PERIOSTIN REGULATES ATRIOVENTRICULAR VALVE MATURATION

Abstract

Cardiac valve leaflets develop from rudimentary structures termed endocardial cushions. These pre-valve tissues arise from a complex interplay of signals between the myocardium and endocardium whereby secreted cues induce the endothelial cells to transform into migratory mesenchyme through an endothelial to mesenchymal transformation (EMT). Even though much is currently known regarding the initial EMT process, the mechanisms by which these undifferentiated cushion mesenchymal tissues are remodeled “post-EMT” into mature fibrous valve leaflets remains one of the major, unsolved questions in heart development. Expression analyses, presented in this report, demonstrate that periostin, a component of the extracellular matrix, is predominantly expressed in post-EMT valve tissues and their supporting apparatus from embryonic to adult life. Analyses of periostin gene targeted mice demonstrate that it is within these regions that significant defects are observed. Periostin null mice exhibit atrial septal defects, structural abnormalities of the AV valves and their supporting tensile apparatus, and aberrant differentiation of AV cushion mesenchyme. Rescue experiments further demonstrate that periostin functions as a hierarchical molecular switch that can promote the differentiation of mesenchymal cells into a fibroblastic lineage while repressing their transformation into other mesodermal cell lineages (e.g. myocytes). This is the first report of an extracellular matrix protein directly regulating post-EMT AV valve differentiation, a process foundational and indispensable for the morphogenesis of a cushion into a leaflet.

Introduction

Valvulogenesis commences with an endothelial to mesenchymal transition (EMT) resulting in a swellings of undifferentiated mesenchymal tissue, known as endocardial cushions (ECs) (de la Cruz and Markwlad, 1998; de la Cruz et al., 1977; Eisenberg and Markwald, 1995; Kinsella and Fitzharris, 1980; Kinsella and Fitzharris, 1982; Krug et al., 1985; Manasek, 1970; Manasek et al., 1973; Markwald et al., 1977; Markwald et al., 1975; Markwald and Smith, 1972; Person et al., 2005). As blood circulates through the heart, these EC swellings function as primitive valves prohibiting retrograde blood flow between the atria and ventricle (atrioventricular canal) and the ventricle and aortic sac (outflow tract). Over the past thirty years much effort has been put forth to determine the inductive signals essential for initiating and sustaining EMT. To date, more than 100+ genes have been linked to this important morphogenetic event (Camenisch et al., 2002; Holifield et al., 2004; Lakkis and Epstein, 1998; Ma et al., 2005; Mjaatvedt and Markwald, 1989; Mjaatvedt et al., 1998; Savagner, 2001; Schroeder et al., 2003; Yamamura et al., 1997). However, the EMT event leading to the formation of prevalvular cushions is just the initial phase in valvuloseptal morphogenesis. The molecular and cellular events that occur after this initial EMT event are quite complex and poorly understood. This “post-EMT” process, which will yield a mature valve leaflet or cusp and associated tension supporting apparatus involves the migration and proliferation of mesenchymal cells to distend the forming cushions into the lumen, followed by further elongation of the tissue and its gradual attenuation (sculpting/thinning) and maturation. This remodeling of cushions into valves is directed and proceeds through the differentiation of cushion mesenchyme into fibroblastic interstitial cells (Aikawa et al., 2006; Goldsmith et al., 2004; Hinton et al., 2006; Kruithof et al., 2007; Lincoln et al., 2004; Lincoln et al., 2006; Schoen, 1999; Weber, 1989). Based on this fundamental change in cell phenotype, the matrix surrounding these fibroblasts matures into highly organized and compacted lamellar arrays of collagen rich fibrous tissue that confers the structural stability to the growing valve necessary for handling the increasing hemodynamic stress and strain of the beating heart (Icardo and Colvee, 1995a; Icardo and Colvee, 1995b; Kruithof et al., 2007). The process by which the valve leaflet remodels and matures during later stages of valvulogenesis is poorly understood. However, the spatio-temporal pattern of periostin expression invokes a critical role for this gene in post-EMT valvulogenesis (Kern et al., 2005; Kruzynska-Frejtag et al., 2001; Lindsley et al., 2005; Norris et al., 2004; Norris et al., 2005). Periostin is a secreted, matricellular protein which is evolutionarily conserved and contains 4-repeated domains related to the Drosophila midline fasciclin-1 gene (Horiuchi et al., 1999; Litvin et al., 2005; Takeshita et al., 1993). The mammalian fasciclin gene family comprises 4 members: periostin, β IG–H3, stabilin-1, and stabilin-2. These genes have been shown to play important roles in cellular processes such as adhesion, migration, and differentiation (Diamond et al., 1993; Elkins et al., 1990; Falkowski et al., 2003; Gillan et al., 2002; Hortsch and Goodman, 1990; Kim et al., 2002; Kim et al., 2000; Nakamoto et al., 2002; Oshima et al., 2002; Park et al., 2004; Snow et al., 1989; Takeshita et al., 1993; Yan and Shao, 2006). In this report, we demonstrate that periostin protein is highly expressed in the post-EMT murine cushions as they undergo remodeling and maturation. Expression is maintained throughout valvulogenesis; being expressed in the mature fibrous AV and OT leaflets, the tendinous chords of the supporting apparatus and the fibrous anchoring annulus into post-natal and adult life.

