© 2003 American Thoracic Society DOI: 10.1165/rcmb.2003-0159SU
Idiopathic Pulmonary FibrosisCONTENTS Introduction. Augustine M.K. Choi S3 To the Pittsburgh International Lung Conference with Love. Thomas L. Petty S4 Histologic Classification of Idiopathic Chronic Interstitial Pneumonias. Sanja Dacic and Samuel A. Yousem S5 Imaging of the Idiopathic Interstitial Lung Diseases: Concepts and Conundrums. Diane C. Strollo S10 The Prognosis of Idiopathic Pulmonary Fibrosis. Andrew Perez, Robert M. Rogers, and James H. Dauber S19 Idiopathic Pulmonary Fibrosis: New Insights into Classification and Pathogenesis Usher in a New Era in Therapeutic Approaches. Paul W. Noble S27 Microarray Analysis of Idiopathic Pulmonary Fibrosis. Naftali Kaminski S32 Pulmonary Fibrosis of Sarcoidosis: New Approaches, Old Ideas. David R. Moller S37 Proteomic and Inducible Transgenic Approaches to Study Disease Processes. Prabir Ray, Li Chen, Vladimir A. Tyurin, Valerian E. Kagan, and Frank A. Witzmann S42 Pulmonary Fibrosis in Families. James E. Loyd S47 Gene Profiling and Kinase Screening in Asbestos-Exposed Epithelial Cells and Lungs. Maria E. Ramos-Nino, Nicolas Heintz, Luca Scapoli, Marcella Martinelli, Susan Land, Norma Nowak, Astrid Haegens, Brian Manning, Nicole Manning, Maximilian MacPherson, Maria Stern, and Brooke Mossmann S51 The Importance of Sarcoidosis Genotype to Lung Phenotype. Jan C. Grutters, Hiroe Sato, Kenneth I. Welsh, and Roland M. du Bois S59 Cytokine Phenotypes Serve as a Paradigm for Experimental Immune-Mediated Lung Diseases and Remodeling. Steven L. Kunkel, Stephen W. Chensue, Nicholas Lukacs, and Cory Hogaboam S63 CXC Chemokines in Vascular Remodeling Related to Pulmonary Fibrosis. Robert M. Strieter, John A. Belperio, and Michael P. Keane S67 Possible Roles for Apoptosis and Apoptotic Cell Recognition in Inflammation and Fibrosis. Peter M. Henson S70 Regulation of Receptor for Advanced Glycation End Products during Belomycin-Induced Lung Injury. Lana E. Hanford, Cheryl L. Fattman, Lisa M. Schaefer, Jan J. Enghild, Zuzana Valnickova, and Tim D. Oury S77 The Role of Heme Oxygenase-1 in Pulmonary Fibrosis. Danielle Morse S82 Fibroblast Phenotypes in Pulmonary Fibrosis. Sem H. Phan S87 The Epithelial/Fibroblastic Pathway in the Pathogenesis of Idiopathic Pulmonary Fibrosis: Tying Loose Ends. Moisés Selman and Annie Pardo S93 Molecular Targets for Drug Discovery in Idiopathic Pulmonary Fibrosis: Work in Progress. Peter B. Bitterman S98 Pittsburgh International Lung Conference at Nemacolin: Summary. Robert M. Strieter S102
IntroductionThe discipline of respiratory medicine is experiencing an era of unprecedented advance in our understanding of the fundamental basis of human lung disease. Exponential growth in basic lung biology, fueled by the elucidation of the human genome sequence, is successfully coupled with unique human disease applications of this knowledge. This fusion of basic genetics and clinical medicine promises to change our understanding of lung disease diagnosis and treatment. With this promise come major challenges. Our brightest investigators are faced with significant pressures related to academic and clinical productivity, financial stability, and competition. In this backdrop, we have created the Pittsburgh International Lung conference. Our goal is to establish a forum for junior and senior lung investigators to isolate themselves, focus on "state of the art" investigation in a selected area of lung disease, and foster collaborative, collegial, and productive interactions.In proposing this conference, we have turned to the gold standard. For over 40 years, the Aspen Lung Conference has served, with great distinction, a unique role for lung investigators. Inaugurated in 1958 as a conference to understand emphysema and chronic bronchitis, this annual meeting has evolved as a premier forum for basic scientists, physician investigators, and clinicians, to consider progress in the full range of lung diseases. Conference topics have included chronic obstructive and interstitial lung disorders, the acute respiratory distress syndrome (ARDS), genetically determined lung diseases such as cystic fibrosis, and environmental lung disease. The Aspen conference provides a unique level of focus, scientific leadership, and collegiality not traditionally available at other major pulmonary conferences. We believe the dramatic growth in lung investigation warrants a second annual conference with aligned goals and spirit. We are truly indebted to Drs. Thomas Petty and Marvin Schwarz for their guidance and direction in the development of the Pittsburgh International Lung Conference. The Pittsburgh Conference planning committee will maintain a close dialog with the Aspen Conference organizers to avoid redundancy and promote synergy of the selected topics for these two lung conferences on an annual basis. The topic chosen for the inaugural year of the Pittsburgh International Lung Conference was Idiopathic Pulmonary Fibrosis. Few lung disorders have seen a renewed investigative focus like IPF. A historical paradigm of lung inflammation leading to fibrosis is being rapidly revised, incorporating an expanding knowledge base in the topic areas of genetics, lung fibrosis, injury, and repair. National clinical trials, often considered impossible in this disorder, now rapidly explore promising yet unproven therapies. Advancing techniques in lung imaging and noninvasive assessment provide clinicians exciting new tools to diagnose and monitor disease progression. We were honored to assemble an international group of leading investigators to focus on the current and future state of our knowledge in IPF. Our motivation for the topic of IPF is also personal. In 2002, the Simmons family of Pittsburgh provided financial support to the University of Pittsburgh to establish the Dorothy P. and Richard P. Simmons Center for Interstitial Lung Disease. This family grew to appreciate the terrible difficulties of IPF from Dorothy's viewpoint during her entirely too brief course with this disease. Recognizing the existing limitations in our treatments, the family has focused their resources on the promotion of investigation and education in IPF. The initial event of the Pittsburgh International Lung Conference this year was the appointment of Dr. Naftali Kaminski to the Simmons Chair for ILD at the University Of Pittsburgh School Of Medicine. Each year we hope to provide a summary of the conference proceedings in the AJRCMB. The presenters for this year's conference were truly outstanding, and we are indebted to their commitment and collegiality in this inaugural year. We were inspired by your scientific creativity, and motivated by your vibrant and selfless interactions. AUGUSTINE M. K. CHOI, M.D. Division of Pulmonary, Allergy, and Critical Care Medicine University of Pittsburgh School of Medicine Pittsburgh, Pennsylvania
To the Pittsburgh International Lung Conference with LoveAn important new International Lung Conference was inaugurated in October by Professor Augustine Choi (Head of the Division of Pulmonary, Allergy and Critical Care Medicine) and his colleagues at the University of Pittsburgh Medical Center. Its goal is to provide a research forum to focus on the challenging problem of idiopathic pulmonary fibrosis (IPF)-related disorders.A major driving stimulus for this new program was the generosity of Mr. Richard Simmons, who lost his dear wife, Dorothy, to IPF recently. It was my deep honor to be asked by Augustine to offer some remarks about the Aspen Lung Conferences, which began as a series of emphysema conferences in 1958. At that time, the goal was to begin to understand emphysema and chronic bronchitis. Considerable progress has been made since then. The Aspen Lung Conferences, as they were later named, evolved as a forum with the purpose of bringing together basic and applied scientists, as well as clinicians, to consider progress in COPD, asthma, and later the acute respiratory distress syndrome (ARDS). Also, interstitial lung diseases, genetically determined lung diseases such as cystic fibrosis, and less focused topics, such as the environment and the lung, were tackled in the resort atmosphere of Aspen, Colorado. The Aspen conferences were successful, and offered something unique and beyond what was available at major pulmonary conferences in North America and Europe. The newly inaugurated Pittsburgh Conference offers to do similar things that will help bridge the gap between known clinical challenges in lung disease, and ultimately help to provide solutions to diseases such as IPF. This is how we may emerge from a bewildering wilderness. Only by understanding genetically determined and other risk factors, and the impact of environmental exposures that conspire to inflict both acute and chronic lung injuries resulting in progressive fibrosis, will we make significant progress. We must discover the basic molecular and biochemical process involved in IPF as the foundation for developing new therapeutic targets. Armed with much-needed facts about mechanisms of lung damage and destruction, the pharmaceutical industry will be able to design and develop new pharmacologic agents that will prevent or forestall the progress of IPF. Thus, new progress can be made and solutions found. On behalf of the Aspen Lung Conference Steering Committee and as its official historian, I welcome Pittsburgh International Lung Conference into the arena of struggle and discovery. "Fac –et spera" means work and hope. It is with this spirit and my personal optimism that I wish this new conference God speed. THOMAS L. PETTY, M.D. University of Colorado Medical Center and Rush-Presbyterian Medical Center
Histologic Classification of Idiopathic Chronic Interstitial PneumoniasHistorical PerspectivesThe diagnosis and management of idiopathic interstitial pneumonia (IIP) have challenged physicians since their description more than a century ago. Significant progress in the understanding of interstitial lung diseases was made in the mid-1960s with recognition of collagen vascular diseases (CVD), drugs, and occupational exposures as potential causes. However, a large group of entities still remained idiopathic, and in 1968 Liebow and Carrington were first to classify chronic IIPs into the following five histopathologic subgroups: usual interstitial pneumonia (UIP), bronchiolitis interstitial pneumonia (BIP), desquamative interstitial pneumonia (DIP), giant cell interstitial pneumonia (GIP), and lymphoid interstitial pneumonia (LIP) (1). In the ensuing twenty years, new entities were described and the original IIPs studied in greater depth. The results were then codified in the classification schema described in 1998 by Katzenstein and Myers (2). Their classification scheme recognized five entities: usual interstitial pneumonia (UIP), desquamative interstitial pneumonia (DIP), respiratory bronchiolitis–associated interstitial lung disease (RB-ILD), nonspecific interstitial pneumonia (NSIP), and acute interstitial lung disease (AIP) (former Hamman-Rich syndrome). The most recent classification by the American Thoracic Society and European Respiratory Society (ATS/ERS) emphasizes the importance of an integrated clinical, radiologic, and pathologic approach to the diagnosis of IIP (3). In particular, it is vitally important that biopsy findings are correlated with high-resolution computed tomography (HRCT), as heterogeneity of lung injury patterns are common in IIP (4–6). This classification expands on the histopathologic terms defined by Katzenstein and Myers, but more precisely defines the relationship between clinico-radiographic findings and histopathology. Fundamental Rules for Pathologic Classification Although the clinical and radiographic diagnosis of IIP can be made in some cases, many patients still require open lung biopsy to determine their underlying histopathologic pattern. Pathologic classification is a very dynamic process requiring close clinico-radiographic correlation. For most practicing pulmonary pathologists, the diagnosis of chronic interstitial lung disease is made at low magnification. Several questions relating to the histologic review of the lung biopsies need to be answered to make a correct diagnosis (Figure 1) .
The first question is whether the disease process is diffuse or patchy. The process is patchy if there are alternating zones of normal and inflammatory/fibrosing lung parenchyma. In contrast, if the entire pulmonary parenchyma appears affected by the inflammatory process and there is very little or no normal lung parenchyma associated with the disease, the process is classified as diffuse. The second important issue is to identify the primary anatomic sites of the lobule/acinus affected by the inflammatory or fibrosing process. The anatomic locations affected by the common chronic inflammatory lung diseases are summarized in the Table 1. Subpleural or paraseptal distribution reflects injury in the distal portion of the lobule and acinus, and is defined by the extension of the inflammation and fibrosis from the subpleural region centripetally into the pulmonary parenchyma. With a bronchiolocentric distribution, the periphery of the pulmonary lobule is relatively spared and the inflammatory process is primarily localized to the bronchovascular bundle with extension into the contiguous peribronchiolar alveolar septa. Alveolar septal distribution is defined by thickened alveolar septa, either by inflammation or fibrosis, throughout the lobule. The process is lymphangitic if the inflammation tracks along the visceral pleura, interlobular septa, and bronchovascular bundles with relative sparing of the septa.
The third basic concept in the understanding of interstitial lung disease is the concept of temporal homogeneity and temporal heterogeneity. Temporal homogeneity indicates that the age of lung injury is approximately the same (acute, subacute, or chronic), and there are no mixtures of all three injury patterns in the same biopsy. In contrast, in temporally heterogenous lung injury one can identify areas of honeycombing (chronic), interstitial or air space fibromyxoid connective tissue (subacute), and alveolar epithelial cell necrosis and hyaline membranes (acute). Finally, one needs to define the overall phase of interstitial injury that may play an important role in predicting responsiveness to therapy: acute, interstitial edema with alveolar pneumocyte necrosis, fibrin, and hyaline membranes; subacute, airspace or interstitial loose fibromyxoid granulation tissue; or scar, remodeled or densely fibrotic lung where architecture is often destroyed, remodeled, or thickened by dense eosinophilic collagen. Once the above described features are identified, histologic classification of IIP should be relatively straightforward in most cases, particularly when correlated with radiographic and clinical findings (Table 2). It is important to emphasize that subclassification of IIP requires exclusion of known causes of these patterns of injury. This mandates close communication between clinicians, radiologists, and pathologists, and could result in reclassification of an interstitial pneumonia if additional information becomes available.
Usual Interstitial Pneumonia/Idiopathic Pulmonary Fibrosis UIP is characterized by patchy subpleural and paraseptal distribution of parenchymal injury. Temporal heterogeneity is seen at low magnification, with alternating areas of normal lung parenchyma, interstitial mononuclear infiltrates, septal fibromyxoid tissue (fibroblastic foci), and honeycomb lung (Figure 2) . Secondary changes, such as pulmonary hypertension and mucous plugs, are frequently present (2, 7, 8).
