Periostin Induces Intracellular Cross-talk between Kinases and Hyaluronan in Atrioventricular Valvulogenesis*

Background: Periostin-null mice exhibit a myxomatous atrioventricular valve phenotype. We propose two mechanisms as follows: periostin binds to collagen and links it to cell-surface receptors; periostin/β-INTEGRIN signaling promotes valve morphogenesis. Results: Periostin/β-INTEGRINs/focal adhesion kinase/PI3K/ERK signals promote hyaluronan synthase-2 activation, matrix remodeling, and valve progenitor cell survival/differentiation. Conclusion: The phenotype of periostin-null valves is consistent with a role for PN cell signaling through INTEGRIN receptors. Significance: Periostin is a valvulogenic signaling morphogen. Periostin (PN), a novel fasciclin-related matricellular protein, has been implicated in cardiac development and postnatal remodeling, but the mechanism remains unknown. We examined the role of PN in mediating intracellular kinase activation for atrioventricular valve morphogenesis using well defined explant cultures, gene transfection systems, and Western blotting. The results show that valve progenitor (cushion) cells secrete PN into the extracellular matrix, where it can bind to INTEGRINs and activate INTEGRIN/focal adhesion kinase signaling pathways and downstream kinases, PI3K/AKT and ERK. Functional assays with prevalvular progenitor cells showed that activating these signaling pathways promoted adhesion, migration, and anti-apoptosis. Through activation of PI3K/ERK, PN directly enhanced collagen expression. Comparing PN-null to WT mice also revealed that expression of hyaluronan (HA) and activation of hyaluronan synthase-2 (Has2) are also enhanced upon PN/INTEGRIN/focal adhesion kinase-mediated activation of PI3K and/or ERK, an effect confirmed by the reduction of HA synthase-2 in PN-null mice. We also identified in valve progenitor cells a potential autocrine signaling feedback loop between PN and HA through PI3K and/or ERK. Finally, in a three-dimensional assay to simulate normal valve maturation in vitro, PN promoted collagen compaction in a kinase-dependent fashion. In summary, this study provides the first direct evidence that PN can act to stimulate a valvulogenic signaling pathway.

is a matricellular protein containing fasciclin gene motifs that encode an ancient family of adhesion proteins (1,2). Like most matricellular proteins, PN can function as a bridge connecting the cell surface to structural extracellular matrix (ECM) glycoproteins, thereby potentially modulating cell/matrix interactions related to adhesion, migration, tissue formation, and growth (3,4). In normal heart development, PN promotes collagen fibrillogenesis and matrix organization. Developmental profiles of PN expression peak immediately following birth when valve primordia complete their morphogenetic transformation from mesenchymal swellings, called "cushions," into mature, sculpted leaflets. During this period of peak PN secretion, the prevalvular ECM is progressively compacted and organized into specific layers or zones. Thereafter, when remodeling is completed, PN falls to baseline levels and does not increase again unless an injury occurs that promotes fibrosis (5,6). At all developmental time points, PN is expressed in valve or ventricular fibroblasts or their mesenchymal (cushion) progenitor cells, which are derived by the transformation of epicardium or endocardium cells (7)(8)(9). In PN-null mice, the atrioventricular (AV) valves of those that survive into early adult life have a myxomatous-like phenotype characterized by interspersed "pockets" of undifferentiated cushion-like tissue, diminished collagen expression, and poorly organized regions of noncompacted ECM (10,11). These alterations in AV valve matrix organization and col-lagen secretion correlate with a significant loss of biomechanical properties (12,13). Thus, based on the phenotype of the AV valves in PN-null mice (14,15), we propose that PN has an important developmental role in the differentiation and maturation/compaction of prevalvular cushion cells into valve fibroblasts and their subsequent remodeling into sculpted leaflets. The following questions are asked. How does periostin do this? Is it by directly binding to extracellular structural proteins like collagen or elastin (promoting cross-linking) or is it through signaling mechanisms triggered by the binding of PN to cellsurface receptors such as ␣v␤3 and ␣v␤5 INTEGRINs (16,17)? The valve phenotype of the periostin-null mouse does not allow us to distinguish between these two candidate mechanisms.
To address this question, we isolated prevalvular cushion cells from embryonic chicks and mice and determined whether PN, normally secreted by prevalvular cushion cells or added to culture medium, activates intracellular signaling kinases through binding to cell-surface receptors and, if so, whether these signaling changes affect cell behaviors related to normal valvulogenesis. Receptor candidates tested were specific ␤-INTEGRINs (␣v␤1, ␣v␤3, and ␣v␤5). Our findings provide evidence that in cultured valve progenitor cells, PN can directly activate intracellular kinases through INTEGRIN binding, including those downstream of focal adhesion kinase (FAK), PI3K, AKT, and ERK. In vivo and in vitro data collectively indicate that PN can promote activation of HAS2 by promoting phosphoserine, and this increase in phosphoserine levels is correlated with an increase in hyaluronan synthesis and the survival of prevalvular progenitor cells. Similarly, PN can promote phosphothreonine, and this activation in phosphothreonine-HAS2 is correlated to down-regulation in HA synthesis. We have also linked PN-induced INTEGRIN/FAK-mediated PI3K and MAPK signaling to changes in morphogenesis of prevalvular cushion cells (adhesion, migration, and survival) and to their differentiation into a valve fibroblastic lineage. Such changes in differentiation into valve fibroblasts are reflected by enhanced collagen 1 (COL1␣1) synthesis and the generation of contractile forces sufficient to compact and align collagen fibrils as occurs in normal valve maturation.

MATERIALS AND METHODS
Animals and Cell Culture-Wild type (WT) mice (C57BL/6 strain) were obtained from the Jackson Laboratory. PN-deficient mice on a C57BL/6 genetic background were provided by Dr. Simon Conway (Indiana University-Perdue University, Indianapolis). Mice at 8 -10 weeks of age were used in experiments as described previously (10). All animal care and experimentation were done in accordance with the institutional guidelines. Adult sheep valve cells were provided by Dr. Norris and Dr. Bischoff (18).
Cell Lysis and Immunoblotting-Prevalvular mesenchymal cells were cultured until they were confluent. Cells were washed twice at 4°C with PBS, harvested with 0.05% Versene, and then washed in cold PBS again as described previously (19 -27). The cells were pelleted by centrifugation at 5000 ϫ g for 2 min at 4°C. The pellets were treated with the lysis buffer containing 1% Nonidet P-40, 0.5 mM EGTA, 5 mM sodium orthovanadate, 10% (v/v) glycerol, 100 g/ml phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 1 g/ml pepstatin A, 1 g/ml aprotinin, and 50 mM HEPES, pH 7.5. The lysates were clarified by centrifugation at 12,000 ϫ g for 10 min at 4°C and then stored at Ϫ80°C as described previously.
