|Home | About | Journals | Submit | Contact Us | Français|
Cardiac fibroblasts are the most populous non-myocyte cell type within the mature heart and are required for extracellular matrix synthesis and deposition, generation of the cardiac skeleton, and to electrically insulate the atria from the ventricles. Significantly, cardiac fibroblasts have also been shown to play an important role in cardiomyocyte growth and expansion of the ventricular chambers during heart development. Although there are currently no cardiac fibroblast-restricted molecular markers, it is generally envisaged that the majority of the cardiac fibroblasts are derived from the proepicardium via epithelial-to-mesenchymal transformation. However, still relatively little is known about when and where the cardiac fibroblasts cells are generated, the lineage of each cell, and how cardiac fibroblasts move to reside in their final position throughout all four cardiac chambers. In this review we summarize the current understanding regarding the function of Periostin, a useful marker of the non-cardiomyocyte lineages, and its role during cardiac morphogenesis. Characterization of the cardiac fibroblast lineage and identification of the signals that maintain, expand and regulate their differentiation will be required to improve our understanding of cardiac function in both normal and pathophysiological states.
Although non-cardiomyocytes constitute the majority of the cell types present in the postnatal heart and form the cardiac skeleton within which the cardiac myocytes reside, relatively little is known about how the cardiac interstitial microenvironment is formed and the source of the cardiac fibroblast (CF) lineage. The signals that trigger a secretory fibroblast phenotype and collagen formation (fibrogenesis) as well as the morphogenesis of the CF lineage are also not well understood. The CF is the most abundant non-cardiomyocyte cell type present within the postnatal mature heart and is chiefly responsible for deposition of the extracellular matrix (ECM). The ECM is considered a dynamic modulatory network due to the continuous changes in secretory activity which alters both cell environment and response throughout development. Indeed, far from inert, the ECM is characterized by constant reorganization in response to endogenous and exogenous stimuli1. The ECM also provides structural support for cardiac myocytes and formation of the elaborate cardiac skeleton. The cardiac skeleton encodes the 3D structure of the heart and is composed of a tough sheet of fibrous connective tissue that electrically isolates the atria from the ventricles, contains all four valves and valvular anchorage tissues, and serves as an attachment site for cardiac muscle fibers.
Balanced synthesis2 and degradation3–5 of this ECM is key to normal cardiovascular development, physiological growth of cardiac muscle (exercise), pathological responses to injury (myocardial infarction, hypertrophy, hypertension and ischemia-reperfusion)6–8, and for optimal heart function9. CFs themselves are a source of paracrine growth factors10, but can also respond to hormones, growth factors, cytokines and mechanical forces. Expression of receptors for ECM11,12 and neurotransmitters13, allow CFs to couple mechanical14, electrical and sympathetic stimuli to functional responses. Excess Transforming Growth Factor beta (TGFβ)-mediated15–19 deposition of cardiac ECM, resulting in fibrosis, has been associated with activation of various signal transduction pathways in utero, postnatally and during pathophysiological heart overload. These TGFβ activated CFs are re-classified as “myofibroblasts” due to their unusual morphology, which is characterized by some features of smooth muscle differentiation (can express actin and/or myosin), and functional characteristics20–22. Cardiac fibrosis, which results in stiffening of the ventricular walls, diminished contractility, and abnormalities in cardiac conductance, is a common consequence of heart disease; thus understanding the role of CFs in sensing, integrating, and responding to stimuli is of both scientific and clinical significance. The role of CFs in pathologic remodeling and heart failure has been extensively reviewed elsewhere23. Periostin (gene Postn), a TGFβ superfamily-responsive matricellular protein, has recently emerged as important for Collagen fibrillogenesis and overall organization of ECM24–28. In this review we summarize the current knowledge regarding Periostin function within CFs and cardiac morphogenesis.
