|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: PDB APP JKB MBH BBM. Performed the experiments: PDB APP JKB ACA MDL AMG. Analyzed the data: PDB APP JKB MBH. Contributed reagents/materials/analysis tools: GSP. Wrote the paper: PDB MBH. Manuscript review: GSP BBM MBH.
In bronchopulmonary dysplasia (BPD), alveolar septae are thickened with collagen and α-smooth muscle actin, transforming growth factor (TGF)-β-positive myofibroblasts. Periostin, a secreted extracellular matrix protein, is involved in TGF-β-mediated fibrosis and myofibroblast differentiation. We hypothesized that periostin expression is required for hypoalveolarization and interstitial fibrosis in hyperoxia-exposed neonatal mice, an animal model for this disease. We also examined periostin expression in neonatal lung mesenchymal stromal cells and lung tissue of hyperoxia-exposed neonatal mice and human infants with BPD. Two-to-three day-old wild-type and periostin null mice were exposed to air or 75% oxygen for 14 days. Mesenchymal stromal cells were isolated from tracheal aspirates of premature infants. Hyperoxic exposure of neonatal mice increased alveolar wall periostin expression, particularly in areas of interstitial thickening. Periostin co-localized with α-smooth muscle actin, suggesting synthesis by myofibroblasts. A similar pattern was found in lung sections of infants dying of BPD. Unlike wild-type mice, hyperoxia-exposed periostin null mice did not show larger air spaces or α-smooth muscle-positive myofibroblasts. Compared to hyperoxia-exposed wild-type mice, hyperoxia-exposed periostin null mice also showed reduced lung mRNA expression of α-smooth muscle actin, elastin, CXCL1, CXCL2 and CCL4. TGF-β treatment increased mesenchymal stromal cell periostin expression, and periostin treatment increased TGF-β-mediated DNA synthesis and myofibroblast differentiation. We conclude that periostin expression is increased in the lungs of hyperoxia-exposed neonatal mice and infants with BPD, and is required for hyperoxia-induced hypoalveolarization and interstitial fibrosis.
Increased survival of very premature infants has been accompanied by an increased incidence of bronchopulmonary dysplasia (BPD) . In the “new BPD,” there are larger and fewer alveoli, as well as poorly formed secondary crests, indicating interference with septation , . Alveolar septa are thickened with collagen and α-smooth muscle actin-, transforming growth factor (TGF)-β-positive myofibroblasts , , , . Adenoviral transfer of the TGF-β gene to newborn rat lungs induces changes consistent with BPD, including excess matrix deposition and large undeveloped pre-alveolar saccules . Overexpression of TGF-β in neonatal mouse lungs induces proliferation of α-actin-positive cells within the alveolar septal walls and hypoalveolarization . Together, these data imply a critical role for TGF-β in the development of BPD.
We have isolated mesenchymal stromal cells from the tracheal aspirates of premature infants . Primary cell colonies produce TGF-β1 and undergo TGF-β-induced myofibroblastic differentiation, suggesting that, in the absence of other signals, myofibroblastic differentiation represents the “default program” for mesenchymal stromal cell specialization . Isolation of these cells is associated with the development of BPD . Gene expression profiling revealed that, compared to lung fibroblasts, mesenchymal stromal cells overexpress the gene POSTN . POSTN encodes periostin, a secreted protein with an N-terminal secretory signal sequence and four fasciclin domains. Periostin, a member of a subset of non-structural extracellular matrix-associated molecules termed “matricellular proteins,” directly interacts with other extracellular matrix proteins including collagen and fibronectin and is a ligand for αvβ3, αvβ5 and α4β6 integrins. In the lung, periostin is expressed in stromal cells surrounding squamous cell carcinoma , rat pulmonary artery smooth muscle cells , primary human lung fibroblasts  and human bronchial epithelial cells , . Periostin treatment increases TGF-β1 mRNA expression in human bronchial epithelial cells as well as TGF-β1-mediated collagen I gene expression in airway fibroblasts . In the heart, periostin is induced by TGF-β but also required for normal TGF-β responsiveness . Periostin promotes myofibroblast differentiation of palmar fascia mesenchymal cells  and is a component of subepithelial fibrosis in asthma . Lysyl oxidase, which crosslinks collagen and elastin, is proteolytically activated by periostin , . Together, these data suggest that periostin plays a significant role in TGF-β-mediated fibrosis and myofibroblast differentiation.
