INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: vasculogenesis; lung development; lung injury

Chronic lung disease of premature infants, particularly bronchopulmonary dysplasia (BPD), is a major cause of long-term hospitalization, slow growth, and recurrent respiratory illnesses. The cause of BPD is complex and likely involves oxidant injury, barotrauma/volutrauma from mechanical ventilation, chronic inflammation, and disordered repair in the immature lung. The pathology of BPD includes inflammation, abnormal alveolarization, fibrosis, and vascular abnormalities (1, 2). Recently, Jobe and Ikegami suggested that BPD represents developmental arrest of the lung (3). Pulmonary microvascular development, which is essential for efficient gas exchange, may be disrupted in BPD, but few studies have examined the alveolar vasculature or factors that may regulate vascular development in this disease.

In mature lung, alveolar capillaries are abundant and in close proximity to the alveolar epithelium, creating a thin air- blood barrier. Formation of the pulmonary microvasculature begins by at least the tenth embryonic day in the mouse, continues in all lung developmental stages, and is directly related to overall lung growth (4, 5). In the canalicular stage the capillaries, which previously formed a loose network within the mesenchyme, begin to arrange around the developing distal airspaces (4). In the saccular stage, bilayer capillaries become embedded in thick primary septa. As the lung matures, thin alveolar septa develop a single capillary network immediately subjacent to the alveolar epithelium. Premature infants born at 24-27 wk gestation, who have a high incidence of BPD, are in the late canalicular stage. Although the number of endothelial cells has not been quantified in human BPD, decreased endothelial cells and microvascular dysmorphogenesis are characteristic of experimental BPD in premature baboons (6, 7). The mechanisms of abnormal microvascular development are not known.

Vasculogenesis, the differentiation of endothelial cells from mesenchymal precursors, and angiogenesis, the sprouting of new vessels from the preexisting vasculature, are regulated by several angiogenic factors, including vascular endothelial growth factor (VEGF) and angiopoietin-1 (Ang-1). VEGF (also known as VEGF-A) is a specific mitogen for endothelial cells that also regulates endothelial cell migration and capillary permeability (8). Heterozygous null mutants for VEGF die in early embryonic development, suggesting that VEGF plays a key role in vascular development (9). Our previous study of murine lung showed that VEGF mRNA increased 10-fold between the Day 13 embryo and the 2-wk postnatal animal (10). Major changes to the developing microvasculature occur during this time, including expansion of the capillary surface area and orientation of capillaries with the alveolar epithelium (4). In developing lung, distal airspace epithelial cells express VEGF, implying that the epithelium may regulate the temporal and spatial development of alveolar capillaries (10, 11). In adult lung, VEGF is expressed by type II cells that have relatively low surfactant protein-C (SP-C) (12).

Hyperoxia and inflammatory cytokines, which are implicated in the pathogenesis of BPD, are toxic to endothelial cells. VEGF is an endothelial cell survival factor that can ameliorate oxygen and cytokine injury (13). Acute oxygen toxicity results in decreased lung VEGF expression (11, 12, 17), but it is not known if decreased VEGF is mechanistically linked to loss of endothelial cells in hyperoxia.

VEGF actions are mediated by two tyrosine kinase receptors, Flt-1 (VEGFR1) and KDR/Flk-1 (VEGFR2). Flt-1 may be involved in capillary network formation whereas KDR/Flk-1 transmits a mitogenic signal in endothelial cells. Null mutants of VEGF receptors are embryonic lethal (18, 19). In developing murine lung, we found a 10-fold increase in message for KDR/Flk-1 that paralleled VEGF mRNA (10).

Little is known about Ang-1 and its tyrosine kinase receptor TIE-2 in lung. Ang-1 mediates vessel maturation from simple endothelial tubes into more elaborate vascular structures composed of several cell types. Vessel maintenance via the cell-cell and the cell-matrix associations are also regulated by Ang-1 (20). Ang-1 and TIE-2 knockout animals die with vascular abnormalities in early embryo development (21, 22).