Open lung biopsies may occasionally show a combination of UIP pattern and subacute (cryptogenic organizing pneumonitis [COP]) or acute diffuse alveolar damage (DAD) lesions (9, 10). If no underlying cause can be determined for such presentation, this histology reflects an accelerated phase or acute exacerbation of UIP (6, 7). Occasionally, moderate number of interstitial or airspace eosinophils may be seen, but they are typically focal, and eosinophilic pneumonia can be excluded (11). It is important to remember that the pattern of interstitial inflammation and fibrosis in patients with CVD, drug-induced interstitial disease, chronic hypersensitivity pneumonitis or asbestosis can be histologically indistinguishable from the idiopathic pulmonary fibrosis (IPF), and the clinico-radiologic–pathologic correlation is essential in such instances. Nonspecific Interstitial Pneumonia NSIP is an idiopathic interstitial pneumonia that does not meet the diagnostic criteria for UIP, DIP, RB-ILD, AIP, or COP. The lung injury is typically diffuse, but may be patchy, and has an alveolar septal pattern. It is characterized by temporally homogenous mild to moderate interstitial mononuclear inflammation (cellular pattern) with dense interstitial fibrosis (fibrosing pattern) (Figure 3) . Some cases may show mixed cellular and fibrosing pattern (12–15). Lymphoid aggregates are common. Fibroblastic foci and honeycombing are absent or inconspicuous.
Histologic patterns of NSIP can be associated with CVD, hypersensitivity pneumonitis, drug reactions, and infections including HIV, and those clinical conditions should be clinically excluded. Desquamative Interstitial Pneumonia DIP is characterized by diffuse, temporally homogenous alveolar septal inflammation and fibrosis with uniform airspace filling by smokers' macrophages (16). The alveolar septa are lined by reactive pneumocytes and are thickened by mononuclear infiltrate and mild increase in septal collagen (Figure 4) . It has the appearance of NSIP with all airspaces filled with alveolar macrophages, often of the smokers type.
Because many patients with other IIP are often current or former smokers, DIP-like pattern can be seen focally in UIP, RB-ILD, NSIP, eosinophilic pneumonia, chronic hemorrhage, and veno-occlusive disease. Respiratory Bronchiolitis-Associated Interstitial Lung Disease The histologic changes of RB-ILD are patchy and bronchiolocentric in distribution. It is characterized by a temporally homogenous peribronchiolar mononuclear infiltrate with rare eosinophils, inconspicuous septal mononuclear cells, and irregular, centrilobular airspace filling by finely pigmented macrophages (Figure 5) . Mild peribronchiolar fibrosis is also seen (17–20).
DIP and RB-ILD, once considered distinct entities, are related lesions which differ only in the severity, distribution, and extent of the histopathologic abnormality. Cryptogenic Organizing Pneumonia COP is a patchy bronchiolocentric temporally homogenous process characterized by fibromyxoid connective tissue plugs in lumens of airways and airspaces (Figure 6) . There is a mild peribronchiolar and interstitial mononuclear inflammatory infiltrate. The lung architecture is relatively preserved (21–29).
The presence of airspace neutrophils, acute bronchiolitis, granulomas, necrosis, hyaline membranes, and prominent eosinophilic inflammatory infiltrate strongly argues against the diagnosis of primary COP. Lymphoid Interstitial Pneumonia LIP is characterized by a dense diffuse temporally homogenous lymphoid infiltration predominantly alveolar septal in distribution (Figure 7) . The lymphoid infiltrate is comprised mostly of T lymphocytes, plasma cells, and macrophages. Some architectural distortion, including honeycombing, nonnecrotizing granulomas, and small foci of organizing pneumonia, may be present. Lymphoid hyperplasia (MALT hyperplasia) is a frequently associated finding (30–34).
The major differential diagnosis from a clinical standpoint is the separation of LIP from low-grade lymphoma, particularly extranodal marginal zone B-cell lymphoma of MALT. Malignant lymphoma usually show a monomorphous lymphoid infiltrate distributed along lymphatic routes, often associated with destruction of alveolar architecture, Dutcher bodies, and pleural infiltration. Immunohistochemical and molecular gene rearrangement studies may be necessary to exclude lymphoproliferative disorder (3). LIP also must be differentiated histologically from follicular bronchiolitis, nodular lymphoid hyperplasia, infection (especially Pneumocystis carinii pneumonia), and other interstitial lung disorders such as NSIP, organizing pneumonia, and UIP. The cited references provide a very detailed description of lymphoid hyperplasia of the lung that is beyond the scope of this very brief summary. Role of Surgical Lung Biopsy ATS/ERS recently published a consensus statement describing major and minor criteria for the clinical diagnosis of IPF. The panel noted that in the absence of surgical lung biopsy findings, the diagnosis of IPF remains unproven, and that a definitive diagnosis of IIP can be established only with the aid of a surgical lung biopsy (3). In addition, the role of HRCT as an integral part of the evaluation of the patient with suspected IIP has been emphasized. The primary role of HRCT is to separate patients with UIP from those with other IIP such as NSIP, RB-ILD, DIP etc. HRCT may also be helpful in identifying patients with other diseases such as sarcoidosis, lymphangioleiomyomatosis, eosinophilic granuloma and hypersensitivity pneumonitis (3). The role of transbronchial biopsy in the diagnosis of IIP in most cases is to exclude sarcoidosis, lymphangitic carcinoma, infections, DAD, and some rare conditions such as alveolar proteinosis, lymphangioleiomyomatosis, and Langerhans' cell histiocytosis (35–41). Most pulmonary pathologists would agree that the assessment of IIP requires a surgical (open or thoracoscopic) lung biopsy. It is important for the surgeon not to biopsy the radiologically or grossly palpable "worst" areas. This is often nondiagnostic and most times shows nonspecific end-stage honeycomb lung. The open lung biopsy should be taken from more than one lobe of the lung. It is still controversial whether to biopsy lingula and right middle lobe, as both of these sites frequently show nonspecific fibrosis (39–44). The biopsy should be large in the size, and in our experience at least 5 cm in greatest dimension. It should be obtained at the edge of the grossly abnormal areas of the lung to include grossly normal lung parenchyma. Most important is that the biopsy must be deep, extending well into the "medulla" of the subpleural lung parenchyma. Shallow subpleural biopsies are frequently nondiagnostic. This allows one to escape nonspecific subpleural scarring and obtain actively injured lung parenchyma to assess the features that are important in making a diagnosis of IIP. Footnotes This section was written by Sanja Dacic and Samuel A. Yousem (Department of Pathology, Division of Anatomic Pathology, University of Pittsburgh Medical Center, Presbyterian University Hospital, Pittsburgh, Pennsylvania). References
Imaging of the Idiopathic Interstitial Lung DiseasesConcepts and ConundrumsThe interstitial lung diseases (ILD) are a diverse group of inflammatory-fibrosing disorders that affect predominantly the pulmonary interstitium rather than the airspaces. The etiology of most ILD is known, such as collagen vascular diseases, drug-induced lung diseases, infectious and noninfectious granulomatous diseases, inhalation of organic and inorganic materials and noxious gases, and proliferative and malignant neoplastic processes (1). The diagnosis of ILD, in some instances, may be established from clinical, laboratory, and radiologic data, without lung biopsy (2). The insult that produces idiopathic ILD remains unknown, and the disease is limited to the lungs and is typically immune mediated. Despite the variety of insults that may induce the initial injury of ILD, the pathogenesis and end-stage morphologic changes may be similar, although the rate of disease progression may vary considerably (1).The current clinico-pathologic classification of idiopathic nongranulomatous ILD includes a disparate group of lung diseases that have at least some degree of interstitial cellular inflammation and may culminate in pulmonary fibrosis (3) (Table 1). This grouping of diseases includes idiopathic pulmonary fibrosis (IPF), nonspecific interstitial pneumonitis (NSIP), respiratory bronchiolitis–interstitial lung disease (RB-ILD), desquamative interstitial pneumonitis (DIP), cryptogenic organizing pneumonia (COP), and acute interstitial pneumonitis (AIP). The etiology is not always idiopathic. Whereas IPF and NSIP are predominantly diseases of the interstitium, RB-ILD and DIP represent a continuum of smoking-related diseases of the small airways, interstitium, and alveoli. COP is an idiopathic inflammation of the small airways and airspaces with minor involvement of the interstitium. AIP is an idiopathic form of diffuse alveolar damage and involves the alveoli and the interstitium. NSIP is a relatively new histologic category, and like DIP, may not represent a distinct clinicopathologic syndrome (4). In some instances, when an ILD may not be easily or concisely characterized, it may be labeled as "unclassifiable ILD."
Imaging of Idiopathic ILD The primary role of imaging is to identify the presence and extent of pulmonary fibrosis, as this portends a less favorable prognosis regardless of etiology. In addition, certain patterns of lung involvement may suggest a specific disease category (5). Diseases that involve the interstitium (IPF, NSIP) may result in pulmonary fibrosis and manifest with reticulations (innumerable interlacing linear opacities typically due to intralobular interstitial thickening), "honeycomb" change (cystic dilatation of distal bronchioles and airspaces that share common thickened walls), and traction bronchiectasis or bronchiolectasis (irregular dilatation of bronchi and bronchioles, typically associated with reticulations or honeycomb cysts) (6). Processes that affect the airspaces (DIP, COP, AIP) may exhibit consolidation, defined as a homogeneous increase of lung attenuation that obscures the margins of vessels and airways. Bronchiolocentric diseases (RB-ILD, COP) may exhibit dilated and/or thickened bronchioles. Any process with active inflammation or fibrosis may exhibit "ground glass" attenuation, defined as hazy increased lung attenuation that does not obscure or distort the underlying lung architecture. Ground glass attenuation may precede the development of pulmonary fibrosis (7, 8). When pulmonary fibrosis is the predominant pattern, associated ground glass attenuation typically reflects microscopic changes of fibrosis (9). Computed tomography (CT) and high-resolution CT (HRCT) are the mainstays of the noninvasive evaluation of ILD and play a critical role in its early detection, characterization, and differentiation from other lung diseases. A normal HRCT does not always exclude early and clinically significant ILD, especially when physiologic testing is abnormal (10). The radiologic findings of pulmonary fibrosis on HRCT correlate strongly with fibrosis on histology (P = 0.0001), and pure ground-glass attenuation in patients with suspected ILD correlates well with interstitial inflammation (P = 0.03) (5, 11). CT may be used to select an optimal site of lung biopsy and to exclude patients with severe end-stage fibrosis who may not benefit from biopsy (5, 12). Radiologic and pathologic features of idiopathic ILD may be identical to those of ILD of known etiology (13).
Idiopathic Pulmonary Fibrosis
On CT, IPF typically exhibits features of pulmonary fibrosis with little active inflammation (Figure 4) . Honeycomb cysts (96%), distorted intralobular reticulations (80%), and traction bronchiolectasis (50%) have a striking predilection for the lung periphery and bases, and may involve all lobes in advanced disease (14, 19). Ground-glass attenuation (75%) is typically admixed with fibrosis and is rarely the dominant pattern (Figures 5 and 6) . Mildly enlarged reactive mediastinal lymph nodes are common, and correlate with greater disease severity (20, 21). Following single lung transplantation of patients with IPF, the native lung may exhibit progressive fibrosis (22).
It has been reported that  10% of patients with IPF, typically older male smokers, may develop lung cancer (23). Squamous cell carcinoma is the most common histologic type. Lung cancer typically arises within areas of fibrosis and manifests as a peripheral pulmonary nodule or mass or as a subtle, poorly defined area of consolidation or asymmetric fibrosis, which may make early detection difficult (23, 24) (Figure 7)
.
Nonspecific Interstitial Pneumonitis NSIP has varying amounts of subacute interstitial inflammation (cellular NSIP) and fibrosis (fibrotic NSIP) that do not meet the histologic criteria of UIP or other ILD (25–27). The histopathologic pattern of NSIP exhibits temporal homogeneity that results from a single lung insult, with all areas of reparation at the same stage of healing. Compared with IPF, patients with NSIP pattern tend to be younger and have milder symptoms of shorter duration, with a more favorable prognosis. The radiologic features of NSIP reflect the varied amounts of inflammation and fibrosis. NSIP has a predilection for the basilar and peripheral portions of the lungs, similar to IPF, but tends to have a greater component of inflammation and potentially reversible disease (14). Chest radiography may be normal, but characteristically reflects decreased lung volumes (70%) and symmetric bibasilar peripheral ground glass attenuation, consolidation, and/or fibrosis (28) (Figure 8) .
CT features of NSIP may have significant overlap with those of IPF/UIP, DIP, and COP, and cases of NSIP are commonly misdiagnosed on CT (25, 29). Cellular NSIP typically exhibits prominent ground glass attenuation (70–100%). Fibrotic NSIP may show fine reticulations, thickened septal and pleural lines, traction bronchiolectasis, and honeycomb cysts, which may be identical to those of IPF (25, 29–31). Bronchocentric consolidation, similar to that of COP, is occasionally detected (32). Radiologic and clinical features of NSIP may improve following corticosteroid therapy (33).
Respiratory Bronchiolitis–Interstitial Lung Disease and Desquamative Interstitial Pneumonitis Imaging features of RB and RB-ILD typically overlap, and may be normal or show subtle areas of ground glass attenuation, fine linear reticulations, and/or emphysematous changes (Figure 9) . CT may also reveal small airways disease with mildly thickened and dilated bronchioles, "soft" centrilobular nodules, and ground glass attenuation that may be centrilobular or diffuse (Figures 10 and 11) . These abnormalities are typically greatest in the upper aspects of the lungs and may improve or resolve following smoking cessation (35). Honeycomb change and traction bronchiectasis are uncommon.
Imaging of patients with DIP may be normal or demonstrate predominantly ground glass attenuation and/or consolidation that may be bibasilar and peripheral, or less commonly, diffuse (36) (Figure 12) . A smaller component of reticulations and honeycomb change reflect pulmonary fibrosis. Emphysematous changes may also be present.
Cryptogenic Organizing Pneumonia COP is a subacute febrile noninfectious disease of the small airways and airspaces that may mimic pneumonia. Patients typically present in the 5th to 6th decades of life with cough and dyspnea of approximately three months duration. COP typically improves dramatically following corticosteroid therapy, but may quickly relapse when the dosage is reduced or discontinued (37, 38). COP is the preferred nomenclature, but is still used interchangeably with the older term of bronchiolitis obliterans organizing pneumonia (BOOP) (37). Injury of the small airways results in mucosal ulcerations that heal with granulation tissue plugs (proliferative "bronchiolitis obliterans") that extend into the alveoli ("organizing pneumonia"), with a smaller component of interstitial inflammation. Proliferative bronchiolitis of COP is a distinct entity from constrictive bronchiolitis obliterans, which is a common injury of lung and bone marrow transplant recipients and results in irreversible scarring of the bronchiolar walls with secondary air trapping. On chest radiography, COP typically manifests as decreased lung volumes and multifocal subsegmental patchy consolidations with a juxta-pleural or bronchocentric distribution (38) (Figure 13) . On CT, COP typically exhibits mixed consolidations and areas of ground glass attenuation that may be triangular, patchy, and peripheral in distribution or extend centrifugally from plugged airways (37, 38). A nodular component, defined by the secondary pulmonary lobule, may have a patent bronchus or feeding vessel (39, 40). Both unilateral and migratory lung involvement have been reported (38). In addition, a subset of patients with COP may exhibit a fulminant clinical course that culminates in severe pulmonary fibrosis and/or death (41). Bibasilar juxta-pleural reticulations are uncommon, and generally correlate with fibrosis and a less favorable outcome (37, 38, 42). Reactive mediastinal lymph nodes may be mildly enlarged.