For SDS-PAGE, the denatured cell lysates were loaded onto a 4 -12% gradient polyacrylamide gel at 15-30 g of protein per lane in an Invitrogen mini-gel apparatus. Proteins were transferred to nitrocellulose membranes and blocked for 1 h with 5% nonfat dry milk in Tris-buffered saline containing 0.1% Tween 20 followed by washing in the same Tris/Tween buffer. The membranes were probed with the appropriate antibody diluted in Tris-buffered saline containing 5% bovine serum albumin (for polyclonal antibodies) or 5% nonfat dry milk (for monoclonal antibodies) followed by treatment with peroxidase-linked secondary antibodies and Luminol reagents. The proteins on the blots were detected with antibodies for PN, ␤3-, ␤1-, and ␤5-INTEGRINs, HSP47, p-ERK, ERK, p-AKT, and AKT (19 -24). ␤-TUBULIN and ␤-ACTIN were used as internal standards. Sizes of proteins were estimated from prestained molecular weight standards electrophoresed in the same gel as the samples. Immunoreactive bands were quantified by densitometry. Each protein sample was analyzed from at least three independent experiments from each set of fibroblasts.
Immunoprecipitation with Anti-HAS2 Antibody and Western Blot Analyses-For immunoprecipitation, all procedures were done at 4°C unless otherwise mentioned. Cell lysates were diluted to 1 g of protein/l using the lysis buffer above. Aliquots of 500 l of lysate were mixed with 5 l of anti-HAS2 antibody or nonimmune IgG as described previously by our group (20,27). The immune complex was captured by adding 80 l of 1:1 (v/v) protein A-Sepharose 4B suspension and incubated for another 1 h. The Sepharose 4B beads were collected by brief centrifugation followed by washing three times in ice-cold lysis buffer, three times with lithium chloride buffer (5 mM LiCl, 0.1 mM sodium orthovanadate, 0.1 M Tris-HCl, pH 7.4), and three times with buffer containing 150 mM NaCl, 5 mM EDTA, 0.1 mM sodium orthovanadate, 10 mM Tris-HCl, pH 7.4. Finally, the immune complexes on the beads were recovered with 500 l of SDS-containing denaturing buffer and heated to 65°C for 5 min. Cell lysates (normalized for protein concentration) were analyzed by immunoblotting with anti-phosphoserine or reblotting with anti-HAS2 as a loading control using SDS-PAGE (4 -12% gradient gels), as described previously (22)(23)(24)(25). For SDS-PAGE, the denatured immunoprecipitates were loaded onto a 10% polyacrylamide gel at 15-30 g of protein per lane as described under "Cell Lysis and Immunoblotting" (20,27).
Cell Adhesion Assay (25)-Antibodies against INTEGRINs were purchased from EMD Millipore. Ninety six-well plates were coated with or without PN (1 g/ml) at 4°C overnight. Plates were blocked with 10 g/ml heat-denatured fatty acidfree bovine serum albumin for 30 min. Chick cushion valve cells (10 5 /ml in M199 with 0.1% fatty acid-free bovine serum albumin) were incubated with function inhibiting anti-INTEGRIN antibodies (␤1-, ␤3-, and ␤5-INTEGRINs) for 10 min before plating. Then 100 l of the cell suspension/well was added onto plates and incubated for 30 min at 37°C under 5% CO 2 with the lid off. After incubation, unattached cells were removed by rinsing with PBS. Attached cells were fixed in 5% glutaraldehyde for 20 min and stained with 0.1% crystal violet. After washing, stains were dissolved in 200 l of 10% acetic acid, and color was read at 575 nm in a Bio-Tek plate reader.
Cell Migration Assay-Cell migration assays were done using 24-well chemotaxis chambers (Nalgene) as described previously (25). Polycarbonate filters with a pore size of 8 m (Nalgene) were coated with 5 g/ml fibronectin (Invitrogen) at 4°C overnight in phosphate-buffered saline (PBS) and dried under sterile air as described previously (25). FGF was diluted to 10 ng/ml in M199 supplemented with 0.5% fatty acid-free bovine serum albumin (Sigma), and 600 l of the final dilution was placed in the lower chamber of a modified Boyden chamber.
PN was overexpressed by PN cDNA expression vector (empty vector was used as a control) in valve cells followed by treatment for 18 h with control antibody or with PN-Ab, or with 10 M LY294002 (PI3K/AKT inhibitor), or with 10 M U0126 (ERK inhibitor). In a companion experiment, cells after PN overexpression were also transfected with control siRNA or FAK siRNA alone for 48 h, or they were first transfected with FAK siRNA followed by incubation with PN-Ab. After the treatments, cells were washed and trypsinized for the minimum time required to achieve cell detachment. Approximately 5 ϫ 10 4 cells suspended in 100 l of M199 with 0.5% bovine serum albumin were placed in the upper compartment. The cells were allowed to migrate for 3 h at 37°C in a humidified chamber with 5% CO 2 . After the incubation period, the filter was removed, and the cells persisting on the upper side of the filter were removed with a cotton swab. The cells on the filters were fixed with 4% formaldehyde and stained with hematoxylin. Migration was quantified by counting cells in three random fields (100 times) for each filter. Alternatively, filters were stained in 0.1% crystal violet, washed, and eluted with 10% acetic acid in 24-well plates. Quantification was done based on absorbance at 575 nm in a Bio-Tek plate reader.
Caspase-3 Assay-Caspase-3 activity was measured in a plate reader at 405 nm by using the kit from Millipore following the manufacturer's instruction. The assay is based on detection of the chromophore p-nitroaniline (p-NA) after cleavage from the substrate DEVD-p-NA. The absorbance of released p-NA is measured at 405 nm. The difference in absorbance between control and treated samples determined the increase in caspase-3 activity. The assay was done in 96-well plates. Cytosolic extracts (1.0 ϫ 10 6 cells in 300 l in cell lysis buffer) were mixed with 10 l of substrate (3 mg/ml DEVD-p-NA). Buffer blanks (without substrate and cell extract) and substrate blanks (with substrate but without cell extract) were included in the assay. The plates were incubated for 1 h at 37°C, and read at 405 nm using Bio-Tek plate reader.
Immunohistochemical Staining in the Heart Sections-Heart sections from WT and PN-null mice were deparaffinized using standard procedures and permeabilized with 0.1% Triton X-100 in PBS. PN, HA, and Ki67 were localized in sections by immunohistochemical staining using PN antibody (10,11), HA by HA-binding protein (Saikagaku Corp.), and a Ki67 antibody (Millipore) following standard protocols. As a negative control, the primary antibody was replaced with nonimmune rabbit IgG, or sections were treated with Streptomyces hyaluronidase followed by treatment with HA-binding protein (in both cases, no staining was observed).
