Lack of periostin leads to suppression of Notch1 signaling and calcific aortic valve disease

Abstract

The Postn gene encodes protein periostin. During embryonic development, it is highly expressed in the outflow tract (OFT) endocardial cushions of the developing heart, which give rise to several structures of the mature heart including the aortic valve. Periostin was previously implicated in osteoblast differentiation, cancer metastasis, and tooth and bone development, but its role in cardiac OFT development is unclear. To elucidate the role that periostin plays in the developing heart we analyzed cardiac OFT phenotype in mice after deletion of the Postn gene. We found that lack of periostin in the embryonic OFT leads to ectopic expression of the proosteogenic growth factor pleiotrophin (Ptn) and overexpression of delta-like 1 homolog (Dlk1), a negative regulator of Notch1, in the distal (prevalvular) cushions of the OFT. This resulted in suppression of Notch1 signaling, strong induction of the central transcriptional regulator of osteoblast cell fate Runx2, upregulation of osteopontin and osteocalcin expression, and subsequent calcification of the aortic valve. Our data suggest that periostin represses a default osteogenic program in the OFT cushion mesenchyme and promotes differentiation along a fibrogenic lineage. Lack of periostin causes derepression of the osteogenic potential of OFT mesenchymal cells, calcium deposition, and calcific aortic valve disease. These results establish periostin as a key regulator of OFT endocardial cushion mesenchymal cell fate during embryonic development.

a family of notch receptors controls a highly conserved signaling pathway that regulates a multitude of biological processes, including cell fate specification, progenitor cell maintenance, tissue boundary formation, proliferation, and apoptosis. Notch pathway plays a crucial role in cardiac development as evidenced by the wide range of cardiac phenotypes in mice with targeted disruption of genes encoding various components of the Notch signaling pathway (15, 28). These data indicate that Notch is involved in cardiac mesoderm differentiation during early embryonic development, atrioventricular canal development, epithelial-to-mesenchymal transition and valve development, ventricular trabeculation, and outflow tract (OFT) development. Human mutations in several Notch signaling components result in congenital cardiovascular malformations such as Alagille syndrome and tetralogy of Fallot (15, 28). Loss-of-function mutations in NOTCH1 cause aortic valve disease (14). Although a consensus has emerged that the Notch1 signaling cascade plays a key role in aortic valve disease (13), two recent genomewide linkage studies for valve calcification susceptibility loci suggested that NOTCH1 mutations do not account for all cases of calcific valve disease and found that other genes are involved (4, 25).

Periostin, a 90-kDa secreted protein, was first isolated from the mesenchymal cell line MC3T3-E1 (16). Subsequent studies revealed that periostin plays an important regulatory role in osteoblast differentiation (37), cancer metastasis (3), and embryonic development of teeth and skeleton (41). We have shown (20, 21, 30) that periostin is also expressed in the embryonic heart. Its expression first appears in the mouse heart at embryonic day (E) 10.5 and is localized to the prevalvular endocardial cushions of the OFT and atrioventricular canal (21). However, the role of periostin in cardiac development is unclear. In this study, we analyzed the role of periostin in OFT development, using periostin-knockout (Postn^−/−) mice. Our data suggest that by modulating Notch1 signaling periostin promotes OFT mesenchymal cell differentiation along a fibrogenic lineage while repressing a default osteogenic program. Lack of periostin causes derepression of the osteogenic potential of OFT cushion mesenchyme and calcific aortic valve disease.

MATERIALS AND METHODS

Periostin-knockout mice.

All animals received water and food ad libitum and were maintained in accordance with a protocol approved by the Medical University of South Carolina's Institutional Animal Care and Use Committee. The colony of Postn^−/− mice was maintained on the C57BL/6J background from original stock described by Oka et al. (35). A total of 73 wild-type, 137 Postn^+/−, and 48 Postn^−/− animals were analyzed. Genotype was determined by PCR analysis of genomic DNA from either tails (adult animals) or placentas (embryos). Genomic DNA was isolated and amplified by PCR using the following primers: 5′-CCTTGCCAGTCTCAATGAAGG-3′ (forward primer) and 5′-TGACAGAGTGAACACATGCC-3′ (reverse primer directed against wild-type allele) or 5′-GGAAGACAATAGCAGGCATG-3′ (reverse primer corresponding to mutant allele).

Morphology.

Gross anatomic examination of the Postn^−/− hearts was carried out under a Leica MZ 7.5 stereomicroscope, and images were captured with a Leica EC3 digital camera (Leica Microsystems). Masson's trichrome staining was performed with a Masson's trichrome stain kit (DakoCytomation). Calcium deposits were identified with a calcium phosphate-specific von Kossa stain kit (Diagnostic BioSystems) and alizarin red S reagent (Sigma-Aldrich). Sections were viewed with a Leitz Ortholux II microscope.

Echocardiographic analysis of cardiac function.

Echocardiography was performed with a Vevo 660 ultrasound system equipped with a 40-MHz scanhead (VisualSonics). Two-dimensional imaging and Doppler interrogation of blood flow was done as previously described (55). Mice were anesthetized with 4% isoflurane, and anesthesia was maintained with 2% isoflurane. Hair was removed from the ventral thorax with a hair removal cream (Nair, Carter-Horner). Body temperature was monitored continuously with a rectal probe (Indus Instruments) and maintained at 36–38°C with a heated platform and a small heat lamp.

