PROX1 Inhibits PDGF-B Expression to Prevent Myxomatous Degeneration of Heart Valves

Abstract

Background: Cardiac valve disease is observed in 2.5% of the general population and 10% of the elderly people. Effective pharmacological treatments are currently not available, and patients with severe cardiac valve disease require surgery. PROX1 (prospero-related homeobox transcription factor 1) and FOXC2 (Forkhead box C2 transcription factor) are transcription factors that are required for the development of lymphatic and venous valves. We found that PROX1 and FOXC2 are expressed in a subset of valvular endothelial cells (VECs) that are located on the downstream (fibrosa) side of cardiac valves. Whether PROX1 and FOXC2 regulate cardiac valve development and disease is not known.

Methods: We used histology, electron microscopy, and echocardiography to investigate the structure and functioning of heart valves from Prox1^ΔVEC mice in which Prox1 was conditionally deleted from VECs. Isolated valve endothelial cells and valve interstitial cells were used to identify the molecular mechanisms in vitro, which were tested in vivo by RNAScope, additional mouse models, and pharmacological approaches. The significance of our findings was tested by evaluation of human samples of mitral valve prolapse and aortic valve insufficiency.

Results: Histological analysis revealed that the aortic and mitral valves of Prox1^ΔVEC mice become progressively thick and myxomatous. Echocardiography revealed that the aortic valves of Prox1^ΔVEC mice are stenotic. FOXC2 was downregulated and PDGF-B (platelet-derived growth factor-B) was upregulated in the VECs of Prox1^ΔVEC mice. Conditional knockdown of FOXC2 and conditional overexpression of PDGF-B in VECs recapitulated the phenotype of Prox1^ΔVEC mice. PDGF-B was also increased in mice lacking FOXC2 and in human mitral valve prolapse and insufficient aortic valve samples. Pharmacological inhibition of PDGF-B signaling with imatinib partially ameliorated the valve defects of Prox1^ΔVEC mice.

Conclusions: PROX1 antagonizes PDGF-B signaling partially via FOXC2 to maintain the extracellular matrix composition and prevent myxomatous degeneration of cardiac valves.

Keywords: aortic valve; endothelial cells; extracellular matrix; mitral valve; prolapse.

Figure 1.

Deletion of Prox1 (prospero-related homeobox transcription factor 1) from valvular endothelial cells (VECs) results in progressive myxomatous valve degeneration and aortic valve (AV) stenosis. A, Representative Movat Pentachrome–stained histological images of the AVs and mitral valves (MVs) of control and Prox1^ΔVEC mice at various ages. Glycosaminoglycans, collagen, and elastin are stained in blue, red, and black, respectively. B, Quantification of AV and MV thickness. For each of the mice, 4 to 7 sections at 100–120 µm intervals were analyzed. The area of the valve leaflets (2 leaflets from base to tip) was measured for every section and their average was used as a measure of valve thickness. Each dot represents an individual mouse, 1-month-old mice: n=8 controls and n=10 Prox1^ΔVEC; 3-month-old mice: n=8 controls and n=7 Prox1^ΔVEC; 6-month-old mice: n=23 controls and Prox1^ΔVEC; 12-month-old mice: n=8 controls and n=11 Prox1^ΔVEC. Data is represented as mean±SEM. C, Echocardiography of 12-month-old control and Prox1^ΔVEC mice. Left, representative echocardiography images show decreased cusp separation (double-headed arrows in the top) and increased aortic peak velocity (bottom) in Prox1^ΔVEC mice. Right, Bar charts of cusp separation and AV peak velocity. No obvious defects were observed in left ventricle (LV) functions in the Prox1^ΔVEC mice as indicated by normal ejection fraction and fractional shortening. n=5 control and n=9 Prox1^ΔVEC mice. Data is represented as mean±SD. Statistical significance was assessed as described in Table S1. A indicates aorta. **P<0.01; *P<0.05; ****P<0.0001. Ctrl indicates control.

Figure 2.