Acute Interstitial Pneumonitis (formerly Hammon Rich Syndrome) AIP is an acute fulminant lung injury due to idiopathic diffuse alveolar damage (43). Patients typically develop rapidly progressive hypoxemia and respiratory failure that require mechanical ventilation. At least 50% of patients die in the weeks following disease onset. Hamman and Rich originally described AIP as a rapidly progressive and typically fatal form of pulmonary fibrosis (44). AIP is now recognized as a distinct entity from IPF. AIP is characterized acutely by alveolar hyaline membrane formation and mild interstitial inflammation, followed by type 2 pneumocyte proliferation, and typically culminates in interstitial and alveolar fibrosis (14, 44). The histopathologic pattern of AIP is identical to that of adult respiratory distress syndrome due to sepsis, shock, multi-system trauma, and multiple other insults. Chest radiographic abnormalities typically lag behind signs and symptoms of respiratory failure by 24–48 h, then manifest with decreased lung volumes and mild diffuse ground glass attenuation that may rapidly progress to symmetric and diffuse or bibasilar air space consolidation (45) (Figure 14) . On CT, AIP is characterized by random extensive ground glass attenuation (100%) and consolidation (67%), with focal sparing of scattered second pulmonary lobules (43, 46). Interlobular septal thickening and traction bronchiolectasis may also be present, but honeycomb change is uncommon (44). In the infrequent survivor, AIP may heal with no residua or with variable degrees of fibrosis (45, 46).
Differential Diagnoses The radiologic features of the idiopathic ILD overlap and may be identical to ILD of known etiology. UIP secondary to connective tissue diseases (Figure 15) and asbestosis (Figure 16) may mimic IPF. Chronic aspiration may be a common cause of limited bibasilar fibrosis (Figure 17) . Granulomatous ILD typically manifests as small nodules, which are "hard" and of lymphatic distribution in sarcoidosis (Figure 18) , versus "soft" and centrilobular in patients with hypersensitivity pneumonitis (Figure 19) . In addition, hypersensitivity pneumonitis may exhibit patchy consolidation and air trapping that mimics RB-ILD or COP. The drug-induced and neoplastic ILD (lymphocytic interstitial pneumonitis and lymphangetic metastases) may also exhibit reticulations and ground glass attenuation. Secondary findings may provide important clues to the etiology of ILD, such as a dilated esophagus with scleroderma, pleural plaques with asbestos exposure, lymphadenopathy with sarcoidosis or lymphangetic metastases, and increased liver attenuation with amiodarone drug-induced lung disease.
Conclusions The primary role of imaging of ILD is to identify the presence and extent of fibrosis, to detect secondary findings characteristic of ILD of known etiology, and to direct the optimal sites of lung biopsy. The diagnosis of idiopathic ILD requires a multidisciplinary approach and integration of the clinical features, to include disease duration and prevalence, with the radiologic pattern and distribution (47). In the clinical setting of progressive chronic ILD, the diagnosis of UIP/IPF can be made with confidence on HRCT. However, in many instances, the clinical and radiologic features of the various ILD overlap, and surgical lung biopsies may be needed. The temporal heterogeneity of UIP pattern is the defining histopathologic feature of IPF; all other ILD have temporal homogeneity. NSIP and DIP may have a similar peripheral and basilar distribution as IPF, but typically have more cellular inflammation on histopathology and ground glass attenuation on CT. RB-ILD is characterized by small airways disease in the upper aspects of the lungs, and smoking cessation with follow-up CT to document disease resolution may pre-empt a lung biopsy. COP clinically may mimic pneumonia of several months duration, is a diagnosis of exclusion, and typically manifests on CT as patchy peripheral nodular consolidations that resolve following corticosteroid therapy. AIP is a rare explosive lung disease that rapidly progresses to respiratory failure and death. Lung biopsy is usually needed to confirm the diagnosis and exclude infectious or other treatable entities. Footnotes This section was written by Diane C. Strollo, M.D. (Dorothy P. and Richard P. Simmons Center for Research and Education in Interstitial Lung Disease, Department of Radiology, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania). References
The Prognosis of Idiopathic Pulmonary FibrosisIdiopathic pulmonary fibrosis (IPF) is a relentlessly progressive disease that usually leads to death within 5 yr of diagnosis (1–4). Because a response to current therapeutic agents is unusual, treatment is unlikely to alter outcome in most patients (5). Nonetheless, prognosis will have an effect on the decision of whether or not to treat. Some individuals with this disease will survive for many years, a fact that highlights the variability of the natural history of this disease (4, 6). Determining factors that predict the outcome has become increasingly important because lung transplantation is now an accepted therapy for selected candidates (7). In addition, patients typically seek advice about how long they might live so that they and their families may plan for the future and establish appropriate supportive care for advanced disease (8). For these reasons, clinicians have attempted to develop indicators that will reliably predict longevity in patients with IPF.Early studies (1, 9–12) identified a number of factors (Table 1) that seem to affect survival, but these studies must be interpreted in light of their limitations. These include: retrospective design, lack of histologic confirmation of usual interstitial pneumonia (UIP), lack of high-resolution computed tomography (HRCT) scans to assess inflammation and fibrosis in a semiquantitative manner, relatively small numbers of subjects, and inadequate duration of follow-up. Recent efforts to clarify the histopathology of idiopathic interstitial pneumonias (13) and the advent of more advanced imaging and lung biopsy techniques allows for a more rigorous approach to the diagnosis of IPF. A consensus statement recently released by the American Thoracic Society and European Respiratory Society details the present understanding of the pathogenesis of IPF and offers rational recommendations for confirming the diagnosis and for therapy (7). This report highlights the challenges in making a confident diagnosis of IPF. Contemporaneous reports emphasize the differences in survival between IPF and the other idiopathic interstitial pneumonias such as nonspecific interstitial pneumonia (NSIP) (3, 13–15). Incorrect inclusion of patients with the latter disease into outcome studies for the former disease will certainly confound the results by overestimating the survival in IPF. With the emergence of the new classification for idiopathic interstitial pneumonias, it is necessary to consider only studies that use the currently accepted diagnostic criteria in determining the factors affecting survival for patients with IPF. The purpose of this short review is to summarize the results of such studies and to suggest additional quantitative methods for determining extent of disease and its impact on outcome.
Assessing Survival by Routine Testing In 1997, Erbes and colleagues (6) published results of a study involving 99 subjects with biopsy-documented IPF who were followed from 1973 to 1988 at their institution in Berlin, Germany. The mean age of the group was 53.2 ± 15.4 yr, and 56 were current smokers, with the remaining 43 being nonsmokers for at least 5 yr. The most common symptoms were cough and dyspnea. These symptoms were noted for a longer time in nonsmokers compared with current smokers, but on average were present for 21 ± 41 mo before diagnosis. Only two subjects did not have abnormal pulmonary function results at the time of diagnosis. Mean values for PaO2, total lung capacity (TLC), and forced vital capacity (FVC) were normal in smokers and abnormally low for the nonsmokers. All of their subjects were treated initially with corticosteroids starting at 0.5–1.0 mg/kg with a maximum dose of 60 mg/d for 1 mo, and then tapering by 10 mg/mo to 15 mg/d. The overall duration of therapy was not specified. Subjects who did not respond to corticosteroids were given azathioprine. The mean for follow-up was 5.5 yr, with a range of 6.6 mo to 18 yr. Inclusion in this study was discontinued in 1993. The 5-yr survival rate was 62%, with the mean survival time from presentation of 41 ± 41 mo for all of the patients who died during the study. The two factors that had a negative impact on survival were (i) age of > 50 yr, and (ii) TLC and FVC < 78% of predicted (Figure 1) . Sex, DLCO/Va and the (A-a)O2 gradient during exercise did not have an impact on survival.
The main criticism that may be leveled against this study is the inclusion of subjects with other forms of idiopathic interstitial pneumonias that have a better prognosis. Because the survival rate was generally higher than what now is thought to be the case with IPF (3, 4, 7, 13, 16), it is possible that not all of the subjects had IPF. This is not surprising, because during the course of the study, the histologic features that typify the various forms of idiopathic interstitial pneumonia were less well clarified than they are today. Despite the new classification that was enunciated in 2000, however, there is still variation in the blind reading of the same slide by different highly experienced pathologists (17). The strengths of the study are the number of subjects that had been well characterized physiologically and treated in a relatively uniform fashion. HRCT and Pathology Gay and colleagues (2) published a study in 1998 that correlated the survival of patients with IPF with the results of HRCT scanning of the chest in addition to pathologic scores on surgical lung biopsy specimens. This study comprised only 38 subjects with histologically confirmed idiopathic pulmonary fibrosis. The mean age of 54 ± 2.2 yr was similar to the study of Erbes and coworkers, as was the mean duration of symptoms before diagnosis (2.6 ± 0.6 yr). Only 11 were never–smokers, and the remainder current or ex-smokers. The mean values for FVC, FEV1 and DLCO were 69.7 ± 2.5% of predicted, 72.9 ± 2.3% of predicted, and 49.9 ± 2.4% of predicted, respectively. Surgical lung biopsy specimens were scored for cellularity, desquamation, granulation, and fibrosis using previously published criteria (18). The authors also employed a method for scoring the severity of ground glass infiltrates and fibrosis in all of the lobes of the lungs by HRCT that they had developed earlier and averaged the score for the individual lobes to arrive a whole lung score (19). All subjects received prednisone at 1 mg/kg/d for at least 3 mo. Those that responded by demonstrating improvement in their clinical, radiologic, and physiologic score (20) were continued on prednisone, but in a tapering dose for up to 18 mo. Prednisone was quickly tapered in subjects who progressed or who did not improve. They were then crossed over to oral cyclophosphamide therapy (2 mg/kg/d) for 6 mo. After 3 mo of prednisone, 10 subjects (26%) responded, 14 (37%) remained stable, and 14 progressed. Of this latter group, seven subjects died within the first 3 mo of therapy. Two subjects in the stable group and one in the responder group died after 3 mo and were put into the nonresponder group. The most relevant parameters that contrasted the responder-stable group from the nonresponder group in long term follow-up were age (48.6 + 0.4 versus 62.3 + 2.1 yr, P = 0.0006), HRCT fibrosis score (1.2 + 0.2 versus 2.0 + 0.1, P = 0.001), and pathology fibrosis score (8.7 + 1.0 versus 13.3 + 0.7, P = 0.001). They also found in ROC analyses that the only factors demonstrating statistical significance in predicting the likelihood of death were the initial HRCT fibrosis score and fibrotic pathology score (Figure 2) . The authors also demonstrated that survival as measured by a Kaplan-Meier analysis was less in subjects with a HRCT score of greater than 2 (Figure 3) .
The major finding of this study is the value of the HRCT fibrosis score for predicting survival. Although the scoring system is at best semiquantitative, in the hands of experienced radiologists there was good agreement between individual readers. These results also suggested that there is a correlation between the pathologic fibrosis score and the HRCT fibrosis score, which validates the value of the HRCT scoring system to the assessment of patients with IPF. The major criticisms of this study are the relatively small number of subjects (n = 38) and the rather large percentage of patients who responded to prednisone. At 26% it is higher that what is generally expected, which is usually only 10–15%. The major strength of this study is the demonstration of the value of HRCT in the evaluation of patients with IPF. This has been supported in the recent study from Flaherty and colleagues (21, 22), where they have studied the ability of disease pattern on HRCT to predict survival. They looked at the HRCTs of 96 patients with histologically characterized UIP or NSIP. They showed that patients with a HRCT pattern of UIP were likely to have UIP on histology, but patients with an indeterminate pattern could have either UIP or NSIP. Furthermore, those patients with histologic UIP and a HRCT pattern consistent with UIP had a worse outcome than those patients with histologic UIP and indeterminate pattern on HRCT. Median survivals were 2.08 yr versus 5.76 yr, respectively. These results further support the utility of the HRCT fibrosis score in determining the outcome in IPF. Clinical Physiologic Scoring Two reports detailing factors influencing survival in IPF appeared in 2001. The first (23) comprised 87 subjects with confirmation of UIP by surgical biopsy. All data were collected prospectively between 1982 and 1993. The clinical, radiographic, and physiological (CRP) score at the time of diagnosis was calculated using standard definitions (20), and 80 subjects were treated with either corticosteroids alone or a combination of corticosteroids and immunosuppressive therapy. HRCT scans were not available at the time of diagnosis on all subjects, because the technology was not available at the time they presented. The surgical biopsies were scored for four features: fibrosis, interstitial cellularity, alveolar space cellularity and granulation, and young connective tissue. Pulmonary function assessment consisted of spirometry, lung volumes, diffusion, lung mechanics, and gas exchange at rest and with maximum exercise. The population was very representative for IPF based on age and severity of lung function abnormalities at the time of presentation. Sixty-three patients died during follow-up. Ten were censored because the cause of death was due to an illness other than IPF (n = 6) or they underwent a lung transplant (n = 4). Median survival was 47.5 mo, with a 95% confidence interval of 33.4–73.4 mo. A multivariate analysis done with factors shown to have an impact on survival in univariate analyses revealed that the following variables were significant: smoking status, granulation/connective tissue factor (P < 0.0001), coefficient of retraction (maximal transpulmonary pressure/total gas volume, P < 0.0001), and to a much lesser extent, the dyspnea score (P = 0.017). Interestingly, current smokers had better survival in the Kaplan-Meier analysis than did former and never-smokers (Figure 4) . The risk of death in current smokers was 22% less than that for never–smokers, whereas former smokers had an 88% increase in the risk of death compared with never-smokers. When factors such as granulation/connective tissue score, level of dyspnea, and coefficient of retraction were held constant, current and former smokers appeared to have a poorer survival rate compared with never-smokers.