This is a vague but convenient term for an ill-defined connective tissue cell derived from the primitive mesoderm29,30. Originally, fibroblasts were described in the late 19th century based solely on their location and morphological criteria29,30. Fibroblasts are typically identified by their spindle-shaped flattened morphology, ability to adhere to culture plates30, and absence of markers of epithelial, smooth muscle, endothelial, perineural, and histiocytic cells30. Fibroblasts synthesize most of the ECM of connective tissue (composed of fibrillar Collagens and Fibronectin). Their nuclei are large, euchromatic and possess prominent nucleoli. Fibroblasts are characterized as being non-vascular, non-epithelial, and non-inflammatory cells29. In all tissues, fibroblasts are usually adherent to the fibers which they themselves lay down and thus can form a 3D-network and become embedded within the fibrillar ECM29,32.
Fibroblasts serve diverse vital functions during embryonic development including synthesis of ECM, instructive epithelial differentiation, inflammation regulation, and wound healing29. However, the lack of a reliable and specific fibroblast marker is a major limiting factor in the study of fibroblasts in vivo and is assuredly why they remain so poorly understood in both molecular and cellular terms29,30. Although there are several established indicators of fibroblast phenotype, none are exclusive to fibroblasts or are present in all fibroblasts29. Fibroblasts are heterogeneous and exhibit topographic differentiation, meaning fibroblasts from different anatomical sites have distinct characteristics and phenotypes which can subsequently be maintained in vitro when fibroblasts are isolated from their surrounding environment and the influence of other cells30. Fibroblasts retain a memory of the number of divisions they have completed, and even if suspended from division by being frozen, they will complete only the remainder of their divisions before arresting33. Additionally, fibroblasts from differing anatomical sites have distinct transcriptional patterns. Of particular interest, ECM gene expression patterns can vary based on location of fibroblast harvest30 as well as genes involved in lipid metabolism, cell signaling pathways that control proliferation, cell migration, and cell fate determination30. Similar to local differentiation of skin fibroblasts, atrial fibroblasts are known to express singular gene expression patterns and exhibit different morphology when compared to ventricular fibroblasts34. These data correlate with studies showing atrial fibrosis is more severe than ventricular fibrosis in congestive heart failure35. These chamber-specific phenotypic differences may arise from the different physiological environments that exist in the atria which are absent within the ventricles, where the CFs originate from or even when they colonize the various chambers of the heart. Accordingly, it remains unclear if all CFs are secretory, whether they have multiple origins, and if they have memory and are pre-specified during development of the hearts chambers.
The main cellular components of the postnatal heart are cardiac myocytes, endothelial cells, vascular smooth muscle cells (VSMCs), and CFs36. Although CFs are the predominant cell type in number, the cardiac myocytes actually occupy the greatest volume37,38. Unlike cardiac myocytes, endothelial cells and VSMCs; CFs never have a basement membrane and display multiple processes39. Thus, CFs can be distinguished from other non-myocyte lineages via use of Laminin or Collagen-IV to assess absence of a basement membrane in CFs. Additional characteristics include extensive rough endoplasmic reticulum, prominent Golgi apparatus, and abundant cytoplasmic granular material39. CFs are found throughout the heart in a 3D-network surrounding myocytes (Fig. 1) and bridging the gaps between myocardial tissues32,40. Myocytes are arranged in laminae bounded by endomysial Collagen, and the CFs lie within this endomysial network38,41. Although there are no fibroblast-restricted or even CF-restricted molecular markers, both in utero and mature CFs have been shown to express Discoidin domain receptor-2 (Ddr2), Vimentin, Fibroblast-specific protein (Fsp-1), Periostin (Fig. 2) and several key ECM molecules36–42 (Table 1).