The qRT-PCR assays were done in three individual experiments with triplicate samples using standard conditions. After sequential incubations at 50°C for 2 min and 95°C for 10 min, respectively, the amplification protocol consisted of 50 cycles of denaturing at 95°C for 15 s and annealing and extension at 60°C for 60 s. The standard curve was made from a series dilution of template cDNA. Expression levels of Col1␣1 and HA synthase mRNAs were calculated after normalization with the housekeeping gene GAPDH. Variation within samples was less than 7%. Statistical analysis was done with the Student's paired t test.
Transient Transfection Using Cushion Prevalvular Mesenchyme-All transfections were done using Lipofectamine (Invitrogen) in cultures at ϳ75% confluence. After each transfection, the cultures were grown for another 48 h for further transfection or treatment and analyzed for each experimental design.
Floating Collagen Gel Cultures and Quantitation of Gel Contraction-Experiments were done essentially as described previously (31). Briefly, 24-well tissue culture plates were precoated with BSA. For PN-induced effects, cells were pretransfected with vector control or PN cDNA expression vector, and transfectants were grown for 48 h. For PN inhibition, the cells were first transfected with control shRNA (scrambled shRNA) or PN shRNA followed by co-transfection with PN cDNA expression vector. These transfectants were then grown for another 48 h. In a parallel set of experiments to inhibit ␤1-INTEGRIN, FAK, or PI3K, cells were preincubated in the presence of 10 M LY294002 (PI3K/AKT inhibitor), 10 M PF573228 (FAK inhibitor), or 5 g/ml blocking antibody against ␤1-INTEGRIN or 5 g/ml blocking antibody against CD44 for 2 h, prior to PN overexpression. Finally, these transfected and treated HH40 chick cushion mesenchyme cells were suspended in Medium 199 with 1% FBS, 100 units/ml penicillin, and 100 g/ml streptomycin and mixed with a collagen solution (1 part HEPES, pH 8.0, 4 parts collagen 3 mg/ml (Vitrogen 100, Cohesion Technologies, Palo Alto, CA), and 5 parts medium (2ϫ M199)), which yielded a final concentration of ϳ80,000 cells/ml and 1.2 mg/ml of collagen. A collagen cell suspension (1 ml) was added to each of four wells (Nunc). After polymerization at 37°C for 2 h, the gels were detached from the wells followed by addition of 1 ml of the medium. For inhibition experiments, after polymerization the lattices were detached from tissue culture plates and treated further with or without 10 M inhibitor or 5 g/ml blocking antibody against ␤1-INTEGRIN for 24 h. Contractions of the gels were quantified by measuring the decrease in gel diameter over a 24 -48-h period.
Statistical Analysis-Each experiment was repeated three times for each set of fibroblasts and pooled for statistical analysis. Western blot analyses, mRNA analyses, caspase-3, migration, and collagen gel contraction experiments for each separate experiment were repeated between 3 and 4 times, depending upon the particular experiment. Data are expressed as means Ϯ S.D. The Student's two-tailed t test (Microsoft Excel software) was used for comparison between two groups. Statistical analysis of the Western blots was done using t test with Mann-Whitney modification or analysis of variance as applicable. When analysis included more than two groups, oneway analysis of variance was used. p values Յ 0.05 were considered statistically significant.

RESULTS
Technically, it is difficult to obtain enough mesenchyme cells from mouse embryos. This inherent problem of small cell numbers is a historic problem for developmental biological studies. Therefore, we did most of the proposed experiments with embryonic chicks to optimize the isolation method and yields. In addition to chicks, we also included fibroblasts isolated from adult sheep mitral valves to obtain sufficient numbers of fully differentiated valve cells to do the relevant kinase signaling experiments prior to using cells isolated from WT and PN-null mice. The aims of this study were as follows: 1) to use biochemical analyses of prevalvular mesenchyme cells isolated as primary cultures from embryonic or adult valves to assess whether endogenous (i.e. cell-secreted) PN or exogenously added PN can activate intracellular signaling pathways; 2) to assess whether INTEGRIN/FAK-induced PI3K/AKT and/or ERK signals modify prevalvular mesenchymal cell morphogenetic behaviors (e.g. adhesion, migration, anti-apoptosis, and synthesis of hyaluronan) necessary for valvulogenesis; and 3) to determine whether PN/INTEGRIN interaction promotes kinase-mediated differentiation of valve progenitor cells into fibroblasts as evidenced by increasing collagen synthesis and the potential to compact (align) collagen in three-dimensional gel assays.

PN Promotes ␤3-INTEGRIN/FAK/AKT/ERK1/2 Activation in Chick and Mouse Embryonic AV Valve Cells and Adult Sheep
Valve Cells-We and others have shown that PN mediates the formation and remodeling of the inlet valves of the AV junction, most likely through an INTEGRIN-dependent process (5, 32). However, candidate mechanisms linking PN and INTEGRIN interaction to the regulation of valvulogenesis are still poorly understood and often attributed to PN acting as an adaptor protein cross-linking ECM fibrils or linking ECM to the cell surface. Here, we tested the hypothesis that PN can signal intracellular changes in valvulogenesis through the autophosphorylation of the cytoplasmic tyrosine kinase domain of FAK following PN-INTEGRIN engagement, which serves to recruit signaling proteins such as Src, and the subsequent activation of multiple downstream signaling pathways, e.g. PI3K/AKT and ERK kinases. These kinases have been associated with migration, survival, and phenotype stability of many cell types (33,34) and thus were logical candidates for testing putative PN signaling in normal valvular morphogenesis.
Results in Figs Fig. 3, A-G, show the results of Western blots using lysates from cultures of embryonic valve interstitial cells harvested from E15.5 mouse mitral valves that were treated using the same protocol as used for the embryonic chick valve cells (Fig.  1). As with the chick valve cells, overexpression of PN in mouse AV cushion cells also induced activation of p-FAK (Fig. 3, A and B) and of p-AKT and p-ERK without altering the total levels of AKT and ERK (Fig. 3, C and D). However, overexpression of PN only had a small inductive effect on ␤3-INTEGRIN expression (Fig. 3, A and B). These results (Figs. 1-3) indicate that mouse prevalvular cushion cells, like those of the embryonic chick cushion mesenchyme, exhibit sensitivity to PN sufficient to activate p-FAK and retain that responsiveness even when fully differentiated.
To further examine the role of FAK signaling network in valvulogenesis, we examined the activation of PI3K/AKT and MAPK signaling pathways in WT E16.5 mouse valve cells (Fig.  3E) in the presence or absence of PF573228, a selective inhibitor of FAK. PF573228 treatment potently suppressed the activation of PI3K/AKT, ERK1/2, and FAK on Tyr-925 without affecting total levels of FAK (data not shown) and ␤-TUBULIN (Fig. 3E). This observation suggests that FAK lies upstream of PI3K/AKT and ERK1/2 proteins.