Previous autopsy studies of BPD in humans have generally focused on abnormalities of precapillary arterioles (23). However, little is known about alveolar microvascular abnormalities in this disease. Coalson found dysmorphic alveolar capillaries in patients who died with BPD (2). We compared lungs from infants dying with the clinical and pathological diagnosis of BPD with lungs from infants dying without lung disease. We hypothesized that premature newborn infants dying with BPD would have abnormal alveolar capillaries and decreased expression of angiogenic factors or their receptors compared with infants dying without BPD.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Collection and Handling of Autopsy Tissue Sample

Lung samples were obtained from infants who died in the Neonatal Intensive Care Unit at the Children's Hospital at Strong, University of Rochester Medical Center between 1995 and 2000. Effort was made to obtain lung samples rapidly after death. The diagnosis of BPD was based on a review of the patient's record by one of the authors (G.S.P.) and examination of lung specimens by a pediatric pathologist (L.A.M.). This diagnosis was restricted to premature infants who had a chest radiograph consistent with BPD and who received supplemental oxygen and ventilator therapy at the time of death. Pathological diagnosis of BPD was based on standard light microscopic criteria. As controls, we collected autopsy samples from term or near-term patients dying without lung disease.

At autopsy the right middle lobe (RML) and a portion of the right lower lobe (RLL) were removed. The RML was immediately inflation fixed at 20 cm H₂O with 10% buffered formalin for 18 h, cut into 1-cm blocks, and dehydrated in an alcohol series. A segment of the RLL was flash frozen in liquid N₂. Trichrome staining was performed by standard techniques.

PECAM-1 and VEGF Immunohistochemistry

Immunohistochemistry was performed as described previously (10). For platelet endothelial cell adhesion molecule-1 (PECAM-1), the slides were incubated with goat anti-mouse PECAM-1 antibody (sc-1506, Santa Cruz). Nonimmune goat immunoglobulin G (IgG) was the negative control. Slides were incubated with biotinylated anti-goat IgG. Vectastain ABC-AP Elite Reagent and Vector Red substrate were then added. Slides were counterstained with methyl green.

For VEGF, tissues were digested with 0.1% trypsin and incubated with 1% H₂O₂ in methanol. The slides were incubated in rabbit anti-human VEGF (sc-152, Santa Cruz). Nonimmune IgG was the negative control. Slides were incubated with biotinylated anti-rabbit IgG, treated with streptavidin-horseradish peroxidase (HRP), and then with tyramide (New England Nuclear). The slides were incubated again with streptavidin-HRP, stained using 3,3'-diaminobenzidine, and counterstained with hematoxylin.

PECAM-1 Immunoblot

Western immunoblot was conducted as previously described (27). Briefly, proteins (10 µg) were separated by gel electrophoresis and transferred to polyvinylidene difluoride membrane. Membranes were incubated in either goat anti-PECAM-1 polyclonal IgG₁ (sc-1506; Santa Cruz Biotechnology) or rabbit antiactin polyclonal IgG₁ antibody (A2066; Sigma). The membranes were incubated in peroxidase-conjugated anti-goat or anti-rabbit secondary IgG (1:5,000; Santa Cruz Biotechnology). Immunodetection was by chemiluminescence (ECL Plus: Amersham Arlington Heights, IL).

RNase Protection Assay (RPA)

Total RNA was isolated using standard methods. VEGF, Flt-1, Ang-1, TIE-2, TIE-1, and PECAM-1 mRNAs were measured by RPA using a human angiogenesis multiprobe template set (hAngio-1; Riboquant, PHARMINGEN, San Diego). [-³²P]UTP-labeled probe was diluted to 3.6 × 10⁵ cpm/µl per hybridization. Probe was added to RNA samples (10 µg), denatured at 90° C, and incubated at 56° C for 12-16 h. After RNase treatment, protected fragments were separated by gel electrophoresis and the radioactivity was quantified by phosphorimaging. L-32, a ribosomal protein, was used as the control for quantification.

Northern Hybridization

Northern hybridization for KDR (VEGFR2) was conducted as described previously (10). The complementary DNA (cDNA), which was made by reverse transcriptase polymerase chain reaction (RT-PCR) of human RNA, is 591 bp and corresponds to bases 2837-3427 of the coding region. The northern hybridization was quantified by phosphorimaging using the ribosomal 18 S band for normalization.