In the second study (4), these authors revised the CRP scoring system for IPF they had reported earlier (20). In devising the new CRP score, they assigned weight to variables found to have a significant impact on survival in a hierarchical multivariate analysis performed on a cohort of 183 patients for whom all pertinent data were available, including a surgical biopsy. These variables included age, smoking status, clubbing, profusion of infiltrates and findings of pulmonary hypertension on plain chest radiograph, TLC as % predicted, and pO2 at maximal exercise. Because it is not always practical to perform maximal exercise testing on subjects with advanced disease, the authors also crafted an abbreviated CRP score in a cohort of 228 subjects, which comprised all of the above variables except for pO2 at maximal exercise. The maximal score in the two models were 100 and 89.5, respectively. In this second study, which contained nearly twice as many subjects as their previous study, they found a similar relationship between survival and smoking status, but in this instance the difference in survival for the never-smokers and past smokers was not as great as noted in the previous study (23). The impact of the new CRP score on survival is shown in Figure 5 . The difference in survival is quite dramatic for subjects on the high and low end of the scoring system. The probability of survival based on the abbreviated score was quite similar to that predicted by the complete CRP score.
The strengths of these studies are the careful diagnosis of the IPF, numbers of subjects, duration of follow-up, and rigor of the statistical analyses. They make the results very convincing and provide clinicians with an approach to predicting survival of patients with this discouraging disease. The drawback of the new "complete" CRP is the requirement for maximum exercise testing, which is not always feasible or safe for patients with advanced disease, particularly when they have a significant oxygen requirement. The lack of this requirement for the new "abbreviated" CRP score and its relatively high predictive ability makes this approach attractive to the average clinician. Substituting the DLCO for the pO2 at maximal exercise in the multivariate analysis likely would strengthen the predictive ability of the abbreviated CRP, because the former is strongly correlated to the latter and can be measured in most patients. The authors also pointed out that inclusion of HRCT results would likely strengthen the scoring system, because assessment of the profusion score and detection of pulmonary hypertension on the plain chest radiograph is subjective. Nonetheless, the authors are to be commended for providing this approach for the assessment of patients with IPF. The recently published study by Wells and coworkers (24) attempts to further the development of composite scoring by correlating their results with HRCT. They studied 212 patients with a clinical/CT diagnosis of IPF. They divided the patients evenly into two groups, using the first group to develop a composite physiologic index (CPI), which was correlated with a semiquantitiative scoring system which they had developed. They determined the appropriate weight of variables from pulmonary function testing (FVC, FEV1, and DLCO). Thus the CPI is based solely on physiologic results. The CPI was tested against the second group of patients and correlated more strongly with extent of disease than the individual PFT variables. The CPI's primary advantage was the ability to account for coexistence of emphysema by including the FEV1, which tends to be higher in patients with coexistent emphysema. Furthermore, when they studied prognostic factors in a subgroup of patients with histologically proven UIP; the CPI was strongly linked to mortality (P < 0.005). Decreased percentage FVC and increases in alveolar–arterial oxygen gradient were also strongly associated; however, DLCO was not shown to have a significant association. The advantages of this scoring system are (i) it does not require that a full exercise test be administered, and (ii) it does not require experienced readers of HRCT. Therefore, it has the potential to stage the severity of disease and predict outcome in most clinical practices. The study also reinforces the utility of HRCT in assessing prognosis. HRCT Predicting Need for Transplant The focus of this study (5) was slightly different than that of the previous studies, which looked at survival in all subjects with IPF. In this instance, the authors limited the study to patients with IPF who were considered to be candidates for transplantation. Their goal was to define the appropriate time for transplantation based on anticipated survival. The study group consisted of 115 patients with well-documented IPF who were also below the age of 65 yr. Only about one third (38%) underwent surgical biopsy for confirmation. The diagnosis in the remainder (62%) was made on clinical grounds and findings typical for IPF on HRCT. Median follow-up was 26.2 mo, with a range of 1–97 mo. Median survival was 55 mo, and the major cause of death was respiratory failure from progression of IPF. Three out of the 46 deaths were due to bronchogenic carcinoma. This population was similar to those described in the previous studies, but they were slightly younger due to the exclusion of patients over the age of 65 yr. In addition, their degree of impairment based on FVC and DLCO expressed as a percent of predicted appeared to be slightly less. The pulmonary fibrosis scores were derived according to the technique of Kazerooni and colleagues (19) that Gay and coworkers used in their study (2). The authors performed univariate analyses based on spirometry, lung volumes, diffusion, and HRCT fibrosis scores. When they performed a multivariate analysis with the parameters found to correlate with survival in univariate analyses, the only factors found to be significant were DLCO percent of predicted and fibrosis score on HRCT. For every 1% decrease in DLCO percent predicted, the hazard of death increases by 8% (CI 1–14%). For each unit increase in the HRCT fibrosis score, the hazard of death increases by 527% (CI 32–2,890%). ROC analyses confirmed the value of combining the DLCO and HRCT fibrosis score in predicting survival compared with FVC, DLCO, or HRCT fibrosis score alone (area under the curves were 0.907, 0.693, 0.802, and 0.863, respectively) (Figure 6) . Finally the authors derived a prediction model for survival of 2 yr based on the DLCO percent predicted and HRCT fibrosis score. They presented a "lookup" table in the article, which is depicted in Table 2.
The strengths of this study include the relatively robust number of patients who had a sound initial characterization and adequate follow-up. In addition, the authors employed standard clinical testing to derive the factors used in their survival prediction model. Finally, standard and well-accepted statistical analyses were employed to derive the survival prediction model. The major weakness is that only 38% had confirmation of UIP by surgical lung biopsy. The authors acknowledge this problem and constructed a survival prediction model based on this smaller subset of patients. They found that only the HRCT fibrosis score was an independent predictor of survival, with the hazard of death increasing by 375% for each unit increase in HRCT fibrosis score. These results further support the utility of the HRCT fibrosis score in determining the outcome in IPF. Summary The recent studies summarized above largely confirm what had been presented earlier regarding factors that influence survival in individuals with IPF (Table 1). The only factor that the recent studies failed to confirm to affect outcome was male sex. Perhaps this reflects under-representation of females in the earlier studies. The advantages of the recent studies over previous studies in predicting survival result from a clearer definition of the histologic abnormalities in IPF. The implications of a significant amount of young granulation tissue on the biopsy, which probably includes "fibroblastic foci," has become much clearer of late (13, 16, 23). There is controversy about the importance of the fibrosis score on surgical lung biopsy (2, 23, 25), but the finding in two studies that survival is related to the HRCT fibrosis score suggests that the total amount of fibrosis is an important prognostic finding. Given the size of the surgical biopsy specimen, it is not surprising that it may not convey accurate information about the total amount of fibrosis. Furthermore, the amount of fibrosis may reflect the timing of the diagnosis (i.e., late in the disease), because symptoms develop only after significant impairment is present. Unfortunately, we do not know the pathology of the early lesions in IPF. It is also reassuring that standard clinical testing such as the DLCO contribute to the accuracy of prediction models. Future Survival Prediction Models Looking to the future, it is important to develop a survival prediction model that uses clinical results that can be obtained at most medical centers. One disadvantage of the DLCO is the lack of standardization in performing the technique, and different sets of normal values (26). The strong predictive ability of this test, however, argues for continued efforts at the development and application of universal standards for the performance of this test and analysis of the results (27). The disadvantage of the pathology fibrosis score is that it requires a pathologist who is skilled in reading and scoring biopsies demonstrating UIP. Most medical centers cannot boast of the presence of pathologists who have expertise with interstitial lung disease. In addition, the scoring is by nature semiquantitative and open to variability between readers. For these reasons, there likely will be less emphasis on the pathology fibrosis score in the future, unless a consortium of skilled pathologists willing to score surgical biopsy specimens is relatively accessible to the majority of physicians caring for these patients. The HRCT fibrosis score may be less prone to variability between readers but is still semiquantitative. Differences between centers may arise for several reasons. First, different imaging algorithms may be used on various models of scanners. Second, the volume of lung assessed on HRCT is usually only 10% of the total. This could lead to under- or overestimation of the fibrosis score. Third, the more volume analyzed the greater the time required, which may discourage busy chest radiologists from undertaking such a task. Nonetheless, a fibrosis scoring system using HRCT may be the approach that is most reliably standardized in the future, but remains a semiquantitative method. Several methods are being developed to quantitatively assess the morphologic features on CT of the entire lung; one such method is CT Morphometry (CTM) (28–30). CT scans can provide information about lung morphometry when analyzed to provide an estimate of specific volume of lung units (29). Because the attenuation of X-rays is linearly related to density within the biological range, any variation in the quantity of lung components (blood, tissue, air) will be reflected in the density of the lung. CTM uses attenuation data to estimate the inflation of the lung, in terms of volume of gas per gram of tissue. Because CTM studies the lung in its entirety, we would have a tool that may be better at recognizing interval changes than traditional HRCT, which usually obtains samples at intervals. CTM is performed by segmenting the lung parenchyma from the chest wall and central vessels for each slice. The volume of the slice is estimated by summing the volume of all voxels in the slice. (A voxel is a three-dimensional pixel.) The density of lung tissue in the voxel is estimated by adding 1,024 to the Hounsfield units of each voxel and dividing the sum by 1,024 (28). Lung weight equals the product of mean lung density, determined from the densities of all voxels, and volume of the entire lung, which is the volume sum of all voxels. The specific volume of the lung is the inverse of density, expressed as milliliters of gas per gram of lung tissue. It is calculated by the following formula: mL (gas)/g (tissue) = Specific Volume (tissue & gas) - Specific Volume (tissue). The volume of lung tissue and gas is calculated from the CT, and the volume of tissue is the inverse of the density of the lung tissue. The density of tissue is assumed to be 1.065 gm/ml (31). The inflation of each lung voxel is categorized by the cutoffs of 2, 4, 6, 8, and 10.2 mL (gas)/gm (tissue). The amount of lung between the cutoffs is determined and is used for comparison between different scans. The scans must be obtained at similar lung volumes to compensate for changes in specific volume related to the degree of inspiration. A distribution of the specific volume for all voxels may be constructed. In the case of fibrotic lung disease, there is a shift to the left toward lower specific volumes in lungs affected by IPF (Figure 7) when compared with normal. The potential value of this technique is that when software is developed to automate the process of segmenting the lung from chest wall and central vessels, the entire analysis will not be confounded by variations in reading and interpretation.
We recently had the opportunity to apply CTM to a patient with a recent onset of interstitial lung disease whose biopsy revealed a pattern consistent with UIP. The initial diagnosis was IPF, but further studies revealed a diagnosis of overlap syndrome of connective tissue disease. Before the detection of his CTD, he was treated with 60 mg/d of prednisone. Fortunately the patient responded to this therapy, essentially normalizing his pulmonary function test results. The extent and density of infiltrates on the CT scan also decreased (Figure 8) . A distribution of specific volumes derived from the pre- and post-treatment CT scans is shown in Figure 9 . Before therapy, the histogram reveals a shift toward an increase in both the low specific volume voxels (< 2 ml/gm tissue) and in the high specific volume voxels (> 12 ml/gm tissue). Presumably, the low specific volume voxels were detected in areas of the lung where connective tissue predominated. The high specific volume voxels may represent areas of hyperinflated lung or even areas of honeycomb fibrosis. Following therapy, the number of high specific volume voxels diminished and the proportion of voxels in the mid-range of specific volume increased. The estimated weight of the lung decreased from 1,865 to 1,558 g, and the mean specific volume from 4.97 to 3.35 ml/gm of tissue. The total volume of lung with specific volume of greater than 8 ml/g decreased dramatically from 1,469 ml to 103 ml. At the present time, the significance of these shifts in specific volume are unknown, but this example demonstrates that sizable shifts do occur in the distribution of specific volume when the lung is involved with interstitial lung disease and when the disease responds to therapy.
