Loss of Endothelial Furin Leads to Cardiac Malformation and Early Postnatal Death

Abstract

In mammals, seven proprotein convertases (PCs) cleave secretory proteins after basic residues, and four of them are called furin-like PCs: furin, PC5, PACE4, and PC7. In vitro, they share many substrates. However, furin is essential during development since deficient embryos die at embryonic day 11 and exhibit multiple developmental defects, particularly defects related to the function of endothelial cells. To define the role of furin in endothelial cells, an endothelial cell-specific knockout (ecKO) of the Furin gene was generated. Newborns die shortly after birth, indicating that furin is essential in these cells. Magnetic resonance imaging revealed that ecKO embryos exhibit ventricular septal defects (VSD) and/or valve malformations. In addition, primary cultures of wild-type and ecKO lung endothelial cells revealed that ecKO cells are unable to grow. Growth was efficiently rescued by extracellular soluble furin. Analysis of the processing of precursors of endothelin-1 (ET-1), adrenomedullin (Adm), transforming growth factor β1 (TGF-β1), and bone morphogenetic protein 4 (BMP4) confirmed that ET-1, Adm, and TGF-β1 are in vivo substrates of endothelial furin. Mature ET-1 and BMP4 forms were reduced by ∼90% in ecKO purified endothelial cells from lungs.

INTRODUCTION

The mammalian proprotein convertases (PCs) form a family of 9 secretory serine proteinases exhibiting similarities to bacterial subtilisin (40, 41, 43). Seven basic amino acid-specific PCs are related to yeast kexin: PC1 (also known as PC3), PC2, furin, PC4, PC5 (also known as PC6), PACE4, and PC7. These cleave after single or paired basic amino acids within the motif (R/K)-X_n-(R/K)↓, where X_n is equal to 0, 2, 4, or 6 variable amino acids separating the two canonical basic residues required for cleavage recognition. The two last members of the PC family are known for their key role in cholesterol homeostasis (41). SKI-1/S1P cleaves membrane-anchored transcription factors after nonbasic amino acids, and PCSK9 triggers the degradation of the low-density lipoprotein receptor independently of its enzyme activity by targeting it to lysosomal degradation (41).

The membrane-bound furin cycles from the cell surface back to the trans-Golgi network (TGN) through endosomes, a pathway regulated by signals in its cytosolic tail (52). Furin processes precursors either in the TGN, at the cell surface, or in endosomes. It is implicated in the activation of a wide variety of proteins such as growth factors, receptors, enzymes, blood coagulation factors, and even glycoproteins of infectious viruses (42, 52). Inactivation of the furin gene in mouse results in lethality at about embryonic day 11 (E11). Embryos fail to undergo the axial rotation and ventral closure needed to form a looping heart tube and a coherent primitive gut (39). In addition, the chorion does not fuse with the allantois, a phenotype observed in embryos lacking bone morphogenetic protein 2 (BMP2) (59), BMP5 and BMP7 (48), vascular cell adhesion molecule 1 (VCAM-1) (15), or its receptor, α4-integrin (58).

Although only a few specific in vivo substrates of furin have yet been identified, the overlap in the distribution of furin mRNA and that of transforming growth factor beta (TGF-β) family members, such as TGF-β1 and BMP4 (1, 39, 54), is striking. A furin knockout (KO) in liver and other tissues from adult mice, using the inducible Mx1-cre transgene, resulted in mice with no phenotype, demonstrating redundancy with other PCs, since cleavage of typical hepatocyte furin substrates was reduced but not eliminated (38). In contrast, in vivo studies showed that furin can uniquely process the Ac45 subunit of the vacuolar type H⁺-ATPase in pancreatic β cells (30). Furthermore, conditional deletion of furin in T cells allowed normal T-cell development but impaired the function of regulatory and effector T cells, which produced less TGF-β1 (36).

Some of the phenotypes observed in total furin KO mice, such as blood island density and yolk sac vasculature defects, suggest endothelial cell-specific functions (39). Accordingly, we generated mice lacking furin specifically in endothelial cells. This resulted in postnatal lethality, likely due to heart ventricular septal and/or valve defects. The data also showed that loss of furin in endothelial cells is associated with significantly reduced processing of endothelin-1 (ET-1), adrenomedullin (Adm), TGF-β1, and bone morphogenetic protein 4 (BMP4) precursors and that furin-deficient endothelial cells cannot grow ex vivo.

MATERIALS AND METHODS

Production of Furin^flox/flox Tg(Tie2-cre) mice.

Conditional Furin^flox/flox mice with a C57BL/6J pure background (at least 10 backcrosses to Jackson Laboratory strain 000664) were generated by flanking exon 2 of the Furin gene with two loxP sites (flox allele), as reported previously (38). These mice were crossed with Tie2-cre transgenic (Tg) mice, which express the Cre recombinase under the control of the Tie2 promoter (25). Heterozygotes carrying one copy of the transgene Furin^flox/+ Tg(Tie2-cre)^+/0 were selected and crossed with Furin^flox/flox mice to obtain Furin^flox/flox Tg(Tie2-cre)^+/0 mice, in which exon 2 is missing on both alleles in endothelial cells, and herein named endothelial cell-specific knockout (ecKO) mice. All procedures were approved by the Clinical Research Institute of Montreal (IRCM) Bioethics Committee for Animal Care.