Although gene-targeted periostin mice have previously been described, a detailed characterization of a heart valve phenotype was not reported (Kii et al., 2006; Norris et al., 2007b; Oka et al., 2007; Rios et al., 2005). As a starting point, we sought to determine if periostin was required for proper maturation of AV valve leaflets where we have found structural defects in a subset of periostin null animals. Commencing at E12.5, pockets of cells abnormally positive for myosin heavy chain were detected in the superior and inferior AV cushions of null animals. This ectopic myocardial staining persisted throughout development and was maintained into adulthood. AV valves in the adult periostin null animals were hypoplastic, failed to completely delaminate, and contained significant presence of myocardial tissue. Additionally, these valves had a significant reduction in collagen expression and organization which resulted in a reduction of biomechanical properties. The chordae tendineae, an area of intense periostin expression, exhibited significant defects in the number of branches and thickness in their stems in the null mice. Rescue experiments demonstrated that periostin not only promoted post-EMT AV cushion migration, but was also essential for the synthesis of collagen while repressing cardiac myosin expression. Thus, the data substantiate a hierarchical role for periostin in the differentiation and remodeling pathways of cushion mesenchymal cells into mature valve fibroblasts which is indispendable for proper leaflet formation and function.

Materials and Methods

Immunohistochemical localization of periostin

Preparation and characterization of affinity purified anti-mouse periostin antibodies were described previously (Kruzynska-Frejtag et al., 2004). Immunohistochemical procedures for mouse embryos and adult hearts were described previously (Sugi et al., 2004).Briefly, mouse embryos from ED9.0-ED13.5, hearts from mouse fetuses at ED16 and adults were collected in Earl’s-buffered saline solution (EBSS, Invitrogen) and fixed with 100% methanol at −20 C overnight. Fixed samples were processed through descending methanol series at 4 C and embedded in Paraplast X-TRA. Serial 6 μm sections were cut and deparaffinized in xylene. Sections were blocked with 10% normal goat serum (MP Biomedicals) in 1% bovine serum albumin (BSA, Sigma) / phosphate buffered saline (PBS, pH.7.4) and processed for double immunohistochemistry. Sections were incubated with anti-mouse periostin antibodies (3μg/ml) followed by treatment with FITC-labeled goat anti-rabbit IgG (MP Biomedicals). Normal rabbit IgG was used as a negative control for anti-mouse periostin antibodies. The sections were then rinsed and incubated with MF20 (Developmental Studies Hybridoma Bank) followed by RITC-labeled goat anti-mouse IgG (MP Biomedicals). Immunostained sections were examined under a Leica BMLB fluorescent microscope.

Periostin null mice analysis

The generation of the periostin null mouse has previously been described (Kii et al., 2006; Norris et al., 2007b; Oka et al., 2007; Rios et al., 2005). For analysis of embryonic time points, matings were set up between heterozygote parents. Following caesarian section, embryo’s at E12.5 were harvested (n=50), genotyped by PCR and analyzed by either immunohistochemistry, or Masson’s trichrome staining (DakoCytomation) per standard procedures. Adult wild-type (n=15), heterozygote (n=20) and periostin null (n=28) mice were analyzed for cardiac pathologies.

3-D Hanging drop assays

AV cushions were dissected free of myocardium from E12.5 wild-type and periostin null mice. Individual AV cushions were placed in 25 μl hanging drops (OptiMEM supplemented with 1% heat inactivated fetal bovine serum, 5 μg/ml insulin, 5μg/ml transferring, 5 ng/ml selenium, 100 units/ml Penicillin, 100 mg/ml Streptomycin). 16 cushions were isolated from periostin null mice of which 8 were randomly chosen for the rescue experiments. For rescue experiments, 10 μg/ml of full-length purified mouse periostin (R&D systems) was added to the hanging drops. 8 AV cushions were isolated from wild-type mice and placed in culture as described above. Cushion explants were grown in culture for 7 days after which total protein lysate was taken and processed for Western analysis. The entire procedure was repeated 3 times resulting in the analysis of 24 null, 24 wild-type, and 24 rescued AV cushions.

Western Blot Analysis

AV cushions were lysed in 1X RIPA cocktail which included protease inhibitors (Sigma). Lysate was mixed with a 2X protein loading dye and loaded onto 4–15% SDS-PAGE protein gel (BIO-RAD). Gels were blotted onto nitrocellulose and probed for various marker antibodies: (MHC-abcam (1:1,000), MLC-abcam (1:1,000), Periostin-abcam (1:10,000), collagen I-abcam (1:1000), actin-chemicon (1:5000). Secondary antibodies used were goat anti-mouse (Sigma-1:10,000) and goat anti-rabbit (Sigma-1:10,000). Immunopositive bands were detected with Visualizer (UpState Biologicals). Western analyses was performed on each of the triplicate set of experiments. Densitometric analyses was performed using Adobe Photoshop CS3 and results are presented graphically. Actin is used as a normalization control.