Differential screening.

Screening for differentially expressed genes in the hearts of Postn^−/− mice was performed with a suppression-subtractive-hybridization-based technique essentially as previously described (50). Poly(A)⁺ RNA was isolated from either 10 wild-type or 10 Postn^−/− E12.5 hearts. Subtracted/normalized cDNA libraries were generated with a PCR-Select cDNA Subtraction Kit (Clontech). Secondary PCR products were cloned into pCRII-TOPO vector with a TOPO TA cloning kit (Invitrogen), and resulting libraries were organized in 384-well plates. Bacterial clones from these plates were manually printed onto Hybond-XL nylon membranes (GE Healthcare) by using a 384-pin tool (V&P Scientific) to prepare two identical sets of bacterial colony filters. Six sets of duplicate filters each containing 384 randomly picked clones of either library (total of 2,304 clones) were hybridized with ³²P-labeled subtracted cDNA probes, and differential clones were selected as previously described (50).

In situ hybridizations.

In situ hybridizations were performed essentially as previously described (53). E12.5 C57BL and Postn^−/− embryos were dissected in ice-cold PBS, fixed in 4% paraformaldehyde in PBS for 30 min, cryoprotected in 30% sucrose in PBS, and sectioned in a cryostat at a thickness of 10 μm. Sections were hybridized with digoxigenin (DIG)-labeled cDNA probes followed by incubation with anti-DIG antibodies conjugated with horseradish peroxidase (HRP) (Roche). The peroxidase activity was localized and amplified with biotinyl tyramide amplification reagent (PerkinElmer) followed by alkaline phosphatase-conjugated streptavidin. The signal was detected with nitro blue tetrazolium (NBT; 0.25 mg/ml) and 5-bromo-4-chloro-3-indolyl phosphate (BCIP; 0.125 mg/ml) (Roche) as substrates. For the double fluorescent in situ hybridizations, the first probe was labeled with DIG while the second probe was labeled with fluorescein. Slides were incubated with either anti-DIG HRP-conjugated antibodies (Roche) or anti-fluorescein HRP-conjugated antibodies (Roche). Detection of the signal was carried out with the Fluorophore Tyramide Amplification System (Molecular Probes) according to the manufacturer's protocol. Periostin was visualized with Alexa Fluor 488 (green), and the second gene was visualized with Alexa Fluor 546 (red).

Real-time PCR.

cDNA synthesized with 1 μg of mRNA isolated from either 10 wild-type or 10 Postn^−/− E12.5 hearts was used for PCR reactions, which were carried out with AmpliTaqGold thermostable DNA polymerase (Applied Biosystems). SuperROX (Biosearch Technologies) was used as a passive reference at a final concentration of 300 nM. The level of gene expression in the hearts was estimated by multiplex real-time PCR using gene-specific primers and TaqMan probes: Ptn: 5′-CCATGAAGACTCAGAGATGTAAGATCC-3′ (forward primer), 5′-CAAGGCGGTATTGAGGTCACATT-3′ (reverse primer), and 6FAM-AGCCTGGAACTGGTACTTGCACTCAGCTC-BHQ1 (TaqMan probe); Dlk1: 5′-CCCTCTGTGACAAGTGTGTAACTG-3′ (forward primer), 5′-GGAGCATTCGTACTGGCCTTTC-3′ (reverse primer), and 6FAM-CCAGGTCCACGCAAGTTCCATTGTTGGC-BHQ1 (TaqMan probe); Notch1: 5′-CCAACTGAGGACAGACGGACT-3′ (forward primer), 5′-GTGAGCACAGCCCTGAACCA-3′ (reverse primer), and 6FAM-CCCAGAAGCAGGCTCCACCCACATTCCAG-BHQ1 (TaqMan probe); Hes1: 5′-GTCCTAACGCAGTGTCACCTTC-3′ (forward primer), 5′-ATGGTCAGTCACTTAATACAGCTCTC-3′ (reverse primer), and 6FAM-TGGCTCCTCGCTCACTTCGGACTCCA-BHQ1 (TaqMan probe); Hey1: 5′-GAGGTGAAGGGAGAAAGGTGTCT-3′ (forward primer), 5′-CAAGTGCAGGCAAGGTCTACAT-3′ (reverse primer), and 6FAM-CTGTCTTGCTTGCTGCCAACGCTCCAA-BHQ1 (TaqMan probe); Hey2: 5′-AGCAATCAGCCCACCCTTGT-3′ (forward primer), 5′-CCGAAATGGACGACATCAGTTACT-3′ (reverse primer), and 6FAM-TCCGCAGCCTCCAGTCCTCAGCAGA-BHQ1 (TaqMan probe); Runx2: 5′-ACTCTTCTGGAGCCGTTTATGTG-3′ (forward primer), 5′-CCTTTGAGAACCGTGTGCATCT-3′ (reverse primer), and 6FAM-ACTAACCAACCCTTCCCTCCACTTCCCTG-BHQ1 (TaqMan probe); Bglap1/2: 5′-ATGTCCAAGCAGGAGGGCAATA-3′ (forward primer), 5′-CTCGTCACAAGCAGGGTTAAGC-3′ (reverse primer), and 6FAM-AGTGAACAGACTCCGGCGCTACCTTGG-BHQ1 (TaqMan probe); SPP1: 5′-CACATGAAGAGCGGTGAGTCTAAG-3′ (forward primer), 5′-GCTTTGGAACTTGCTTGACTATCG-3′ (reverse primer), and 6FAM-AGTCCCTCGATGTCATCCCTGTTGCCCA-BHQ1 (TaqMan probe). Gapdh was used as an internal control for normalization: 5′-CCAATGTGTCCGTCGTGGATC-3′ (forward primer), 5′-TGTAGCCCAAGATGCCCTTCA-3′ (reverse primer), and VIC-CCCTCAGATGCCTGCTTCACCACCTTCTT-MGBNFQ (TaqMan probe). PCRs were carried out on the 7300 Real-Time PCR System (Applied Biosystems), and data were processed with the manufacturer's software.