Based on the above evidence that TGF-β plays an important role in the pathogenesis of BPD, we hypothesized that periostin expression is required for the myofibroblastic differentiation and alveolar simplication in hyperoxia-exposed neonatal mice, a commonly used animal model for this disease. We also examined the effects of periostin on mesenchymal stromal cell myofibroblastic differentiation in vitro, as well as the expression of periostin in BPD lungs.
Two-to-three day-old wild-type C57BL/6J mice were exposed to air or 75% oxygen for 14 days, as described . This period corresponds to the alveolar stage of mouse lung development, corresponding to the 28–32 week human fetus . As shown previously , compared to air-exposed mice (panels 1A–D), hyperoxic exposure caused the development of fewer and larger airspaces (Figure 1E). Analogous to BPD, lungs of hyperoxia-exposed mice showed increased elastin and collagen-I deposition by α-actin-positive myofibroblasts (Figure 1F). Hyperoxia also increased lung periostin expression. In the lungs of air-exposed mice, immunohistochemical staining revealed a modest amount of periostin expression in the airway tissues, as well as the tips of developing alveolar walls (Figure 1C). In the lungs of hyperoxia-exposed mice, there was increased periostin expression in the alveolar walls, particularly in areas of interstitial thickening (Figure 1G). In some instances, periostin co-localized with α-smooth muscle actin, suggesting synthesis by myofibroblasts (Figure 1H). Sections stained with isotype control showed no signal (not shown).
We also analyzed whole lung extracts for periostin protein expression by immunoblotting (Figure 2). The full-length periostin protein is 90 kD in size, but there are also products of splice variants . Hyperoxic exposure increased periostin protein abundance over two-fold (p<0.05, one-way ANOVA).
Periostin staining of lung sections from infants dying of BPD were compared with those from full-term infants. Immunohistochemical staining in the lungs of full-term infants was localized to the airway subepithelium (Figure 3A). Staining in BPD lungs was increased, and was most prominent in the subepithelium of thickened alveolar walls (Figures 3B–D). There appeared to be some staining in fibroblastic foci. Using florescence microscopy, lungs of full-term infants showed periostin expression in the airway subepithelium which was distinct from the adjacent smooth muscle (Figure 3E). In BPD lungs, there is colocalization of α-smooth muscle actin and periostin in interstitial fibroblasts and at the tips of secondary crests (Figures 3F–H).
Two-to-three day-old B6;129-Postntm1Jmol/J mice were exposed to air or 75% oxygen for 14 days. Air-exposed wild-type mice showed a normal alveolarization pattern (Figure 4A). In contrast, hyperoxic exposure of wild-type mice caused alveolar simplification (Figure 4B). Air-exposed periostin null mice showed a normal alveolarization pattern. Close examination appeared to show a slight increase in the cellularity of alveolar walls, and increased alveolar macrophages (Figure 4C). Unlike wild-type mice, hyperoxia-exposed periostin null mice did not show larger air spaces (Figure 4D). In hyperoxia-exposed wild-type mice, hypoalveolarization was associated with an increase in mean alveolar chord length (Figure 4E).
Air-exposed wild-type mice showed α-smooth muscle actin staining (red) in the airway and arterial walls, but no expression in the alveolar interstitia (Figure 5A). Hyperoxia-exposed wild-type mice showed thickening of the interstitial space with α-smooth muscle-positive myofibroblasts (Figure 5B, α-actin, red; periostin, green; collagen I, blue; colocalization is white). Neither air- nor hyperoxia-exposed periostin null mice showed α-actin-positive myofibroblasts, interstitial thickening or lung periostin expression (Figures 5C and 5D).