Atomic Force Microscopy Measurements

Chordae tendineae (dissected from the right AV tricuspid valve apparatus) and mural tricuspid valve leaflets were isolated from adult periostin null and wild-type mice. The isolated mouse chordae and leaflets were placed on individual glass slides. The tissues were rinsed of any residual salts in double distilled water and maintained in a moist condition within the isolation chamber. An Asylum MFP-3D Series AFM instrument (Asylum Research) was used to scan the sample. The scanning head used to make the measurements has a 100 μm scanning area. The cantilever used was silicon nitride coated with a diving-board tip at 0.01 N/m constant force (Asylum Research). These tips have a radius of curvature of less than 20 nm. To obtain stiffness, the AFM was used in the “indentation mode”. First, the sample was imaged to ensure that the tip was on the sample. The tip was then brought toward the surface until it contacted the sample then was retracted a few nanometers. A force distance curve was taken at this point by allowing the tip to approach the surface and retract, in quick pulses; this then relayed the static deflection of the tip on the surface. An average of 6 independent spots had a force distance curve performed on them. These were then averaged and normalized against the glass slide surface from which the force distance curve was generated. The result was a plot with two averaged curves, one for the approach and one for the retraction demonstrating pico-Newton forces along the nanometer scale of approach and retraction. In order to corroborate that the data was reliable, the curves generated were analyzed for similar slopes on both the approach and retraction. Stiffness, by definition, is calculated as the force/distance (F/D) traveled which is the slope of the curve obtained from the force distance curve.

Results

Periostin is predominantly expressed during post-EMT valvulogenesis

The expression of periostin RNA during cardiac development has been previously documented in the chick and the mouse (Kern et al., 2005; Kruzynska-Frejtag et al., 2001; Norris et al., 2004; Norris et al., 2005). However, the expression and localization of periostin protein has not been rigorously addressed, especially in post-EMT development of the mouse heart. Therefore, specific anti-mouse periostin antibodies were developed and utilized in immunohistochemical analyses of various timepoints during cardiovascular development. Periostin protein is first detected within the atrioventricular (AV) canal during the EMT process at embryonic day (ED) 9.5 (Fig. 1A, B). At this stage, cytoplasmic and extracellular expression/secretion of periostin is seen within transformed mesenchymal cells (Fig. 1B). Periostin expression is not detected in the myocardium of the AV canal. Intense periostin immunostaining is also detected in the umbilical vessels in ED9.5 embryos (Fig. 1A). As EMT progresses, periostin expression becomes predominantly extracellular with intense staining in the enlarging AV and outflow tract (OT) cushions at ED 11 and ED12.5 (Fig. 1 C, D, E, F). By E12.5, the endothelium lining the ventricular trabeculae and the epicardial epithelium express periostin protein (arrow heads and arrows, respectively in Fig. 1 E, F).

Figure 1. Periostin protein localization during mouse embryonic cardiac development.

A Sagittal section of an ED 9.5 mouse embryo shows onset of periostin expression (green) in the atrioventricular (AV) canal (arrows). Note the strong periostin expression within umbilical vessels (UV). MF20 staining is confined to the myocardium (red). B Higher magnification view of the AV canal in A. Arrows indicate subendocardial expression of periostin. Periostin expression is also detected in the cytoplasm of forming cushion mesenchymal cells (arrows). C Sagittal section of an ED11 mouse embryo shows periostin expression in the AV and outflow tract (OT) cushion mesenchyme. D Higher magnification view of the cushion mesenchyme in C. Both AV and OT cushion mesenchyme show intense periostin expression. E Frontal section of an ED 12.5 mouse embryo shows intense periostin expression in the OT cushion. Note, periostin expression extends beyond the border of the OT covered by the myocardium and is also expressed in the epicardium (arrow heads). F Frontal section of an ED 12.5 mouse embryo shows intense periostin expression in the AV cushion. Examination by serial sections did not reveal any MF20 staining in the AV cushion mesenchyme at ED12.5. Also note, expression within the epicardium (arrow heads) and endothelial lining of the ventricular trabeculae (arrows). A, atrium; DA-Dorsal Aorta; 1st, first branchial arch; IC, inferior cushion; SC, superior cushion; V, ventricle. Scale Bars: A,C–F=100μm; B=50 μm

After the initial EMT process has been completed, strong expression is detected within the developing aortic and pulmonary valve leaflets at E13.5 and E16 (Fig. 2A, E). Expression within the aortic valve appears to be widespread whereas the pulmonary valve exhibits more intense expression on the ventricular aspect of the leaflet. Weak expression is also evident within the aorta and the anterior aspect of the interventricular septum, representing the ventral mesenchyme of the fused AV cushions (arrows in Fig. 2A). Additionally, intense expression is seen in the mesenchymal cap of the atrial septum that attaches to the fused AV cushions (Fig. 2 B,C-arrow head). The developing mitral and tricuspid AV valves also exhibit intense periostin staining at E13.5 and E16 (Fig. 2A–C, EG). Within these AV valves, expression is predominantly localized to the atrial aspect of each valve. Additionally, periostin is expressed in the endothelial lining of the ventricular trabeculae and epicardium at E13.5 (Fig. 2D) and can also be seen in epicardial derived cells in the myocardial wall as well as the AV sulcus at E16 (arrows in Fig. 2H and asterisk in 2F, respectively). At E16, on either side of the AV septum, strands of periostin expression are seen associated with fibrous components of the developing conduction system (Fig. 2G-arrows). Periostin expression is more intense where the future annulus will develop that serves to anchor the base of the free AV valve leaflets to the myocardial wall (asterisks in Fig. 2E–G). In the adult mouse, periostin expression is also intense within the developing tendinous cords of the valve supporting apparatus (Fig 2 I–L). Within the AV valves (tricuspid and mitral), expression is more pronounced on the ventricular aspect of the leaflets. Periostin expression is diminished within the epicardium epithelium after ED16 and absent in the adult heart (data not shown).