Once phosphorylated on Tyr-925, FAK serves as a docking site for p85 regulatory subunit of PI3K (35-38) through recruitment of growth factor receptor-bound protein 2 (Grb2) (36), which is known to be involved in Raf/MEK/ERK signaling (35,36). We then focused on the activation of PI3K/AKT and MAPK/ERK in response to exogenous addition of PN at various time points. In cultured E16.5 WT mouse valve cells, stimulation with PN results in immediate activation of PI3K and MAPK/ERK signaling (Fig. 3F). An important distinction from the results of the experiment in Fig. 3F is that MAPK/ERK activation in cultured cells is transient (typically lasting at a high level for only a few hours (Fig. 3F, 4 h), after which a sustained small increase in ERK activation persists). In contrast, persistent detectable levels of activated AKT are observed over sev-eral hours (Fig. 3F, 24h) in WT mouse valve cells. Such dramatic differences in signaling kinetics may have important functional implications when comparing biological responses in valve morphogenesis.
Next, to examine the role of PI3K and MAPK signaling networks in valve morphogenesis, we examined the activation of PI3K/AKT and MAPK in the presence and/or absence of a PI3K inhibitor (LY294002) and an ERK inhibitor U0126. The LY294002 treatment potently suppressed the activation of PI3K/AKT and ERK1/2 (Fig. 3G) without affecting total levels of p-FAK (data not shown), whereas U0126 treatment did not have any effect on AKT activation but significantly suppressed ERK activation (Fig. 3G). Thus, these results suggest that FAK accounts for the ERK1/2 responses through PI3K/AKT-dependent or -independent mechanisms and that PI3K regulates ERK activation (Fig. 3G) as has been seen in other cell types (39). Thus, to our knowledge, these results are the first to indicate that PN can activate specific intracellular kinases that are known to be associated with adhesion, migration, and survival and that PN activation of these kinases was mediated through binding to ␤3-INTEGRIN. As noted below, PN signaling is not restricted to this particular INTEGRIN receptor alone as other signaling responses were observed for ␤1and ␤5-INTEGRINs.
Of potential importance for future studies, as shown in Figs. 1 and 2, we did not detect similar changes in PN or INTEGRIN expression after a one-time treatment of valve progenitor cells with exogenous addition of PN compared with transfecting PN expression vectors, which require endogenous protein synthetic mechanisms to make and secrete PN. This suggests to us that adding exogenous PN in this system does not always replicate the cellular distribution or molecular interactions that result from increased endogenous synthesis of PN. One explanation for this is that PN interacts with transmembrane INTEGRINs whose kinase-active binding sites are situated on the cytoplasmic side of the plasma membrane. Upon activation, INTEGRINs can interact with PN or other matricellular proteins or assemble into extracellular matrices. Simple addition of exogenous PN cannot duplicate all of these configurations. This is reflected in results in activation of intracellular kinases in Figs. 1-3 for p-FAK, p-AKT, and p-ERK1/2 signals. Thus, in the subsequent experiments, we used PN expression vectors to promote overexpression of endogenously synthesized PN to promote a more accurate, physiologically relevant representation of signaling than a simple addition to the culture medium.

PN-INTEGRIN Interactions Promote Cell Behaviors Associated with Valvular Morphogenesis-Valve morphogenesis
involves an active elongation phase in which the primitive prevalvular cushions extend from the inner margins of the myocardial walls into the lumen of the AV junction to form the mesenchymal models of the future mitral and tricuspid valve leaflets (40 -42). These phases begin at E15.5 and continue into early postnatal life and are accompanied by changes in cell growth (cell number and size), adhesion, migration, matrix synthesis, and alignment into linear histological arrays (36).
Adhesion-As indicated in Figs. 1 and 2, exogenously added and endogenously secreted PN activated INTEGRIN-mediated FAK, a kinase generally recognized for its potential to promote adhesion and migration in a number of cell systems, including metastatic tumor cells (43). To determine whether PN can affect prevalvular cushion cell adhesion (as a potential first step toward migration) through an INTEGRIN-dependent mechanism, we cultured chick HH40 valve cells during their period of active elongation with blocking antibodies for ␤1-, ␤3-, or ␤5-INTEGRINs prior to plating on culture wells coated with either BSA or PN (Fig. 4, A and B). The cells exhibited robust adhesion to culture wells coated with PN (Fig. 4A), which was substantially inhibited by blocking antibodies against ␤3-INTEGRIN, and to a lesser extent with ␤1-INTEGRIN, but not with ␤5-INTEGRIN (Fig. 4B, lanes 2 and 3 versus lanes 1). An isotypematched control antibody did not inhibit the adhesion of these cells at this concentration (Fig. 4B, lane 5 versus lane 1). Adhesion to BSA-coated wells was less than for PN and essentially invariant among the antibody treatments (Fig. 4B).
In a companion experiment, the effect of PN on adhesion of these cells was variably blocked by all three INTEGRIN antibodies in a dose-dependent manner with ␤3-INTEGRIN antibodies having the most effect (data not shown). The optimum concentration of ␤3-INTEGRIN antibody for inhibition of adhesion of these cells to PN-coated plates was 5 g/ml.
Cell Growth-To determine whether PN promoted growth of cushion primordia through an effect upon increased cell growth, we assessed the expression of Ki67 in three sections of AV cushions from WT and PN-null 16.5 ED mouse hearts. Only a 2-fold difference in the number of Ki67 ϩ cells was observed in WT cushions compared with those of PN-nulls (Fig. 5A). These in vivo cell proliferation data were consistent with BrdU incorporation assays using in vitro WT mouse cushion cell cultures with or without silencing of PN, which indicated a modest decrease in cell proliferation when endogenous PN is silenced (Fig. 5B, lane 4 versus lane 3). To what extent proliferation drives cushion growth and elongation remains an open question. Although growth factors, such as bone morphogenic protein, TGF␤, and EGF, certainly may come into play (42,44), and based on our findings, PN signaling would seem to contribute partly to the answer.