Use of a methodology such as CTM may not improve our diagnostic accuracy with interstitial lung disease. Using a method which does not depend upon visual analysis and the reader's interpretation of interval change would allow widespread uniform application at multiple centers. It is clear that further development of this method using textural or three-dimensional analyses may lead to the ability to distinguish honeycombing from septal thickening or emphysema, rather than just relying upon volumetric numbers. Quantification of the extent of such patterns may have potential in predicting prognosis. But development of a method which uses information from the entire lung is important, rather than relying upon semiquantitative assessment. Such an approach could prove important in staging and the prediction of outcome, particularly when coupled with the composite scoring methods outlined previously. Further investigation is warranted and ongoing. Footnotes This section was written by Andrew Perez, Robert M. Rogers, and James H. Dauber (Simmons Center for Interstitial Lung Diseases; and University of Pittsburgh, Pittsburgh, Pennsylvania). References
Idiopathic Pulmonary FibrosisNew Insights into Classification and Pathogenesis Usher in a New Era in Therapeutic ApproachesIdiopathic pulmonary fibrosis (IPF) is a fascinating and poorly understood disease with devastating consequences to those afflicted that has undergone several revisions in the last 30 yr. IPF is a clinical, physiologic, radiographic, and pathologic entity that has been difficult to characterize as a homogenous "disease" (1). A major reason for the difficulty in defining IPF has been the lack of a gold standard in the diagnosis. Unlike cancer, infection, or pulmonary embolism, in which it is possible to unequivocally obtain documentation of disease, it has been debated whether lung biopsy provides a gold standard in IPF. There are (at least) three major reasons for this ambiguity: (i) not all patients with radiographic suspicion of pulmonary fibrosis have a lung biopsy performed, (ii) the lung biopsy only samples small regions of the lung, and (iii) pathologists have differed in their description of the pathologic findings in IPF. The pathologic description of IPF was originally delineated by Leibow and colleagues, and termed "usual interstitial pneumonia (UIP)" (2). Over the last 30 yr, there has been a refinement in the pathologic description of Idiopathic Interstitial Pneumonias, and the most recent classification represents a consensus of leading experts in the field (2). The major refinement has been that the mere presence of fibrosis in the lung biopsy does not necessary mean that the patient has IPF. What has been reproducible in several large retrospective series of patients with presumed IPF is that the mean survival after recognition is between 2 and 5 yr, with the most recent studies suggesting a mean of 2.8 yr (3). UIP has been recognized as a distinct pathological entity characterized by the presence of (i) patchy chronic interstitial inflammation; (ii) oldest disease (fibrosis) is peripheral in the lung acinus or lobule; (iii) transitions to uninvolved lung in the biopsy, i.e, temporal heterogeneity; and (iv) a leading edge of fibroblastic foci and microscopic honeycombing. In studies in which this description has been identified by skilled observers in lung biopsies from patients with the clinical, physiologic, and radiographic findings characteristic of IPF, the majority of patients have demonstrated progressive disease and died despite treatment with high doses of corticosteroids and/or potent cytotoxic immunosuppressants such as cyclophosphamide and azathioprine. This is in contrast to patients that have distinct pathologic descriptions as well as distinguishing clinical, physiologic, and radiographic features that allow for the characterization of nonspecific interstitial pneumonia (NSIP), desquamative interstitial pneumonia (DIP), respiratory bronchiolitis–associated interstitial lung disease (RBILD), or acute interstitial pneumonia (AIP). NSIP has been subdivided into cellular and fibrotic forms that differ in prognosis (4). With the exception of AIP, the other idiopathic interstitial pneumonias (IIPs) have a better prognosis and improve with corticosteroid/immunosuppressive more frequently than observed with UIP. Important observations from Flaherty and colleagues have suggested that patients with IIP may have both UIP and NSIP-fibrotic patterns in different regions of the lung (5). However, the presence of UIP appears to be dominant in terms of clinical outcome. That is, patients with both NSIP-fibrotic and UIP patterns on biopsy have a poor prognosis relative to those with only NSIP-fibrotic (5). This work raises the interesting question as to the relationship between NSIP and UIP. Future studies will be needed to determine if these are a continuum or distinct pathologic and clinical entities.The combination of the unique pathologic features of UIP on biopsy, the inexorable progression to death, and resistance to anti-inflammatory therapy constitute the cardinal manifestations of what is now termed IPF/UIP, and have led to recent suggestions that new therapies should be directed at regulating fibroblast functions rather than targeting the inflammatory response per se (6). New Paradigms in the Pathogenesis of IPF Recent observations have led to new concepts in the biology of progressive pulmonary fibrosis. These observations have occurred both in patients with IPF and in animal models of fibrosis. The concept that dominated the field in the 1970s and 1980s has been described the "inflammatory" concept of pulmonary fibrosis. The paradigm was based largely on the observation that bronchoalveolar lavage fluid from patients with IPF had increased numbers of inflammatory cells (mostly neutrophils and eosinophils) relative to normal individuals (7). The concept that permeated the literature in that era was that IPF resulted from an unremitting inflammatory response to an exogenous insult, culminating in progressive fibrosis. By targeting the inflammatory response, the belief was that the fibrosis could be limited and/or prevented. Unfortunately, it now appears that the data are more likely explained by structural abnormalities in lung architecture (traction bronchiectasis) such that inflammatory cell trafficking is altered. That is to say, the airway inflammation is likely a result, rather than a cause, of the fibrosis. Several key observations have led to a revised hypothesis of the key elements in the pathogenesis of progressive pulmonary fibrosis. Although the role of inflammation in the pathology of IPF remains controversial, it is difficult to ignore the lack of efficacy of corticosteroids. Epithelial Cell Apoptosis
An emerging body of literature has accumulated in recent years to suggest that alveolar type II cell injury and apoptosis may be an important early feature in the pathogenesis of pulmonary fibrosis. Ultrastructural studies have demonstrated alveolar type II cell injury and apoptosis in lung biopsies from patients with IPF (8). Studies from Hara and colleagues have demonstrated increased expression of pro-apoptotic proteins in alveolar epithelial cells from patients with IPF (9). In addition, proof of principle experiments using the bleomycin model of lung injury and fibrosis in animal models have suggested that inhibiting epithelial cell apoptosis, with a variety of approaches including inhibiting the Fas-Fas ligand pathway, inhibiting production of angiotensin, and blocking caspase activation all abrogate the development of experimental fibrosis (10). Important studies from Uhal and colleagues have suggested that IPF fibroblasts produce angiotensin peptides that promote epithelial cell apoptosis (11). More recently, transforming growth factor-ß has been demonstrated to promote epithelial cell apoptosis (12). An additional mechanism proposed to explain epithelial cell apoptosis is increased production of oxidants in IPF (13). Several studies have shown excessive oxidant production in IPF, as well as deficiencies in glutathione production (14–15). These observations have led to an ongoing clinical trial in Europe currently evaluating the role of antioxidant therapy with N-acetyl cysteine (U. Costabel, personal communication). In addition to the potential regulators of epithelial cell apoptosis listed above, tumor necrosis factor (TNF)-
Many observers agree that an essential and unique element in the progressive pulmonary fibrosis of IPF is loss of the integrity of the subepithelial basement membrane (8). This is a unique feature of IPF. Loss of the alveolar epithelial protective barrier could lead to exposure of the underlying basement membrane to oxidative injury, resulting in degradation of key constituents of basement membrane. The loss of basement membrane integrity could be an important signal for epithelial cell regeneration. Hyperplastic alveolar type II cells are a common feature of the pathology of UIP (8). "Frustrated" epithelial cell generation could be an essential proximal signal for mesenchymal cell recruitment. The rationale for this is that a variety of growth factors accumulate following epithelial cell injury to promote epithelial cell proliferation. These include keratinocyte growth factor, transforming growth factor Angiogenesis
Parallels have been drawn between the biology of IPF and the biology of cancer. The unremitting recruitment and maintenance of the altered fibroblast phenotype with generation of myofibrobasts that fail to die is reminiscent of the transformation of cancer cells. A hallmark of tumorigenesis is the production of new blood vessels to facilitate tumor growth. A number of therapies targeting angiogenesis are in varying stages of clinical development for cancer. The concept that an important aspect of progressive fibrosis is angiogenic activity has been championed by Strieter and colleagues (18). They have demonstrated increased angiogenic activity in the lung tissue of IPF and experimental fibrosis (18). This increased angiogenic activity has been attributed to an imbalance of pro-angiogenic chemokines (interleukin [IL]-8) and anti-angiogenic C-X-C chemokines (IP-10). IP-10 is induced by interferon (IFN)- In contrast to the concept that progressive IPF is associated with increased angiogenesis are the recent reports that there is decreased expression of expression of VEGF and endothelial cell proliferative indices in IPF. In particular, it has been suggested that there is a paucity of expression of pro-angiogenic proteins in the fibroblastic foci in UIP in comparison to the granulation tissue in organizing pneumonia (21–23). It may be that there is enhanced angiogenesis in the earlier stages of the development of UIP, whereas there is a loss of blood vessels in the more advanced stages. These will be important areas of focus in the future. Abnormal Matrix Turnover The essential hallmark of IPF is an exorbitant production of extracellular matrix molecules including collagen, tenascin, and proteoglycans. There is clearly an imbalance between the production and degradation of extracellular matrix. Data from Selman and Pardo have suggested that there is increased production of inhibitors of matrix degradation (TIMPs), accounting for the inability to degrade matrix (24). One of the properties of transforming growth factor-ß (TGF-ß) is promoting matrix production while inhibiting TIMP production. It is therefore unclear if the decreased production of TIMPs is an inherent defect in IPF (such as a polymorphism) or a consequence of excessive TGF-ß. Th1 versus Th2 Cytokines
One of the rationales for corticosteroids and immunosuppressive therapy for IPF has been to target the immune system. This therapy has proven effective in autoimmune disorders such as Wegener's granulomatosis and systemic vasculitides, but IPF does not appear to be in the same category. However, data have been obtained suggesting that a cytokine imbalance may exist in IPF. RT-PCR studies have suggested that there is increased Th2 cytokines (IL-4, IL-5, and IL-13) in lung tissue of patients with IPF (25). In addition, preliminary data suggesting that IPF may represent a relative IFN- Growth Factor Production A variety of growth factors that influence fibroblast and myofibroblast functions have been shown to be produced in the lung tissue of patients with IPF, and to mediate the pathogenesis of experimental fibrosis. TGF-ß1 has been shown to be a critical mediator of lung fibrosis in animal models (28). A number of studies have shown that antagonizing TGF-ß1 prevents the development of tissue fibrosis (28). However, concerns have been raised about potential consequences of TGF-ß1 blockade because of the finding that TGF-ß1 knockout mice die of unremitting inflammation (29). In addition, failure to activate TGF-ß1 following fibrotic lung injury has been shown to not only prevent fibrosis, but result in unremitting lung inflammation. However, recent data in mice have shown that long-term treatment with a TGF-ß1 antagonist did not result in significant immune disturbances (30). Targeted overexpression of TGF-ß1 has been shown to produce progressive fibrosis (31). Therefore, targeting growth factor–signaling pathways, such as TGF-ß1, PDGF, or IGF-1 with small molecules such as perfenidone, or tyrosine kinase inhibitors such as imatinib mesylate, are important potential therapeutic strategies for IPF. A phase II trial to evaluate Gleevec (imatinib mesylate) in patients with IPF is underway. A number of other growth factors such as IGF-1, PDGF A and B, and CTGF are expressed in IPF lung tissue, but the direct contribution of these mediators to progressive fibrosis is unknown. In addition to effects on fibroblast proliferation, growth factors such as IGF-1 may promote fibroblast (and myofibroblast) survival. IGF-1 has been shown to inhibit apoptosis by activating the Akt pathway. This may have important consequences for maintenance of a profibrotic environment. Altered Fibroblast Phenotypes The concept that fibroblasts from patients with IPF have a unique phenotype is generally accepted, although the specifics of the phenotype have been different in various studies (32–34). Raghu and colleagues made the important observation that fibroblasts from different regions of the lung corresponding to new versus old fibrosis had different growth rates (33). Subsequent studies have shown altered production of TIMPs and other mediators, suggesting that IPF fibroblasts have different properties than normal lung fibroblasts. Some discrepancies exist on whether IPF fibroblasts proliferate more or less in comparison with normal lung fibroblasts (35). In addition, some studies have suggested increased rates of apoptosis consistent with rapidly turning over subpopulations (35). Studies examining the response of IPF fibroblasts to apoptotic stimuli have not been reported. In addition, the pattern of expression of pro-apoptotic proteins and inhibitors of apoptosis (IAP) have not been examined in IPF fibroblasts. Myofibroblast Recruitment and Maintenance
Much attention has been focused recently on the role of the myofibroblast in the pathogenesis of IPF. Kuhn and McDonald described myofibroblasts in a contractile phase in fibroblastic foci from IPF lung biopsies in 1991 (36). Recent attention was generated by the observation from three different groups that the frequency of fibroblastic foci in lung biopsies from patients with IPF correlates with poor prognosis (37). The defining characteristic of the myofibroblast in the fibroblastic foci is the production of new collagen and fibronectin at the leading edge of existing scar. Myofibroblasts have contractile properties and stain positive for The Paradox of Pulmonary Fibrosis: Increased Apoptosis in Epithelial Cells and Decreased Apoptosis in Myofibroblasts The challenge of future targets for therapeutic intervention is to reconcile the two potentially opposing observations that, on the one hand, increased epithelial cell apoptosis can contribute to fibrosis, while at the same time decreased fibroblast or myofibroblast apoptosis promotes fibrosis. This is where experimental model systems can be useful. It is important to distinguish studies that demonstrate a decrease in bleomycin-induced pulmonary fibrosis achieved by a loss-of-function intervention (either gene deletion or inhibitor studies) from studies where a gain-of-function intervention causes pulmonary fibrosis. Bleomycin in particular is a very interesting but puzzling model system for studying pulmonary fibrosis. Bleomycin is not a good model of IPF because the fibrosis is not progressive. The initial pathologic lesions in bleomycin lung injury are focal areas of diffuse alveolar damage. These lesions subsequently heal into self-limited foci of collagen deposition. Myofibroblasts are present transiently. Although it is important to acknowledge the limitations of the bleomycin model and be cognizant that inhibiting fibrosis after bleomcycin treatment is not curing IPF, it is also an informative model because the injury appears to mimic the microscopic injury pattern that is believed to characterize the stepwise progression in fibrosis observed in IPF. Bleomycin is a product of fungi, and bleomycin hydrolase exists in both mouse and humans. One cannot help but wonder if there is not some significance to this naturally occurring antimetabolite of the environment for which processing enzymes exist. It is therefore reasonable to use bleomycin injury as a "mimic" of the acute injury pattern that occurs in patients with IPF. A useful model system for IPF would be to identify genetic mutations in mice that lead to progressive rather than self-limited fibrosis after bleomcyin injury. Model systems to dissect the relative contribution of epithelial apoptosis and myofibroblast survival will help unravel the relative contributions of these opposing forces in the overall pathogenesis of pulmonary fibrosis. Summary We appear to be entering a new era in the understanding of the classification, pathogenesis, and biology of IPF. New therapeutic targets include the epithelial cell, myofibroblast, and chronic inflammation. The acceptance that "standard" therapy of corticosteroids and cytotoxic medications is rarely effective, coupled with new insights into pathogenesis, has led to new therapeutic approaches that hold out great hope for the future. Footnotes This section was written by Paul W. Noble, M.D. (Yale University School of Medicine, Section of Pulmonary and Critical Care Medicine, New Haven, Connecticut). References
Microarray Analysis of Idiopathic Pulmonary FibrosisIdiopathic pulmonary fibrosis (IPF) is a refractory and lethal interstitial lung disease characterized by fibroblast proliferation, extracellular matrix deposition, and progressive lung scarring. The incidence of IPF is estimated at 15–40 cases per 100,000 per year, and the mean survival from the time of diagnosis is 3–5 yr regardless of treatment (1). Much of our understanding of the mechanisms of pulmonary fibrosis comes from studies of the classical model of bleomycin-induced lung injury and fibrosis in animals, as well as several newer models that have yet to gain widespread acceptance. Additionally, the application of these models to genetically modified animals greatly enhanced our understanding of the mechanism of lung fibrotic responses. However, the relevance of some of these studies to IPF or other human fibrotic lung diseases is still unclear, and there are many gaps in our understanding of the human disease (2).The advent of high throughput genomic profiling technologies such as microarrays, combined with advanced computational approaches, provides scientists with the ability to create high-resolution expression profiles of distinct disease states and to dissect molecular networks that underlie a diseased phenotype. Microarrays are now routinely used in almost every line of biomedical research. The most impressive examples are in the area of cancer research, where gene expression patterns obtained from microarray experiments allowed identification of new classes of lymphoma (3), prediction of metastasis in breast cancer (4), and prognosis determination in lung cancer and pediatric leukemias and lymphomas (5–7). Furthermore, many molecular pathways have been better characterized, and new targets for therapeutic intervention have been identified. The aim of this section is to provide an overview of the few works that applied microarrays to the study of pulmonary fibrosis; they range from studies of animal models of disease to analysis of human tissues. The specific challenges inherent to microarray analysis of lung disease are presented, and future directions are discussed. Testing the Water: Analysis of Animal Models of Lung Fibrosis
In our initial set of experiments using microarrays, we analyzed changes in gene expression in response to bleomycin using oligonucleotide microarrays that contained probe sets for Different insights were derived from comparing gene expression patterns in response to bleomycin of 129 mice homozygous for a null mutation of the integrin ß6 subunit gene (ß6-/-) and wild-type 129 mice. We had previously shown that ß6-/- mice developed inflammation, but not fibrosis, in response to bleomycin (13). This current experiment was aimed to use this feature to identify genes that are specifically characteristic of the fibrotic process. A simple hierarchical clustering procedure identified two clusters: one that contained mainly inflammatory genes, and another that contained mostly genes related to fibrosis, suggesting that the other genes in the cluster were also fibrosis-related genes (8). Applying self-organizing maps to the same dataset, we recently obtained a finer distinction of the clusters (Figure 1) . The genes induced by bleomycin fell into three distinct clusters. A cluster of 75 genes preferentially induced in wild-type mice and indeed contains many fibrosis-related genes (Figure 1, Cluster A). A cluster of 34 genes induced similarly in both mouse strains (Cluster B) potentially reflects a specific response to bleomycin that does not lead to fibrosis. A cluster of 55 genes that are induced by bleomycin but are higher in the ß6-/- compared with wild-type in every time point, contains many genes that are associated with exaggerated inflammatory infiltrate seen in ß6-/- both at baseline and after bleomycin. The data is available at http://genechip.ucsf.edu/. Katsuma and coworkers analyzed bleomycin-induced response in C57BL/6 mice using a lung cDNA array that they have created from a normalized lung library (14). They identified 89 nonredundant genes that changed; 25 changed after 2 d, and the rest changed at later stages. They identified a group of 12 genes that behaved in a manner similar to those discussed our previously published results, including osteopontin, whereas the rest of the genes did not change concordantly. It is unlikely that this difference is related to mouse strain, because we did get a 63% concordance rate between the genes increased in 129 mice and C57BL/6 (8). Potential causes for this disagreement may be the different RNA extraction methods used, the different microarray platforms, or potentially a different analytic scheme. In our experience, experiments that involve analyzing the injury and an intervention (inhibitor, gene knockout, etc.), are more productive in terms of the ability to generate biologically meaningful hypotheses. These two very different papers represent a common problem with microarrays, which is the difficulty in translating data across different microarray platforms. Microarray platforms differ in the type of the probes on the arrays (fragments of cDNA, different size of oligonucleotides), the design of the probes (probes that are 3' biased, probes that utilize alternative sequence resources, multiple or single probes per gene), and the labeling techniques (single-dye, competitive dual-dye, direct and indirect labeling). Any of these technical differences could lead to significant variability. The improvement in sequence information availability, the introduction of standards for data storage and publication, and most importantly the move away from cDNA arrays and the more widespread use oligonucleotide probes, should improve the ability of researchers to perform cross platform analysis. A simple strategy to overcome the cross-platform problem is to share samples between groups that use different platforms and to generate parallel datasets in more than one platform. This has the additional benefit of providing a global verification of the results. Naturally, as it is beyond the resources of a single group to analyze all models for pulmonary fibrosis and all relevant knockouts and transgenic mice, an effort must be made to deposit all microarray data in standard formats in public data repositories.