Figure 2. Periostin protein localization during mouse fetal cardiac development and within the adult heart.

A Frontal section of an ED 13.5 heart, showing intense periostin immunostaining (green) in the developing aortic valves (AoV) and aorta (Ao). Periostin expression is also detected within the inter ventricular septum (IVS, arrows). B Frontal section of an ED 13.5 heart, showing intense periostin immunostaining in the developing tricuspid (TV) and mitral valves (MV) and the mesenchymal cap of the dorsal mesenchymal protrusion (DMP) (arrow). C Higher magnification view of the TV, MV, and DMP in B reveals fibrous immunostaining of periostin. D Periostin expression is detected within the forming epicardium (Ep) covering the ventricle and atrium and is also detected in the subendocardial space of the trabeculae (Tb). E Frontal section of an ED16 heart, showing intense periostin immunostaining in the forming pulmonary valves (PV), aorta (Ao), pulmonary trunk (PT), and forming fibrous annulus (arrow heads). F Frontal section of an ED16 heart, showing intense periostin immunostaining within the developing tricuspid valves (TV), epicardium (Ep), fibrous annulus (arrow head) and AV sulcus (asterisk). G Frontal section of an ED16 heart, showing intense periostin staining in the developing mitral valves (MV). Periostin staining is also detected near the luminal edge of the inter ventricular septum (arrows), and the fibrous annulus (arrow head). H Higher magnification view of the epicardium at ED16, showing robust epicardial and EPDC staining (arrows). I Image of an adult heart showing intense periostin immunostaining within the aortic valves (AoV) and aorta (Ao). Note, the fibrous staining of periostin in the aortic wall. J Intense immunostaining is detected in the tendinous chord (TC), attached to the papillary muscle (PM). K, L Intense periostin immunostaining is detected in the adult tricuspid (TV) and mitral valves (MV). LA, left atrium; LV, left ventricle; RA right atrium; RV, right ventricle. Scale Bars: A,C–I, K,L=100μm, B,J=500μm.

Periostin null mice exhibit embryonic lethality coincident with AV cushion anomalies

The periostin gene targeted mouse and over-expressor mouse have been previously described by us and others (Kii et al., 2006; Norris et al., 2007b; Oka et al., 2007; Rios et al., 2005). Using these mouse models, results on tooth development and adult myocardial remodeling in the ventricles have been published. However, to date, a thorough investigation of potential cardiac valve defects have not been reported. An analysis of mice generated from het X het matings (n=136) have indicated that the expected Mendelian numbers of offspring were not statistically different between genotypes (Fig 3). However, it was noted that het X het matings generated less offspring than the wild-type matings. This suggested that a sub-population of the null animals were probably dying during development. Figure 3 shows that the viability of the embryos is significantly reduced when both of the periostin alleles are removed; lethality being most apparent after the E10.5 timepoint. Based on the intense expression of periostin during post-EMT cushion morphogenesis we hypothesized that knocking out this gene may have resulted in valve defects that were not compatible with survival. At E10.5, no morphological defects were observed (data not shown). However, by E12.5, periostin null mice exhibited small islands of MF20 positive cells within both the superior and inferior cushions (Fig. 4). MF20 is a well-established marker for contractile cardiac sarcomeric myocytes and is not normally expressed within either the superior or inferior AV cushions of wild-type mice. This suggested that periostin may function to repress aberrant myocardial differentiation within the AV cushion mesenchyme. To further test this possibility, AV cushions from periostin null and wild-type mice were dissected and placed in hanging drop cultures. Consistent with in vivo findings, periostin null AV cushions were positive for the cardiac muscle markers myosin heavy chain (MHC) and myosin light chain (MLC) (Fig. 5).

Figure 3. Embryonic lethality of periostin deficient mice.

Numbers of viable neonates at P1 were counted from heterozygote matings (n=136) and compared to the theoretical Mendelian numbers. No statistically significant differences were observed between each of the genotypes. However, a comparison between litter size of Het X Het and WT X WT matings demonstrated some degree of lethality. This was further substantiated in the null X null matings which have less than one-half the litter size of the wild-type mice. An analysis of embryo numbers at E10.5 indicated that it was after this timepoint when the embryo’s began to die.

Figure 4. Periostin null mice exhibit aberrant differentiation of AV cushion mesenchyme.

A. Sagittal sections through E12.5 wild-type mouse hearts were analyzed for MF20 (green) and periostin (red) co-expression. B. MF20 expression is absent within the AV cushions of wild-type mice. C. Periostin null mice exhibit ectopic, de novo, sarcomeric myosin expression indicating aberrant differentiation of AV cushion mesenchyme into a cardiac myocyte lineage.

Figure 5. Western analysis of myocyte and fibroblast markers on periostin null, wild-type, and “rescued” AV cushion mesenchyme.

AV cushions from periostin null mice were placed in hanging drop cultures and incubated with either purified periostin protein (10μg/ml) or PBS. AV cushions from wild-type mice were used as control tissues. After 7 days tissues were harvested and analyzed for differentiation markers by immunoblotting. A. Periostin null cushions exhibit high levels of the myocardial markers: myosin heavy and light chains (MHC, MLC) and low levels of fibroblastic markers: collagen 1α1 and 1α2. Addition of periostin to the culture medium induces expression of collagen 1α1 and 1α2 while repressing MHC and MLC (arrow heads). B. Graphical representation of Western analyses presented in A. Densitometric analyses were performed as described in Material and Methods. Values obtained were compared against wild-type values (baseline) and represented as relative percent change.