RESULTS

Loss of periostin leads to calcific aortic valve disease.

To explore the role of periostin in heart development, we analyzed the cardiac phenotype in mutant mice with targeted deletion of Postn gene (Postn^−/− mice). Gross anatomic examination of adult hearts revealed that the hearts of the 10-mo-old Postn^−/− mice had a pronounced dilation of the left ventricle and the aortic root (Fig. 1, A and B). A closer examination revealed that aortic valves in Postn^−/− mice were severely deformed and had a bicuspid-like morphology (Fig. 1, C and D). These pathological features, which often accompany severe cases of calcific aortic valve disease (5), were fully penetrant. Therefore, we investigated whether aortic valve leaflets of Postn^−/− mice undergo calcification. The calcium-specific von Kossa staining revealed that the leaflets of the aortic valve in 6-mo-old Postn^−/− mice have extensive calcium deposits, while no such deposits were observed in the age-matched wild-type animals (Fig. 1, E and F; Fig. 2). Occasional small calcium deposits were observed in 3-mo-old Postn^−/− animals (data not shown), while no calcifications were found in neonatal Postn^−/− mice. Since backflow of the blood through a leaky calcified aortic valve (aortic regurgitation) represents one of the main pathological features of calcific aortic valve disease (5), we analyzed cardiac blood flow in Postn^−/− mice, using Doppler echocardiography that revealed aortic regurgitation in 6-mo-old Postn^−/− animals (Fig. 1, G and H). One of the hallmarks of pediatric aortic valve disease is a reduced survival rate; therefore, we analyzed the survival rate of the offspring from Postn^+/− × Postn^+/− matings and found that the survival rate of the Postn^−/− animals was significantly reduced compared with both Postn^+/+ and Postn^+/− animals (Table 1). We did not detect any pathological changes in the Postn^+/− hearts. Thus only Postn^−/− mice exhibited pathological features characteristic of aortic valve disease.

Fig. 1.

Postn^−/− mice exhibit pathological features reminiscent of calcific aortic valve disease. A and B: left ventricle (LV) and aortic root were severely dilated (white arrowheads) in 10-mo-old Postn^−/− mice (B) compared with age-matched wild-type (WT) control mice (A). C and D: aortic valve in 10-mo-old Postn^−/− mice was severely deformed and had a bicuspid-like morphology (D) compared with age-matched WT control mice (C). E and F: calcification of the aortic valve leaflets (black precipitates, red arrowheads) in 6-mo-old Postn^−/− mice (F) revealed by von Kossa stain; no calcification was observed in age-matched WT control mice (E). G and H: Doppler echocardiographic examination of the cardiac blood flow in 6-mo-old Postn^−/− mice (H) revealed aortic regurgitation (red arrows) compared with WT control mice (G). AO, aorta; bpm, beats/minute; ECG, electrocardiogram; RV, right ventricle. Scale bars, 50 μm.

Fig. 2.

Calcification of the aortic valve leaflets in 6-mo-old Postn^−/− mice. A: Masson's trichrome staining of the aortic valve from 6-mo-old Postn^−/− mouse. B–E: high magnification of insets in A depicting aortic valve leaflet stained with Masson's trichrome stain (B and D) and adjacent section stained for calcium deposits with von Kossa stain (black precipitates, red arrowheads; C and E). F: aortic valve leaflet stained for calcium deposits with alizarin red (orange-red precipitates). G: high magnification of inset in F depicting aortic valve leaflet stained with alizarin red (calcium deposits, orange-red precipitates, black arrowheads). Scale bars, 100 μm.

Table 1.

Analysis of Postn^−/− offspring survival rate

Mating (n)	Offspring			Litter Size (means ± SD)
	Postn−/−	Postn+/−	WT
WT × WT (7)	0	0	36	5.1±1.5
Postn+/− × Postn+/− (69)	48 (19%/25%)	137 (53%/50%)	73 (28%/25%)	3.7±1.9
Postn−/− × Postn−/− (3)	8	0	0	2.7±0.6

Number of offspring with Postn^−/− genotype from Postn^+/− × Postn^+/− matings was significantly reduced compared with normal Mendelian distribution [P < 0.022, Postn^−/− vs. wild type (WT), n = 69]. As evidenced by the litter size, the survival rate of Postn^−/− mice was significantly lower compared with both Postn^+/− (P < 0.05) and WT (P < 0.005) animals. n, No. of matings. Percentages of offspring per genotype are shown at postnatal day 1 compared with expected frequency from Mendelian distribution (bold).