CFs appear coincidently with ventricular compaction around embryonic day (E) 12.5 (Fig. 1) and increase in number steadily through postnatal day one43. CFs have recently been shown to play an important role in proliferation during heart development43. The mammalian heart undergoes a major change in physiological pressures in order to transition from a fetal to neonatal circulation. Cessation of flow through the ductus arteriosis and increased pulmonary return cause an elevation of ventricular pressure. A robust CF response to the increased neonatal circulatory demands are seen during the first two neonatal weeks when the CF population increases from ~10% to ~20–70% within a relatively short time36,44. Most studies agree that the first two weeks of murine cardiac growth result in at least a doubling of the CF population. However, less consensus is found regarding the relative CF make-up of the adult heart. Early studies on adult rat left ventricle estimated that 65–70% of the cells were non-cardiomyocytes45,46, while recent studies analyzing the total mouse heart by FACS and confocal microscopy estimate a much higher number of cardiomyocytes ~56% and a ~44% non-myocyte content, with only 27% of those staining positive for Ddr236,41. The densest population of CFs in healthy adult hearts is found around the sinoatrial node47,48, thus providing complete electrical insulation. It remains unclear whether CFs are evenly distributed throughout the developing heart and whether they emerge via a uniform or clustered spatiotemporal manner, as quantitative studies describing embryological populations in specific regions of the developing heart have yet to be reported. Intriguingly, confocal staining with Ddr2 in E16 mouse hearts shows localized epicardial surface, atrial, and incomplete ventricular free wall and septal expression49. In most species the cardiac system continues to develop through prenatal and postnatal life, but there are wide interspecies differences in timing and duration of specific events, different electrophysical properties, and distinct TGFβ ligand requirements50–55.
It has been estimated that every cardiomyocyte is in direct contact with one or more fibroblasts38 (Fig. 1). Thus CFs are perfectly positioned to be able to contribute to the structural, biochemical, mechanical, and electrical properties of the working myocardium32. The primary function of CFs is the synthesis and maintenance of a 3D scaffold for cardiomyocytes which insures the functional integrity of the myocardium56,57. This mechanical scaffold integrates the contractile activity of individual cells to coordinate the pump function of the heart40,58. In addition to providing a scaffold, the CF fashioned ECM serves to connect cellular structures, distributes mechanical forces throughout the myocardium, transmits signals via cell surface ECM receptors, and its density regulates fluid movement within the extracardiac environment37. In vitro, interstitial flow has also been shown to regulate TGFβ expression, myofibroblast differentiation and ECM alignment in a TGFβ-dependent process59. In addition to its primary structural role, the ECM can also play an instructive role; by acting as a repository of growth factors, incorporating them into the matrix to regulate ligand availability, and possibly helping to establish morphogenetic gradients during cardiovascular development. A heterodimer of the α-subunits of Collagen-I is the major collagenous product of CFs, accounting for 80% of the total content15. Significantly, CFs generate essential autocrine and paracrine factors that control muscle cell growth43,60,61. A study that examined the effect of medium derived from CFs on isolated ventricular cardiomyocytes indicated that CFs could induce alterations in both myocyte structural and functional characteristics62. Co-culture of CFs isolated from E12.5–13.5 murine hearts with cardiomyocytes resulted in significantly higher proliferation of cardiomyocytes than was seen in co-culture with adult-derived CFs. Embryonic fibroblasts have elevated levels of Tenascin C, Fibronectin1, Periostin (Fig. 2,,3),3), Hyaluronan and Proteoglycan link protein-1, Heparin-binding EGF-like growth factor, Pleiotrophin and several types of Collagen as compared to adult-derived CFs. Embryonic knock-down of Collagen-IIIα1 and Fibronectin1 using siRNA phenocopies cardiomyocyte proliferation rates seen with adult CFs43. In addition to stimulating myocyte growth, CFs also have the ability to sense mechanical stress through multiple pathways, including integrins, ion channels, and second messenger responders32. Mechanical stimulation of cultured CFs results in ECM gene expression, growth factor production, and collagenase activity60. CFs have clearly been shown to affect myocardial development and remodeling through direct contact with cardiomyocytes, but their ability to alter cardiomyocyte behavior through noncontact, profibrotic signals is less well understood62.
CFs are connected to one another via specific Cadherins and Connexins (Cx-40), to the ECM via integrins, and to the myocytes by connexins (Cx-45)41,47. In addition to their role in forming insulating barriers, studies have suggested CFs may electrically couple to cardiac myocytes. Through electronic interactions, CFs may synchronize and possibly relay electrical activity of multicellular cardiac tissue over distances upto 300μm47,56. Therefore, it is possible that CFs (particularly myofibroblasts) could provide bridges that connect regions of myocytes that otherwise would be electrically isolated by connective tissue. Even though most of the data is from in vitro work, CFs have been shown to synchronize contraction among individual cardiomyocytes, with accompanying membrane potential fluctuations56. Although they do not respond to electric stimulus with the generation of an action potential (non-excitable), CFs do contribute to myocardial electrophysiology32,63. Coupling of myocytes and CFs would lead to significant changes in action potential duration and upstroke velocity63. When CFs are electrically coupled to myocytes, they could act as current sinks and consequently decrease conduction velocity and maximum depolarization rate of the action potential58,64. Gap junctional communication between CFs and myocytes is thought to be established for short range interaction at the single cell level56. However, the strength of the coupling, how widespread it may be in vivo and its potential impact on action potential characteristics remain unknown.