Apoptosis-An alternative mechanism to proliferation to promote growth would be to inhibit apoptosis. To determine whether PN potentially affects apoptosis, E16.5 mouse AV cushion cell cultures were transfected for 48 h with a PN-silencing vector (PN shRNA), a control vector, or full-length sense PN cDNA expression vector to overexpress PN or treated with INTEGRIN blocking antibodies. The treated cultures were subsequently assayed by ELISA for caspase-3 levels, an active apoptotic marker. The knockdown of PN substantially increased the caspase-3 level compared with cells transfected with the vector alone (Fig. 6, lane 4 versus lane 3). In contrast, overexpression of PN cDNA expression vector essentially kept the level of the apoptotic marker at a basal level (Fig. 6, lane 2  Migration-The initial expression of PN in prevalvular cushion cells occurs immediately after their transformation from the AV endocardial epithelium as they migrate to colonize the acellular cardiac jelly (consists primarily of hyaluronan (45)) of the primitive AV cushion swellings. Expression of PN continues as waves of prevalvular mesenchyme extend distally into the cushions as they elongate into the lumen to form the primitive leaflets. These findings suggest that PN may promote the movements of prevalvular mesenchyme cells as they progressively colonize the elongating cushions during their morphogenesis. To determine whether PN affected migratory behavior through an INTEGRIN-linked signaling mechanism, we used explant cultures from E15.5 mouse valves, the period of active elongation. Fig. 7 shows that treatment of explant cultures with PN cDNA expression vector stimulated migration ϳ4-fold, which was prevented by the following:     1-6). In the same experiment, cells after PN overexpression were also transfected with control siRNA (ContsiRNA) or FAK siRNA alone for 48 h, or they were first transfected with FAK siRNA followed by incubation with PN-Ab after PN overexpression. After the treatment, cells were trypsinized and analyzed for migration assays as described under "Materials and Methods." Migration was quantified by counting cells in three random high power fields for each filter. Data are expressed as the number of cells per field from four independent fields and are expressed as the means Ϯ S.D. of four independent experiments. Statistical analyses were done using t test with Mann-Whitney modification as applicable. p Յ 0.05 (*) was considered statistically significant. U0126 (10 M) did not affect migration (Fig. 7, lane 9). These results indicate that PN/␤3-INTEGRIN/FAK/PI3K/AKT signaling pathway promotes AV cushion cell migration. The lack of engagement of ERK is consistent with recent findings that ERK pathways may be involved in other biological functions, e.g. proliferation or collagen synthesis (46 -49).

PN Can Induce Collagen 1␣1 (Col1␣1) Gene Expression in Mouse Cushion Cells by Both PI3K/AKT and ERK Pathways-
There is evidence that PN elevates collagen synthesis severalfold in adult remodeling diseases that lead to fibrosis (30). Whether PN can also directly up-regulate type 1 collagen (collagen 1) expression during development has not been demonstrated for prevalvular mesenchyme, even though collagen becomes the key structural protein of the inlet valves.
Because PN supported and enhanced cushion migration through a ␤3-INTEGRIN/FAK interaction (Fig. 7), we asked whether PN signaling could also induce or promote collagen synthesis. Expression of collagen is normally associated with differentiation of valve progenitor cells into interstitial fibroblasts. Fig. 8A shows that Col1␣1 mRNA expression in PN-null cushion cells is significantly lower than for WT cushion cells. In  have inhibited ECM remodeling of AV valve primordia (10,(52)(53)(54)(55). FLNA is an ACTIN-binding protein that anchors various transmembrane proteins to the cytoskeleton, and it provides a scaffold for many cytoplasmic signaling proteins involved in ACTIN cytoskeleton remodeling. Thus, we hypothesize that during valve progenitor cell growth/cell migration, ␤1-INTEGRIN may have a nonredundant role for regulating valve maturation and differentiation, and work is in progress to determine whether FLNA may be the downstream target of PN/␤1-INTEGRIN signaling. In this study, we tested whether PN and its interaction with ␤1-INTEGRIN receptors affect the potential of cushion prevalvular cells to compact and organize collagenous ECM using three-dimensional collagen contraction assays. In this assay, the collective contractile force of cells embedded within a collagen gel lattice was determined by measuring changes in the diameter of the gel over time (the smaller the diameters, the greater the contractile forces). Fig. 9 shows that control cell cultures contracted the diameter of the gel ϳ50% after 2 days in culture, whereas cultures overexpressing PN contracted the diameter of the gel ϳ80 -90%. The ability of the seeded cells to contract collagen gels was abolished by the following: (i) silencing PN using PN shRNA (lane 3); (ii) treatment with blocking antibodies to ␤1-INTEGRIN (lane 4); and (iii) pharmacological inhibitors for FAK (PF573228) (lane 5) and PI3K/AKT (LY294002) (lane 6) prior to PN overexpression. Because HA can signal through its receptor CD44 (56), we determined whether the cross-talk between PN/␤1-INTEGRIN/PI3K signaling and HA-CD44 signaling regulates three-dimensional collagen contraction by treating the cushion cells with blocking antibody against CD44 prior to PN overexpression (Fig. 9, lane 7 compared with lane 1 and lane 2). These treatments inhibited compaction even when PN was being overexpressed (Fig. 9, lanes 3-7 compared with lane 2). These results provide strong evidence that cross-talk between PN signaling mediated through the downstream targets of INTEGRINs (FAK and PI3K) can induce contractile forces in target cushion cells sufficient to contract (compact) collagen in a manner consistent with homeostatic valve maturation as indicated by the progressive attenuation and histological zonation of leaflets. In addition, HA/CD44 signaling plays a key role in cushion cell differentiation and maturation (Fig. 9). Because HA/CD44 signaling, like that of PN/INTEGRIN signaling, has been linked to promoting cell survival and migration (57), PN/INTEGRIN signals are amplified by a stimulatory effect on HA/CD44 signaling through an up-regulation of HAS2 expression/activity. The putative cross-talk between these two signaling pathways might serve to enhance signaling effects of both PN and HA.

Comparison of HA in Developing AV Cushions of WT and PN Null Mice Indicates That HA Synthesis Is a Target of PN
Signaling-One of the most abundant ECM components in developing AV cushions is hyaluronan (HA) (45,58). HA is an extracellular glycosaminoglycan that, along with chondroitin sulfate proteoglycans like versican, has osmotic properties that can expand extracellular spaces in developing AV cushion growth. This potentially facilitates cushion cell migration during the normal period of cushion growth and elongation (ED14.5 to ED18.5). However, as shown in Fig. 10A, HA is greatly reduced when the PN gene is knocked out, which correlates with failure of the cushions to grow, fully elongate, and mature. These findings led us to determine whether HA syn-   . PN regulates Has2 phosphorylation and HA synthesis in mouse valve cells. A, WT cushion cells were transfected with empty vector, or with PN cDNA expression vector, or with control siRNA, or with PN siRNA. Cell lysates from cells that were grown in the presence of 5% serum were directly probed with antibody against HAS2 and ␤-ACTIN. B, HA secreted in the medium of the experiment in A was measured by an ELISA-like assay and was expressed as percent of untransfected control cells. The data are represented as means Ϯ S.D. (n ϭ 3 independent experiments for each set of cultures). C and D, WT cushion cells and PN-null cushion cells were transfected with empty vector or with PN cDNA expression vector. After 48 h of transfection, the cells were kept in Opti-MEM media for 16 h to bring all the cells to G 0 phase to avoid unnecessary phosphorylation due to serum and then were lysed in the presence of protease and phosphatase inhibitors (from Sigma). The cell lysate were then immunoprecipitated with HAS2 antibody, and the blot was prepared from the immunoprecipitate fraction, washed with lithium chloride/PBS, and probed first with anti-phosphoserine Ab (Poly-Z-PS1, Invitrogen) or anti-phosphothreonine Ab (Z-PT1, Invitrogen) (to detect proteins containing phosphoserine or phosphothreonine residues) or reblotting with anti-HAS2-Ab (C-5, sc-365263) as a loading control. Data are representative of three experiments. WB, Western blot. E, total RNAs were extracted from the valve cells isolated from PN-null, WT mice, and from serum-starved WT control cushion cells. qRT-PCR for Has1, Has2, Has3, and GAPDH was done to determine relative mRNA levels of Has1, Has2, and Has3 as described under "Materials and Methods." The data are represented as means Ϯ S.D. (n ϭ 3 independent experiments for each set of cultures) after normalization with GAPDH levels and expressed as relative mRNA levels compared with serum-starved WT control cushion cells. p Յ 0.05 (*) was considered statistically significant. F, phosphoserine-HAS2 was detected in HAS2 immunoprecipitates of cell lysates from E16.5 WT mouse mitral prevalvular cushion cells treated with or without 10 M LY294002 or 10 M U0126 for 1 h. Data are representative of three experiments. G, HA secreted in the medium of the experiment in F was measured by an ELISA-like assay. Data are expressed as means Ϯ S.D. Statistical analyses were done using t test with Mann-Whitney modification as applicable. p Յ 0.05 (*) was considered statistically significant.