Purified periostin was administered to half of the null tissues in culture and after 7 days the tissues were analyzed by Western blotting for MHC, MLC and collagen I expression (Fig. 5). Both myocardial markers, MHC and MLC, were decreased whereas the expression of the fibroblast markers, collagen Iα1 and collagen Iα2, were increased in null AV cushion mesenchyme that received purified periostin. This addition of exogenous periostin significantly repressed the muscle markers in the nulls while stimulating fibroblast markers, indicating during normal development periostin in undifferentiated cushion mesenchyme correlated with expression of fibroblast markers, whereas its absence correlated with expression of myocardial markers.

Interatrial septal defects observed in the periostin null mice

AV cushion tissue is also crucial for proper formation of septal structures (de la Cruz and Markwlad, 1998). Normal AV and atrial septation requires the fusion of the dorsal mesenchymal protrusion (DMP), or spinal vestibuli, and the mesenchymal cap on the leading edge of the primary atrial septum with the mesenchyme of the AV septum (formed by the fusion of the superior and inferior AV cushions). The formation of this AV mesenchymal complex anchors the atrial septum and progressively closes the primary atrial foramen (Snarr et al., 2007). Due to the intense expression of periostin in the AV cushions and in the mesenchymal cap of the interatrial septum, periostin null mice were examined for defects in the closure of the primary atrial foramen. 100% of the periostin null mice exhibited large primary atrial septal defects (Fig. 6). Sections were taken from comparable planes and regions of wild-type mice and no defects were seen. Due to the change in overall shape of the periostin null hearts (being smaller and rounder), it was not possible to get the exact plane. However, through a whole mount view, it is possible to see the extent of the ASDs which were “probe patent” (Fig. 6C).

Figure 6. Periostin null mice exhibit interatrial septal defects.

Frontal sections of adult wild-type (A) and periostin null (B) animals stained for Masson’s trichrome. Periostin null mice have a failure of fusion between the dorsal mesenchymal protrusion and the extending primary atrial septum resulting in atrial septal defects. The hole formed by this failed fusion results in an interatrial communication between the right and left atrium evidenced by an insect needle on gross examination (C). RA-right atrium, RV-right ventricle, LA-left atrium, IVS-interventricular septum, IAS-interatrial septum, IAC-interatrial communication.

Periostin is essential for the proper differentiation and maturation of the AV valves

As the mitral and tricuspid valves elongate, attenuate and maturate during fetal and post-natal development, the mesenchymal cells differentiate into fibroblasts and secrete collagen (Kruithof et al., 2007). Periostin is most intensely expressed during these post-EMT stages of valvular modeling. Thus, adult periostin null mice were examined for defects in post-EMT cardiac valve maturation. The periostin null mice did exhibit significant defects in the mitral and tricuspid valve architecture and cellular content. In wild-type mice, these valves stained exclusively blue with Masson’s trichrome stain, which indicates a primarily fibrous, collagen composition. However, the periostin null mice have a significant reduction in blue staining, indicating a lack of organized fibrillar collagen containing matrix. Quantitative Western analyses indicated that there is a ~70% reduction in collagen 1α1 and 1α2 protein in the heterozygote and null leaflets. Additionally, the tricuspid septal leaflet of the periostin null mouse had extensive myocardial tissue (red) associated with the ventricular aspect of the leaflet. This finding was also present on the ventricular surface of the left and right mural AV valve leaflets indicating a potential defect in delamination. The mural leaflets normally develop upon a template/substrate of myocardial tissue that is later remodeled to remove or replace myocardial cells during delamination of the leaflet from the ventricular wall, ultimately leaving a purely fibrous leaflet as shown in the wildtype heart (Gaussin et al., 2005). Periostin is most intensely expressed in the boundary between AV cushion tissue and associated myocardial tissue, suggesting that periostin may play a role in the remodeling and/or separation (delamination) of developing AV mural valves from myocardium. Finally, the leaflets of the periostin null mouse had mesenchymal-like swellings containing abundant ECM and cells resemebling post-EMT cushion mesenchyme. The cell type(s) within these swellings could not be specifically identified. But due to their cell shape (stellate), size (small) and lack of Masson’s staining, these cells did not appear to be either myocytes or fibroblasts. However, we cannot rule out that other cell types may contribute to these swellings.

Periostin is required for chordae tendineae formation and maturation

Another area of intense periostin expression is the tendinous chords, or chordae tendineae (CT). The chords are fibrous structures that ramify through several generations of branching that connects the mural leaflets of the mitral and tricuspid valves to papillary muscles muscles (Morse et al., 1984). Periostin is expressed throughout the chordae tendineae with expression being most intense at the myotendinous junction where the stem or trunk of tree-like chords inserts into the papillary muscle. The periostin null mice exhibit striking defects in chordal development as depicted in Fig. 8. The chordal trunks of the tricuspid valve in periostin nulls have few if any branches compared to neonatal or adult wild-type mice. In wild-type development, the trunks of the murine tricuspid valve apparatus have 2–3 generations of branching at birth. As the mouse ages, the tendinous chords continue to divide and branch, giving rise to ~10–15 mature fibrous chords (Fig. 8A,D). In null hearts, the persisting chordal trunks were shortened and abnormally thickened in diameter (compared to wild-type chordae tendineae). This defect was seen in nearly 40% of all neonates and 20% of all adult mice (data not shown). In addition, a gene dosage affect was seen in the percentage of chordal truncks that were not branched, demonstrating a correlation between both periostin alleles and the chordal branching process.