We examined the effects of periostin knockout on lung mRNA expression, as measured by quantitative PCR. Consistent with the observed increase in periostin-positive interstitial myofibroblasts, hyperoxic exposure significantly increased the expression of α-smooth muscle actin, elastin and periostin (Figure 6, each p<0.05, one-way ANOVA). There was no significant increase in α-actin, elastin or periostin expression in the hyperoxia-exposed periostin null mice. Periostin KO mice did not show a reduction in hyperoxia-induced collagen type Iα1 expression. Hyperoxia also significantly increased the mRNA expression of three chemokines – CXCL1, CXCL2 and CCL4 – in wild-type but not periostin null mice. Three genes related to angiogenesis – those encoding vascular endothelial growth factor (VEGF)-A, KDR/VEGF-receptor 2/Flk1 and platelet endothelial cell adhesion molecule (PECAM)-1/CD31 – were significantly downregulated by hyperoxia treatment. However, the expression of these genes was significantly reduced in air-exposed periostin knockout mice, and hyperoxia had no effect on the expression of these genes.
We also examined right ventricular thickness in air- and hyperoxia-exposed wild-type and periostin null mice (Figure 6). As reported previously , hyperoxia increased right ventricular wall thickness, suggestive of pulmonary hypertension (p<0.05, one way ANOVA). There was no increase in wall thickness in periostin null mice.
TGF-β treatment has been shown to increase periostin expression in cardiac myocytes, aortic smooth muscle, and bronchial epithelium. We treated neonatal lung mesenchymal cells with 10 ng/ml TGF-β for 72 h and assessed periostin mRNA and protein expression by qRT-PCR and ELISA, respectively. TGF-β treatment tended to increase periostin mRNA levels, but the changes were not statistically significant (Figure 7A). Mesenchymal stromal cells expressed pg levels of periostin protein at baseline, and there was a small but statistically significant increase with TGF-β treatment (Figure 7B).
We examined the interactive effects of periostin (0–500 ng/ml) and TGF-β (0–10 ng/ml) on cellular [3H]-thymidine incorporation by two-way ANOVA (Figure 8). We employed concentrations of periostin found to present in fibrotic lesions in vivo . Cells treated with 10 ng/ml TGF-ß in the presence of 500 ng/ml periostin showed an increase in DNA synthesis, whereas cells treated with 10 ng/ml TGF-β in the absence of periostin did not show an increase. Similarly, cells treated with 500 ng/ml periostin in the presence of 10 ng/ml TGF-β showed an increase in DNA synthesis, whereas cells treated with 500 ng/ml periostin in the absence of TGF-β did not.
The effects of periostin (50 ng/ml) on TGF-β-induced α-smooth muscle actin and elastin mRNA expression were examined in mesenchymal stromal cells, using two-way ANOVA for statistical analysis (Figure 9A). TGF-β concentrations of 0–10 ng/ml were used. 10 ng/ml TGF-β significantly increased α-actin and elastin expression in both the presence and absence of periostin. However, 50 ng/ml periostin significantly increased α-actin and elastin expression only in the presence of TGF-β. Immunocytochemical stains showed that periostin increases TGF-β-induced α-smooth muscle actin, elastin and collagen I protein expression (Figures 9B–9E).
We previously characterized the gene expression profile of neonatal lung mesenchymal stromal cells . Among other genes, we found that, compared to lung fibroblasts, mesenchymal stromal cells showed increased expression of POSTN. POSTN encodes periostin, a secreted non-structural extracellular matrix protein which regulates TGF-β-mediated fibrosis  and myofibroblast differentiation . Previous studies have shown that periostin is also expressed by human bone marrow-derived mesenchymal stem cells , .