Despite the identification of fibroblasts in the late 19th century and the advent of elegant lineage mapping tools such as DiI labeling, quail-chick chimeras, zebrafish photoactivatable (caged) fluorescein marking, and murine loxP/Cre recombinase-mediated genetic cell-marking techniques; still relatively little is known about the origin and development of the CFs. Fundamental to understanding organogenesis is the ability to determine when and where specific cell types are generated, the ancestry of each cell, and how cells move to reside in their final position. The mesenchymal cells that give rise to the CFs are believed to be derived principally from the embryonic epicardium65–67. However, other sources such as during in utero endocardial cushion epithelial-to-mesenchymal transformation (EMT) and valve morphogenesis68, and via the postnatal recruitment of circulating bone marrow cells of hematopoietic origin28 has been proposed. Todate, lineage mapping in developing mouse and chick embryos has not revealed any significant contribution via the neural crest53,69, second heart field70,71 or endothelial lineages72. Although lineage tracing experiments using endothelial-specific genes such as Tie2−Cre73, Flk1−Cre74, and VE-Cadherin75 do not show significant contribution to the in utero mouse cardiac fibroblast lineage, novel data has emerged that show CFs can be derived from endothelial-to-mesenchymal transition in damaged tissues40,76. These pathological CFs appear to be key for pathogenesis of fibrosis and the facilitation of tumor progression. Due to the current lack of a robust CF-specific molecular marker, the late arrival of CFs after the majority of heart morphogenesis is completed, the difficulty in discriminating between de novo expression of the various transgenic drivers and when derivative cells are labeled71, as well as the inherent variability observed among the various lineage mapping tools; it is still conceivable that there may be multiple spatiotemporal sources for subpopulations of the pleiotropic CF lineage. Nevertheless, epicardially-derived cells (EPDCs) are currently considered to be the major source of CFs.
The epicardium is the last layer of the vertebrate heart to form, and originates from the transient proepicardium77. The proepicardium consists of an accumulation of finger-like vesicular protrusions of the pericardial coelomic mesothelium that forms close to the venous pole of the embryonic heart. As a result of proepicardial EMT, the proepicardial ECM becomes populated with mesenchymal cells that migrate and cover the embryonic heart to form the epicardium (reviewed in 77). Following epicardial EMT, the EPDCs are thought to give rise to the majority of interstitial CFs, undifferentiated subepicardial mesenchyme, coronary endothelium and coronary VSMCs. EPDCs can give rise to CFs, either through EMT from the ventricular surface78–81 or by invading the ventricular and atrial walls and migration via the fibrous annulus80. Epicardial cells additionally, play important modulatory roles in myocardial development81,82 and some controversial data suggests that EPDCs might even differentiate into myocardial cells83,84. However, this theory has recently been challenged85, and illustrates the limitations of relying solely on lineage mapping without robust lineage-restricted molecular markers and clear-cut morphological identification criteria.