Genotyping analyses.

Genomic DNA was extracted from mouse or embryo tail and analyzed by PCR using a pair of oligonucleotides, 5′-ATGCTCAAGGCCAGAAGATC-3′ and 5′-AATCTGTTCCCTGCTGAGGA-3′, that generate 390-bp and 482-bp fragments for wild-type (WT) and flox alleles, respectively. The presence of the Tg(Tie2-cre) was detected by using Cre-specific oligonucleotides: 5′-TGCCAGGATCAGGGTTAAAG-3′ and 5′-TGCATGATCTCCGGTATTGA-3′ (402-bp product).

Primary endothelial cell culture.

Lungs from WT and ecKO littermates at E18.5 were collected and placed in RPMI 1640 medium (Invitrogen, Burlington, ON, Canada) containing 0.1% collagenase A (Roche, Laval, QC, Canada) and 1% penicillin-streptomycin, cut into small pieces, and further homogenized using a 16-gauge needle on a 10-ml syringe. Cells were cultured in an endothelial cell-specific (ec) medium composed of Dulbecco modified Eagle medium (DMEM; low glucose)–F-12 medium (Invitrogen) containing 20% fetal bovine serum (FBS; Invitrogen) and 4% endothelial cell growth serum (BD Bioscience, Mississauga, ON, Canada). Cells were plated on 0.1% gelatin-precoated dishes for 3 to 4 days. Confluent cells were split once and grown again to confluence (2 to 3 days). Endothelial cells were then selected using Dynabeads sheep anti-rat IgG (Invitrogen) and rat anti-mouse CD102 (intercellular adhesion molecule 2 [ICAM-2]) monoclonal antibody (BD Bioscience) for further analyses.

DNA and RNA analyses.

Genomic DNA and total RNA from purified endothelial cells as well as RNA from lung and heart were extracted using TRIzol reagent (Invitrogen), as indicated by the manufacturer's manual. Genomic DNAs were analyzed by PCR and quantitative PCR (qPCR). For cDNA synthesis, 250 ng of total RNA was used in a total volume of 20 μl using SuperScript II reverse transcriptase, 25 μg/ml oligo(dT)_12-18, 0.5 mM 2′-deoxynucleoside 5′-triphosphates, and 40 U of RNaseOUT. For qPCR, each cDNA sample was submitted to two PCR amplifications, one for the gene of interest and the second for the normalizing protein TATA binding protein (TBP), using PerfeCTa SYBR green supermix (Quanta Bioscience, Gaithersburg, MD) and an Mx3000 system (Agilent Technologies, Mississauga, ON). Oligonucleotides used were 5′-CATGACTACTCTGCTGATGG-3′ and 5′-GAACGAGAGTGAACTTGGTC-3′ for furin, 5′-ACTCTTCAGAGGGTGGCTA-3′ and 5′-GCTGGAACAGTTCTTGAATC-3′ for PC5AB, 5′-GCTGGCTAAACAAGCTTTCGA-3′ and 5′-CAAAAATGGAGCCCAGACCTT-3′ for PACE4, 5′-GTTATCAGGGATGTAGGAGA-3′ and 5′-AAGGGTCTTGAGTGTGTTAG-3′ for PC7, 5′-GACTTTCCAAGGAGCTCCAGAA-3′ and 5′-CAGCTCCGGTGCTGAGTTC-3′ for ET-1, 5′-CCAGATACTCCTTCGCAGTT-3′ and 5′-CTGGGTAGGAACTGTCGTCT-3′ for Adm, 5′-CCCTGGATACCAACTATTGC-3′ and 5′-GGACCTTGCTGTACTGTGTG-3′ for TGF-β1, 5′-TGAGCCTTTCCAGCAAGTTT-3′ and 5′-CTTCCCGGTCTCAGGTATCA-3′ for BMP4, and 5′-GCTGAATATAATCCCAAGCGATTT-3′ and 5′-GCAGTTGTCCGTGGCTCTCT-3′ for TBP.

Immunocytochemistry.

Endothelial cells were plated on a 35-mm glass-bottom dish (MatTek Corp, Ashland, MA) precoated with 0.1% gelatin. Once attached, cells were rinsed with PBS, fixed with 4% paraformaldehyde for 15 min, and incubated with 150 mM glycine to neutralize the excess of aldehydes. Cells were blocked with 1% bovine serum albumin (BSA) solution for 30 min at room temperature and incubated with a rat anti-mouse platelet endothelial cell adhesion molecule 1 (PECAM-1; CD31) antibody (1:100; Chemicon International, Billerica, MA) in 1% BSA overnight at 4°C. On the next day, cells were incubated with fluorescent-tagged secondary antibody, anti-rat antibody–Alexa 488 (1:200; Invitrogen), in the dark for 30 min at room temperature and finally covered with 90% glycerol and 10% phosphate-buffered saline (PBS) containing 50% (wt/vol) 1,4-diazabicyclo[2.2.2]octane for imaging. Negative and positive PECAM-1 cells were counted using the software Northern Eclipse (version 7.0) in 3 different preparations before selection and 6 preparations after selection with CD102-coupled magnetic beads. For each preparation, a field was randomly selected, and total and PECAM-1-positive cells were counted by bright-field and UV illumination, respectively.

Bone and cartilage staining.