Dealing with Reality: Microarray Analysis of Human IPF Tissues To determine the relevance of our results to human fibrotic lung disease as well as the feasibility of analyzing human fibrotic lung disease we analyzed gene expression patterns in 5 lungs of patients with IPF (15). For controls, we used normal histology lung samples resected from patients with cancer, and a pool of RNA obtained commercially. Despite the variability of the patients, we observed an impressive difference in gene expression patterns. To determine the most informative genes (genes that best characterize the IPF samples), we applied scoring methods previously reviewed by us (16). Among the most informative increased genes we identified genes encoding for proteins expressed in smooth muscle differentiation and muscle contractile machinery, potentially representing the transcriptional signal of myofibroblasts and fibroblasts in myofibroblast/fibroblast foci typical of the disease. In concordance with our results in mouse lungs, we observed an increase in genes that encode for extracellular matrix proteins in fibrotic lungs, including Collagen I, III, tenascin C, and fibronectin. Surprisingly, we identified a coordinated increase in the levels of several matrix metalloproteases (MMPs). MMP-1, MMP-2, MMP-7, and MMP-9 were all significantly increased in fibrotic lungs. MMP-7 was the most significantly upregulated gene in our dataset (15). Cosgrove and colleagues presented their results in two conference abstracts and reported similar impressive increase in MMP-7; however, they did not observe an increase in other MMPs (17), and focused on angiogenic signaling in the fibrotic lung (18). Because MMP-7 has also been implicated in cancer progression (19, 20), and patients with IPF have a higher risk of cancer (21), we looked at MMP-7 levels in lung tumors, but did not find a significant increase (Figure 2) . Recently, we have repeated the experiments using a different microarray platform and more samples, and verified that MMP-7 was highly upregulated in most IPF lungs (data not shown). The mechanism by which of MMP-7 plays a role in pulmonary fibrosis is unknown. Some possible mechanisms recently reported that may play a role in fibrosis include regulation of neutrophil migration into the lung through MMP-7 effect on syndecan shedding (22), an effect on epithelial repair through MMP-7 effect on e-cadherin ectodomain shedding during epithelial injury (23), or even an effect on TGF-ß1 activation (G. Cosgrove, personal communication).
One of the benefits of having a dataset in mice and humans is that it allows for attempts at cross-species transcriptional comparisons. Simply put, it is now possible to identify the genes that behave similarly in mice and humans presuming that they are more fundamental to the studied process. We identified the genes that were increased in IPF lungs and that were on the mouse array. Most of these genes were concordant in their expression levels, including collagens I and III, tenascin C, IGFBP 2 and 5, and osteopontin. Osteopontin was the top increased gene in both mouse and humans. Tenascin X, IGFBP-5, and MN-SOD were discordant in their expression patterns. Data Verification and Generated Hypotheses Contrary to the current trend of validating microarray results by real-time RT-PCR, we prefer to do a biologically meaningful verification by determining the protein level of the gene of interest. If we need to verify global changes in gene expression, we advocate repeating at least a part of the experiment with another microarray platform (e.g., if the main platform is single-dye verify by a two-dye platform or vice versa). A more important validation step is an experiment designed to study the role of the protein in the process studies. This was the strategy that we applied to the study of the human IPF lungs. We verified the protein levels of MMP-2, MMP-7, and MMP-9 (see http://www.pnas.org/cgi/content/full/99/9/6292/DC1 and Ref. 15), and chose to evaluate the role of MMP-7 in fibrosis by exposing MMP-7 knockout mice to bleomycin. The knockout mice expressed a marked reduction in the fibrotic response in the lung, suggesting that MMP-7 had a role in pulmonary fibrosis (15). The complete dataset is available at http://fgusheba.cs.huji.ac.il/new_page_1.htm.
One of the most exciting aspects of microarrays is that they can be used as tools for actively introducing serendipity to one's research. For instance, we identified MMP-12 as one of the genes that was substantially increased in ß6-/- mice at baseline. We then analyzed the lung alveolar architecture in these mice and observed that they developed age-dependent spontaneous emphysema. Following up on this observation, a role for TGF-ß activation by the integrin Challenges and Future Directions In this review, we described the few studies that applied microarrays to the study of pulmonary fibrosis. The biologically meaningful information that we derived from our microarray experiments exceeded and continues to exceed our expectations. Furthermore, the free availability of our complete datasets serves as a continuous resource for new hypotheses and discoveries as well as new collaborative projects. In our view, to maximize the impact of microarrays on the research of fibrotic lung diseases, several challenges must be addressed. One challenge, previously mentioned in this section, is the need for data sharing and public availability of raw microarray data in a standardized format that will allow researchers from every discipline to mine the data (25). This will allow the generation of new global models of gene networks in pulmonary fibrosis that are based on multiple experimental models in a variety of organisms and are not limited to a single model or time point. It will also promote the generation of new and unexpected insights overlooked by the original investigators. Examples of such repositories are the ArrayExpress hosted by the European Bioinformatics Institute (http://www.ebi.ac.uk/arrayexpress/) and the Gene Expression Omnibus hosted by the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/geo/). Availability of well characterized human tissue (diseased and control) is a major challenge typical to microarrays experiments that use human lung samples. Large numbers of well-characterized samples are required in order reach statistically significant conclusions and to identify markers of early disease, response to therapy, and prognosis. These samples need to be obtained from patients with a variety of interstitial lung diseases, from patients in different stages of the disease, and from the same patient at diagnosis and transplantation. The controls most often used are normal histology lung samples from lungs resected for cancer. Despite a normal appearance on histologic examination, many of these samples express abnormal gene expression patterns reflecting a response to the presence of cancer in the lung or even infiltration by tumor cells. Special attention should be given to obtaining controls from a variety of resources. To address these issues, tissue consortia for tissue banking and tissue sharing between institutions must be established, and large multi-institutional datasets need to be created. Methods for obtaining, storing, and characterizing tissues, and for isolating RNA, DNA, and protein must be standardized. This will reduce the variability and provide a framework for comparing results between groups from different institutions. An additional challenge typical to the lung is the plasticity of the lung cellular content. Increases in leukocytes or alveolar macrophages, or changes in fibroblast proliferation rate or in epithelial cell differentiation, may seem like "real" transcriptional changes. Therefore, studies using primary cells generated from the lungs, or high-powered technologies such as laser capture microscopy, should be more widely applied. Several groups, including mine, are now in the process of characterizing IPF fibroblasts (C. Feghali Bostwick, D. Morse, and M. Selman, personal communication) using microarrays. Other investigators are successfully applying laser capture microscopy to IPF tissues (K. Gibson, personal communication). We believe that the creation of large multi-institutional gene expression datasets, application of more refined, cell-specific approaches, and the performance of cross-species comparisons will lead to better understanding of the molecular networks underlying pulmonary fibrosis. Such understanding will lead to identification of new targets for therapeutic interventions in lung fibrosis and design of better and more efficient drugs.
Acknowledgments Dr. Kaminski's work was supported in part by the Tel-Aviv Chapter of the Israel Lung Association. Footnotes This section was written by Naftali Kaminski (Dorothy P. and Richard P. Simmons Center for Interstitial Lung Disease, Pulmonary, Allergy and Critical Care Medicine, University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania). References
Pulmonary Fibrosis of SarcoidosisNew Approaches, Old IdeasThe typical features of pulmonary fibrosis in sarcoidosis are different from those of idiopathic pulmonary fibrosis (IPF)/usual interstitial pneumonia (UIP). In sarcoidosis, pulmonary fibrosis often begins in the mid and upper lung zones, in a slowly progressive process that results in upper lobe volume loss with hilar retraction, traction emphysema, and fibrocystic changes. Honeycombing may occur but is uncommon. Obstructive impairment is often more severe than the restrictive component, though both are usually present. Resting hypoxemia tends to occur late in the disease, unlike in IPF, in which hypoxemia is an early manifestation of the disease. These findings can largely be explained by the location of the granulomatous inflammation in pulmonary sarcoidosis, in which the granulomas are typically found in or around the bronchovascular bundles and perilobular spaces, consistent with a lymphangitic distribution (1).
There are few statistics on the frequency of pulmonary fibrosis in sarcoidosis. The prevalence of sarcoidosis ranges from 10–40/100,000 population in both the United States and Europe, with slightly higher rates in women than men (2). The lifetime risk has been calculated to be Pathogenesis of Granuloma Formation in Sarcoidosis
Granuloma formation begins with the tissue deposition of poorly soluble antigenic material. This material is phagocytosed or internalized by receptor-mediated endocytosis by mononuclear phagocytic cells, and processed into peptides that are then bound within the
In sarcoidosis, abundant evidence supports the concept that the granulomatous inflammation involves a highly polarized Th1 immune response, at least in the initial years of known disease (11). The Th1 cytokines IFN-
Evidence that this inflammatory response is driven by antigenic stimulation is provided by studies that demonstrate oligoclonal expansions of T cells expressing specific T cell receptor (TCR) gene segments at sites of disease in sarcoidosis (17). The best-studied example involves the expansion of V In addition to stimulation of Th1 cells, sarcoidosis is characterized by B cell stimulation. Immune complexes are detected in a majority of patients with sarcoidosis (21). Importantly, Lofgren's syndrome, the only subset of sarcoidosis whose onset of disease is known, is associated with formation of circulating immune complexes essentially 100% of the time (21). One hypothesis is that this humoral response may be key to the fact that Lofgren's syndrome is associated with remitting sarcoidosis in 80–90% of sarcoidosis patients by enhancing clearance of pathogenic antigens (22). B cell hyperactivity is also demonstrated by the hypergammaglobulinemia found in many cases of sarcoidosis. Pulmonary Fibrosis of Sarcoidosis Perhaps 50% or more of patients with known sarcoidosis undergo remission. In these patients, the granulomas may resolve, often leaving behind some residual scar tissue. In patients with persistent inflammation, the granulomas may develop fibrotic changes. The fibrosis usually begins at the periphery of the granuloma with gradual envelopment of fibrosis toward the center with hyalinization and collagen deposition (23). The determinants of this outcome are not understood.