Loss of PROX1 (Prospero-related homeobox transcription factor 1) from aortic valvular endothelial cells (VECs) results in valve thickening, damaged endothelium, thrombus formation, and downregulation of FOXC2 (Forkhead box C2 transcription factor) in VECs. A, Representative scanning electron microscope (SEM) images of the tri-cuspid aortic valve leaflets of Prox1^f/f and Prox1^ΔVEC mice. SEM images show that Prox1^ΔVEC mice have thicker aortic valves, disrupted endothelial layer (arrows in the middle row) with infiltration of platelet-like cells (arrows in the bottom row). n=5 for the Prox1^f/f and n=6 for Prox1^ΔVEC mice. B, Representative Movat Pentachrome–stained histological images demonstrated thrombus-like structure in the aortic valve of a Prox1^ΔVEC mouse (arrow). The graph shows the quantification of the percentage of mice with thrombi in the aortic valves. Valves were analyzed by either SEM or Movat Pentachrome-staining of sections. n=10 control mice (5 by SEM, 5 by staining) and n=16 Prox1^ΔVEC mice (6 by SEM and 10 by staining). C, VWF (von Willebrand Factor) expression was increased in the downstream side VECs of Prox1^ΔVEC mice (arrows). n=5 for control mice; n=7 for Prox1^ΔVEC mice. D, FOXC2⁺ VECs (arrows) were reduced in Prox1^ΔVEC mice. n=4 for control mice; n=5 for Prox1^ΔVEC mice. Data are represented as mean±SEM. Statistical significance was assessed as described in Table S1. Each dot represents the average of 3-5 sections from a mouse. *P<0.05; ****P<0.0001. CD31 indicates cluster of differentiation 31; Ctrl, control; and DAPI, 4′,6-diamidino-2-phenylindole.

Figure 3.

Increased proteoglycan expression and disrupted collagen and elastin fibers are observed in the aortic valves of Prox1^ΔVEC mice. A through C, Representative immunofluorescence images and semiquantitative measurements of aggrecan (A), versican (B), and cleaved versican (anti-DPEAAE peptide in C) expression in the aortic valve sections from 6-month-old control and Prox1^ΔVEC mice. Aggrecan and versican were significantly increased and cleaved versican was significantly reduced in Prox1^ΔVEC mice. Each dot in the graph represents the average of 3–5 sections from a mouse. n=8 control and n=13 Prox1^ΔVEC in A; n=6 control and n=10 Prox1^ΔVEC in B; n=7 control and n=8 Prox1^ΔVEC in C. D, Aortic and mitral valve tissues were pooled together from five 3-month-old control or 5 Prox1^ΔVEC mice for generating 1 pooled sample. RNA was extracted from the pooled samples and quantitative real-time polymerase chain reaction (qRT-PCR) was performed to quantify the expression of selected genes. Each dot in the graph represents data from one pooled sample. The dotted line represents the mean expression in control samples. n=3 control and n=3 Prox1^ΔVEC pooled samples. qRT-PCR showed that 3 out of the 4-candidate proteoglycan degrading enzymes had a trend towards reduced expression in Prox1^ΔVEC samples. E, Collagen binding peptide staining followed by semiquantitative measurement suggested comparable expression levels in the aortic valves of 6-month-old control and Prox1^ΔVEC mice. However, the collagen fibers in the aortic valves of Prox1^ΔVEC mice appeared to be disrupted (arrowhead). n=8 control and n=7 Prox1^ΔVEC mice (F) Representative transmission electron microscope (TEM) images show valvular endothelial cells (VECs) (pseudocolored in brown) and valvular interstitial cell–like cells (pseudocolored in blue) in the aortic valves of 12-month-old mice. Well-organized collagen fibers (black arrows) were seen in control mice. In contrast, collagen fibers appeared to be disrupted in Prox1^ΔVEC mice (red arrows) with gaps in between the bundles (red asterisk). G, Resorcin-Fuchsin staining showed normal elastin expression in the upstream side of aortic valves (green arrowhead) in 6-month-old control mice. In contrast, normal elastin distribution was reduced and abnormal distribution within and in the downstream side of aortic valves (orange arrowhead) was observed in 6-month-old Prox1^ΔVEC mice. The graph shows normal and abnormal elastin expression in control and Prox1^ΔVEC mice at various ages. Data are represented as the percentage of total mice. 1-month-old mice: n=8 controls and n=9 Prox1^ΔVEC; 3-month-old mice: n=7 controls and n=8 Prox1^ΔVEC; 6-month-old mice: n=11 controls and Prox1^ΔVEC; 12-month-old mice: n= 5 controls and n=8 Prox1^ΔVEC. Data are represented as Mean ±SEM in (A, B, C, and E). Data are represented as mean±SD in D. Statistical significance was assessed as described in Table S1. *P<0.05, ****P<0.0001. CD31 indicates cluster of differentiation 31; Ctrl, control; and DAPI, 4′,6-diamidino-2-phenylindole.