Mouse embryos or newborns were collected in cold PBS, and excess tissue was removed. Embryos were then fixed in 100% ethanol for 4 days and in acetone for 3 days to remove fat. After rinsing in water, skeletons were stained in a mixture of 0.14% alcian blue 8GX (Sigma-Aldrich, St. Louis, MO) in 70% ethanol, 0.12% alizarin red S (Sigma-Aldrich) in 95% ethanol, glacial acetic acid, and 70% ethanol for 10 days. After staining, skeletons were transferred into a cleaning solution of 1% KOH in 20% glycerol at 37°C overnight and then kept at room temperature until they were completely leached. The specimens were finally placed in a 2:2:1 mixture of ethanol-glycerol-benzyl alcohol for imaging.

Embryo and heart histology.

Embryos or newborns and their hearts were fixed in Bouin's solution at 4°C for 24 h, washed in 70% ethanol, and embedded in paraffin. Sections were cut at a 4-μm thickness and stained with hematoxylin-eosin (Sigma-Aldrich) for imaging.

Magnetic resonance imaging.

Embryos were collected at E15.5, placed for 20 min in Hanks balanced salt solution with 5 mM EDTA at 37°C to bleed, and put on ice. Embryos were then fixed in 4% paraformaldehyde at 4°C for at least 3 days and analyzed as previously described (51).

Echocardiography.

Echocardiograms were obtained with an ultrasound imaging system, Vevo 770 (VisualSonics, Toronto, ON, Canada). The pregnant mother was anesthetized with 2% isoflurane and placed on a warm station to maintain the body temperature at 37°C and heart rate at 450 to 550 beats per min. Abdominal hair was removed and embryos were marked on the mother's abdomen for identification. M-mode images were recorded using a 30-MHz transducer, and heart rates were calculated. Embryos were collected from the exact position to follow the previously marked numbers for each embryo, and the tails were collected for genotyping. Three pregnant females with 25 embryos at E18.5 were analyzed.

ELISA.

Tissues and plasma were collected from embryos and newborns at E18.5 and postnatal day 0 (P0), respectively. Primary endothelial cells were cultured as described above and collected 24 h after the selection. Tissue, plasma, or primary endothelial cell proteins were extracted and analyzed using enzyme-linked immunosorbent assay (ELISA), as recommended by the manufacturer (nonplasma ET-1; Immuno-Biological Laboratories Co. Ltd., Japan; ET-1 in plasma, R&D Systems, Minneapolis, MN; Adm, Phoenix Pharmaceuticals, Belmont, CA; TGF-β1, Invitrogen, Burlington, ON, Canada). Seven to 8 mice per genotype were used for ET-1 in tissues and cells/media, and 4 mice per genotype were used for ET-1 in plasma, Adm, and TGF-β1.

Western blot analyses.

Proteins were extracted from organs or purified endothelial cells in 50 mM Tris-HCl, pH 8, 150 mM NaCl, 0.1% SDS, 1% Nonidet P-40, and 0.25% Na deoxycholate (1× radioimmunoprecipitation assay buffer) in the presence of a cocktail of protease inhibitors (Roche). Protein concentrations were estimated by a Bradford assay. After SDS-PAGE and blotting onto polyvinylidene difluoride membranes, the latter were incubated with mouse BMP4 monoclonal (1:1,000; R&D Systems, Minneapolis, MN), rabbit phospho-Smad1/5/8 polyclonal (1:1,000; Cell Signaling, Danvers, MA), or rabbit polyclonal actin (1:1,000; Sigma-Aldrich, St. Louis, MO) antibodies and finally with appropriate secondary antibodies. Immunoreactive species were detected using enzymatic chemiluminescence (ECL; Amersham Bioscience, Baie d'Urfe, QC, Canada) and quantified using ImageJ (version 1.41) software.

Primary endothelial cell rescue.

In order to rescue the impaired growth in ecKO endothelial cells, medium from WT endothelial cells or 24-h-conditioned medium from transfected COS-1 cells was mixed with fresh endothelial cell medium (1:1) and added to the cultures every 2 days for 13 days. Purified furin (1.5 units/ml, a quantity similar to that produced by furin overexpression in COS-1 cells) (13) was added to medium from COS-1 cells transfected with pIRES2-enhanced green fluorescent protein (EGFP). The cell-permeant PC inhibitor decanoyl-Arg-Val-Lys-Arg-chloromethylketone (dec-RVKR-CMK; Bachem BioScience, Torrance, CA) was used at a final concentration of 25 μM. Viable cells were measured by alamarBlue (optical density [OD] at 570 nm; Invitrogen), as recommended by the manufacturer, 48 h after the conditioned medium had been added onto cells. Three individual cultures were analyzed for each genotype, and statistical significance between two curves was calculated by two-way analysis of variance (ANOVA) test.

Plasmids.

The cDNA encoding human furin and WT mouse BMP4 were cloned into the pIRES2-EGFP expression vector, as described previously (11). S1 mutant BMP4 was generated by mutating Arg₂₈₉ and Arg₂₉₂ in the S1 cleavage site (RAKR₂₉₂↓SP) to Ala (AAKA₂₉₂↓SP). A two-step PCR was used to generate a PstI-BamHI mutant fragment using the sense (S) and antisense (AS) oligonucleotide pairs (oligonucleotide S1, CAGAGCGGCCAAAGCTAGTCCCAAGCATC [underlining indicates coding triplets for replacing Arg289 and Arg292 with Ala residues]; oligonucleotide AS1, CTTCGGCCAGTAACGTTAGGGG; oligonucleotide S2, CTGGACACCAGACTAGTCCATCAC; oligonucleotide AS2, CTTGGGACTAGCTTTGGCCGCTCTGCGGGTC), and the mutant fragment was substituted into the original PstI-BamHI fragment. All cDNAs were verified by DNA sequencing.