In Situ Hybridization

The cDNA for VEGF is 94% homologous with human VEGF and hybridizes to all known VEGF mRNA splice variants (12). The human SP-C cDNA was obtained from Jeffery Whitsett (University of Cincinnati). The cDNAs for Ang-1 and TIE-2 were obtained by RT-PCR of human lung mRNA. In situ hybridization was performed as described previously (11).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Characteristics

The patients with no lung disease (NLD group) were significantly more mature at birth than the BPD group (Table 1; 39 ± 2.0 wk versus 27 ± 1.8 wk; p < 0.001). The BPD group had significantly longer exposure to the fraction of inspired oxygen (FI_O2) > 0.50 and mechanical ventilation. The chronologic age at death was significantly greater in the BPD group, but there was no difference in the postconceptional age at death (Table 2), suggesting that the potential lung developmental stages at death were similar. Autopsies were performed at 6.5 ± 4.5 hr (range 3-14 hr; p > 0.05) after death in both groups. Causes of death in the NLD group were congenital anomalies or hypoxic-ischemic encephalopathy. Of the infants with BPD, two were exposed to prenatal corticosteroids, all were treated with exogenous surfactant for respiratory distress syndrome (RDS), and four of five received postnatal corticosteroids for BPD. All patients with BPD received bronchodilators, diuretics, and antibiotics during their course.

Microvascular Changes in BPD

Immunohistochemistry for PECAM-1 (CD31) was used to identify endothelial cells. Infants with no lung disease had extensive staining in alveolar regions, as expected in this highly vascularized tissue (Figure 1A-1D). Vascular smooth muscle cells (Figure 1A, arrowhead) and airway epithelium (Figure 1A, arrow) did not stain. High power showed intertwining of the alveolar capillaries in the alveolar septa and the close proximity of the microvasculature to the airspaces (Figure 1B). In contrast, lung sections from infants with BPD showed overall decreased PECAM-1 immunostaining (Figure 1E-1H). The alveolar capillaries were often located in the interior of thickened septa. They appeared dilated and lacked extensive network organization. Minimal background was observed in tissue sections incubated with nonimmune goat IgG (not shown).

View larger version (158K): [in this window] [in a new window]   Figure 1.   PECAM-1 immunohistochemistry. Lung sections from patients with NLD (A-D) and BPD (E-H) were immunostained with a polyclonal anti-PECAM-1 antibody. Patients with NLD had extensive staining of cells in alveolar walls. Higher power (B) shows complex intertwining of alveolar capillaries that are in very close proximity to the alveolar space. BPD lung had thickened alveolar septa with dilated vessels frequently in the interior of the septa. The intensity of immunostaining was decreased in the BPD lungs. Each image is a different patient. Original magnification: ×100 for all except B, which is ×400.

Abnormal Capillaries and Increased Collagen in BPD Lungs

Trichrome staining of NLD lung (Figure 2A and 2B) showed thin alveolar septa and minimal staining for collagen in the alveolar walls. The red blood cells, which are probably in alveolar capillaries, are directly subjacent to the alveolar epithelium. In BPD lung (Figure 2C-2F), subepithelial alveolar capillaries were sparse; the dilated capillaries were frequently noted in the thickened alveolar septa. The alveoli were often lined with cuboidal epithelial cells (Figure 2D, arrow). Fine strands of collagen (stained light blue) or bundles of collagen in the thickened alveolar septa were noted in BPD lung. Clusters of elongated cells underlying the epithelium suggest fibroblasts within surrounding extracellular matrix.

View larger version (139K): [in this window] [in a new window]   Figure 2.   Trichrome staining of NLD and BPD lung. NLD lung had thin alveolar septa and minimal collagen staining (A, B). Red blood cells are noted in subepithelial capillaries (A, arrow). BPD lung (C-F ) had fine strands or thick bundles of collagen (stained light blue) in the thick alveolar septa. The alveolar epithelium had numerous cuboidal cells and minimal subepithelial capillaries. Capillaries were frequently in the thickened alveolar septa. Original magnification: ×300.

Decreased PECAM-1 Protein and mRNA in Patients with BPD

PECAM-1 protein in whole lung homogenates was evaluated by Western immunoblot (Figure 3A). In NLD patients abundant PECAM-1 was detected at the appropriate molecular weight (250 kD; Figure 3A, arrow). The patients with BPD had decreased PECAM-1 protein. Quantification of these data, expressed as a ratio of PECAM-1 to actin, showed a significant (p < 0.03) decrease in the relative abundance of PECAM-1 in the patients with BPD (Figure 3B).

