced EndMT pathway in aortic valve endothelium through Akt/NFκB, and further support that this pathway is conserved during embryonic valve development. These findings collectively suggest that an inflammatory reactivation of embryonic-like EndMT may be a mechanism of early aortic valve dysfunction.

Identifying mechanisms conserved between embryonic and adult valve endothelium remains challenging because there is currently no species for which both embryonic and adult valve endothelial cell cultures have been obtained. Porcine aortic valve endothelial cells are the most characterized and studied adult valve endothelial population, and can be obtained from animals without disease^29,52-54. Human aortic valve endothelial cells have been obtained from recipient hearts from cardiac transplantation surgeries or rejected valve donors. While no gross calcification may be present, the valve endothelium exhibits non-negligible inflammatory activation^55-58 Paruchuri et al. identified clonally isolated sub-populations of human pulmonary VEC that were capable of EndMT in response to TGFβ, which supports our findings in pure whole VEC populations⁵⁹. Holliday et al. recently showed that in vitro cultured human aortic VEC isolated from heart transplant surgeries express endothelial phenotype markers and elevated α-SMA, which is in contrast to in vivo expression patterns in non-diseased human and porcine aortic valves^16,60-62. These co-expression findings suggest that human aortic valve endothelium from these patients may have an increased propensity or capacity for EndMT⁶⁰. Furthermore, several studies demonstrate that inflammatory endothelial activation and calcific degeneration occurs preferentially on the fibrosa side of the valve^4,63,64. Our analysis of calcified human aortic valves found invaded cells co-expressing α-SMA and CD31 on the fibrosa surface, suggesting that EndMT correlates with side-specific propagation of AVD⁶⁵. In situ mRNA profiling of valve endothelial cells determined hundreds of genes that are differently expressed between sides, with many further modified in experimental hypercholesteremia^52,53. The inflow or ventricularis surface is exposed to pulsatile unidirectional shear stress, while the outflow fibrosa surface experiences nearly oscillatory shear stress⁶⁵. When either endothelial surface is exposed to opposite-sided flow profiles, they similarly upregulate inflammatory receptors⁶⁶. Recently, Wu and colleagues demonstrated that both sides of valve endothelium have the same genetic origin⁶⁷. Our results support and extend these findings to suggest that fibrosa-sided environmental conditions are more conducive to inflammatory activation and EndMT, but additional studies on side-specific endothelial populations are warranted.

Cellular and molecular analysis of embryonic valve formation is almost exclusively conducted with avian and mouse models. Though there are some differences in cardiac anatomy between avians and mammals, the morphogenesis, remodeling, and final anatomy of the clinically important left side of the heart is more similar between chicks and humans than in mice^68-71. Biological assays with avian and mouse valve progenitors demonstrate identical molecular mechanisms and participants, though which isoforms used may differ^72,73. Importantly, chick embryos are inexpensive, can be obtained without additional animal sacrifice, and are very tolerant of experimentation. Our findings of progressive upregulation of inflammatory receptors in valve forming in embryonic chicks is supported by observations in fetal human aortic valves¹⁶. Our result that EndMT and matrix invasion occurs at least through TGF–1 in adult porcine cells and TGF–3 in embryonic quail is supported by several studies^23,43. Therefore, we have confidence that the general molecular regulation we identify is likely conserved across age, but we cannot definitively rule out species differences in the isoforms utilized.

The downstream fate of EndMT-derived mesenchyme in adult valves is of great interest, but was not a focus of this study. Bischoff and Aikawa recently proposed that EndMT creates a type of progenitor-like cell that renews the valve interstitial cell (VIC) population and maintains tissue homeostasis⁷⁴. Both embryonic and adult valvular mesenchymal cells have the potential for osteogenic differentiation^75,76, but whether these cells are directly EndMT-derived or from subsequent differentiation processes is not yet known. TNF-α and NFκB have been linked to the mesenchymal cell molecular and cellular mechanisms that mediate valvular calcification^9,77. TNF-α expression and nuclear localization of NFκB is present in stenotic human aortic valves⁷⁸, and TNF-α treatment of valve interstitial cells in 2D culture induces calcium nodule formation and expression of runx2 independent of osteogenic media^9,79,80. We found EndMT derived subendothelial cells in calcified human aortic valves, but were always located distal to calcific lesions. This was not unexpected, as valve calcifications are associated with endothelial loss⁸¹. Taken together, these results support that inflammatory EndMT is a component of early stage AVD, but how these progeny contribute to and/or regulate downstream calcific progression is still unclear.

In conclusion, our results establish an inflammatory mechanism of early aortic valve pathogenesis that is, in part, developmentally conserved. Age dependent use of canonical TGFβ signaling may be an important mode of inflammatory disease pathogenesis distinct from potentially healthy tissue formation and remodeling. Selective molecular inhibition of the AKT/NFκB pathway in valve endothelium may therefore be a potential strategy for impacting early stage aortic valve disease.