Experimental studies in both chick and mouse have proved that EPDCs have important roles in heart development67. The major consequence of the absence of epicardium is a thin myocardial compact layer which results in poor cardiac function and usually leads to in utero lethality86. In addition, there are often significant abnormalities within cardiac looping, septation, and coronary morphogenesis associated with EPCD dysfunction/ablation87,88. Given that a subpopulation of EPDCs can differentiate into a variety of different cell types (including coronary endothelium, coronary VSMCs, interstitial cardiac fibroblasts, and atrioventricular cushion mesenchymal cells), the EPDCs have even been called the ultimate ‘cardiac stem cell’67,77,89. When EMT is blocked via antisense targeting of the Ets1/2 transcription factors, epicardial organization is disturbed; and there is a lack of epicardial mesenchyme, coronary VSMCs, and normal myocardial morphology90. In this experiment, the proepicardium formed normally, and the epicardial cells were able to migrate and cover the cardiac surface, but due to a loss of EMT, the formation of the subepicardial mesenchymal was hindered90. Both the Wilms’ Tumor gene and Erythropoietin growth factor are highly expressed in the epicardium and knockouts of both these genes result in ventricular hypoplasia, pericardial bleeding and mid-gestation lethality91,92. Finally, it has been shown that there is cross-talk between the epicardium and myocardium, and this interaction is essential for normal cardiac development93–97. Vascular cell adhesion molecule (VCAM-1) is a surface protein that mediates adhesion via α4 Integrin94,98,99. VCAM-1 is expressed in the myocardium, while α4 is expressed in the epicardium94,95,100. Both gene knockout mutants have early placental defects in ~50% of the null embryos while the other half have an absent epicardium and embryonic lethal heart defects94,95. Another essential interaction is between Gata4 and its co-factor, Friend of Gata2 (Fog2)96,97. The transcriptional activity of Gata4 is modulated through a physical interaction with the transcription factor Fog296. Targeted disruption shows that Fog2 null embryos die midgestation. Although Fog2 nulls form an intact epicardial layer that properly expresses epicardium-specific genes, EMT is disrupted and the vascular network in the myocardium never forms96. Significantly, Fog2 expression in the myocardium under the αMHC promoter can rescue coronary vasculature96. This transgenic rescue experiment is consistent with the idea that signals from the myocardium control epicardial EMT79,93,96,89,101. Thus, there are several lines of evidence that collectively indicate that when either the epicardium is absent/abnormal or when myocardial crosstalk is compromised, there is a resultant thin myocardial compact layer or “non-compaction of ventricular myocardium” CHDs102. Thus, signaling between the epicardium, cardiomyocytes and EMT-derived CFs is multidirectional, and probably alters as the developing heart undergoes morphogenesis. Although it is presently unclear whether a thin myocardial layer is incapable of compaction or even if CF morphogenesis is affected in the various non-compaction and thin myocardial heart phenotypes, it is interesting to note that primitive zebrafish and newt hearts contain few CFs (mainly confined to the subepicardial layer) and have a spongy (not compact) myocardium with an absence of coronary vessels103.
In addition to the interstitial CFs of the myocardium, the valvular interstitial cells (VICs) are also classed as fibroblasts. However, VICs are different from CFs of the cardiac skeleton/fibrous annulus and interstitial CFs as they are largely thought to be derived from endothelial cells that have undergone EMT72,104. VICs are considered ‘fibroblast-like’ but unlike CFs, may vary their phenotype in response to ECM, mechanical force, and soluble factors in their microenvironment105. Morphologically and functionally, VICs have characteristics of both CFs and smooth muscle cells106. During development, VICs maintain normal valve structure by producing, secreting, and degrading the ECM in which they are embedded107. Maintenance of ECM architecture provides the mechanical characteristics essential for perpetuating the unique behavior of the valve108.
Periostin is a 90kDa secreted protein involved in cell adhesion and contains four repetitive fasciclin domains that are similar in sequence to the insect protein Fasciclin-1, which is involved in neuronal cell-cell adhesions109. Periostin is expressed in the periosteum and periodontal ligament110, injured vessels111, metastatic cancer cells112, and in cells undergoing EMT112,113. Although the vast majority of Periostin made by cells is secreted and deposited extracellularly26, some intracellular staining has is observed, but is limited to the secretory apparatus26. As a secreted ECM protein that associates with areas of normal fibrogenesis or pathological fibrosis, Periostin can directly interact with other ECM proteins such as Fibronectin, Tenascin-C, Collagen-I/V, and Heparin26. Indeed, ultrastructural studies demonstrate that Periostin directly interacts with Collagen-I and can regulate fibril diameter25. Periostin can also serve as a ligand for select integrins, such as αvβ3, αvβ5, and α4β6, where it can affect the ability of cancer cells to migrate and/or undergo a mesenchymal transformation112–115. However, it remains unclear whether this ligand-receptor association also occurs during normal heart development.