There are no data on cytokine profiles in late stages of fibrotic sarcoidosis to assess the contribution of Th1 or Th2 cytokines to the fibrotic process. If the dominant Th1 responses seen in sarcoidosis are central to disease pathogenesis, then determining the role of these cytokines becomes critical to understanding the fibrotic outcome. This assessment is complicated by the pleomorphic effects of many of the cytokines involved, including the prototypical Th1 cytokine, IFN- Potential Role of Th1 Cytokines in Pulmonary Fibrosis of Granulomatous Lung Disease
Although there are no animal models that recapitulate the type of chronic progressive granulomatous inflammation seen in sarcoidosis, experimental models may be used to assess the potential role of cytokines in pulmonary fibrogenesis. In the murine model of bleomycin-induced pulmonary fibrosis, the inflammatory and fibrotic outcome has a genetic basis. "Bleomycin-susceptible" C57BL/6 mice develop intense inflammation and fibrosis following intratracheal bleomycin; interestingly, this strain tends to express Th1 cytokines (IFN- Overall, a role for Th1 cytokines should be considered in the pulmonary fibrosis of Th1-associated interstitial lung diseases such as sarcoidosis, hypersensitivity pneumonitis, pneumoconiosis, and chronic beryllium disease. Human studies are required to determine the net effects of these cytokines in regulating the fibrotic outcome in pulmonary sarcoidosis. Newer Approaches to Understanding Lung Fibrosis Insights into potential pathogenic mechanisms in the evolution of fibrogenic processes in sarcoidosis may be gleaned from new information on pathways relevant to lung fibrosis in experimental murine models and human IPF. For example, using oligonucleotide microarray gene expression analysis, matrilysin was identified as a major mediator of fibrosis in both murine and human lung fibrosis (29). Consistent with this finding, extensive nuclear accumulation of ß-catenin indicating activation of the Wnt signaling pathway was found along with upregulation of two of its downstream targets, matrilysin and cyclin-D1, by immunohistochemical analysis of IPF tissues (30). A role for abnormal re-epithelialization and lung remodeling in IPF was suggested by finding expression of truncated isoforms of the p63 gene (which counteract the apoptotic and cell cycle inhibitory functions of p53 after DNA damage) in IPF lungs (31). A role for TNF-enhanced TGF-ß1 expression in fibroproliferative lung disease was confirmed by a recent study that employed a replication-deficient adenovirus to overexpress TGF-ß1 in the lungs of TNF-receptor knockout mice that are resistant the fibrogenic effects of bleomycin, silica, and inhaled asbestos (32). The cell-surface adhesion molecule and hyaluronan receptor CD44 was shown to play a role in resolving lung inflammation in a murine model of bleomycin-induced pulmonary toxicity, suggesting that the CD44 pathway is important in moderating lung inflammation and subsequent fibrotic outcomes (33). Given the importance of these pathways in lung fibrogenesis, research to determine their role in fibrotic pulmonary sarcoidosis is clearly indicated. Treatment Approaches to Pulmonary Fibrosis in Sarcoidosis
Most clinicians view sarcoidosis as a treatable disorder that is usually responsive to corticosteroid therapy or other anti-inflammatory or immunosuppressive medications. Although unproven by rigorous controlled trials, corticosteroid therapy can improve symptoms and organ function over weeks to months and often years in most patients with sarcoidosis (2). The implications of these observations support the notion that granulomatous inflammation is responsible for the resultant fibrosis; i.e., fibrosis is not an independent process, but progresses as a result of ongoing inflammation and tissue injury. These clinical observations are valid whether Th1 or Th2 cytokine production is dominant in fibrocystic sarcoidosis. There are no data that support the effectiveness of other putative direct antifibrotic agents such as colchicine or perfenidone in fibrocystic sarcoidosis, and antifibrotic biologic agents such as IFN- A different strategic approach to suppress the chronic, progressive granulomatous inflammation in sarcoidosis is to reduce antigen deposition, enhance antigen clearance, or inhibit antigen processing and presentation. Therapies that enhance or induce remission of sarcoidosis would, of course, be an ideal way of preventing the complications of chronic progressive pulmonary fibrosis. Recent studies provide for a genetic and immunologic basis of remitting sarcoidosis that could potentially be exploited to enhance the likelihood of disease remission (22). HLA class II genes, particularly HLA-DR3 haplotypes and the linked DQB1*0201 haplotype, have been associated with presentations of sarcoidosis with favorable clinical outcomes or protective against severe sarcoidosis or disease progression (36, 37). In contrast, the DQB1*0602 allele and the closely linked DRB1*1501 have been associated with severe disease (37). Remitting sarcoidosis has been associated with downregulation of the immune responses with reduced TNF and enhanced TGF-ß production by sarcoidosis alveolar macrophages (38). These findings suggest that targeting specific MHC–T cell interactions may be a beneficial therapeutic strategy. In a similar context, it is reasonable to speculate that an outcome of remitting disease depends on clearance or tolerance of pathogenic tissue antigens. A dominant Th1-driven granulomatous response in the initial stages of sarcoidosis may be not be effective in clearing exogenous granuloma-inducing antigens unless accompanied by an effective humoral response. If this is true, then the immune response to pathogenic tissue antigens in remitting sarcoidosis may be associated with an effective disease-specific or healthy humoral (B cell) response, whereas chronic sarcoidosis may be associated with a dominant Th1 immune response that is ineffective in clearing relevant pathogenic tissue antigens. The possibility of regulating antigen-specific T and B cell responses to enhance the probability of disease remission is encouraged by the study of Grunewald and colleagues (39). These investigators found that higher proportions of AV2S3+ lung T cells in DR17(3)+ Scandinavian sarcoidosis patients are associated with an acute disease onset and a short (< 2 yr) disease duration (39). Given that these TCR-specific lung T cells may be specific for pathogenic sarcoidosis antigens, it may be possible to regulate this (or other) antigen-specific T cell subsets to either enhance Fc-mediated clearance of exogenous pathogenic antigens or T cell tolerance of disease-relevant autoantigens. However, until the specific pathogenic peptides are identified, it will be difficult to determine whether blockade or stimulation of the specific antigen/MHC/TCR-specific trimolecular complex will be therapeutically useful for enhancing disease remission. Identification of Candidate Pathogenic Antigens in Sarcoidosis The feasibility of identifying pathogenic tissue antigens in sarcoidosis has both a clinical and scientific basis in the Kveim-Siltzbach reaction (40). In this reaction, a nodular eruption develops 1–3 wk following the intradermal injection of sarcoidosis tissue homogenates in patients with sarcoidosis. Ansgar Kveim noted that histologic examination of skin biopsies of this reaction showed epithelioid granulomas that were essentially identical to granulomas in affected tissues. Louis Siltzbach used splenic suspensions and demonstrated the specificity of the reaction in sarcoidosis. The granulomatous inflammation in the Kveim-Siltzbach reaction site is infiltrated by CD4+ T cells, histiocytes, and mononuclear cells, similar to sarcoidosis tissues. More recently, these CD4+ T cells have been shown to be oligoclonal, consistent with an antigen-driven response (20). The granuloma-inducing component remains unknown. Since the reaction was first studied, investigators have hypothesized the granuloma-inducing compound in the Kveim-Siltzbach reagent was derived from an infectious agent. The kinetics of the reaction is similar to the Mitsuda reaction to lepromins in tuberculous leprosy, but no mycobacterial or other microbial remnants have been identified in validated Kveim suspensions (41). Studies have shown that the biologic activity is enriched in post-nuclear membrane fractions containing endolysosomes (40). The granuloma-inducing component is insoluble to neutral detergents and has relative resistance to heat, acidity, denaturing detergents, organic solvents, nucleases and proteases. Importantly, the granuloma-inducing activity is abrogated by potent denaturants (urea + ßME) suggesting a protein component is responsible for the reaction (42). These studies suggest that it may be possible to biochemically concentrate disease-relevant antigens in sarcoidosis tissues in at least a limited way. Using this information, we have initiated studies in an attempt to detect potential pathogenic antigens in sarcoidosis tissue. Our hypothesis is that sarcoidosis is caused by linked T and B cell immune responses to insoluble protein aggregates of microbial and/or endogenous origins. Our approach is based on using a limited proteomic approach together with immunoassays of sarcoidosis tissue extracts. Specifically, we hypothesized that antigenic proteins in sarcoidosis tissues can be enriched by biochemical means and detected using a combination of protein immunoblot techniques and mass spectroscopy. Frozen sarcoidosis tissues were biochemically processed to concentrate for tissue proteins that were insoluble in neutral detergents. Protein immunoblots of these extracts were analyzed using IgG from patients with sarcoidosis and control subjects. We have identified protein bands from sarcoidosis tissue that bind sarcoidosis but not control IgG. These bands have been excised and subjected to matrix-associated laser desorption/ionization-time of flight (MALDI-TOF) mass spectroscopy. Our preliminary results suggest that some of these protein bands contain microbial peptides. One of the proteins identified is the mycobacterial catalase-peroxidase gene (43). The B and T cell immune responses to recombinant protein derived from this mycobacterial gene are currently being analyzed to determine whether there are disease-specific responses to this protein in sarcoidosis. Conceivably, disease-specific immune responses to persistent microbial antigens may be associated with either disease remission or chronic progressive disease; these findings could then be used to design specific immunotherapy to amplify or suppress these responses (22). The approach outlined above offers one potential approach to the identification of pathogenic antigens in sarcoidosis tissues. This approach might also be generalizable to other granulomatous disorders of uncertain etiology based on the premise that poorly soluble proteins form a central nidus of granuloma formation and are also targets of an immune response in these disorders. Whether these proteins derive from remnants of microbial organisms or endogenous proteins, and whether they induce pathogenic immune responses, could be experimentally determined. Footnotes This section was written by David R. Moller (The Johns Hopkins University School of Medicine, Baltimore, Maryland). The work in this section of the supplement was supported in part by Grant No. HL68019 from the National Heart, Lung and Blood Institute, the Hospital for the Consumptives of Maryland (Eudowood), and the Life and Breath Foundation. References
Proteomic and Inducible Transgenic Approaches to Study Disease ProcessesLower respiratory track disorders often result in alteration of the normal architecture of the lung. A possible reason for this is abnormal repair of the injured lung. One of the hallmarks of abnormal lung repair is pulmonary fibrosis. Pulmonary fibrosis results from interactions between different cell types and biomolecules. Idiopathic pulmonary fibrosis (IPF) can be classified as a collection of fibrotic lung disorders of unknown etiology, and involves sequential biological processes (1). In early IPF, there is alveolitis dominated by macrophages and fewer numbers of neutrophils, lymphocytes, and eosinophils, and there is also an increase in the number of type II cells in the epithelium. In the intermediate phase of IPF, alveolitis persists with the thickening of the alveolar walls with fibrosis. In the late phase, there is a marked change in the normal architecture with inflammation and widening of the alveolar walls with fibrosis. Very little is known about the mechanism of this disease process. As with other complex diseases, there is no easy way to dissect the mechanism of this disease process. Here we have discussed two approaches that can be used in conjunction to study complex disease processes. We consider oxidative lung injury to be one potential insult that can initiate this disease process.There are three essential components in the approach that we have discussed here: (i) identification of differentially expressed proteins between normal lung and during development of lung disease by proteomic analysis; (ii) identification of the specificity of the differentially expressed protein in IPF in comparison with other inflammatory lung diseases; and (iii) identification of the biological consequence of overexpression or functional deletion of that protein using lung specific inducible transgenic mouse model. Proteomic Analysis of a Disease State Proteomics refers to the characterization of the proteome. The term proteome was first discussed in print by Wasinger and coworkers as the total protein complement of a genome (2). Although the genome of an organism is the same in all somatic cells, the proteome is quite diverse in different cell types in different tissues. Also, the same cell may have a different proteome under different physiologic conditions, or the same protein may undergo functional changes due to biochemical modifications like phosphorylation and acetylation. In a disease process, normal cells behave abnormally because of altered proteome or functional alterations in some proteins. Although characterization of this change is challenging, once this is accomplished it will have enormous impact on different areas, from diagnosis of diseases to drug development. The first step in proteome analysis is isolation of proteins from the source. The source could be cell-free fluids obtained from the body such as bronchoalveolar lavage (BAL) fluid (BALF) and serum, specific pathologically altered areas of a tissue and nearby unaffected areas of the tissue isolated by laser capture microscopy (LCM), or a specific cell type isolated by a specialized technique such as fluorescence-activated cell sorting (FACS). Depending on the source and method of separation, one can optimize the method of protein isolation. If one uses the same method for normal and diseased tissue with careful attention to enzymes that can either degrade or modify proteins such as proteases, phosphatases, and kinases, artifactual protein modifications stemming from isolation procedures can be minimized. The second step in this process is the separation of a complex mixture of proteins into individual proteins. Although this can be achieved by different methods, two-dimensional polyacrylamide gel electrophoresis (2-DE) can resolve very complex mixtures of proteins in a gel (3). In this method, the protein mixture is first separated in an isoelectric focusing (IEF) gel. IEF gels are cast with ampholytes to create a pH gradient within the gel and without any denaturing agents. During electrophoresis, proteins migrate to their isoelectric point (pI), the pH at which their positive and negative charges are equal. Because of this, all proteins with the same pI will focus at the same position in the IEF gel. In the second dimension, electrophoresed proteins in the IEF gel are further resolved in a second gel (polyacrylamide) according to their molecular weight. By selecting the appropriate IEF gel and the second dimension gradient gel, one can achieve excellent separation of a complex proteome. One critical point in this regard is that proteins with the same molecular weight and same pI will migrate to the same position. Also, the same protein modified by post-translational modification can be separated from the unmodified protein. After electrophoretic separation, proteins can be stained by a fluorescent stain such as SYPRO Ruby protein gel stain (Molecular Probes, Eugene, OR) for visualization. Once different samples are separated under identical conditions, differential expression or modification of proteins between samples can be assessed by laser scanning of 2-DE spots followed by computer-assisted spot recognition and characterization. After this analysis, potentially interesting protein spots can be removed by a robot using automated spot excision system followed by in-gel protease digestion, elution, and spotting on matrix-assisted laser desorption ionization time of flight mass spectrometer (MALDI-TOF MS) targets for the analysis of peptide mass. The peptide mass data is then matched against different data banks to get information about the protein from which the peptide is derived. Also, tools are available to obtain amino acid sequence of the peptides (4). The following is an example of proteomic analysis of a pathologic condition experimentally induced in mice. Proteomic Analysis of Oxidative Lung Injury Oxidant-induced lung injury initiates a cascade of events involving different cell types that induce widespread epithelial and endothelial injury and cell death together with airway inflammation, edema, and hemorrhage. To combat the toxic effects of inhaled oxidants, the respiratory tract elaborates a complex mixture of molecules that maintains homeostasis. Despite the presence of this innate network of antioxidant defense mechanisms, the lung can be overwhelmed by environmental oxidative stresses culminating in fulminant inflammation and tissue destruction. To analyze the changes in protein expression profile during oxidative injury, we exposed mice either to 100% oxygen or to room air as described previously (5). We then compared the proteins present in the BALF of animals exposed to hyperoxia with those present in the BALF derived from control animals exposed to room air. Identical lavage protocols were used for each set of animals.
Sample Preparation
2-DE
Image Analysis
Peptide Mass Fingerprinting
Inducible Protein Expression System in Transgenic Mice to Study Protein Function Once differentially expressed proteins are implicated in a disease process, the relevance of differential expression of specific proteins in the development of the disease phenotype or in the protection of the tissue from the disease process is an important issue. Inducible overexpression or repression of the protein in transgenic mice can provide some answers in this regard. The following is the description of an inducible tissue-specific gene expression system to study the effector function of a protein x (Figure 2) .