Conditioned medium preparation.

COS-1 cells were transfected with 4 μg of DNA, using 2 μg of each plasmid and 0, 2, or 4 μg of the pIRES2-EGFP vector for double, single, and control transfections, respectively. A Lipofectamine 2000 kit (Invitrogen) was used as recommended by the manufacturer. At 24 h posttransfection, medium was replaced by low-glucose DMEM (Invitrogen) containing 20% (vol/vol) FBS (Invitrogen). Conditioned medium was collected after 24 h for rescue.

RESULTS

The majority of the ecKO mice die within 12 h after birth.

To generate mice that lack furin specifically in endothelial cells, we crossed Furin^flox/flox mice that carry conditional flox alleles of the Furin gene (38) with transgenic Tg(Tie2-cre) mice that express the Cre recombinase under the control of the endothelial cell-specific Tie2 promoter (25). In endothelial cells, Cre is expected to recombine the two loxP sites that flank exon 2, which encodes the signal peptide and most of the prosegment of furin (38), thereby generating endothelial cell-specific furin knockout (ecKO) mice.

Genotyping of litters 3 to 4 weeks after birth (126 mice) revealed that no ecKO mice survived (Table 1). To identify the lethality stage, embryos were genotyped at E15.5, E18.5, and P0. In all cases, close to Mendelian ratios were obtained (Table 2), suggesting that lethality occurred after birth. Survival assessment revealed that only 37%, 16%, and 3% of the ecKO newborns survived for 12, 36, and beyond 60 h, respectively (Fig. 1). The unique mouse out of 32 (3%) that survived with no apparent phenotype suddenly died at 3 months of age. Natural death in wild-type and heterozygote (only one Furin allele inactivated in endothelial cells) newborns was also observed, although to a much smaller extent (79% and 68% survival by 60 h after birth, respectively). Note that in this study, Furin^flox/+ or Furin^flox/flox mice that do not carry the transgene Tie2-cre were assimilated to WT mice.

Table 1.

Absence of ecKO mice at weaning^a

Mouse typeb	No. (%) of mice
Furinflox/+
Cre negative (WT)	39 (31)
Cre positive (HTZ)	46 (36.5)
Furinflox/flox
Cre negative (WT)	41 (32.5)
Cre positive (ecKO)	0 (0)
Total	126 (100)

^aPups were weaned at 21 to 27 days of age.

^bWT, wild type (or assimilated to) Furin^flox/+ or Furin^flox/flox mice; HTZ, heterozygote Furin^flox/+ Tg(Tie2-cre)^+/0 mice; ecKO, endothelial cell-specific furin knockout Furin^flox/flox Tg(Tie2-cre)^+/0 mice.

Table 2.

Mendelian distribution of the genotypes^a at E15.5, E18.5, and P0

Stage	No. (%) of miceb
	Furinflox/+		Furinflox/flox		Total
	Cre negative (WT)	Cre positive (HTZ)	Cre negative (WT)	Cre positive (ecKO)
E15.5	11 (22)	13 (25)	16 (31)	11 (22)	51 (100)
E18.5	31 (33)	19 (20)	24 (25)	21 (22)	95 (100)
P0	26 (24)	28 (26)	21 (20)	32 (30)	107 (100)

^aMendelian distribution of the genotypes (25% expected for each) was observed in the progeny of Furin^flox/+ Tg(Tie2-cre)^+/0 × Furin^flox/flox at E15.5, E18.5, and P0.

^bWT, wild type (or assimilated to) Furin^flox/+ or Furin^flox/flox mice; HTZ, heterozygote Furin^flox/+ Tg(Tie2-cre)^+/0 mice; ecKO, endothelial cell-specific furin knockout Furin^flox/flox Tg(Tie2-cre)^+/0 mice. Observed numbers were not significantly different from the expected values according to Pearson's χ² test.

Fig 1.

Lethality stage of ecKO mice. Forty-seven WT, 28 heterozygous (HTZ), and 32 ecKO newborns (P0) (Table 2) were placed under close surveillance. Survival rates were counted at 12, 36, and 60 h after birth. WT, Furin^flox/+ or Furin^flox/flox; heterozygous, Furin^flox/+ Tg(Tie2-cre)^+/0; ecKO, Furin^flox/flox Tg(Tie2-cre)^+/0.

Furin is efficiently inactivated in endothelial cells.