Ptn and Dlk1 are overexpressed in Postn^−/− hearts.

Since we have shown (21) that periostin is highly expressed in the prevalvular endocardial cushions of the OFT during embryonic development as early as E10.5, we hypothesized that adult valve calcific disease can have its roots in embryonic development. To determine whether the adult aortic valve disease has a developmental basis and to unravel the molecular mechanism of aortic valve calcification in Postn^−/− mice, we analyzed alterations of gene expression in the hearts of Postn^−/− E12.5 embryos (vs. wild-type embryos), i.e., 2 days after the onset of Postn gene expression in the heart. To do so, we used a strategy for a gene expression profiling based on a combination of suppression subtractive hybridization and reverse Northern analysis (50). Two subtracted cDNA libraries were generated, one enriched in clones upregulated and the other enriched in clones downregulated in the hearts of Postn^−/− embryos. One thousand one hundred and fifty-two clones from each library were organized on nylon filters. After these filters were hybridized with subtracted cDNA probes, we identified 69 genes with a difference in expression between the hearts of Postn^−/− mice and wild-type animals (Table 2). Our data suggested that two genes, pleiotrophin (Ptn) and delta-like 1 homolog (Dlk1), among others, were upregulated in the heart of Postn^−/− mice (Fig. 3A).

Table 2.