Periostin is considered a “matricellular protein”. Matricellular proteins are ‘matrix’ proteins which regulate cell function and cell-matrix interactions, but do not contribute directly to the physical properties or organization of structures such as fibrils or basal laminae116. Thus, matricellular proteins do not have a direct structural role. In addition to Periostin, the family includes Thrombospondin-1, Osteonectin, Osteopontin, Tenascin-C, and Tenascin-X117. Matricellular proteins are thought to function via binding matrix proteins and cell surface receptors, as well as modulating expression of cytokines, proteases and growth factors118. Several mouse models in which matricellular proteins have been knocked-out survive embryogenesis and only show mild phenotypes after birth116. Furthermore, the phenotypes of these gene targeted mice are consistent with their minimal contribution to structural integrity and suggest a redundancy during development.
To investigate the requirement of Postn, several groups generated mice that lack Postn24,26,119. In order to ensure a null allele and facilitate efficient spatiotemporal reporter expression analysis, we replaced the Postn translation start site and first exon with a lacZ reporter gene (PostnLacZ)120. Analysis of the endogenous Postn mRNA, protein and PostnLacZ reporter (Figs. 2–4) spatiotemporal expression patterns in wildtype and heterozygotes (in the case of the PostnLacZ reporter), all reveal that Postn is initially detected in E10–10.5 CFs as well as the nascent endocardial cushions. In older E15 hearts, Periostin is robustly expressed in CFs and the epicardial cells that cover the embryonic heart (Fig. 2). Within the CF lineage, Periostin is expressed in vivo in all the Epicardin and Ddr2-postive CFs (Fig. 2) and in vitro with Collagen-I and Ddr226. Periostin is co-expressed intracellularly with Collagen-I in the cytoplasmic ER/Golgi, indicative of active ECM synthesis26. Neonatal and adult CFs continue to robustly express endogenous Postn mRNA, protein and the PostnLacZ reporter (Figs. 3,,4).4). Thus, the PostnLacZ reporter provides a useful molecular marker of CFs as they colonize the fetal heart and during adult homeostasis (Fig. 4). Although Periostin is widely expressed, the majority of the PostnLacZ null mice survived well into adulthood, but show smaller overall body weights. While ~12% of the nulls die before weaning due to structural valvular anomalies26, the remaining PostnLacZ nulls all develop an early-onset periodontal ligament (PDL)-like phenotype and craniofacial ECM anomalies120. Analysis of the CF spatiotemporal distribution with Postn null hearts revealed that CF numbers were unaltered, indicating that Periostin is not required for CF formation but may affect CF function. Significantly, when Postn nulls were subjected to pressure overload stimulation and myocardial infarction, they exhibited reduced fibrosis24, which resulted in aneurysm and rupture of the ventricular wall, especially if subjected to changes in hemodynamic pressure or to a heart attack. Conversely, over-expression of Periostin in the heart protected from rupture in infarcted regions24. Studies of Postn-deficient mice also revealed that Periostin can regulate Collagen-I and viscoelastic properties of connective tissue25. In fact, Postn is significantly upregulated by both mechanical stretch121 and pro-fibrotic TGFβ signaling26,110. Adult hearts injured via myocardial infarction or exhibiting myocarditis and calcification, exhibit Periostin upregulation specifically within the CF lineage (Fig. 6). Thus, loss of Postn results in ECM alterations that are reflected as structural defects26,120 and the quantitative amount of Periostin (and perhaps other ECM components) can alter normal physiological interactions between CFs and myocytes that can in turn affect Collagen, fibrosis and scar mechanics122.