To evaluate the efficiency of furin knockout, we isolated endothelial cells from lungs of E18.5 embryos. Lungs were treated with collagenase, and cells were plated and grown to confluence. Cells were passaged once and grown again to confluence. Endothelial cells were purified using beads coated with CD102 antibodies recognizing ICAM-2 and collected or replated for various analyses (Fig. 2A). Genomic DNA qPCR analysis of individual endothelial cell preparations showed that 90% of the furin flox alleles were inactivated by Cre (90% of Δ2 alleles and 10% of remaining flox alleles; Fig. 2B and C). Immunocytochemistry of PECAM-1, an endothelial cell-specific marker, showed that 94% of purified cells were PECAM-1 positive (open arrowheads), whereas before selection 47% were PECAM-1 positive (Fig. 2D). This indicates that the 10% remaining flox alleles partly originated from the 6% nonendothelial cell-contaminating cells. Thus, furin ecKO was estimated to be 96% efficient. Furin and other furin-like PC mRNA levels were also assessed. Surprisingly, total furin expression levels dropped by only 72% in purified cells. The remaining 28% may have been due to the ∼10% contaminating cells, possibly expressing ∼3-fold higher levels of furin than endothelial cells. PC5 and PACE4 expressions were not affected by furin deficiency in purified cells, whereas PC7 expression was upregulated by 65%. Importantly, the levels of expression of PC5, PACE4, and PC7 represented only 0.1%, 1%, and 2% of the furin levels, respectively.

Fig 2.

Efficiency of Furin gene inactivation in endothelial cells. (A) Primary endothelial cells (primary ec) were isolated from WT or ecKO lungs at E18.5. Lung cells were cultured for 3 to 4 days, passaged, further grown for 2 to 3 days, and finally, purified using CD102-coupled magnetic beads. Cells were immediately collected for DNA and RNA analyses and cultured one more day for protein analyses and immunocytochemistry. PCR (B) and qPCR (C) were performed on genomic DNA to visualize and quantify flox (intact) and Δ2 (Cre-recombined) alleles in purified primary endothelial cells (90% of recombined alleles) or tail extracts (93% of intact alleles; n = 5 per genotype). (D) Immunolabeling of the endothelial cell-specific marker PECAM-1 was performed to evaluate the percentage of PECAM-1-positive cells (open arrowheads) before (47%; n = 3) and after (94%; n = 6) immunopurification of endothelial cells. PECAM-1-negative cells are indicated by filled arrowheads. (E) Furin, PC5, PACE4, and PC7 mRNA levels in purified endothelial cells were measured by RT-qPCR (n = 5 per genotype; *, P < 0.05). Bars represent the mean + SEM, and P was determined by a bilateral Student t test. WT, Furin^flox/+ or Furin^flox/flox; ecKO, Furin^flox/flox Tg(Tie2-cre)^+/0. a.u., arbitrary units.

ecKO mice do not exhibit visible defects.

No gross morphological differences were detected between WT and ecKO embryos at E18.5 and P0 (Fig. 3A). At these time points, body size, eyes, limbs, and tail were normal, and at P0, WT and ecKO bone structure and cartilage formation were similar (Fig. 3B to D). Hematoxylin-eosin staining of whole embryos at E18.5 did not reveal any abnormality in their structure. The alveoli of lungs (Fig. 3E), rich in endothelial cells, and kidney structures (Fig. 3F) of ecKO mice also seemed normal.

Fig 3.

Absence of overt phenotypes in ecKO mice. (A) WT and ecKO embryos at E18.5 and newborns at birth (P0) are shown. (B) Skeletons were obtained from WT and ecKO mice at P0 and stained with alizarin red and alcian blue, and closer views of skulls (C) and vertebrae (D) are shown. Paraffin sections showing lungs (E) and kidneys (F) from WT and ecKO embryos at E18.5 were stained with hematoxylin-eosin. Bars, 0.5 mm.

Ventricular septal defects are observed in ecKO heart.

Examination of E15.5 embryos using magnetic resonance imaging confirmed that most organs in ecKO embryos were normal. Severe edema was seen in the lumbar regions of 2 out of 16 WT and heterozygote embryos and in 6 out of 9 ecKO embryos, with one additional ecKO embryo showing dramatic edema throughout the body (Fig. 4A).

Fig 4.

Edema and heart defects in ecKO embryos. (A) Magnetic resonance imaging sagittal sections revealed severe subcutaneous edema in ecKO embryos at E15.5. LL, left lung; Liv, liver; Oed, edema. Bars, 0.5 mm. (B) Magnetic resonance imaging transverse sections (top) and three-dimensional reconstructions (bottom) revealed ventricular septal defects in ecKO hearts at E15.5. LA/RA, left atrium/right atrium; LV/RV, left ventricle/right ventricle; VS(D), ventricular septum (defect); D, dorsal; V, ventral; R, right; L, left; Tr, trachea; RCCA/LCCA, right common carotid artery/left common carotid artery; AoA, aortic arch; DA, ductus arteriosus. Bars, 0.5 mm. (C) Paraffin sections of heart at E18.5 were stained with hematoxylin-eosin. Bars, 1 mm. (D) Magnetic resonance imaging transverse sections revealed atrioventricular tricuspid and mitral valve thickening in ecKO hearts at E15.5. TV(D), tricuspid valve (defect); MV(D), mitral valve (defect). Bars, 0.5 mm. (E) In utero echocardiography (M mode) of embryo hearts at E18.5 was performed using a 30-MHz transducer (n = 7 to 9). P < 0.0005 (***) was determined by a bilateral Student t test, and bars represent the mean + SD. WT, Furin^flox/+ or Furin^flox/flox; heterozygous (HTZ), Furin^flox/+ Tg(Tie2-cre)^+/0; ecKO, Furin^flox/flox Tg(Tie2-cre)^+/0.