The isolation of mesenchymal stromal cells from neonatal tracheal aspirates confers a nearly 23 fold increase in the odds of developing BPD . Overexpression of TGF-β in neonatal mouse lungs induces proliferation of α-actin-positive cells within the alveolar septal walls and hypoalveolarization , a phenotype analogous to BPD. On this basis, we hypothesized that periostin may play a prominent role in BPD pathogenesis. To test this, we measured periostin expression in mice exposed to hyperoxia during the first two weeks of life, a period which corresponds to the alveolar stage of human lung development . We also examined periostin expression in the lungs of human infants with BPD. We found that periostin expression is increased in the lungs of hyperoxia-exposed neonatal mice. Periostin was present in the alveolar walls, particularly in areas of interstitial thickening associated with α-smooth muscle actin-positive myofibroblasts. In human BPD lungs, the distribution of staining was similar, with prominent staining in the subepithelium of thickened alveolar walls and fibroblastic foci. A similar distribution was recently described in patients with usual interstitial pneumonia and fibrotic-type nonspecific interstitial pneumonia . The colocalization of periostin and α-smooth muscle actin suggests that periostin was expressed in part by alveolar myofibroblasts.
To further examine the possible contribution of periostin to BPD pathogenesis, we compared the responses to hyperoxia of wild-type mice to periostin null mice. As expected. wild-type mice show hypoalveolarization and interstitial fibrosis when exposed to hyperoxia during the alveolar period of lung development. Lung mRNA expression of α-actin, elastin and periostin was also increased. In contrast, periostin null mice showed normal alveolar development. Further, in contrast to control mice, the alveolar walls of periostin null mice did not show thickening with α-smooth muscle-positive myofibroblasts or deposition of collagen or elastin. Lung mRNA expression of α-actin and elastin was not increased in hyperoxia-exposed periostin null mice. These results suggest that periostin is required for hypoalveolarization and interstitial fibrosis in hyperoxia-exposed neonatal mice.
TGF-β overexpression in neonatal mouse lungs induces proliferation of α-smooth muscle actin-positive cells within the septal walls . On this basis, we examined the effects of periostin on neonatal lung mesenchymal stromal cell proliferation and myofibroblastic differentiation. TGF-β and periostin had synergistic effects on DNA synthesis. Further, treatment with periostin enhanced TGF-β-induced α-smooth muscle actin, elastin and collagen type I expression, outcomes indicative of myofibroblast differentiation. These data are consistent with the notion that the observed effect of TGF-β on myofibroblast proliferation and differentiation, which contributes to interstitial thickening and fibrosis in vivo , is mediated in part by periostin expression. Since periostin null mice were protected from hyperoxia-induced hypoalveolarization, we speculate that abnormal myofibroblast proliferation and differentiation, mediated by aberrant periostin expression, is incompatible with secondary crest development.
Matricellular proteins are non-structural proteins that are secreted and sequestered in the extracellular matrix, where they interact with integrins, growth factors, proteases, cytokines and other extracellular matrix proteins. Among various functions, they regulate TGF-β activation, adhesion, migration, fiber deposition and angiogenesis, as well as cell proliferation, differentiation and survival. All are involved in fibrosis and increased matrix deposition, and most are expressed at low levels in normal adult tissue but upregulated during development, wound healing and tissue remodeling. Thus, many of the effects of TGF-β may be mediated by matricellular proteins. Few studies have examined the role of matricellular proteins in neonatal lung injury. Overexpression of connective tissue growth factor (CTGF) during development results in a lung phenotype analogous to BPD (33). CTGF expression is also increased after high tidal volume ventilation in newborn rat lungs (34). Lysyl oxidase, which crosslinks collagen and elastin, is proteolytically activated by CTGF and periostin , , and increased in the lungs of infants with BPD and hyperoxia-exposed mice (35). Since overstabilization of the extracellular matrix by excessive lysyl oxidase activity might impede the normal matrix remodeling that is required for pulmonary alveolarization, it is conceivable that the periostin knockout mice are protected from hypoalveolarization due to reduced lysyl oxidase activation.