View larger version (14K): [in this window] [in a new window]   Figure 3.   Decreased PECAM-1 protein and mRNA in BPD lung. (A) Western immunoblot of four patients with NLD and four patients with BPD using anti-PECAM-1 and antiactin antibodies (arrow indicates PECAM-1). (B) Quantification of immunoblot. Expression of PECAM-1 relative to actin shows the patients with BPD had significantly decreased PECAM-1 protein. Each point represents an individual patient and the horizontal bar is the group mean. (C ) RPA for PECAM-1 mRNA expressed relative to L32 mRNA shows a significant decrease in PECAM-1 mRNA in the patients with BPD. Each point is an individual patient. The horizontal line is the group mean. Means were compared by Student's t test.

PECAM-1 mRNA was measured in patients with NLD and BPD by RPA using a multiprobe human template set. Signal intensity was expressed relative to the housekeeping gene L32 (Figure 3C). The ratio of PECAM-1 mRNA to L32 mRNA was significantly decreased in the BPD group compared with the NLD group (p < 0.05). These data suggest a decrease in the relative number of endothelial cells in the patients with BPD.

Decreased mRNA for VEGF and Its Receptor Flt-1, and the TIE-2 Receptor in BPD

RPA for VEGF, Ang-1, Flt-1, and TIE-2 was used to measure these angiogenic factors and their receptors. The abundance of each message was expressed relative to L32. Patients dying with BPD had a decreased relative abundance of VEGF mRNA compared with the NLD group (p < 0.05, Figure 4A). Expression of Flt-1 mRNA (VEGFR1) was also significantly decreased in patients with BPD (Figure 4B). BPD lung had no change in the relative abundance of Ang-1 mRNA (Figure 4C), but had significantly decreased abundance of mRNA for TIE-2, the Ang-1 receptor (p < 0.02, Figure 4D). The relative abundance of KDR/murine homologue fetal liver kinase (Flk-1) (VEGFR2) mRNA, TIE-1 mRNA, an orphan receptor in the TIE family, and Flt-4 mRNA, a receptor for VEGF-C on lymphatic endothelial cells, was not altered (not shown).

View larger version (16K): [in this window] [in a new window]   Figure 4.   Quantification of RPAs for VEGF (A), its receptor Flt-1 (B), Ang-1 (C ), and its receptor TIE-2 (D) in patients with NLD and BPD. The data are expressed relative to L32 mRNA. The patients with BPD had significantly decreased mRNA for VEGF, Flt-1, and TIE-2. Each point is an individual patient and the horizontal bar is the group mean. Means were compared by Student's t test.

Changes in VEGF Immunostaining in BPD Lung

In patients with NLD, the alveolar regions were uniformly immunostained for VEGF (Figure 5A-5E). Vascular smooth muscle that was subjacent to the endothelium had staining, whereas more peripheral smooth muscle around the same artery was lightly stained (Figure 5A). Epithelial cells of conducting airways were also immunostained in control lung (not shown). BPD lung had an overall decrease in staining intensity (Figure 5F-5J), but staining intensity was heterogeneous in different parts of the lung. Lung areas with thickened alveolar septa and simple alveolar structures had light VEGF immunostaining. Other areas of BPD lung that had relatively normal architecture had more abundant VEGF. The cellular distribution of VEGF in BPD lung was similar to NLD lung, including in smooth muscle and airway epithelium. BPD lung had sparse immunostaining in the interior of the thickened septa but some detectable VEGF protein in alveolar epithelial cells. Macrophages frequently had intense immunostaining (Figure 5J). A semiquantitative assessment of VEGF immunostaining was obtained by grading staining intensity from 0 (background staining) to 4 (most intense staining of all samples) in the alveolar regions, smooth muscle, and airway epithelium. In the alveolar regions (Figure 6), VEGF immunostaining in the NLD lung was generally greater than in the BPD lung. Staining intensity of the smooth muscle and airway epithelium was not different in NLD and BPD samples (not shown).