PERIOSTIN is reduced in pediatric patients with bicuspid aortic valve26, but PERIOSTIN levels were increased123 in Marfan syndrome (MFS) patients with thickened heart valves and primary myocardial dysfunction124. In diseased pediatric cardiac valves, PERIOSTIN was largely absent from regions where ECM stratification was lost, Collagen fibrils were disorganized, and elastin content reduced26. In mouse models of MFS (Fibrillin-1 mutations), there was loss of tissue integrity due to dysfunctional microfibril assembly and function but an elevation in TGFβ superfamily signaling123. Significantly, in adult mice hearts, Periostin is strongly up-regulated before and during remodeling resulting from long-term pressure overload stimulation and myocardial infarction24. These studies demonstrated that levels of Periostin correlated with the amount of Collagen produced within the adult heart, and that Periostin expression is restricted to the non-cardiomyocyte lineage24,26. Previous controversial studies indicated that Periostin might also be expressed by cardiac myocytes125 and may directly mediate cardiomyocyte proliferation126. However, these studies were equivocal as in vitro studies using neonatal myocytes are rarely free of fibroblasts122. Moreover, these studies used de-differentiating adult rat myocytes cultured for 9 days and a recombinant truncated form of Periostin to ascertain their biological effects27. Nevertheless, over-expression of Postn in the mouse heart results in the proper accumulation of Periostin protein in the ECM without detectable intracellular myocyte retention. Furthermore, Postn over-expression did not increase cardiac myocyte number at baseline, nor did it augment incidence associated with cardiac repair after myocardial infarction injury24. Similarly, more recent studies have unequivocally demonstrated that genetic manipulation of mouse Postn expression in the heart does not affect adult myocyte content, cell cycle activity, or cardiac repair127. Periostin may also play an essential role during healing in response to acute myocardial infarction via FAK-Integrin mediated recruitment of activated CFs128.
It is likely that the cardiovascular developmental function of Periostin may include facilitating proper organization of the ECM and in affecting cellular trafficking of CF cells through an injured or reorganizing area. Whereas Ddr2 can act as a collagen receptor 129, Periostin itself may act as a scaffold-binding protein that enables collagen realignment in response to TGFβ, increasing cross-linking and ensuring normal fibrillogenesis25,122. There is currently no data to support a role for Periostin in stimulating ECM secretion; rather, its absence may result in an ECM that does not sustain a reorganized laminar structure. Indeed, collagen fibrils from Postn null mice are reduced in size, somewhat disorganized, and less efficiently cross-linked25,128. The close association of Periostin and Collagen-I25, which is one of the most abundant ECM proteins in the cardiac skeleton that provides tensile strength, suggests that Periostin may play a role in CF stretch-sensitive signaling and absorption of mechanical stresses. This is born out by the finding that pressure overloaded Postn null hearts exhibit rupture of the ventricular wall, but Postn over-expression protects against rupture24. Similarly, Postn nulls exhibit a PDL-like phenotype120 that can be ameliorated via removal of masticatory forces121. Finally, Periostin (in conjunction with Serum Response Factors) may act as nodal switch that tips the balance of cardiac skeleton differentiation from a fibroblastic to a myocardial/smooth muscle phenotype, as adult Postn null epicardial cells can ectopically express myocardial markers130. This suggests that secreted Periostin may act at several levels during cardiovascular morphogenesis and postnatal cardiac homeostasis. Thus, Periostin provides a useful molecular marker of the CF population and future characterization of the Postn promoter elements may generate valuable lineage mapping/conditional targeting reagents.
Notwithstanding the complex and intriguing correlation of deregulated Postn expression levels in both normal and pathological hearts, very little is known about how Periostin is regulated. In addition to the in vivo and in vitro Postn null data described above26, several studies have suggested a link between Periostin and elevated TGFβ signaling110,111,131–133. Exogenous addition of TGFβ1 to isolated CFs and cardiomyocytes results in a significant upregulation of Periostin expression within only the CF lineage26. Coupled with in vivo analysis of wildtype and Postn null hearts24,26, these in vitro approaches demonstrate that it is uniquely the CF lineage that expresses Periostin and that Periostin is responsive to TGFβ activation. Furthermore, not only is Periostin directly induced via TGFβ signaling, but Postn itself may be required for normal TGFβ-responsiveness26. Using isolated E14 PostnLacZ null and wildtype embryonic fibroblasts within a 3D-Collagen lattice formation assay system, it was shown that PostnLacZ null embryonic fibroblasts exhibited reduced 3D-lattice formation and TGFβ-responsivness was blunted26. In order to directly test in vivo TGFβ-responsiveness in mature hearts, Periostin expression was examined using cardiac-restricted αMHC-TGFβ1 constitutively active mice. Significantly, TGFβ-induced Periostin upregulation occurs, even in the absence of fibrosis26. Inversely, when TGFβ activity is abrogated using an inducible dominant-negative TGFβ type-II receptor131 or TGFβ– neutralizing antibodies134, expression of Periostin is reduced. Although the mechanism causing this response is presently unknown, this suggests Periostin deposition may be required to enhance the fibrotic remodeling effects of TGFβ and stabilize microfibril networks within the cardiac skeleton26.