Genes differentially expressed in heart of Postn^−/− mouse

Clone Name	GB_ACC	Gene Name	Gene Symbol	UniGene ID	Expression Postn−/−/WT
NTA2A14	EF437324	Expressed in nonmetastatic cells 2, protein	Nme2	Mm.1260	↓
NTA2B16	EF437325	Phosphoglycerate kinase 1	Pgk1	Mm.336205	↓
NTA2B18	EF437326	Eukaryotic translation elongation factor 1 ε1	Eef1e1	Mm.36683	↓
NTA2C11	EF437327	Leucine rich repeat containing 41	Lrrc41	Mm.260786	↓
NTA2C12	EF437328	Iron responsive element binding protein 2	Ireb2	Mm.208991	↓
NTA2C17	EF437329	Integrin β3 binding protein (β3-endonexin)	Itgb3bp	Mm.257094	↓
NTA2D10	EF437330	Pleiomorphic adenoma gene-like 1	Plagl1	Mm.287857	↓
NTA2D14	EF437331	Unknown			↓
NTA2D15	EF437332	RD RNA-binding protein	Rdbp	Mm.279907	↓
NTA2D16	EF437333	Solute carrier family 25 (mitochondrial carrier, adenine nucleotide translocator), 5	Slc25a5	Mm.371544	↓
NTA2E1	EF437334	ATPase, Na+/K+ transporting, β1 polypeptide	Atp1b1	Mm.4550	↓
NTA2E20	EF437335	Coenzyme Q2 homolog, prenyltransferase (yeast)	Coq2	Mm.260661	↓
NTA2F8	EF437336	RIKEN cDNA 4921517L17 gene	4921517L17Rik	Mm.435448	↓
NTA2G11	EF437337	Septin 2	Sept2	Mm.428652	↓
NTA2H11	EF437339	Ribosomal protein L14	Rpl14	Mm.289810	↓
NTA2H12	EF437340	Nuclear autoantigenic sperm protein (histone-binding)	Nasp	Mm.257181	↓
NTA2H2	EF437338	RIKEN cDNA 1810026J23 gene	1810026J23Rik	Mm.39704	↓
NTA2J19	EF437341	Aldolase 1, A isoform	Aldoa	Mm.275831	↓
NTA2K23	EF437342	Selenoprotein	Sep15	Mm.29812	↓
NTA2L11	EF437343	Phosphoglucomutase 2	Pgm2	Mm.217764	↓
NTA2L12	EF437344	Coiled-coil domain containing 132	Ccdc132	Mm.45539	↓
NTA2L13	EF437345	NOP17	Nop17	Mm.181689	↓
NTA2L17	EF437347	Ubiquitin A-52 residue ribosomal protein fusion product 1	Uba52	Mm.297372	↓
NTA2L18	EF437348	A disintegrin and metallopeptidase domain 19 (meltrin β)	Adam19	Mm.89940	↓
NTA2M16	EF437351	RNA binding motif protein, X chromosome retrogene	Rbmxrt	Mm.24718	↓
NTA2M18	EF437352	F-box protein 11	Fbxo11	Mm.386857	↓
NTA2M6	EF437349	Myeloblastosis oncogene-like 2	Mybl2	Mm.4594	↓
NTA2M7	EF437350	Expressed sequence AI506816 (AI506816)	AI506816	Mm.205125	↓
NTA2N13	EF437353	5-aminoimidazole-4-carboxamide ribonucleotide formyltransferase	Atic	Mm.38010	↓
NTA3B12	EF437354	Zinc finger protein 289	Zfp289	Mm.43636	↓
NTA3B13	EF437355	Myosin, light polypeptide 6, alkali, smooth muscle and nonmuscle	Myl6	Mm.337074	↓
NTA3C8	EF437356	ATPase, H+ transporting, lysosomal V1 subunit B2	Atp6v1b2	Mm.249096	↓
NTA3I8	EF437357	Transmembrane protein 126A	Tmem126a	Mm.28963	↓
NTA3J20	EF437358	KRIT1, ankyrin repeat containing	Krit1	Mm.32368	↓
NTA3J21	EF437359	Exocyst complex component 7	Exoc7	Mm.22530	↓
NTA3K14	EF437360	Eukaryotic translation initiation factor 3, subunit 5 (ε)	Eif3 s5	Mm.182962	↓
NTA3K20	EF437361	Lysosomal-associated protein transmembrane 4A	Laptm4a	Mm.30071	↓
NTA3K21	EF437362	Zinc finger, AN1-type domain 5	Zfand5	Mm.292405	↓
NTA3K23	EF437363	Farnesyltransferase, CAAX box, β	Fntb	Mm.151174	↓
NTA3K24	EF437364	Ras homolog gene family, member U	Rhou	Mm.168257	↓
NTA3L6	EF437365	Diazepam binding inhibitor	Dbi	Mm.2785	↓
NTA3L7	EF437366	Pyrophosphatase (inorganic) 1	Ppa1	Mm.28897	↓
NTA3L8	EF437367	Glutamate dehydrogenase 1	Glud1	Mm.10600	↓
NTA3O4	EF437368	Eukaryotic translation elongation factor 1 α1	Eef1a1	Mm.435662	↓
PTA4A4	EF437369	Myeloid/lymphoid or mixed lineage-leukemia translocation to 3 homolog (Drosophila)	Mllt3	Mm.288898	↑
PTA4B19	EF437370	Mitochondrial cytb gene for cytochrome b	Cytb		↑
PTA4B20	EF437371	Zinc finger, AN1-type domain 3	Zfand3	Mm.220972	↑
PTA4B21	EF437372	Low-density lipoprotein receptor-related protein 10	Lrp10	Mm.28465	↑
PTA4B22	EF524068	Stathmin 1	Stmn1	Mm.378957	↑
PTA4C2	EF437373	Insulin-like growth factor 2 (Igf2)	Igf2	Mm.3862	↑
PTA4C6	EF437374	Procollagen, type VI, α1	Col6a1	Mm.2509	↑
PTA4D10	EF437375	SCY1-like 1 (Saccharomyces cerevisiae)	Scyl1	Mm.276063	↑
PTA4D22	EF437376	Nucleoporin 205	Nup205	Mm.261208	↑
PTA4F21	EF437377	Integrator complex subunit 2	Ints2	Mm.171304	↑
PTA4G24	EF437378	Potassium voltage gated channel, Shab-related subfamily, member 1	Kcnb1	Mm.387390	↑
PTA4I7	EF437379	Pleiotrophin	Ptn	Mm.279690	↑
PTA4J20	EF437380	ADP-ribosylation factor-like 6 interacting protein 1	Arl6ip1	Mm.29924	↑
PTA4K23	EF437381	UDP-glucose ceramide glucosyltransferase-like 1	Ugcgl1	Mm.261022	↑
PTA4L14	EF437382	Zinc finger protein 422, related sequence 1 (Zfp422-rs1)	Zfp422-rs1	Mm.276296	↑
PTA4N24	EF437383	SAC1 (suppressor of actin mutations 1, homolog)-like (S. cerevisiae)	Sacm1l	Mm.273671	↑
PTA4P11	EF437384	Metastasis associated lung adenocarcinoma transcript 1	Malat1	Mm.392400	↑
PTA5H10	EF437386	Troponin T2, cardiac	Tnnt2	Mm.247470	↑
PTA5H6	EF437385	Purine rich element binding protein B	Purb	Mm.296150	↑
PTA5K10	EF437387	Neuroguidin, EIF4E binding protein	Ngdn	Mm.21214	↑
PTA5M16	EF437388	Hexosaminidase (glycosyl hydrolase family 20, catalytic domain) containing	Hexdc	Mm.275130	↑
PTA5N12	EF437389	Dehydrodolichyl diphosphate synthase	Dhdds	Mm.100290	↑
PXL2L15	EF437346	Delta-like 1 homolog (Drosophila)	Dlk1	Mm.157069	↑
PXL3J19	EF524069	Tafazzin	Taz	Mm.268483	↑
PXL3N17	EF524070	Ras homolog gene family, member G	Rhog	Mm.259795	↑

Reverse Northern analysis using subtracted probes resulted in identification of 141 differential clones that gave a strong hybridization signal with 1 of the probes and at least 10-fold weaker signal with the probe subtracted in the opposite direction. Sequencing analysis of the corresponding cDNAs identified a total of 69 differentially expressed genes, thus revealing an average redundancy of 2.0 that indicates a very good coverage of genes differentially expressed in embryonic day 12.5 Postn^−/− hearts.

Fig. 3.