Figure 4.

Hyperactivation of PDGF-B (platelet-derived growth factor-B)/PDGFRβ (PDGF receptor β) signaling causes myxomatous degeneration of aortic valve. A, Aortic and mitral valve tissues were pooled together from five 3-month-old control or 5 Prox1^ΔVEC mice for generating 1 pooled sample. Following RNA extraction and cDNA synthesis from the pooled samples quantitative real-time polymerase chain reaction (qRT-PCR) was performed for the expression of selected cytokines and growth factors. Each dot represents data from 1 pooled sample. n=2 control and n=2 or 3 Prox1^ΔVEC pooled samples. The dotted line represents the mean expression in control samples. Data are represented as mean±SD. B, Representative RNAscope images for Pdgfb expression (red dots) in the aortic valves of 3-month-old control and Prox1^ΔVEC mice. The lower parts are enlarged images of the boxed areas in the upper parts. Quantification of the RNAscope results showed that Prox1^ΔVEC mice had higher Pdgfb expression. Each dot represents the average from 3 sections of a mouse valve. n=4 mice per genotype. C, Representative Movat Pentachrome–stained images of aortic valves from 6-month-old control and Prox1-2A-Cre;R26^+/LSL-PDGFB mice. Thicker aortic valves were observed in Prox1-2A-Cre;R26^+/LSL-PDGFB mice when compared with control valves. n=6 control and n=9 Prox1-2A-Cre;R26^+/LSL-PDGFB mice. D, Representative Movat Pentachrome stain images of the aortic valves of 6-month-old control and Tie2-Cre;Pdgfrb^+/D849V mice. Tie2-Cre;Pdgfrb^+/D849V mice had enlarged aortic valves when compared with control valves. n=5 for control and Tie2-Cre;Pdgfrb^+/D849V mice. The area of valve leaflets was measured from 4 to 6 stained sections per valve. The average of these areas is represented as the thickness of valve in that mouse. Statistical significance was assessed as described in Table S1. *P<0.05. **P<0.01. CD31 indicates cluster of differentiation 31; Ctrl, control; and DAPI, 4′,6-diamidino-2-phenylindole.

Figure 5.