thesis is a direct binding target of PN signaling in prevalvular AV cushions.
To do so, cushion cell cultures were prepared from E16.5 embryonic heart AV valves microdissected from wild type and PN-null mice and examined for HA synthase-2 (HAS2) and secretion of HA (Fig. 11, A-G). Cells from PN-null cushions were transfected with PN cDNA expression vector as part of a "rescue" experiment. Immunoprecipitates were prepared from the lysates using monoclonal anti-HAS2 antibodies and analyzed for phosphoserine and phosphothreonine residue(s) within the native HAS2 protein (59). This procedure will determine the extent of Has2 protein phosphorylation at serine/ threonine sites of Has2, which play critical regulatory roles in the activation of Has2 and its potential to secrete HA. The synthesis of HA is an energy-consuming process where the ratio of ATP/AMP-activated kinase controls the synthesis of HA (59,60). During energy stress, the ratio of AMP/ATP increases with consequent activation of AMP kinase. Activated AMP kinase phosphorylates threonine in HAS2, which inactivates the anabolic process to synthesize HA (59,60). Inhibiting anabolic processes and inducing catabolic pathways to restore ATP levels back to normal would be needed to synthesize the UDP-GlcNAc substrate (61,62). On the other hand, phosphorylation of a serine residue would be critical for the normal activation of HAS2 to maintain HA synthesis (59 -62).
In Fig. 11A, Western blots demonstrate that phosphoserine-Has2 was reduced in the PN-null cells but restored or increased when PN-null valves were transfected in culture with PN cDNA expression vector to the same extent as seen in WT cells. Interestingly, the phosphothreonine level is high in the PN-null primary cell cultures, and the level was decreased in PN-null valve cell cultures transfected with PN cDNA expression vector (Fig.  11D). These results indicate that threonine phosphorylation of HAS2 correlates with reduction in the HA synthetic activity in PN-null cell cultures and that serine phosphorylation of HAS2 correlates with higher synthesis of HA in PN/WT mouse cushion progenitor cell cultures. From the results in Fig. 11D, we cannot rule out the presence of other proteins in the HAS2 immunoprecipitate. To know which of the HA synthases in the primordia and cells are affected by PN, we therefore investigated Has1, Has2, and Has3 transcript levels in WT and PNnull cushion progenitor cells. Results in Fig. 11E shows that HA synthase isoforms are differentially localized in post-EMT cushion cells. The endogenous mRNA level of Has2 remains significantly increased, whereas Has3 is only modestly increased, and the Has1 level remains very low (Fig. 11E). Because Has2-null mice die early in cushion development (prior to EMT) (45,63), the observed high level of Has2 (Fig.  11E) may play a major role in the post-EMT stage when the level of PN remains high (Fig. 10).
A previous study reported that ERK-mediated serine phosphorylation of all three HA synthases increased their specific activity for HA synthesis in cancer cells (64). In Fig. 3, we showed that PN/INTEGRIN/FAK-dependent ERK1/2 responses could be modulated by the PI3K/AKT signaling. We next explored the possibility that PN-dependent HAS2 activation and synthesis could be modulated by the PI3K/AKT and/or ERK signaling. Results in Fig.  11, D and E, show that both PI3K and ERK inhibitors attenuated the phosphoserine-HAS2 level and decreased HA synthesis in WT mouse valve cell cultures. In addition, because PI3K in this mouse valve primary cell culture regulates the ERK activation (Fig. 3), the effect of PI3K on inhibition of phosphoserine-HAS2 level and HA synthesis is higher compared with that of ERK inhibition alone (Fig. 11, D-E).
Although these in vivo and in vitro findings support the hypothesis that PN can promote activation of HAS2 and thereby increase synthesis of HA, they did not necessarily prove that PN directly induced signaling for these events. Thus, we determined whether activation of HAS2 (and subsequent HA synthesis) is mediated by PI3K, a downstream kinase target of PN/INTEGRIN interaction as shown in Figs. 1-3. We focused on PI3K because our studies with cancer cells show that the HA synthesis is highly dependent on PI3K signaling (20,21), and these results can be comparable with our findings in Fig. 11, D and E. Cultures of chick prevalvular cushion cells were pretreated with the PI3K inhibitor LY294002 (10 M) for 45 min, then transfected with expression vectors of full-length PN cDNA to promote PN expression, or with plasmids that constitutively activated PI3K (CA-PI3K) (20), and cultured for an additional 48 h. Fig. 12A shows that the effect of overexpressing PI3K on the induction of PN synthesis is greater than the response to PN cDNA expression vector with CA-PI3K (Fig.  12A, lane 2 compared with lane 4). Pretreatment of the cells with the PI3K inhibitor LY294002 almost overrode the effects of PN overexpression and PI3K overexpression on both AKT phosphorylation and PN expression (Fig. 12A, lanes 3 and 5  compared with lanes 2 and 4). In a parallel set of experiments, the ELISA-like assay for HA (Fig. 12B) also showed that the effect of overexpressing PI3K on the induction of HA synthesis is greater than the response to PN cDNA expression vector with or without CA-PI3K (Fig. 12B, lanes 2-4 compared with lane 1). HA synthesis also reverted to control levels (Fig. 12B, lane 1) when cushion cells were treated with LY294002 (Fig. 12B, lanes  5-7 compared with lanes 2-4, respectively). These results further emphasize that PN stimulates HAS2 expression/activation (Fig. 11) and HA synthesis (Figs. 10 and 12B) and that this stimulation is induced by PN/␤3-INTEGRIN/FAK signaling mediated through PI3K. In contrast, the MEK1/2 inhibitor U-0126 (20 M) did not alter the combined stimulatory effect of CA-PI3K and PN cDNA overexpression (Fig. 12B, lane 8 compared with lane 4). Taken together, these data indicate that the PN-induced intracellular signaling pathways directly stimulate HA synthesis and secretion in a time frame that correlates with the elongation of AV cushions into the future mitral or tricuspid leaflets (Fig. 10). Interestingly, an unexpected finding was that by inhibiting PI3K the expression of PN itself was inhibited (Fig. 12A, compare lane 3 with lane 2), indicating that a positive feedback loop may exist between PN and PI3K and/or ERK and HA to sustain PN expression and increase phosphoserine-HAS2 expression and HA production and its downstream signaling potential through INTEGRINs.