Analysis of gene expression in embryonic day (E) 12.5 Postn^−/− hearts. A: differential gene expression analysis identified Ptn and Dlk1 as genes overexpressed (arrowheads) in E12.5 Postn^−/− hearts. B: real-time PCR confirmed significant increase in expression of both Ptn and Dlk1 and revealed downregulation of Notch1 gene expression in E12.5 Postn^−/− hearts. C: real-time PCR analysis of the expression of Notch1 downstream target genes revealed that expression of Hes1, Hey1, and Hey2 (direct downstream targets of Notch1) were diminished in E12.5 Postn^−/− hearts. Conversely, expression of Runx2 and 2 main effectors of osteoblast mineralization, osteopontin and osteocalcin, were strongly upregulated in E12.5 Postn^−/− hearts. Bars represent means ± SD (n = 9).

Ptn and Dlk1/Notch1 signaling cascades have been implicated in Runx2-mediated osteoblast differentiation (2, 14, 49, 54). Therefore, we analyzed the expression of Ptn, Dlk1, and Notch1 in E12.5 Postn^−/− and wild-type hearts with real-time quantitative PCR. Both Ptn and Dlk1 expression were significantly upregulated in Postn^−/− hearts, while Notch1 expression was reduced (Fig. 3B; Table 3). Together, these data suggested that Ptn and Dlk1 upregulation in Postn^−/− mice may trigger changes in the Notch1 signaling cascade causing activation of osteoblast-specific gene expression program in the mesenchymal cells of the endocardial cushions of the OFT and calcification of the aortic valve.

Table 3.

Real-time PCR data

Gene	Ratio Postn−/−/WT	SD	P Value (n = 9)
Ptn	2.1	0.25	0.02283
Dlk1	3.7	0.35	0.00455
Notch1	−1.7	0.13	0.00879
Hes1	−1.3	0.03	0.00198
Hey1	−1.4	0.04	0.00181
Hey2	−2.1	0.08	0.00066
Runx2	4.8	0.14	0.00055
Osteocalcin (Bglap1/2)	1.4	0.11	0.03879
Osteopontin (SPP1)	1.8	0.11	0.00954

n, No. of samples per group.

Periostin functions upstream of Ptn, Dlk1, and Notch1.

To test the hypothesis that lack of periostin may activate Notch1-mediated calcification of the aortic valve, we determined the patterns of expression of Ptn, Dlk1, and Notch1 in the heart during embryonic development and investigated how they are affected by the absence of periostin. We analyzed expression of these genes in E12.5 wild-type and Postn^−/− embryos, using in situ hybridization focusing on the OFT. All three genes were expressed in the endocardial cushions of the OFT (Fig. 4). Specifically, in wild-type embryos, Ptn expression was restricted to the mesenchyme and the endocardial cells of the proximal (preseptal) endocardial cushions (Fig. 4A). Dlk1 was expressed in the OFT endocardium and the mesenchyme of both the proximal and distal cushions (Fig. 4C). Notch1 mRNA transcripts were mostly restricted to the mesenchyme of the distal portion of the OFT (Fig. 4E). Consistent with the real-time PCR data, we observed an increase in the abundance of Ptn and Dlk1 mRNA transcripts and a decrease in the abundance of Notch1 mRNA transcripts in the OFT of Postn^−/− embryos, with the biggest difference in the distal endocardial cushions, which give rise to the leaflets of the outlet (arterial) valves (Fig. 4). Remarkably, we also observed that in the Postn^−/− embryos the Ptn expression domain, normally restricted to the proximal (nonvalvular) OFT cushions, extended into the distal (valvular) cushions (Fig. 4, A and B). These findings further suggested that Ptn, Dlk1, and Notch1 expression are developmentally regulated by periostin.

Fig. 4.

In situ hybridizations revealed abnormal expression of Ptn, Dlk1, and Notch1 in E12.5 Postn^−/− outflow tract (OFT). Nonradioactive in situ hybridization on frontal sections of E12.5 OFT is shown. A and B: Ptn expression pattern. Ptn expression in WT embryos (A) was restricted to the proximal endocardial cushion (PC) and endocardium (EC). In Postn^−/− embryos (B), Ptn expression was markedly increased and the expression domain extended into the distal endocardial cushion (DC). C and D: Dlk1 expression pattern. Dlk1 mRNA transcripts were found throughout the OFT cushion mesenchyme and endocardium in both WT (C) and Postn^−/− (D) embryos. Note a significant increase of Dlk1 expression level in Postn^−/− embryos. E and F: Notch1 expression pattern. Notch1 expression was reduced in Postn^−/− embryos (F) and was mostly restricted to the mesenchyme of the distal cushion (E).

Since the in situ hybridization data and real-time PCR data clearly suggested that there was a link between Postn expression and expression of Ptn, Dlk1, and Notch1, we performed double fluorescent in situ hybridizations to determine expression domains of these genes in relation to Postn in the wild-type E12.5 OFT. Consistent with the in situ data (above), Ptn and Postn were expressed in mutually exclusive domains (Fig. 5, A–C). The Dlk1 gene was expressed throughout the OFT mesenchyme and coexpressed with Postn in the distal cushions (Fig. 5, D–F). The Notch1 expression domain completely overlapped that of Postn except for the endocardium, where no Postn expression was detected (Fig. 5, G–I). Ptn expression was restricted to the proximal cushion mesenchyme, where it was coexpressed with Dlk1, marking a mesenchymal cell population that appears to be distinct from the valvulogenic mesenchymal cells of the distal cushions and lacks both Notch1 and Postn expression (Fig. 5).