Increasing evidence suggests that periostin plays a role in lung inflammation. Subepithelial periostin deposition and fibrosis are present in the bronchial tissue of both ovalbumin-sensitized and ovalbumin-inhaled mice and patients with asthma . Periostin mRNA expression is upregulated in bronchial epithelial cells of asthmatic subjects , . Periostin null mice show defect in allergen-induced eosinophil recruitment to the lungs . In the present study, the lungs of periostin null mice showed attenuated expression of the neutrophil and monocyte chemoattractants CXCL1, CXCL2 and CCL4. CXCR2, the receptor for CXCL1 and CXCL2, is required for double-stranded RNA-induced neutrophil sequestration and hypoalveolarization in newborn mice , and NF-κB signaling in fetal lung macrophages disrupts airway morphogenesis . Together, these data are consistent with the notion that periostin promotes hyperoxia-induced lung inflammation in neonatal mice.
There are important limitations to our study. First, the periostin null mice we used were developed in the 129 strain and backcrossed to C57BL/6. Since neither the 129 or C57BL/6 strain are perfect controls, we employed C57BL/6 mice for this purpose. Nevertheless, strain differences in the response to hyperoxia could have contributed to the protective effect of the periostin knockout , . However, a detailed comparison of C57BL/6 and 129/Sv strains showed similar susceptibility to hyperoxia, including changes in lung neutrophils, IL-6, and expression of collagen type Iα2 and fibronectin. Further, expression of collagens type IIIα1 and IVα3 was increased after hyperoxic exposure in the 129/S strain, suggesting that these mice are capable of a fibrotic response when exposed to hyperoxia. Second, we have not completely characterized the periostin knockout mice, which appear to have a subtle pulmonary phenotype. At post-natal age 16–17, air-exposed animals showed increased alveolar macrophages and cellularity, as well as decreased expression of angiogenesis-related genes. Nevertheless, periostin null mice did not show pulmonary hypertension, as evidenced by a normal right ventricular wall thickness. This uncoupling of alveolar development and expression of angiogenesis-related genes remains unexplained, but could relate to an acceleration of alveolarization in periostin null mice.
In conclusion, we have shown that periostin expression is increased in the lungs of hyperoxia-exposed neonatal mice and infants with BPD, and is required for hyperoxia-induced hypoalveolarization and interstitial fibrosis. Our data also demonstrate that neonatal lung mesenchymal stromal cells are not only potent sources of periostin, but also respond to periostin treatment by differentiation into myofibroblasts. These data significantly extend previous results showing that an excess of TGF-β and CTGF each induce the BPD phenotype in naïve animals , , . Since periostin is a downstream effector of TGF-β, periostin may represent a promising therapeutic target for BPD.
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Institutional Animal Care and Use Committee of the University of Michigan Medical School (protocol ID#10397). All surgery was performed under sodium pentobarbital anesthesia, and all efforts were made to minimize suffering.
Two-to-three day-old wild-type C57BL/6J and B6;129-Postntm1Jmol/J were purchased from Jackson Laboratories (Bar Harbor, ME). The periostin null mice were developed in 129 mice and backcrossed once to C57BL/6 . Mice were exposed to air or 75% oxygen for 14 days using a polypropylene chamber coupled to an oxygen controller and sensor (BioSpherix, Lacona, NY) . Dams were exchanged between air and hyperoxia daily. Because of periodontal disease in periostin null mice , wild-type and knockout animals were fed with powdered chow. After exposure, genotyping was performed by Transnetyx (Cordova, TN). Animal work was approved by the Institutional Animal Care and Use Committee.
Lungs were perfused with 5 mM EDTA and inflated to 30 cm H2O pressure with 4% paraformaldehyde (Sigma-Aldrich, St. Louis, MO). Sections were stained with hematoxylin and eosin and mean alveolar chord length determined . Slides were probed with fluorophore-labeled mouse anti-α-smooth muscle actin (clone 1A4, Sigma-Aldrich), rabbit anti-collagen I, mouse anti-elastin and rabbit anti-periostin (Abcam, Cambridge, MA). Other sections were probed with anti-periostin and stained with a biotinylated anti-rabbit IgG-avidin horseradish peroxidase and diaminobenzidine detection system (Vector Labs, Burlingame, CA).