DISCUSSION
Coordinated regulation of valve progenitor cell migration, increased cell growth, adhesion with secretion, assembly, and alignment of ECM are essential for remodeling of the AV mesenchymal primordia (cushions) into mature valve leaflets. The process begins at midgestation (about E14.5) when PN expression begins to increase, and it finishes postnatally as PN expression declines to a low, barely detectable baseline level (51,52). This suggested a potential role for PN in valvulogenesis that was confirmed by globally deleting the PN gene (10,65). The PNnull AV valves exhibited a noncompacted myxomatous-like phenotype, which indicated that differentiation and remodeling had been adversely affected (7,10,11). At least two not necessarily mutually exclusive mechanisms could account for this abnormal phenotype. First, PN, like other matricellular proteins, could directly bind to collagen fibrils, cross-linking them or linking them to binding sites of the cell surface (15). The second mechanism, and the hypothesis of this study, is that interaction of PN with cell-surface receptors induces downstream signaling activities in prevalvular mesenchymal cells that regulate their differentiation and morphogenesis. Our study tested this mechanism using well defined explant culture and gene transfection models. The results shown in Figs. 1-3 demonstrate that PN in embryonic mouse AV cushion cells increased activation of p-FAK, p-AKT, and p-ERK. However, the regulation of ERK signaling in a PI3K-dependent manner via FAK (Fig. 3, E and G) has not been previously described. Although our data suggest that ERK is one key downstream effector of PI3K-dependent cell survival activity in these cells, we cannot exclude the possibility that PN-␤1/3-INTEGRIN may also regulate signaling pathways in addition to ERK that contributes to the survival of these cells.
Silencing RNA against ␤3-INTEGRIN inhibited these responses, indicating that activation of these kinases was medi-ated through binding of PN to ␤-INTEGRIN heterodimeric receptors (Figs. 4 -6), which are well known for their binding to ECM proteins (53). Using blocking antibodies against ␤1-INTEGRIN, ␤3-INTEGRIN, and ␤5-INTEGRINs, our data in Fig.  4 show that ␤3-INTEGRIN and ␤1-INTEGRIN are necessary to attach and spread prevalvular cushion cells on a migratory surface. Under these conditions, PN-induced cushion cell migration was mediated through downstream FAK and PI3K/AKT signaling (Fig. 7). These data strongly suggest that PN interaction with INTEGRIN receptors cooperates with FAK and PI3K/ AKT signaling to promote cellular activities required for normal valvulogenesis. Interestingly, blocking ERK has little or no effect on migration (Fig. 7). However, PN activation of ERK appears to correlate with expression of collagen ( Figs. 1 and 2). Although our data do not indicate whether PN directly regulates collagen mRNA and protein expression in valve progenitor cells, we observed reduced Col1␣1 mRNA production in cultured PNnull cushion cells (Fig. 8A), which is consistent with the reduction in collagen expression observed in vivo in PN-null valves (15). Also, enhanced expression of Col1␣1 mRNA was observed following overexpression of PN in cultured WT cushion cells, which was prevented by inhibiting ␤3-INTEGRIN, PI3K, or ERK (Fig. 8C), indicating that PN most probably induces COL1␣1 expression directly through the ␤3-INTEGRIN/FAK/PI3K/AKT and/or ERK pathways. Because PN can also bind to ␤5and ␤1-INTEGRINs, it is possible that these INTEGRINs also might induce Col1␣1 expression through these same pathways.
Our data in Fig. 9 suggest the interaction of PN and ␤-INTEGRIN can also generate forces that compact collagen in three-dimensional gel assays, which we used to simulate the morphogenetic role of PN in compacting prevalvular cushions into attenuated valve leaflets. How PN-INTEGRIN binding generates this contractile force is not known, but the data in Fig.  9 suggest that it is likely related to kinase activation of a downstream effector mechanism such as the ACTIN cytoskeleton through activation of FAK/PI3K. Changes in the ACTIN cytoskeletal organization are well known to occur at sites of FAK/ PI3K expression (57,66). Although our compaction assays strongly link a PN-INTEGRIN/FAK/PI3K signaling pathway with generating contractile forces sufficient to compact collagen, there is no clear downstream molecular target that might link a PN-induced signaling pathway with the ACTIN cytoskeleton. One possible candidate is FLNA. Additionally, because the cross-talk between PN and HA-CD44 signaling that plays an important role in cushion cell contraction (Fig. 9), HA-CD44 signaling may also target FLNA to activate contractility as has been seen in other cell types (67). We have found similarity both in the expression patterns of PN and FLNA and in the phenotypes of PN and FLNA knock-out mice, each of which have inhibited ECM remodeling of primordial AV valve (10,53). This phenotype is also characteristically seen in patients with mitral valve prolapse, including a nonsyndromic subset of patients with point mutations in FLNA (52,54,55). Filamin A is an ACTIN-binding protein that anchors various transmembrane proteins to the cytoskeleton, and it provides a scaffold for many cytoplasmic signaling proteins involved in ACTIN cytoskeleton remodeling. We recently demonstrated that the lossof-function mutations in FLNA that are associated with human mitral valve prolapse can reduce cell spreading and migratory potential and disrupt a signaling network that balances RhoA and Rac1 GTPases that regulate FLNA binding capabilities (68). Future studies are needed to determine whether FLNA is a downstream target of PN signaling pathways and, if so, whether PN/FLNA are part of a central regulatory network that generates contractile forces required to compact and maturate cushions into leaflets.