Fig. 5.

Colocalization of Postn with Ptn, Dlk1, and Notch1. Double fluorescent in situ hybridization on frontal sections of WT E12.5 OFT is shown. A–C: colocalization of Ptn and Postn gene expression; Ptn and Postn have mutually exclusive expression domains. D–F: colocalization of Dlk1 and Postn gene expression; Dlk1 was expressed in both distal and proximal endocardial cushions and was coexpressed with Postn in the distal cushion. G–I: colocalization of Notch1 and Postn gene expression; Notch1 and Postn expression domains completely overlapped.

Thus Dlk1 and Notch1 are coexpressed with periostin in the distal cushions of the OFT (Fig. 5, D–I). Absence of periostin expression in the E12.5 cushions, i.e., 48 h after the onset of periostin expression in the heart, leads to a 3.7-fold increase in Dlk1 expression (Fig. 3B, Fig. 4, C and D; Table 3). Since Dlk1 acts as a negative regulator of Notch1 (2), overexpression of Dlk1 combined with diminished expression of Notch1 should lead to a strong repression of Notch1 signaling in Postn^−/− hearts. Indeed, real-time quantitative PCR analysis revealed that the expression of Hes1, Hey1, and Hey2, direct downstream targets of Notch1 (18, 27), were downregulated in Postn^−/− hearts (Fig. 3C; Table 3). This is a direct indication that Notch1 signaling is repressed in Postn^−/− hearts.

Postn and Ptn are expressed in mutually exclusive domains in E12.5 OFT (Fig. 5, A–C). Absence of periostin expression in the Postn^−/− E12.5 OFT, i.e., 48 h after the onset of periostin expression in the heart, results in expansion of Ptn expression domain into the distal OFT cushions (Fig. 4, A and B) (i.e., where periostin is normally expressed in WT mice) and a 2.1-fold increase in its overall expression (Fig. 3B; Table 3), which indicates that periostin has a suppressive effect on Ptn expression. Thus these data suggest that periostin functions upstream of Ptn, Dlk1, and Notch1 and its absence leads to suppression of Notch1 signaling.

Suppression of Notch1 signaling in Postn^−/− hearts leads to mesenchymal cell differentiation along an osteogenic lineage.

Our data strongly suggested that lack of periostin causes repression of Notch1 signaling in E12.5 hearts. Therefore, we analyzed expression of the central transcriptional regulator of osteoblast cell fate, Runx2, which was implicated in aortic valve calcification downstream of Notch1 (14) and was found to be upregulated in patients with valvular calcification (39). Real-time PCR analysis revealed that Runx2 was strongly upregulated in the hearts of E12.5 Postn^−/− embryos (Fig. 3C; Table 3). Moreover, we also found that the expression of two main effectors of osteoblast mineralization, osteopontin and osteocalcin, were also significantly increased (Fig. 3C; Table 3). This suggests that even at this early stage of embryonic development a lack of periostin leads to activation of an osteogenic program in prevalvular endocardial cushions that is manifested as calcific aortic valve disease in early adult life.

DISCUSSION

Origin and formation of aortic valve.

During cardiogenesis, aortic valve leaflets develop from the endocardial cushions of the distal OFT as a result of a highly coordinated process, which involves an endothelial-to-mesenchymal transformation, remodeling accompanied by differentiation of cushion mesenchyme into cells of a fibrogenic lineage, and cushion maturation (8, 17). Several signaling pathways and transcription factors that control early stages of semilunar valve development, i.e., cushion formation, have been identified. Loss-of-function mutations in various components of the bone morphogenetic protein (BMP)2 signaling pathway in mice result in anomalies of semilunar valves, suggesting that the BMP2 signaling pathway plays an important role in development of the OFT endocardial cushions (6, 10, 12). The transforming growth factor (TGF)-β2 signaling pathway also seems to be critical for this process, since deletion of TGFβ2 in mice leads to the loss of semilunar valves (44). This is highlighted by the observation of severe OFT defects in embryos deficient for Pinch1, an adaptor protein acting upstream of TGFβ2 (22). In addition to BMP2 and TGF-β2 signaling pathways, several transcription regulators, including Sox4 (45), Nfatc1 (9, 40), Foxp1 (52), and Tbx20 (48), were shown to be essential for early stages of semilunar valve development. However, the genetic network guiding late stages of cushion remodeling and maturation is largely unknown.

Genetic basis of calcific aortic valve disease.

Calcific aortic valve disease is identified by thickening and calcification of the aortic valve leaflets and often leads to aortic stenosis (47), which represents the second most common indication for cardiac surgery (43). It affects up to 25% of the population over 65 yr of age (47) and up to 2% of pediatric patients (42).

Age-related aortic valve calcification has many features of an active pathobiological process, including inflammation (36, 38), lipoprotein deposition (34), and calcification (33, 39). Certain alleles of APOE and ESR1 were found to be associated with increased risk of age-related aortic valve calcification in adults, indicating that genetic factors influence disease development (29, 32).