Seven out of 9 ecKO hearts exhibited ventricular septal defects (VSDs), which were confirmed by hematoxylin-eosin staining of heart sections (Fig. 4B and C), whereas the 2 remaining ecKO embryos exhibited only atrioventricular tricuspid and mitral valve thickening (Fig. 4D), with no apparent VSD. However, 2 embryos cumulated the two types of defects. Since VSD is associated with cardiac arrhythmia (16), especially tachycardia, we used echocardiography (M-mode; 30-MHz transducer) to measure the heart rate at E18.5 in utero and found that heart rates in ecKO mice were 22% higher than those in WT mice (Fig. 4E).

ET-1, Adm, and TGF-β1 processing is reduced in the absence of endothelial cell furin.

We analyzed the contribution of endothelial furin to the in vivo processing of major endothelial factors that are putative furin substrates and for which sensitive ELISA kits are commercially available.

ET-1 is a vasoconstrictor peptide mainly produced by endothelial cells (17, 18, 57). The ET-1 precursor has to be cleaved at two typical PC-like sites to produce large amounts of ET-1 (2, 9, 23), which is then cleaved by endothelin-converting enzyme to generate active and mature ET-1 (56). As measured by ELISA at E18.5, ecKO lungs exhibited 57% less mature ET-1 than WT lungs (Fig. 5A), whereas ET-1 mRNA levels increased by 3.9-fold in ecKO lungs, suggesting the existence of a feedback upregulation in this tissue (Fig. 5B). The remaining 43% ET-1 in ecKO lungs is likely due to pro-ET-1 production by other cell types and/or processing by furin from neighboring cells. In ecKO hearts, a nonsignificant 33% decrease in mature ET-1 was observed (P = 0.07). The most dramatic drops were observed in newborn plasma (71%), ecKO purified endothelial cells (89%), and media (74%) (Fig. 5A), indicating that endothelial pro-ET-1 is essentially processed by furin and largely contributes to circulating ET-1 levels. Importantly, ET-1 mRNA levels were not affected by genotype in heart tissue or primary endothelial cells (Fig. 5B).

Fig 5.

ET-1, Adm, and TGF-β1 expression in tissues, plasma, and primary endothelial cells (prim ec). (A) Mature protein levels were measured by ELISA in tissues at E18.5, plasma of newborns, or primary endothelial cell extracts and media (n = 4 to 8). (B) mRNA levels were measured by qPCR in lung, heart and/or adrenal and primary endothelial cells and normalized to WT levels (n = 4 to 5). P values were determined by a bilateral Student t test: *, P < 0.05; **, P < 0.005; ***, P < 0.0005. Bars represent the mean + SEM.

Adm is a vasodilator peptide constitutively secreted by vascular endothelial and smooth muscle cells (49, 50). It was hypothesized by Kitamura et al. that Adm was produced by cleavage at two typical furin sites, RVKR₉₄↓ and YGRR₁₄₉↓ (27). However, because a tetrabasic site followed by a P1′ Ser and a P2′ Leu is optimal for furin (29, 37), we predict that Adm C-terminal cleavage would rather take place at RRRR₁₅₂↓SL. Carboxypeptidase trimming of Arg residues would still allow amidation at Gly₁₄₇ (26). However, this remains to be determined. Our data showed that mature Adm was reduced by 37% in ecKO adrenals at E18.5, 43% in plasma of ecKO newborns, and 60% and 67% in primary endothelial cells and media, respectively (Fig. 5A).

TGF-β1 was reported to be a potent inhibitor of endothelial cell proliferation in two-dimensional cultures but showed no effect on growth in three-dimensional cultures (31). It is expressed by a wide variety of cell types, including endothelial cells (33). It is cleaved at a unique S1 site, and furin was shown to be its major enzyme ex vivo (10) as well as in vivo in T cells (36). As measured by ELISA, mature TGF-β1 was reduced by 16% (P = 0.051) in ecKO lung but not heart, in agreement with the fact that TGF-β1 is well synthesized by many cell types in these tissues (33, 53). TGF-β1 levels significantly dropped by 13% in ecKO plasma and by 64% and 37% in primary endothelial cells and media, respectively (Fig. 5A). As for Adm, residual levels of mature product were high. This may be due to other endothelial cell furin-like PCs but also to a substantial production of Adm or TGF-β1 by the 6% contaminating cells in endothelial cell preparations. No significant change of Adm or TGF-β1 mRNA levels was observed in ecKO primary endothelial cells (Fig. 5B).

BMP4 is dramatically reduced in ecKO primary endothelial cells.

BMP4 is a member of the TGF-β superfamily. It is first synthesized as a precursor protein (pro-BMP4), which undergoes proteolytic processing by members of the PC family to produce active BMP4 (4, 7). It is initially cleaved at a primary S1 site (RRRAKR₃₀₄↓SP) immediately upstream of the C-terminal BMP4 domain and subsequently at a secondary S2 site (RISR₂₆₉↓SL) within the prodomain (6). The first cleavage at S1 releases mature BMP4, which remains associated with its prodomain (8). The second cleavage at S2 releases the prodomain and further increases the activity of BMP4 and its signaling range (14).