Knock down of FOXC2 (Forkhead box C2 transcription factor) results in enlarged aortic valves, and over expression of FOXC2 rescues the aortic valve defects of Prox1^ΔVEC mice. A, Representative images of Movat Pentachrome staining that was performed using aortic valves of 12-month-old control and Foxc2^ΔVEC mice with graph showing the semiquantitative measurement of valve thickness. n=8 controls and n=8 Foxc2^ΔVEC mice. B, Representative images of Resorcin-Fuchsin staining that was performed using the aortic valves of 12-month-old control and Foxc2^ΔVEC mice. Green arrowhead shows the normal expression of elastin on the upstream side of valves. The orange arrowheads show abnormal elastin expression within and in the downstream side of valves. The graph shows the distribution of normal and abnormal elastin expression in control and mutant mice. n=11 controls and n=7 Foxc2^ΔVEC mice. C, Representative immunofluorescence images for versican expression in the aortic valves of 12-month-old control and Foxc2^ΔVEC mice, followed by semiquantitative measurement. n=10 control and n=7 Foxc2^ΔVEC mice. D, Representative RNAscope images for Pdgfb expression (red dots) in the aortic valves of 6-month-old control and Foxc2^ΔVEC mice (arrows indicate Pdgfb⁺ valvular endothelial cells [VECs]). Upon quantification, Pdgfb expression appeared to be increased in Foxc2^ΔVEC mice (mean H-score=16.0 in control versus 31.5 in Foxc2^ΔVEC). Each dot represents the average of 3 sections from a mouse valve. n=3 control and n=2 Foxc2^ΔVEC mice. E through H, Over expression of FOXC2 in Prox1^ΔVEC mice (Prox1^ΔVEC;Foxc2^GOF) ameliorates valve thickness (E), abnormal elastin distribution (F), increased proteoglycan expression (G) and Pdgfb expression (H). N=13 control, n=11 Prox1^ΔVEC and n=7 Prox1^ΔVEC;Foxc2^ΔGOF mice in (E); n=13 control, n=11 Prox1^ΔVEC and n=9 Prox1^ΔVEC;Foxc2^ΔGOF mice in (F); n=8 control, n=12 Prox1^ΔVEC and n=7 Prox1^ΔVEC;Foxc2^ΔGOF mice in (G). n=3 for all 3 genotypes in (H). Mice were 6-month-old in (E–G) and 3-month-old in (H). Data are represented as mean±SEM. Statistical significance was assessed as described in Table S1. Each dot represents the average of 3 to 5 sections from a mouse. *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001. Ctrl indicates control.

Figure 6.

PDGF-B (platelet-derived growth factor-B)/PDGFRβ (PDGF receptor β) signaling promotes proteoglycan expression in valvular interstitial cells (VICs) in a SOX9 (SRY-related HMG-box 9)-dependent manner. A, Representative immunohistochemistry images and quantification showing increased number of SOX9⁺ cells in the aortic valves of 6-month-old Prox1^ΔVEC mice when compared with control littermates. Arrows indicate SOX9⁺ valvular endothelial cells (VECs) and arrowheads indicate SOX9⁺ VICs. n=7 control and n=6 Prox1^ΔVEC mice. B, SOX9 expression and quantification in the valves of 6-month-old control and Tie2-Cre;Pdgfrb^+/D849V mice. Arrowheads indicate SOX9⁺ VICs. n=7 for control and n=6 for Tie2-Cre;Pdgfrb^+/D849V mice. C, Representative Western blotting images using lysates prepared from porcine VICs (pVICs) transfected with siCtrl (control siRNA) or siSOX9 (siRNA targeting SOX9) and treated with the indicated amount of PDGFB for 24 hours. D, Quantification of the Western blotting data shows that PDGF-B increases the expressions of aggrecan, SOX9 (SRY-related HMG-box 9), and pSOX9 (phosphorylated SOX9) and that PDGF-B increases the expression of aggrecan in a SOX9-dependent manner. Protein expression was first normalized to actin (internal control) and then compared with expression in pVICs without PDGF-B treatment. Each dots represents an independent experiment. n=4 replicates. E and F, qRT-PCR for the expressions of candidate ECM (extracellular matrix) degradation enzymes. RNA was collected from pVICs treated with vehicle or 100 ng/ml of PDGF-B for 24 hours in the presence (E) or absence (F) of siSox9 siRNA. Dotted line represents expression in vehicle-treated cells. Each dots represents an independent experiment. n=6 replicates in (E) and n=5 in (F). Each dot represents the average from 4 to 6 staining sections per mouse valve (B and D). Data is represented as mean±SEM in (A and B) and mean±SD in (E and F). Each dots represent an independent experiment. Statistical significance was assessed as described in Table S1. *P<0.05, **P<0.01, ***P<0.001. Veh indicates vehicle.