View larger version (63K): [in this window] [in a new window]   Figure 5.   VEGF immunohistochemistry. Lung sections from each patients with NLD (A-E ) and BPD (F-J ) were immunostained with a monoclonal antibody to VEGF. The NLD lung had VEGF protein identified in alveolar septal cells and in vascular smooth muscle cells that were adjacent to the endothelium. In BPD lung, the overall staining intensity was decreased, particularly in areas with thickened alveolar walls. There was a paucity of VEGF noted in the interior of the thickened walls and some epithelial staining was noted. The cellular distribution of VEGF protein was similar to the NLD lung. Each image is a different patient. Original magnification: ×300.

View larger version (6K): [in this window] [in a new window]   Figure 6.   Semiquantitative assessment of VEGF immunostaining. Lung sections immunostained for VEGF were assigned a relative value from 0 (minimal staining similar to control antibody) to 4 (most intense immunostaining) in the alveolar regions, smooth muscle, and air space epithelium. In the alveolar regions, the BPD lungs (circles) had decreased immunostaining compared with the NLD lungs (squares). The staining intensity of smooth muscle and airway epithelium was similar in patients with BPD and NLD (not shown). Each point is a different patient.

Cell-specific Expression of VEGF, Ang-1, and TIE-2 mRNAs

We identified VEGF-expressing cells by in situ hybridization on lung tissue from each group. The patients with NLD had abundant VEGF-expressing cells (white grains) that were located mainly in the thin alveolar septa (Figure 7A-7D). Higher power suggested that VEGF expressing cells were likely alveolar epithelial cells that resembled type II cells (not shown). Little or no message was noted in airway epithelium or smooth muscle cells. The results from BPD lung were more variable (Figure 7F-7J). Overall, VEGF message was decreased, particularly in areas with thickened septa. Areas of lung with well preserved architecture had nearly normal VEGF message. The message was mainly in alveolar epithelial cells in the BPD lung with little message in the interior of thickened alveolar septa, smooth muscle cells, or airway epithelium. In contrast to VEGF immunostaining, little VEGF mRNA was detected in alveolar macrophages. Sense control showed minimal nonspecific grains (Figure 7E).

View larger version (92K): [in this window] [in a new window]   Figure 7.   In situ hybridization for VEGF mRNA. NLD lung (A-D) had scattered alveolar cells that expressed VEGF mRNA. BPD lung (F-J ) had decreased VEGF mRNA, particularly in lung regions that had thickened alveolar walls. Sense control (E ) showed minimal grains. Each image is a different patient. Original magnification: ×150.

Ang-1 mRNA was detected in scattered cells in the alveolar regions of patients with NLD (Figure 8A-8D). Little or no message was found in bronchial epithelial cells, smooth muscle cells, or endothelial cells of large vessels. In the patients with BPD, Ang-1 mRNA was in cells of the thickened alveolar septa (Figure 8E-8H). Unlike VEGF mRNA, which was mainly in alveolar epithelial cells, Ang-1 message was expressed largely by mesenchymal cells. This is the first report of lung cell-specific expression of this angiogenic factor.

View larger version (117K): [in this window] [in a new window]   Figure 8.   In situ hybridization for Ang-1 mRNA. In NLD lung (A-D), Ang-1 mRNA was detected in scattered cells in the alveolar regions. Little or no message was found in bronchial epithelial cells, smooth muscle cells, or endothelial cells of large vessels. In the BPD lung (E-H ), Ang-1 mRNA was in cells in the interior of the thickened alveolar septa. Original magnification: ×200.

Message for TIE-2, the Ang-1 receptor, in NLD lung was noted particularly in endothelial cells of large vessels (Figure 9A). Scattered grains in the alveolar septa probably represent TIE-2 mRNA in capillaries. Little or no signal was found in smooth muscle or airway epithelium. In BPD lung, grains were found in a similar distribution to NLD lung (Figure 9B). Many areas of BPD lung had decreased signal intensity, whereas it was similar to NLD lung in areas with normal morphology.

View larger version (141K): [in this window] [in a new window]   Figure 9.   In situ hybridization for TIE-2 mRNA. In NLD lung (A) mRNA for TIE-2 was noted particularly in endothelial cells of large vessels (arrow). The grains in the alveolar septa are probably in capillaries. In BPD lung (B), there was a decrease in the overall signal, particularly in areas with thickened alveolar septa. Original magnification: ×150.