Figure 8. Periostin is required for proper formation of the chordae tendineae.

A. AV apparatus of an adult wild-type mouse showing (1) anterior tricuspid leaflet, (2) tendinous chords, and (3) papillary muscles. B&C. Comparison of adult wild-type (Pn +/+) and periostin null (Pn−/−) tendinous chordal structure (asterisks). The periostin null mice exhibit thickened, non-branched, severely shortened tendinous chords. D&E. Graphical representation of defects in chordal structure in the periostin null mouse. D. The number of chordae tendineae normally doubles from neonate to adult. This doubling of the chordae tendineae is not evident in the periostin null mice. E. The chordae tendineae of the adult periostin null tricuspid valve apparatus has a significantly thicker diameter than that of the wild-type mouse. This defect appears to be nearly as prevalent in the heterozygote as in the null further suggesting the importance of having both intact periostin alleles.

Biomechanical properties of tricuspid leaflets and chordae tendineae are altered in periostin null mice

In vivo and in vitro analyses demonstrated that periostin is necessary for fibroblast differentiation, collagen synthesis, attenuation and compaction of cushions into leaflets and the formation of their suspensory tension apparatus. How periostin ultimately affected the integrity and strength of these fibrous tissues was further evaluated. The small size of valve tissue made utilization of a conventional materials testing system (MTS) not possible. However, the biomechanical properties for mouse valves could be attained using atomic force microscopy (AFM). AFM is a technique that provides high-resolution scanning and cantilever measurements of micromechanical properties along a tissue surface. This technique has been previously used to generate topographical and biomechanical profiles of cardiac cells and tissues, including human and bovine heart valves (Brody et al., 2006; Jastrzebska et al., 2007; Jastrzebska et al., 2006; Mathur et al., 2001; Merryman et al., 2007). We compared both the surface structure and the mechanical properties of freshly isolated wild-type and periostin null tricuspid valve leaflets and associated tendinous chords. Results demonstrated that the surface of the tricuspid valve leaflet and tendinous chords of the periostin null animals were significantly smoother than that of their wild-type counterparts. Organized collagen bundles were only observed within the wild-type tendinous chords and valve leaflets and never within the periostin null tissues (Fig. 19A–E). Importantly, the cantilever force indentation data collected from 50–100 randomized locations within the wild-type and null tricuspid leaflets and tendinous chords indicated that these tissues were mechanically less rigid (i.e. reduced tensile properties) in the periostin null mice (Fig. 9F). Thus, periostin is not only essential for regulating collagen synthesis and valve differentiation, but also for promoting and maintaining the tissue properties of mature cardiac valves.

Figure 9. Atomic force microscopy (AFM) on mouse chordae tendineae and valves.

Chordae tendineae and valve leaflets were removed from wild-type (A, C) and periostin-null mice (B, D) and processed for AFM. (A) A surface scan at high resolution was performed on a wild-type mouse chordae tendineae. Fibrillar structures (arrows) are apparent at the surface of the chordae. (B) A similar scan of a chordae from a periostin-null mouse shows a smoother surface with less distinct fibril-like structures. (C) A control valve leaflet was scanned at high resolution with numerous pore-like depressions observed (arrows). In comparison, periostin null leaflets were again smooth with minimal surface depressions. The scan sizes were 4 μm² in A and B and 1 μm² in C and 2.5 μm² in D. (E) Demonstrates the view prior to scanning the tissues taken from the AFM. The cantilever is shown as the triangular structure with the tip placed on the center of the chordae tendineae. For reference the cantilever is 160 μm in size. (F) Represents the force measurements from the tissues illustrating that there is a significant decrease in stiffness of chordae tendineae (red and green lines) as compared to the leaflets (blue and gold lines). In addition, there is a dramatic decrease in stiffness in the periostin-null leaflet. The control leaflet (gold line) is more than twice as stiff as the periostin-null valve leaflet (blue line). Relative stiffness numbers are represented in μN.

Discussion

The identification of genes related to normal and abnormal valvuloseptal development remains largely unknown due to the lack of appropriate model systems to study. Most genes important for the EMT phase of valvulogenesis results in an early embryonic lethality and shed little light on how post-EMT mesenchymalized cushions become mature cusps of leaflets. Post-EMT valvulogenesis, as defined in this study, involves three fundamental processes: (i) the differentiation of cushion mesenchymal cells into valvular interstitial fibroblasts, (ii) modeling and attenuation of the valve leaflet, and (iii) formation of valve-associated fibrous structures such as the annulus fibrosae and chordae tendineae. It is these processes that act in concert with fluid dynamics to ultimately sculpt the rudimentary cushions into a mature valve apparatus (Butcher and Markwald, 2006; Butcher et al., 2007a). The lack of functional data and appropriate gene knock-out animals have made the post-EMT process difficult to understand. In this report, we describe the detailed expression and null phenotype of periostin, a candidate, post-EMT valvulogenic protein expressed during intrauterine and postnatal life. Periostin null mice described herein display defects in the three fundamental processes (as described above) required for the normal maturation and development of the cardiac valve apparatus. Importantly, a large subset of the periostin null mice is viable and thus may provide a powerful tool to begin understanding mechanisms underlying post-EMT valvuloseptal maturation.