Figure 7.

PDGF-B (platelet-derived growth factor-B) and SOX9 (SRY-related HMG-box 9) are increased in the myxomatous valves from human patients. A and B, Protein and RNA samples were obtained from the mitral valves of mitral valve prolapse patients (n=77) who underwent mitral valve replacement surgery. Genes and protein expressions were quantified by quantitative real-time polymerase chain reaction and ELISA respectively. Pearson correlation method was used for calculating r and p. C, Aortic valve leaflets from 3 patients with aortic valve insufficiency (AVI) were histologically analyzed. The relatively normal-looking leaflet was used as the internal control. The normal and pathological leaflets were analyzed using Movat Pentachrome stain and by immunohistochemistry for PDGF-B, SOX9, and PROX1. White arrows indicate VECs and yellow arrows indicate PROX1⁺ cells of unknown identity in the valve interstitium. D through F, Expression of PDGF-B and SOX9 was increased in the diseased leaflet, but no obvious difference was observed in the number of PROX1⁺ cells. Data is represented as mean±SEM. Quantification was performed by averaging the data from 5 sections. PDGF-B staining was quantified as a percentage of PDGFB⁺CD31⁺cells to total CD31⁺ cells. Statistical significance was assessed as described in Table S1. CD31+ indicates cluster of differentiation 31⁺; and PROX1, Prospero-related homeobox transcription factor 1.

Figure 8.

Receptor tyrosine kinase (RTK) inhibitor imatinib (Imb) partially rescues aortic valve function in Prox1^ΔVEC mice. A, Schematic of the experimental design. Imatinib (50 mg/kg-body weight/day) or PBS (vehicle [Veh]) was orally administrated to control and Prox1^ΔVEC mice daily for 24 weeks starting from when they were 1-month-old. At the end of the treatment aortic valves were evaluated by echocardiography and histology. B, Echocardiography showed that Imb enhanced cusp separation and lowered aortic peak velocity in Prox1^ΔVEC mice when compared with untreated Prox1^ΔVEC mice. There were no significant differences in left ventricle (LV) functions (ejection fraction and fractional shortening). n=7 Veh-treated controls, n= 8 Veh-treated Prox1^ΔVEC mice, n=5 Imb-treated controls and n=5 Imb-treated Prox1^ΔVEC mice. C, Representative Movat Pentachrome–stained sections from the aortic valves of control and Prox1^ΔVEC mice treated with Veh or Imb. The graphs show the quantification of valve thickness and the presence of thrombi. Arrow points to a thrombus-like structure in an untreated Prox1^ΔVEC mouse. The number of thrombus-like structures was reduced in Prox1^ΔVEC mice treated with Imb. The area of valve leaflets was measured from 4 to 6 sections per valve. The average of these areas is represented as the thickness of valve in that mouse. Each dot represents an individual mouse. Valve thickness in Prox1^ΔVEC mice was not reduced by Imb treatment. n=12 Veh-treated controls, n=10 Veh-treated Prox1^ΔVEC mice, n=5 Imb-treated controls and n=5 Imb-treated Prox1^ΔVEC mice. D, Abnormal elastin distribution was rescued in the aortic valves of Imb-treated Prox1^ΔVEC mice. n=10 Veh-treated control, n=10 Veh-treated Prox1^ΔVEC, n=5 Imb-treated control and n=5 Imb-treated Prox1^ΔVEC mice. E, Immunohistochemistry and quantification indicated that abnormal versican production in Prox1^ΔVEC mice was inhibited by Imb treatment. n=6 Veh-treated controls, n=10 Veh-treated Prox1^ΔVEC mice, n=4 Imb-treated controls and n=5 Imb-treated Prox1^ΔVEC mice. Data are represented in Mean ±SD in (B) and Mean ±SEM in (C and E). Statistical significance was assessed as described in Table S1. Each dot represents the average of 3–5 sections from a mouse. *P<0.05; **P<0.01, ***P<0.001, ****P<0.0001.