Targeted deletions of each individual TGFβ ligand reveals both distinctive and overlapping functions135–137. Tgfβ1 null mice are largely unaffected and in utero viable, but exhibit postnatal bone and hair follicle maldevelopment138. Tgfβ3 nulls are also in utero viable but exhibit cleft palate139. Only Tgfβ2 nulls show evidence of congenital heart defects, including thickened valves and double-outlet-right ventricle with concomitant interventricular septal defects140. Since Postn null mice exhibit reduced TGFβ activity (i.e., pSmad2/3), it is possible that Periostin may also be transcriptionally regulated through one or more of the TGFβ ligands. In situ hybridization reveals that only Tgfβ2 mRNA colocalizes with Postn, as Tgfβ1 is restricted to the adjacent endothelial lineage and Tgfβ3 is largely confined to the endocardial cushions/valve and annulus (Fig. 5). Significantly, measurement of Postn levels in isolated Tgfβ1, β2 and β3 null ventricles reveals that Postn is unaffected in Tgfβ1 and β3 null hearts, but is significantly reduced in Tgfβ2 null hearts (Fig. 5). Furthermore, Periostin protein expression is also significantly diminished in Tgfβ2 null ventricles (Fig. 5). Given that several studies have shown exogenous non-physiological levels of TGFβ1 and pathological TGFβ1 overexpression can both induce Postn110,111,123,132, these data suggest Periostin is responsive to TGFβ1 but requires TGFβ2 for normal spatiotemporal expression within the in utero heart. Given the diminished TGFβ1 responsiveness observed in Postn null embryonic CFs, it is tempting to speculate that the relative composition of the ECM imparts contextual specificity to Periostin-TGFβ signaling by either concentrating the ligands at sites of intended function (positive regulation) or by inhibiting their bioavailability (negative regulation).
The past several years have yielded remarkable insights and progress in identifying and mapping the various cell lineages that initially give rise to the developing heart, and deciphering many of the key morphological events that are required for both normal heart development and the underlying causes of congenital heart defects. Despite this recent progress, our understanding of the mechanisms of induction and lineage specification of early non-cardiomyocyte cell fate is still rudimentary, and the signals that instruct epicardial precursors to select a CF cell lineage remain unclear. Progress towards understanding the molecular, cellular and morphological events required to generate the hearts interstitium and the cardiac skeleton that provides its 3D-form, have been hampered via the lack of suitable molecular markers and appropriate lineage analysis tools. We propose that the matricellular Periostin protein provides a useful system with which to probe the functional role of the CF lineage and non-cardiomyocyte heart development.
A challenge now facing investigators is characterization and isolation of the key progenitor cells within the developing embryo that give rise to the CF lineage. The identification of cell surface markers of putative CF progenitor/stem cells would be invaluable for the investigation of the potential stem cell niches in the developing heart. Isolation of these progenitor cells would allow delineation of their molecular profile and would accelerate discoveries of the signals instructing these cells to select cardiovascular cell fates. Understanding the biology of CF progenitor cells as the heart forms would also allow a detailed search of the adult heart to assess whether progenitors of similar phenotype remain in the post-natal heart or may be reactivated during repair after cardiac injury.
We thank Goldie Lin and Ralston Barnes for immunohistological aid and members of the Riley Heart Research Center and our colleagues for their important insights and many useful suggestions. Whilst we have tried to comprehensively review as much of the current cardiovascular development literature as possible, we regret that our review cannot include all of the field’s many exciting findings.
Sources of Funding: This work was supported, in part, by the Riley Children’s Foundation, P01 HL085098 (S.J.C), R01 HL092508 (T.D. and S.J.C.), T32 HL079995 Training Grant in Vascular Biology and Medicine (P.S. and K.S.), and by the Indiana University Department of Pediatrics/Cardiology (S.J.C.).