Periostin is an evolutionary conserved extracellular matrix (ECM) protein that has significant similarity with its other family members:βIG–H3, stabilin-1, and stabilin-2. The expression of the stabilin and βIG–H3 genes have been evaluated throughout development (Lindsley et al., 2005; Norris et al., 2005). Whereas expression of the stabilin genes do not overlap with periostin, there is partial overlap of periostin with βIG–H3 which might confer sufficient compensation to permit survival. A periostin/βIG–H3 double knock-out mouse would help address this possibility. Periostin has been demonstrated to play an important role in cellular processes such as adhesion, migration/cell sorting, and differentiation.

It is well known that periostin expression is significantly increased upon TGFβ or BMP stimulation in a variety of cell types, but is also known to be transcriptionally regulated during development by the basic helix-loop-helix gene, twist1 (Horiuchi et al., 1999; Li et al., 2005; Lindner et al., 2005; Afanador et al., 2005; Connerney et al., 2006; Oshima et al., 2002). Periostin can interact in vivo with other ECM scaffold proteins, such as collagens, fibronectin, and tenascin as well as integrin receptors at the cell surface including: α_V/β_3, α_V/β_5, and β1 (Butcher et al., 2007b; Gillan et al., 2002; Norris et al., 2007b; Takayama et al., 2006). Recently, we have found that periostin binding to integrins α_V/β₃ and α₁ on cardiac mesenchymal progenitor cells can initiate signaling related to migration and collagen crosslinking or compaction transduced through Rho and PI3 kinases (Butcher et al., 2007b; Norris et al., 2007b). Thus, there is an integrin-based receptor mechanism for mediating the potential effects of periostin on valvulogenesis.

Although RNA in situ expression of periostin has been well documented in mouse and chick, a detailed analysis of the protein expression in the mouse has yet to be reported. Our analysis has revealed temporal-spatial patterning of periostin not previously appreciated. For example, while expression of periostin commences at E9.5, which coincides with the ongoing EMT process, the majority of expression at this time remains cytoplasmic suggesting that even though the protein is produced, it is not secreted or fully functional. By E10.5, periostin is expressed and secreted by the cushion mesenchyme of the AV and OFT regions as seen by ECM staining. As post-EMT valvuloseptal morphogenesis commences, periostin expression intensifies in areas that are undergoing active morphogenetic remodeling such as the fibrous annulus, the AV and OFT valve leaflets, the chordae tendineae, the mesenchymal cap of the septum primum, and the epicardium. These areas express high levels of periostin are fated to normally become fibrous tissue. Thus, it is our wrking hypothesis that periostin expression is required for fibrogenesis. In fact, our recent data have demonstrated that periostin regulates collagen fibrillogenesis and the biomechanical properties of connective tissues (Norris et al., 2007b). Consistent with this observations, periostin has been shown to be integral to the formation of a fibrous scar following a myocardial infarct (MI) (Iekushi et al., 2007; Oka et al., 2007). Inducing an MI in the periostin null mice gave precisely the opposite affect: a decrease in fibrous tissue and formation and/or survival of cardiomyocytes (Oka et al., 2007; Norris et al., 2007a). However, it should be pointed out that these data conflict with what has recently been described by Kuhn et al. who reported that periostin promoted cardiac regeneration and reduction in scar tissue following the addition of “purified periostin” to the heart after a myocardial infarction (Kuhn et al., 2007). The periostin used in these studies was a carboxyl-truncated form of periostin lacking 163 amino acids of the carboxyl terminus. It has been shown that the carboxyl terminus encodes a functional domain that is important for migration, invasion, and adhesion (Kim et al., 2005). Thus, one alternative interpretation is that absence of this domain may act as a dominant-negative protein when applied in vivo. Our data, which includes deleting expression of the entire protein, strongly suggests that periostin is profibrogenic and important for the differentiation of cushion mesenchyme into fibroblastic tissues.

By E12.5, the periostin null mouse displays defects in AV cushion mesenchymal differentiation as pockets of cells that are positive for the myocardial marker, MF20, within the AV cushions. In the case of the proximal (conal) cushions of the outflow tract, this process happens naturally whereby the cushion mesenchyme normally undergoes muscularization (termed “myocardialization”) (van den Hoff et al., 2001). Periostin expression decreased in the conus mesenchyme prior to myocardialization (Norris et al., 2004). This finding combined with the observation that mesenchyme in the periostin null AV cushions can differentiate into myocardial tissue suggested to us that this matrix factor is functioning to promote fibrogenesis of the AV cushions, and/or block aberrant differentiation into a non-fibroblastic lineage (e.g. myocardial). Furthermore, in culture, periostin null AV mesenchymal cells expressed cardiomyocyte markers, whereas expression of fibroblasts markers was low or absent. The addition of exogenous periostin to null cushion mesenchyme significantly reduced myocardial marker expression while enhancing fibroblast differentiation markers. To date, this is the first example of a secreted, extracellular protein that can directly, or indirectly, alter the differentiation pathway of AV cushion mesenchyme. We interpret these findings to indicate that periostin may function as a “binary or hierarchical switch” promoting firbogenesis (if present) vs. cardiomyogenesis (if absent). What is clear is that these altered patterns of differentiation seen during intrauterine, post-EMT valve development in periostin null animals is sustained into postnatal and adult life. Thus, periostin is a candidate valvulogenic protein whose expression correlates with the normal differentiation of cushion mesenchyme.