Western blot analysis of purified ecKO lung endothelial cells revealed a strong decrease (∼95%) of mature BMP4 levels (Fig. 6A). In agreement, the phosphorylation levels of Smad1/5/8, which reflect BMP4 signaling, were reduced by 40% in ecKO cell extracts. Although furin, PACE4, and PC7 overexpressed in COS-1 cells can all cleave pro-BMP4, with furin being the most efficient PC (4), furin is likely the exclusive PC to process pro-BMP4 in primary endothelial cells. In addition, PACE4 and PC7 mRNA levels were 50- to 100-fold lower than those of furin mRNA (Fig. 2E). The mRNA level of BMP4 also remained unchanged in ecKO cells (Fig. 6B).

Fig 6.

BMP4 and pSmad1/5/8 expression in purified endothelial cells. (A) Mature BMP4 and phosphorylated Smad1/5/8 levels were analyzed by Western blotting in 5 pooled endothelial cell extracts per genotype. (B) mRNA levels of BMP4 were measured by qPCR in primary endothelial cells and normalized to WT levels (n = 4). Bars represent the mean + SEM.

Furin is essential for ex vivo endothelial cell growth.

Although ecKO newborns exhibited no gross vessel abnormalities, ecKO purified endothelial cells failed to grow and died within a week. These cells thus seem to lack essential factors, even in the presence of 20% FBS and 4% endothelial cell-specific growth serum. In order to rescue their defective growth, 24-h-conditioned medium from WT endothelial cells was mixed with fresh endothelial cell medium (1:1) and added to cultures every 2 days. Viable cells were measured by alamarBlue. The conditioned medium had no effect on WT endothelial cell cultures (data not shown) but completely rescued ecKO endothelial cell growth to the level of WT endothelial cells (Fig. 7A; right).

Fig 7.

Growth rescue in ecKO endothelial cells. (A) Conditioned media of COS-1 cells either transfected with full-length furin (COS:furin) or not transfected but containing 1.5 units/ml of purified furin (COS+pur.furin) were transferred onto WT or ecKO primary endothelial cells, without or with prior addition of the furin inhibitor dec-RVKR-CMK (cmk). Fresh media (ec media) or conditioned WT primary endothelial cell medium (WT ec medium), as well as medium from COS-1 cells transfected with the empty vector pIRES2-EGFP (COS:pIR), was used as the control. (B) Conditioned media of COS-1 cells transfected with BMP4 alone, WT (COS:BMP4), or mutant (COS:BMP4*) or together with furin (COS:BMP4:furin or COS:BMP4*:furin) were transferred onto WT or ecKO primary endothelial cells. Stars point at rescuing media. P values were determined by a two-way ANOVA test: **, P < 0.005; ***, P < 0.0005. n = 3 per genotype. Error bars represent ±SEMs.

The ability of furin to rescue ecKO endothelial cell growth was next assessed. Conditioned media from COS-1 cells either transfected with full-length furin or not transfected but supplemented with purified soluble furin were applied onto ecKO cells, in the presence or absence of the furin inhibitor dec-RVKR-CMK, and the furin activity in the media was measured using the fluorogenic substrate pyroglutamic acid-Arg-Thr-Lys-Arg-7-amido-4-methylcoumarin (data not shown). An efficient rescue was obtained with both transfected furin, which is released in the COS-1 medium by shedding, and purified soluble furin, with the latter being slightly more efficient (Fig. 7A, right). Importantly, rescue, as well as furin activity in the medium (data not shown), was lost in the presence of the furin inhibitor. However, WT endothelial cell conditioned medium was the most efficient at the rescue of growth.

Because BMP4 processing was drastically reduced in ecKO endothelial cells, we verified whether BMP4 could rescue their deficient growth. Media from COS-1 cells expressing BMP4 or its S1 mutant form (RAKR₂₉₂ into AAKA₂₉₂; BMP4*) alone or with furin were tested (Fig. 7B). BMP4 alone seemed to allow cell survival, but not growth, as the BMP4 curve was significantly improved compared to that obtained with the medium with COS-1 cells expressing the empty pIRES2-EGFP vector. However, the major effect of furin on ecKO endothelial cell growth was slightly, but significantly, improved by BMP4 (two-way ANOVA, P < 0.0001). In agreement with a specific role of BMP4, its noncleavable form generated BMP4* and BMP4*-furin curves that were similar to those of the empty vector and furin alone, respectively.

DISCUSSION

The absence of furin causes death at ∼E11 with various developmental malformations (39). To bypass early embryonic death and to focus on vascular phenotypes, we inactivated the furin gene specifically in endothelial cells (ecKO). Herein, we demonstrated that endothelial furin is essential for (i) survival after birth, (ii) proper heart development, and (iii) ex vivo endothelial cell growth. Furthermore, (iv) tissue and primary endothelial culture analyses showed that ET-1, Adm, and TGF-β1 are in vivo substrates of endothelial furin, whereas evidence for BMP4 came exclusively from primary endothelial cells.

Furin, PC5, and PACE4 were shown to cleave numerous TGF-β-like growth factors, and the absence of PC5 or PACE4 leads to severe defects in anteroposterior and/or left-right axes (5, 12, 51). Furin-related defects cannot be examined due to the early death of furin KO embryos (39). However, WT and ecKO littermates had no overt phenotypes, indicating that endothelial furin does not contribute to axial patterning. Similarly, PC5 ecKO mice did not show any axial defects (32), suggesting that endothelial cell PCs do not play a major role in embryonic axial patterning.