Calcification of the aortic valve in pediatric patients is commonly associated with bicuspid aortic valve (BAV) (24). The heritability of BAV was estimated at 89%, suggesting that pediatric aortic stenosis is almost entirely determined by genetic factors (7). Loss-of-function mutations in NOTCH1 were shown to be associated with both BAV and age-related aortic valve calcification (14). However, two recent genomewide linkage studies for valve calcification susceptibility loci have identified multiple other chromosomal loci linked to calcific aortic valve disease, in addition to 9q harboring NOTCH1. Strong linkage was found to chromosomes 1q42, 2q37, 5q21, 13q34, 16q22, 17q23-24, 18q22, and 19q13, suggesting that multiple other genes are involved in the development of aortic valve disease (4, 25).

Although the upstream events may differ in different forms of aortic valve disease, they all lead to the activation of Runx2 (1, 14, 39), a central transcriptional activator of osteoblast differentiation that directly activates expression of several osteoblast-specific genes (11). Overexpression of Runx2 and its direct downstream targets osteopontin and osteocalcin was found to be associated with calcification of the aortic valve in patients with aortic stenosis (14, 39) and in a mouse model of the aortic valve disease (1).

Periostin controls OFT mesenchymal cell fate by modulating Notch1 signaling.

Periostin has recently become a popular target of investigation for its role in cell migration, cancer metastasis, myocardial remodeling, and atrioventricular valve maturation (23, 31, 35), but the discovery that periostin plays a key role in aortic valve development and OFT mesenchymal cell fate determination represents a new and very intriguing avenue of research. Our data suggest that periostin plays a critical role in the morphogenetic remodeling of the OFT endocardial cushions, i.e., the process that ultimately leads to the formation of muscular septal tissue from proximal cushions and mature aortic valve from distal cushions (8, 51). Our data indicate that periostin regulates Dlk1/Notch1 and Ptn signaling cascades in the OFT during embryonic development. Dlk1 is a negative regulator of Notch1 (2). Deactivating mutations in Notch1 can lead to Runx2-dependent overexpression of osteocalcin and calcification of the aortic valve (14). Ptn is a growth/differentiation factor that has been implicated in early osteoblast differentiation (54). Overexpression of Ptn resulted in an increase in bone thickness and mineralization (49). Binding of Ptn to its receptor, the receptor-type protein tyrosine phosphatase β/ζ (RPTPβ/ζ), increases phosphorylation and subsequent degradation of β-catenin (26). Activated (dephosphorylated) β-catenin was shown to inhibit Runx2-mediated transcriptional activation of the osteocalcin promoter (19). Therefore, Ptn-mediated β-catenin degradation can lead to activation of osteocalcin transcription. Under normal circumstances, periostin appears to negatively regulate Ptn and Dlk1 expression in the distal OFT cushions, while it seems to be necessary for maintaining adequate levels of Notch1 in these cushions. Our data suggest that this suppresses the default mesenchymal cell potential to differentiate into osteoblast-like cells and promotes their differentiation along a fibrogenic lineage necessary for normal valvular development. Thus we propose that the lack of periostin causes derepression of the osteogenic program by causing overexpression of Dlk1 and misexpression of Ptn in the distal OFT cushions, which in turn leads to suppression of Notch1 signaling and upregulation of the central regulator of osteoblast cell fate, Runx2. As summarized in Fig. 6, activation of Runx2 combined with suppression of its corepressors Hey1 and Hey2 would lead to activation of osteopontin and osteocalcin transcription and ultimate calcification of the aortic valve leaflets in early adulthood.

Fig. 6.

Periostin signaling cascade: model for proposed periostin signaling cascade in the cardiac OFT. We propose that periostin binds to an integrin receptor. This leads to suppression of Ptn and Dlk1 expression that results in activation of receptor-type protein tyrosine phosphatase β/ζ (RPTPβ/ζ) and Notch1. This causes activation of Runx2 corepressors β-catenin, Hey1, and Hey2 and, combined with suppression of Runx2 transcription, leads to suppression of osteogenic program in the mesenchymal cells of the distal OFT cushions and their subsequent differentiation along a fibrogenic lineage. Lack of periostin triggers a reversal of the cascade, activation of osteopontin and osteocalcin transcription, and ultimate calcification of the aortic valve leaflets (see text).

Recently, a marked reduction in periostin expression was observed in pediatric patients with aortic stenosis (46), suggesting that our mouse developmental findings may extend to the human disease. Thus delivering periostin to calcific valves at the early stages of the disease may have a therapeutic potential because it may prevent disease progression. The intriguing possibility of β-catenin phosphorylation (and degradation) as a result of RPTPβ/ζ inhibition by Ptn and subsequent further enhancement of Runx2-mediated activation of osteopontin and osteocalcin promoters needs to be further investigated. Thus periostin has emerged as an important regulator of cardiac valve development and a potential therapeutic target.

GRANTS

This work was supported by National Institutes of Health Grants HL-077342 (to S. J. Conway) and HL-33756, NCRR C06, and COBRE 2P20-RR-016434 (to R. R. Markwald).