In wild-type hearts, by ED 16, periostin is expressed predominantly within the mural leaflets at sites where the developing tricuspid AV valve is separated from the ventricular myocardium. Conversely, in the adult periostin null mouse, the delamination or separation of the leaflet from myocardial tissue appears incomplete and the valve retains primitive characteristics (swollen foci of ECM containing undifferentiated mesenchyme). To date, the mechanism(s) by which the myocardium is removed from the tricuspid leaflet is not known. However, there is intense expression of periostin at sites where cushion prevalvular mesenchyme directly contacts myocytes (e.g. ventricular trabeculae and the AV junctional myocardium) which correlates with the normal removal or “disappearance” of the associated myocardial tissue (Gaussin et al., 2005). This suggests to us a possible role for periostin as an anti-myocardal factor affecting the fate or survival of myocytes. A more speculative alternative would be that specific myocytes in direct contact with periostin secreting fibroblasts are capable of “transdifferentiating” into myofibroblasts or fibroblasts. Such a mechanism has been proposed for myocytes during pathological remodeling in adult life (d'Amati et al., 2000). This theory is further supported by recent work by Kolditz et al. who find that periostin is expressed around the myocardial accessory pathways of the conduction system during quail development. By adulthood, the expression of periostin within these myocardial accessory pathways eventually result in inhibition of the myocardial phenotype and transdifferentiation of the myocytes into fibrous tissue (Kolditz et al., 2007). Additionally, this theory is consistent with data presented by Katsuragi et al. whereby transfection of periostin into adult rat hearts not only induced ventricular fibrosis but also resulted in a loss of ventricular cardiomyocytes (Katsuragi et al., 2004) and recent work involving the Alk3/Bmpr1A knockout mouse also implicated periostin as a putative mediator in regulating myocyte-fibroblast interactions during tricuspid leaflet delamination and maturation (Gaussin et al., 2005). The tricuspid leaflets of the Alk3/cGATA6-Cre conditional knockout (CKO) mouse exhibit a reduced level of periostin expression. In this mouse, Alk3 was specifically deleted in myocytes, a cell type that does not secrete periostin, suggesting that there is dynamic signaling and communication between the myocytes and fibroblasts that could affect myocyte survival, adhesion, and/or migration (Gaussin et al., 2005). Whether this interaction is mediated by periostin remains to be determined.

Congenital heart defects represent the most common form of birth defects in humans, with those attributable to post-EMT valvulogenesis among the most prevalent (Hinton et al., 2006). These include structural (anatomical) defects such as Ebstein’s anomaly, mitral valve stenosis, mitral valve arcade, parachute asymmetric valve, myxomatous valves, Down syndrome, Ehlers-Danlos syndrome, and polyvalvular disease (Anderson et al., 1979; Bonnet et al., 1997; Castaneda et al., 1969; Chesler and Gornick, 1991; Gittenberger-de Groot et al., 2003; Oosthoek et al., 1997; Ruckman and Van Praagh, 1978; Zuberbuhler et al., 1979). It is important to note that defects in post-EMT valvulogenesis (except severe mitral regurgitation) are compatible with embryonic and post-natal life and may not be immediately recognized at birth but over time cause morphological or functional changes. Thus genes that encode ECM proteins, like periostin, may not produce an immediate phenotype. It is their regulation, organization, alignment, and turn-over which, over time, accumulate to adversely affect valve function, resulting in preogression of prolapsed valves, ectopic fibrosis, formation of myxomatous lesions, and calcifications. Also, hemodynamic responses to aberrant valvular ECM composition can have secondary consequences such as myocardial hypertrophy, dilated cardiomyopathies, ectopic cardiac fibrosis, and cardiac failure. Collectively, structural and ECM based functional defects contribute to 100,000+ valve replacement surgeries per year in the United States. Thus, a greater understanding of how developmental process, such as post-EMT valve maturation contribute to adult diseases is necessary which will provide the foundation for the generation of better therapeutics to combat these diseases. Here, we have described the function of periostin during post-EMT valve maturation. To date, this is the first known candidate valvulogenic protein capable of controlling cell fate, either directly or indirectly, within the maturing valve apparatus, and as such has provided us with a potential novel tool for understanding mechanisms underlying cardiac valvular diseases.

Figure 7. Periostin null mice exhibit defects in delamination, fibroblastic differentiation, and remodeling of associated myocardium.

A–E Frontal section of an adult periostin null mouse heart stained with Masson’s trichrome. A significant reduction of collagen matrix (blue staining) is seen in the tricuspid mural and septal leaflets (B–E) compared to wild-type mice (F). Failure of complete delamination of the mural tricuspid leaflet is evident in the periostin null mouse as are large foci, or swellings (B–D). In the septal leaflet of the tricuspid leaflet there is failure of normal remodeling/removal and/or transdifferentiation of associated myocardium (B,E). Collagen 1α1 and 1α2 are reduced by ~70% in the periostin heterozygote and null mice as shown by Western analysis. Actin is used as a loading control