Increased Cells That Express SP-C mRNA in BPD Lung

In previous studies in vivo and in vitro we found that type II cells that express VEGF have a relatively low abundance of SP-C mRNA (12). A possible mechanism of decreased VEGF message in BPD may be an increase in the population of type II cells that express abundant SP-C. In situ hybridization for SP-C message in NLD lung showed scattered SP-C expressing cells in the alveolar regions (Figure 10A-10D). In contrast, the BPD lung showed greatly increased expression of SP-C mRNA (Figure 10E-10H). Cells expressing SP-C frequently lined the entire alveolar epithelium in ring-like structures, particularly in alveoli that were close to airways or large vessels.

View larger version (135K): [in this window] [in a new window]   Figure 10.   In situ hybridization for SP-C. In NLD lung (A-D) SP-C mRNA was found only in scattered cells of the alveolar walls. BPD lung (E-H ) had an overall increase in SP-C expressing cells that were limited to alveolar epithelial cells. In some alveolar structures in BPD lung, the majority of epithelial cells had SP-C mRNA. Original magnification: ×200.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Abnormal lung development is a hallmark of chronic lung disease of premature infants (3). Our study, which focused on the alveolar vasculature and angiogenic factors in human infants who died with BPD, found that lung vascular development was disrupted in BPD, possibly due to impaired expression of angiogenic factors or their receptors.

Patients with BPD had an abnormal pulmonary alveolar vasculature, including dilated capillaries and abnormal placement in the alveolar septa. Dysmorphic capillaries in thickened alveolar septa suggest injury or developmental arrest of lung vascularization. Our findings of decreased PECAM-1 mRNA and protein are consistent with a decreased number of endothelial cells in BPD. Furthermore, we found decreased expression of messages for the angiogenic growth factor VEGF and its receptor Flt-1, and for TIE-2, the receptor for Ang-1. Increased SP-C expression by alveolar epithelial cells suggests a shift in alveolar epithelial cell phenotype that may result in decreased VEGF expression.

This study compared premature infants who died with BPD with term infants who died from nonlung causes and who had brief exposures to supplemental oxygen and mechanical ventilation. A better comparison group would have been premature infants without BPD who died from nonlung causes at a postconceptional age similar to the BPD group. Such patients are very rare and none was accumulated during 4 yr of the study. Our comparison of preterm infants with BPD with term NLD infants is justified by the similarity in their postconceptional ages at death. Thus, the NLD group represents the potential stage of lung development that the BPD infants may have achieved if they did not develop BPD.

Most studies of BPD have described changes in pulmonary arteries and arterioles, but did not address pulmonary microvascular changes or angiogenic growth factors. Those findings included periarteriolar thickening, degeneration of elastica, and medial muscular hypertrophy (23, 24). Tomashefski and coworkers found increased medial thickness of muscular preacinar and intraacinar pulmonary arteries and an increased number of intraacinar arteries in eight patients with BPD (25). A postmortem study of infants who were not treated with exogenous surfactant showed endothelial edema, mild medial thickening, and increased elastic tissue (26). A recent autopsy study of BPD showed abnormalities in alveolar capillaries similar to our findings (2). In experimental BPD in extremely immature baboons, Coalson and coworkers found decreased capillaries and dysmorphic capillary morphology (7). Newborn mice exposed to 80% oxygen had decreased lung capillary surface area (28), consistent with toxic effects of hyperoxia on endothelial cells. Our finding of decreased PECAM-1 immunostaining intensity and decreased PECAM-1 protein and mRNA suggests decreased endothelial cells in BPD.

A potential mechanism of alveolar capillary dysmorphogenesis in BPD may be decreased expression of angiogenic growth factors or their receptors. VEGF, which has important roles in angiogenesis and vasculogenesis, is expressed early in organ development, including lung development (29, 30). Abnormal blood vessels in heterozygous VEGF-deficient (VEGF+/) embryos, resulting in death at mid-gestation, underscore an essential role for VEGF in embryonic vessel development (9). Antibody inhibition of VEGF in developing mouse kidney resulted in abnormal glomeruli that lacked capillary tufts (31), whereas VEGF injection in embryonic chick stimulated myocardial vascularization (32). Overexpression of VEGF in fetal murine lung enhanced pulmonary vasculogenesis but also resulted in abnormal alveolar development (33).