Mouse lungs were washed in PBS and homogenized in a buffer containing 50 mM Tris (pH 7.5), 100 mM NaCl, 50 mM NaF, 40 mM β-glycerophosphate, 2 mM EDTA, 200 µM Na3VO4, 1% Triton X-100, 1% SDS and 1% sodium deoxycholate containing complete protease inhibitors (Roche Diagnostics, Indianapolis, IN). After centrifugation for 30 min at 4°C at 10,000×g, supernatant fluids were matched for protein and 30 µg/lane processed for SDS-PAGE and nitrocellulose blotting. Nitrocellulose membranes were probed with anti-periostin (Abcam, Cambridge, MA).
Human lung tissue was obtained from the University of Rochester Lung Biorepository under a protocol approved by the Institutional Review Board of Strong Memorial Hospital (Rochester, NY). Specimens were obtained from 3 infants who died in the intensive care nursery (Table 1). The diagnosis of BPD was based on premature delivery, need for chronic respiratory support, a requirement for supplemental oxygen after 36 weeks' gestation, and consistent chest radiographs, as well as pathologic tissue diagnosis at autopsy. Specimens were also obtained from infants succumbing to non-pulmonary disorders. Samples were processed within 6 hours of death.
Total RNA was extracted using the RNeasy Plus Mini kit (Qiagen, Valencia, CA), then transcibed to first-strand cDNA using Taqman Reverse Transcription Reagents (Applied Biosystems, Foster City, CA). First-strand cDNA was then used to quantify the expression of periostin (POSTN), α-smooth muscle actin (ACTA2), elastin (ELN), collagen type Iα1 (COLIA1), CXCL1, CXCL2, CCL4, VEGF-A, KDR/VEGFR2/Flk1, PECAM1/CD31 and the housekeeping gene glyceraldehyde dehydrogenase (GAPDH). The primers used are shown in Table 2. All primers were from IDT (Coralville, IA) except for VEGF-A, which were obtained as an assay mix from InVitrogen (Carlsbad, CA).
Isolation of neonatal lung mesenchymal stromal cells was approved by the University of Michigan Institutional Review Board. Cells were isolated from tracheal aspirates of premature infants undergoing mechanical ventilation (Table 2), as described . Passage 3 and 4 cells were studied.
Hearts from normoxic or hyperoxic mice were removed, fixed in formalin, and processed for hematoxylin and eosin staining. Images of the right ventricular outer wall were made at 10× magnification and measured using NIH ImageJ. A hemacytometer was used to standardize the number of pixels/micrometer.
Mesenchymal stromal cells were grown on fibronectin-coated glass slides (BD Biosciences), fixed in 1% paraformaldehyde and probed with antibodies against α-smooth muscle actin, collagen I and elastin.
Mouse lung mesenchymal cells were plated into 12-well dishes at 100,000 cells/well, allowed to attach overnight, and incubated in serum-free medium supplemented with 500,000 cpm [3H]-thymidine. After 8 h, cells were treated with TGF-β1 or mouse periostin. Cells were maintained in the medium for 48 h. At the end of this time, the radioactive medium was removed, wells were washed three times with PBS, and the DNA precipitated with 10% trichloroacetic acid at 4°C. After 30 min, the acid was removed and the precipitate was washed three times with 70% ethanol at 4°C and solubilized at room temperature for 30 min with 1 mL 0.5 N NaOH and 1% Triton X-100. Solubilized DNA was counted in a liquid scintillation counter.
Cell supernatants were analyzed for soluble periostin by ELISA (Adipo Bioscience, Santa Clara, CA).
All data were described as mean±SEM. For simple comparisons of two or more groups, we used one-way analysis of variance (ANOVA). When examining the influence of two factors, we used two-way ANOVA. Significant differences between groups were pinpointed by Fisher's least significant difference multiple comparisons test.
Competing Interests: The authors have declared that no competing interests exist.
Funding: This work was supported by NIH grants HL79339 and HL90134 (MBH) and HL07749 (PDB). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.