Finally, a link was found between PN signaling pathways and the activation of HA synthesis (through Has2) and HA secretion. Has2 is expressed by most, if not all, cells and is essential for life. Although the Has1 and -3 null mice are developmentally normal, the Has2 null mouse dies at an early embryonic stage prior to the period when endocardial endothelial cells normally undergo epithelial-mesenchymal transition to form prevalvular mesenchyme. This suggests that Has2 expression is important for EMT; however, the level of Has2 continues to remain high after EMT when cushion mesenchyme normally expands and elongates distally into the lumen of the heart beginning at E15.5 and continuing into early postnatal life (36). Organ cultures of the wild type heart tube endothelial cells undergo EMT to form cardiac cushion mesenchyme at the initiation of HA synthesis by HAS2. In contrast, organ cultures of the Has2-null heart endothelium do not undergo any EMT, but they do so if hyaluronan is added to the culture medium (45,63,69). In the initial experiment using the cell lysate, we found that HAS2 antibody recognizes two bands associated between ϳ50 and 60 kDa (Fig. 11A). We also found that the slow moving band of the doublets of HAS2 changes with overexpression of PN and with silencing of PN (Fig. 11A). The faster moving band close to ϳ50 kDa remains fairly constant (Fig. 11A), indicating that HAS2 must remain inactive (likely the ϳ50-kDa band) until it is inserted into the plasma membrane where it is activated to synthesize and extrude the hyaluronan chains. Moreover, as shown in the Fig. 11B, stimulation of hyaluronan synthesis correlated with both an increase and a decrease of the expression of the slow moving band close to ϳ60 kDa of the doublet HAS2 expression in E16.5 WT cushion cells. These data indicate that HAS2 phosphorylation may occur for the activation of HAS2 protein, and this activation may be correlated to hyaluronan synthesis. Within a protein, phosphorylations can occur on several amino acids. Phosphorylation on serine is the most common, followed by threonine. Tyrosine phosphorylation is relatively rare in most of the eukaryotes. The correlative evidence in Fig. 11, A and B, led us to perform the immunoprecipitation experiment with anti-HAS2 antibody and then to probe the HAS2 immunoprecipitates with phosphoserine and phosphothreonine antibodies to provide further evidence that phosphoserine and phosphothreonine on HAS2 are involved in regulation of enzyme activity in responses to PN stimulation and inhibition. Our findings demonstrate that in the absence of PN (e.g. in null mice), there is a change in the balance of serine versus threonine phosphorylation of HAS2. In PN-null prevalvular cells, phosphoserine-HAS2 is low, whereas phosphothreonine-HAS2 is high (Fig. 11, C and D). Such a ratio correlates with low ATP/AMP ratios (70) that are known to inhibit HA synthesis by enhancing phosphorylation of HAS2 at threonine sites that decrease its HA synthetic activity as indicated in Fig. 10. In Fig. 11, we show that the levels of HAS2-serine phosphorylation (Fig. 11C) and the levels of HAS2-threonine phosphorylation (Fig. 11D) parallel the changes observed in HA expression/synthesis in WT versus null mouse valves (Figs. 10 and 11E). In addition, our study found that Has2 mRNA had highest expression in WT cushion tissues and had significant statistical difference when compared with that of PN-Null (Fig. 11E). The expression of HA synthases in ex vivo cushion cultures from both WT and PN-null mice and the expression of HA and PN in WT and PN-null valve sections (Fig. 10) indicate that these two molecules are closely associated. Moreover, the expression of HAS2 protein and Has2 mRNA and its activation in response to PN cDNA in PN-null cells (Fig. 11, C-E) also indicate that PN is one of the stimulators of expression/and activation of HAS2 and HA secretion. However, it is unlikely that any effect of PN on promoting HA secretion is related to the EMT process as mesenchyme formation is not affected by deletion of the PN gene. Thus, whatever role PN has in promoting HA synthesis, it is likely to be a post-EMT function, and this process requires PI3K and PI3K-regulated ERK (Figs. 11, F and G, and 12, A and B).
Although our findings indicate that ERK activation is also enhanced by a PI3K-dependent mechanism through Tyr-925-FAK (Fig. 3F), we do not have an explanation as to the mechanism. Normally Tyr-925-FAK serves as a docking site for the p85 regulatory subunit of PI3K that strongly activates PI3K (35)(36)(37)(38). Overall, our studies reveal a unique mechanism of the dramatic differences in signaling kinetics of PI3K over ERK (Fig. 3F), and PI3K-dependent regulation of ERK activation (Fig. 3F) may provide some insight as to why prevalvular cells when induced to overexpress PI3K (using CA-PI3K) have less sensitivity to a single agent ERK inhibitor U0126 for HA production (Fig. 12B, lane 8 compared with lane 4).
Summary-The results presented in this study provide a more comprehensive picture of the role of PN in valve morphogenesis by testing the hypothesis that PN secreted by developing valve progenitor cells induces signaling pathways that change biological functions related to morphogenesis via binding to cell-surface receptors, ␤3and ␤1-INTEGRINs. The interaction between PN and ␤-INTEGRINs was found to activate FAK and downstream PI3K and MAPK pathways that lead to the phosphorylation of AKT and ERK, factors that are well known for their effects upon survival, adhesion/migration, and collagen synthesis. Because PN can promote HA synthesis and HA, in turn, can activate intracellular signaling pathways through CD44 receptors, it is possible that one consequence of PN/INTEGRIN activation of HAS2 is to indirectly amplify its own signaling potential. HA/CD44 signaling, like that of PN/INTEGRIN signaling, has also been linked to promoting cell survival and migration (57). Thus, a putative positive feedback loop between these two signaling pathways might enhance activation of PI3K/cell survival signaling. This would also be consistent with our previous study (20,71) where PI3K, and its downstream signaling component AKT (20), control Has2 expression at the molecular level. Furthermore, this study demonstrates for the first time that signaling pathways activated by PN/␤-INTEGRIN interaction in developing valve cells can also stimulate HA synthesis and activation of HAS2 protein at a phosphoserine site. There is increasing evidence that phosphorylation of serine and threonine residues in HAS2 control hya-luronan synthesis whether or not it is activated (59,72). The immunoprecipitates in Fig. 11, C, D, and F, correlate well with the HA synthesis results in Fig. 11B, i.e. phosphoserine increases when HA synthesis increases and phosphothreonine increases when HA synthesis decreases, as is expected from the data discussed by Hascall et al. (72). Thus, our results provide the evidence that stimulation (and inhibition) of hyaluronan synthesis may be correlated with an increase (and a decrease) of the expression of the slow moving band of the doublet of HAS2 expression in E16.5 WT cushion cells in response to stimulation (or inhibition) of periostin. A detailed characterization of the serine and threonine sites of phosphorylation may lead to the understanding of the role of serine/ threonine kinases in the regulation of cellular functions in response to PN. Work is in progress to determine whether PI3K/AKT, as well as ERK, regulates feedback stimulation of hyaluronan and involves transcription or other post-translational modifications of HAS2. Our findings in Figs. 1-12 lead to the hypothesis that PN INTEGRIN signaling pathways might be coordinately linked to HA signaling pathways through its adhesion receptor, CD44, for valve cell differentiation/maturation through collagen compaction and may be for collagen production as has been seen in our recent study in fibrogenic lung fibroblasts (19). This hypothesis is included in Fig. 13, which shows a model summarizing our findings in which we propose that by binding to specific INTEGRINs PN activates intracellular kinase signaling pathways (e.g. FAK/PI3K/MAPK), and this signaling cooperatively works with HA-CD44 signaling to promote fibrogenic differentiation and maturation of AV prevalvular mesenchyme into mitral and tricuspid valve leaflets.