We found decreased VEGF protein and mRNA in BPD lung, particularly in areas with thickened alveolar septa. The cell type expressing VEGF mRNA was not changed. These data support the hypothesis that decreased VEGF may have a role in developmental arrest of pulmonary microvasculature in BPD. Studies of human tracheal aspirates also found decreased VEGF in patients with BPD (34). In previous studies, we and others found decreased lung VEGF in acute hyperoxia (11, 17). Decreased VEGF in BPD lung has implications for endothelial cell survival because both hyperoxia and inflammatory cytokines may be toxic to endothelial cells. Intraocular injection of VEGF during hyperoxia preserved the retinal vasculature (13). Similarly, VEGF inhibited tumor necrosis factor- (TNF-) induced apoptosis in endothelial cells (16). These data suggest that VEGF is a survival factor for endothelial cells. Decreased lung VEGF in BPD may result in endothelial cell death or apoptosis induced by hyperoxia or inflammatory agents.

We noted variations in VEGF immunostaining and in situ hybridization. Although normal alveolar septal cells have intense VEGF immunostaining and abundant mRNA, smooth muscle, airway epithelium, and alveolar macrophages have little message but detectable protein. It is likely that alveolar macrophages take up VEGF, which is shed into the airspace, accounting for the intense immunostaining and relatively little message. Other cell types, such as smooth muscle, may have VEGF mRNA that is too low to detect by in situ hybridization but accumulate cell-bound isoforms of VEGF protein that are detected on immunostaining. Supporting this speculation, we found previously that lung has a relatively high proportion of VEGF₁₈₉, a cell-bound isoform (35). Our semiquantitative assessment of VEGF immunostaining suggests that distal airspace cells, but not airway epithelium or smooth muscle, had altered VEGF in BPD.

BPD lung had decreased messages for Flt-1 (VEGFR1) and for TIE-2, the Ang-1 receptor, but not KDR/Flk-1 (VEGFR2). Mice with targeted disruption of TIE-2 die by embryonic Day 9.5 to 10.5, exhibiting growth retardation and malformations in the vascular network consisting of a dilated vasculature with limited branching and capillary sprouting (36). A similar phenotype was described in mice lacking Ang-1 (21). Defects in vessel architecture in TIE-2-deficient mice suggest roles in stabilizing vascular structures. The decrease in Flt-1 and TIE-2 mRNA that we found may be due to decreased capillary endothelial cells in BPD. Alternatively, because VEGF increases TIE-2 expression, low VEGF may have contributed to decreased TIE-2 in BPD. Other endothelial-specific tyrosine kinase receptors, TIE-1 and KDR/Flk-1, were not changed in BPD, suggesting heterogeneity in regulation of members of these receptor families.

A potential mechanism of decreased VEGF expression in BPD is a change in the cell phenotype of alveolar epithelial cells as a result of chronic epithelial injury and repair. In previous work, we found that type II cells in vitro and in vivo that express high levels of SP-C have low levels of VEGF (12). In this study we found that BPD lung had increased numbers of type II cells that express SP-C. VEGF-expressing type II cells may have been displaced by high SP-C-expressing cells as part of the injury-repair process. Confirmation of this speculation awaits further investigation.

Our study supports the importance of studying fresh postmortem human tissue to understand human disease. We collected lungs at an average of 6.5 h after death. The importance of using relatively fresh tissue was demonstrated by our inability to detect VEGF mRNA by in situ hybridization on archived autopsy specimens from human newborns (not shown). Problems in studying fresh autopsy tissue include a relatively low mortality from BPD, a declining autopsy rate, difficulty in coordinating lung inflation fixation at off hours, and the potential of RNA and protein degradation. In 5 yr we collected five suitable patients with BPD and five with NLD from an NICU with over 1,100 admissions annually. In addition to the small sample size, our study has several important limitations, particularly the effects of complications or therapies on microvascular development or angiogenic factors. For example, four of five patients in the BPD group were treated postnatally with dexamethasone, which may affect VEGF expression (10, 37, 38). We do not know the effects of premature birth, infection, antibiotics, or nutrition on vascular development or expression of angiogenic factors and we cannot assign a precise cause for our results. However, our findings are the first to document changes in the microvasculature and angiogenic factor expression in human BPD.

In conclusion, infants dying with BPD have abnormal alveolar microvessels that are consistent with disrupted alveolar vascular development. These abnormalities may result from impaired expression of VEGF and angiogenic endothelial receptors.