All ecKO embryos displayed ventricular septal defects and/or valve defects. Although we cannot exclude other causes, the observed sudden death of newborns may be due to cardiac failure. In agreement with this hypothesis, tachycardia and subcutaneous edema, which are common characteristics in heart failure (16, 35), were observed in ecKO embryos.

The processing of four candidate substrates of endothelial furin was assessed. In vivo evidence for reduced cleavage was obtained in ecKO tissues and/or plasma for ET-1, Adm, and TGF-β1, with drops from 13% to 71%. The strongest reductions were obtained for ET-1 in ecKO lungs (57%) and plasma (71%), reflecting its high level of production by endothelial cells. The mildest reductions were obtained for TGF-β1, reflecting its strong expression in other cell types of lung and heart (33, 53).

ET-1 and BMP4 are likely exclusive substrates of endothelial furin. Their residual mature forms, estimated to be present at 11% and 5% in endothelial cell preparations, respectively, may originate from the small percentage of endothelial cells with incomplete Cre-mediated recombination or contaminating cells. In contrast to ET-1 and BMP4, Adm and TGF-β1 exhibited higher levels (∼40%) of residual mature forms. Whether these residual forms originate from contaminating cells heavily expressing Adm and TGF-β1 or from partial processing in endothelial cells by other PCs remains to be determined.

The lack of ET-1 or BMP4 was implicated in VSD (19, 28). However, the absence of ET-1 processing by endothelial furin is not expected to be responsible for the early death of ecKO newborns, as mice lacking endothelial ET-1 were recently reported to be viable (24). Valves and septa originate from the endocardial cushion resulting from the interaction between the endocardium and myocardium. However, at this stage, BMP4 is expressed only in cardiomyocytes and not in endothelial cells (19, 21). Accordingly, specific inactivation of BMP4 in cardiomyocytes leads to heart malformations, including VSD (19). Thus, our data that showed a lack of BMP4 processing in purified endothelial cells isolated from ecKO lungs do not support any implication of BMP4 processing in ecKO heart phenotypes. Finally, neither Adm nor TGF-β1 should be responsible for ecKO heart phenotypes, as TGF-β1 KO leads to death 20 days after birth without cardiac malformations (44) whereas Adm KO embryos die at ∼E14 with hydrops fetalis (extreme subcutaneous edema) and overdeveloped ventricular trabeculae but no VSD (3). However, we cannot exclude the possibility that the combination of partial deficits in two or more factors may lead to heart defects.

An enzyme-substrate association may depend on the exclusive ability of a given PC to cleave a given substrate, as it was shown for PC5 and Gdf11 (12), or on the enzyme-substrate spatiotemporal distribution. If furin is the unique cognate PC for a given substrate, then furin KO is expected to at least recapitulate the phenotypes generated by the loss of the substrate. The fact that mice lacking furin (39) exhibited less severe phenotypes than mice lacking BMP4 (54) thus indicates that other PCs cleave pro-BMP4 during early development. In addition, upon overexpression in COS-1 cells (4; unpublished data), furin, PC5, PACE4, and PC7 were all able to cleave pro-BMP4. Thus, the unique ability of endothelial furin to cleave BMP4 seems to be due to its 50- to 1,000-fold higher expression compared to that of the other furin-like PCs. Because the KOs of ET-1, Adm, and TGF-β1 KOs all lead to death at later stages than the KO of furin, it remains difficult to evaluate the exclusivity of furin in their processing.

The absence of furin resulted in ecKO endothelial cell death in primary cultures. However, growth was completely rescued by conditioned medium from WT primary endothelial cells. COS-1 cell media containing expressed or purified soluble furin led to an efficient, but not optimal, rescue that was slightly improved upon BMP4 coexpression. However, growth was still not optimal, suggesting that intracellular cleavage of some other endothelial cell-specific factor(s) by furin is required for optimal growth. BMP4 alone acted as a survival factor rather than a growth factor in ecKO primary endothelial cells. Because the ecKO vasculature seemed normal, we hypothesize that, in vivo, neighboring cells likely contribute to endothelial cell growth by providing mature growth factors and/or processing enzymes. Although we focused on lung endothelial cells that may not reflect all endothelial cell properties, our ecKO primary endothelial cells constitute a valuable model that will allow future confirmation of the in vivo role of furin on the cleavage of other substrates, such as endothelial lipase (20), platelet-derived growth factor A/B (PDGF-A/B) (46, 47), vascular endothelial growth factor C/D (VEGF-C/D) (34, 45), insulin-like growth factor 1 (IGF-1) receptor (22), and C-type natriuretic peptide (55).

Three out of four furin-like PCs (furin, PACE4, and PC5) play key roles in development through processing of growth factors and/or their receptors. Due to low endogenous levels and nonsensitive commercial antibodies, determination of the implication of a PC in the in vivo processing of a given substrate often relies exclusively on phenotypic analyses (12). The present study provides a new genetic model of furin deficiency resulting in abnormal ventricular septation as the primary defect. This underlines the importance of endothelial furin in cardiogenesis via the activation of critical growth factors. Our work confirms that ET-1, Adm, and TGF-β1 are in vivo substrates of furin and that furin is required for ex vivo endothelial cell growth.