Generation of Lung-Specific Inducible Transgenic Mice Using a Dual Repressor-Activator System In this method, we generate transgenic animals by simultaneously microinjecting three constructs into the pronuclei of C57BL/6xSJL F2 mouse embryos. One is a construct encoding the tet transrepressor tTR, a hybrid protein containing the class B DNA-binding domain and class E dimerization domain of TetR (11) and the KRAB (Kruppel-associated box) repressor domain of the mammalian Kox1 protein, the second is the reverse tet transactivator rtTA (12), and the third is a construct containing full-length cDNA of the protein of interest linked to tetO/P. The tTR is expressed under the control of the ß-actin or ROSA26 promoter (13, 14), rtTA is expressed under the control of the Clara cell–specific CC10 promoter, which is active in lung epithelial cells, and the cDNA of the protein of interest is linked to a minimal CMV promoter and tet O/P sequences. All constructs are confirmed by sequencing. Linearized minigenes are separated from vector DNA and used for microinjection (detailed information of constructs is available from the author). Transgenic mice are characterized by PCR amplification of DNA isolated from tail biopsies. Effect of Keratinocyte Growth Factor Expression in Mice Exposed to 100% Oxygen The expression of endogeneous keratinocyte growth factor (KGF) is increased in oxidative lung injury (15). To address the function of increased expression of KGF, we have generated transgenic mice expressing KGF in an inducible, lung-specific fashion using the system described above. Using these mice, we are currently investigating several potential functions of KGF in the oxidative injury model. One question that we have asked in this context is "does KGF upregulate production of antioxidant molecules during oxidative exposure to protect the lung?" To address this question, transgenic mice or control nontransgenic mice were exposed to hyperoxia and the lungs of animals were lavaged. To characterize the redox status of the BALF, ESR detection of ascorbate radicals, the one-electron oxidation intermediate of ascorbate, generated upon addition of albumin/Cu complexes, was followed.
Electron Paramagnetic (Spin) Resonance Assay of Ascorbate Radicals in BAL Figure 3A shows the typical continuous repetitive recordings of EPR signals of ascorbate radicals generated by albumin/Cu from endogenous ascorbate in BAL samples from control mice and mice exposed to 100% oxygen. In the presence of albumin/Cu, a distinctive ascorbate radical EPR signal was detectable with a characteristic doublet splitting aH = 1.79 G (Figure 3A, traces 2–4). As expected, the Cu/albumin complex exhibited a remarkable catalytic redox-cycling activity toward ascorbate, as evidenced by the formation of its one-electron oxidation product, ascorbate radical. The greatest magnitude of the EPR signal and the shortest life-span of the radical signal was observed in BAL samples from KGF(-)/Dox(-) mice exposed to hyperoxia (Figure 3A, trace 4). The effect of hyperoxia was significantly less pronounced in BAL from KGF(+)/Dox(+) mice exposed to hyperoxia. In this case, the magnitude of the signal was significantly smaller and the decay of the ascorbate radical signal was much slower (Figure 3A, trace 3). In BAL samples from both KGF(-)/Dox(-) and KGF(+)/Dox(+) mice exposed to room air, Cu/albumin catalyzed oxidation of ascorbate and formation of its radicals proceeded at a slow rate resulting in a relatively weak signal that did not significantly change its magnitude over 4 min of recording (Figure 3A, trace 2). No detectable EPR signals were observed from BAL samples in the absence of albumin/Cu (Figure 3A, trace 1). A summary of measurements performed on different BAL samples from several mice is presented in Figure 3B. The rate of ascorbate radical decay after hyperoxia was significantly higher in the KGF(-)/Dox(-) BAL samples as compared with BAL samples from KGF(+)/Dox(+) mice. An even lower rate of ascorbate radical decay was detected in BAL samples of animals exposed to room air.
In conclusion, to understand the complexity of biological systems or the development of disease processes, characterization of the expressed proteins is an essential step. Establishment of protein expression profiles in different classes of fibrotic disorders will not only facilitate better diagnosis, but will also help identify potential drug targets. Expression of potential disease-related genes in mice, on the other hand, will aid in the development of murine models of human diseases. Ultimately, these murine models will be valuable tools for studying mechanisms of disease processes. Acknowledgments The writers of this section thank Drs. M. Gossen and H. Bujard for the plasmid pUHD172-neo containing the reverse tetracycline transactivator, Dr. W. Hillen for the plasmid pCMV-TetR(B/E)-KRAB encoding TetR, and Dr. J. Whitsett for the plasmid pCC10CAT-2300 containing the CC10 promoter. This section was supported by grants HL 69810 and HL 60207 (to P.R.) from the National Institutes of Health. Footnotes This section was written by Prabir Ray, Li Chen, Vladimir A. Tyurin, Valerian E. Kagan, and Frank A. Witzmann (Dorothy P. and Richard P. Simmons Center for Research and Education in Interstitial Lung Disease, Division of Pulmonary, Allergy, and Critical Care Medicine and Department of Immunology, Department of Environmental and Occupational Health, University of Pittsburgh School of Medicine and School of Public Health, Pittsburgh, Pennsylvania; and Department of Cellular and Integrative Physiology, Indiana University, Indianapolis, Indiana). References
Pulmonary Fibrosis in FamiliesFamilial idiopathic pulmonary fibrosis (FIPF) is identified by confirming IPF in two or more members of the same family. FIPF was described in the English literature fifty years ago (1–3), but has received scant attention since. Families with IPF have been described from all around the world, in forty-three reports mostly from Europe and North America (1).A Family with IPF with Sixteen Patients One of the earliest reports described twin sisters who both died of IPF at age 46, after living in separate geographic regions for the prior 25 years, suggesting that genetic factors were more important than environmental exposure (2). The following year, another report described a mother and adult daughter who both died from IPF (3). It was subsequently learned that these two studies reported different branches of the same family, and in 1965 Dr. Bonanni described a comprehensive investigation of this large family (4). Eight patients had developed IPF in this family by the time of his sentinel report, and there have been eight new cases since that time, making it the largest reported (5). Vertical transmission in four generations, along with father-to-son transmission, are present and indicate that FIPF is caused by a single autosomal dominant gene (Figure 1) . Ages of onset of IPF in this family ranged from 38–52 yr.
Familial IPF/Cryptogenic Fibrosing Alveolitis in the United Kingdom
The clinical and epidemiologic findings of 25 families identified in the United Kingdom were recently described (6). Adult pulmonary physicians in the United Kingdom were asked to identify all families under their care in which two or more individuals had been diagnosed with fibrosing alveolitis of unknown cause. Twenty-five families were identified, comprising 67 cases. The male:female ratio was 1.75:1. The mean age at diagnosis was 55.5 yr. Fifty percent of cases had been smokers and 18% had been diagnosed as asthmatic. Exposure to known fibrogenic agents was recorded by 36% of patients. Familial patients were younger at diagnosis, but otherwise indistinguishable from nonfamilial cases. The clinical, radiographic, histopathologic, and treatment outcomes are identical between sporadic and familial IPF (6). It is not known what percentage of all IPF is familial, but current reports estimate that it is Familial IPF Database at Vanderbilt University In our database of families with IPF we have collected clinical information from 76 families. One family has had 16 patients with IPF (Figure 1), another family has had 14 patients (Figure 2) , and 14 other families have had 5 or more patients with IPF. There have been 222 patients with IPF recognized among 2,134 individuals at risk in these families. Nearly 20% of IPF deaths in these families occurred before age 50 yr. The ethnic origins are predominantly Caucasian and Hispanic.
Vertical transmission, in up to four consecutive generations, is seen in the majority of our IPF pedigrees, and indicates an autosomal dominant mode. Father-to-son transmission is present in many families, and excludes X linked inheritance. These characteristics indicate that there is at least one autosomal dominant gene that causes IPF. It is possible that there may be more than one locus or more than one allele, or that IPF is more than one disease. In the Vanderbilt lung transplant program, the cause for end stage lung disease was IPF for 47 lung recipients. Nine of the 47 (19%) transplanted for IPF have a family history positive for interstitial lung disease. Prior Reports of Linkage of IPF to Chromosome 14
Two remote studies suggested that IPF may be related to genes on chromosome 14, using associations between IPF and Our Studies Exclude FIPF Linkage to Chromosome 14 In a recent collaborative study, we examined markers spaced across chromosome 14 to establish or to exclude the suggested linkage. We genotyped 70 individuals from 11 families with IPF for 14 markers spanning the length of chromosome 14 (unpublished data). The power of this dataset was marginal to confirm linkage, but it was sufficient to exclude linkage with confidence. Our results revealed that all two-point and multi-point LOD scores on chromosome 14 were negative or near zero. These results suggest that linkage of the total 11 families to a putative gene on chromosome 14, within 5 cM on either side of all, and within 10 cM on either side of eight of the 14 markers, can be excluded with a LOD of < -2. Surfactant Protein C Mutations as a Cause of Interstitial Lung Disease A recent report described a mutation in the surfactant protein C gene (SFTPC) that was associated with nonspecific interstitial pneumonitis (NSIP) in an infant whose mother had desquamative interstitial pneumonitis (DIP) (11). Heterozygous G to A transition of the first base of intron 4 (IVS4+1 G to A) was present in both patients, and caused skipping of exon 4 with deletion of 37 amino acids. Symptoms were not present at the time of birth in either patient, but both developed interstitial lung disease (ILD) later as infants. The IVS4+1 G to A mutation in SFTPC was identified on only one allele of both patients, consistent with an autosomal dominant pattern (11). The mutation resulted in the production of an abnormal proprotein, and the levels of transcripts encoding normal SP-C precursor protein were similar to those of transcripts encoding the abnormal protein. Mature surfactant protein C is derived through the proteolytic processing of a 197–amino acid proprotein (or a 191–amino acid proprotein with alternative splicing). Surfactant protein C precursor protein is an integral membrane protein that is anchored in the membrane by the hydrophobic core of mature surfactant protein C. The IVS4+1 G to A mutation caused skipping of exon 4 and the deletion of 37 amino acids in the carboxy-terminal domain of surfactant protein C precursor protein (11). Deletions in this domain have been shown to disrupt the intracellular transport of surfactant protein C precursor protein. Interactions between normal and abnormal surfactant protein C precursor protein could impair the transit of normal surfactant protein C precursor protein through the processing pathway or enhance its degradation. The lack of mature surfactant protein C in lung tissue and bronchoalveolar lavage fluid from the patient supports the idea that the precursor protein was not being processed and secreted normally (11). A Different Mutation in SFTPC Causes IPF in 14 Patients in One Family Because of the report of an SFTPC mutation causing NSIP and DIP in a newborn child and mother (11), we chose SFTPC as a candidate gene, and then tested it in the largest IPF family in our database (12). This kindred (Figure 2) spans five generations, contains 97 total members, including 11 adults with IPF, including 6 with biopsy-proven usual interstitial pneumonitis (UIP), and 3 affected children (III:5, V:1,V:2) with cellular nonspecific interstitial pneumonitis. We screened this multiplex FIPF kindred for mutations in SFTPC. A microsatellite marker (SFTP2) located 9 kilobases from SFTPC was used to perform linkage analysis on this family. LOD scores were calculated assuming an autosomal dominant mode of inheritance of a single gene with a disease allele frequency of 0.0001. Analysis was done assuming that unaffected individuals were not disease gene carriers, using phenotype information on only affected individuals. A LOD score of 4.33 at a recombination fraction of 0.00 was generated between FIPF and the marker, which is highly significant and confirms a definite relationship of this marker to the disease in this family (12). We screened patient DNA specimens by dideoxyfingerprinting (ddF) using a primer scanning approach, which yielded an abnormal ddF pattern from polymerase chain reaction (PCR) fragments containing exon 5 sequences derived from DNA template of three affected family members (12). Sequencing of these DNA fragments revealed a heterozygous exon 5 +128 T to A transversion that substitutes glutamine for leucine at the highly conserved amino acid position 188 of the carboxy-terminal region of SP-C precursor (proSP-C) protein. Restriction analysis then confirmed the mutation was present in DNA from all available affected family members (12). The mutation was not present in 88 control chromosomes. Immunohistochemistry was performed by immunostaining for pro SP-C; in normal adult alveolar Type II cells, it showed focal brown staining of the cytoplasm adjacent to lamellar bodies (12). Lung from affected patients with IPF in this family when immunostained for pro SP-C showed a very abnormal distribution of staining, with diffuse brown cytoplasmic staining in cuboidal type II cells, and absence of identifiable lamellar bodies (12). Mouse lung epithelial cells were transfected with plasmids containing normal and mutant SFTPC (12). Pooled stable lines were grown, then supernatants and lysates of 105 viable nontransfected, wild-type, or mutant SFTPC-transfected cells incubated for 24 h were assayed for cellular toxicity. Mutant SFTPC-transfected cells displayed sluggish growth rates compared with nontransfected and wild-type SFTPC-transfected cells, taking several days longer to grow to confluence on culture plates (12). The percentage of nontransfected, wild-type SFTPC-transfected, and mutant SFTPC-transfected cells displaying cytotoxicity (percentage of cytotoxicity = LDH activity supernatant/LDH activity cell lysate) was 5.8 ± 1.4%, 4.6 ± 2.3%, and 11.9 ± 1.9%, respectively (P < 0.05 for comparison between both nontransfected and mutant cells and between wild-type and mutant cells). The carboxy terminal domain plays a critical role in trafficking and processing of proSP-C (13), because deletional mutants of this region remain localized in the endoplasmic reticulum/Golgi body without proper proteolysis (14). Thus, we predict the carboxy-terminal SFTPC mutation we discovered may cause misfolding of SP-C and/or SP-C deficiency with subsequent type II cell injury and alveolar instability. Mechanisms of Disease from SFTPC Mutations It appears that severe interstitial lung disease can result from either the absence of SP-C, or alternatively from production of an abnormal proSP-C protein (15). Selective absence of expression of pro SP-C and the active SP-C protein was described in a family with interstitial lung disease recently described (16). Also, an inbred strain of SP-C (-/-) mice developed severe ILD with features similar to those seen in patients with ILD (15). SP-C deficiency also causes lung disease in Belgian White and Blue cattle, a strain in which some newborn calves may develop respiratory distress associated with selective deficiency of SP-C (17). The deficiency of SP-C in surfactant could cause abnormal shear forces in the alveoli, thereby causing mechanical injury of the respiratory epithelium, which in turn may contribute to the pathogenesis of IPF. Missense or short deletion mutations, as seen in the studies by Nogee (11) and in our studies, result in the production of a stable mRNA that produces an abundance of a misfolded protein that may escape from protein quality control systems. Accumulation of the abnormal pro SP-C protein or protein complexes may cause type II epithelial cell injury. Further, the expression of a mutant SP-C protein directly caused a lethal lung disorder in transgenic mice (18), providing support for the concept that mutations in the SP-C gene-caused misfolding and misrouting of pro SP-C, may contribute to the pathogenesis of lung disease in mice and patients expressing mutant proSP-C peptides. Thus, both the presence and absence of proSP-C can be associated with lung disease. Clinical Features of SFTPC Mutations Information describing the clinical expression of SFTPC mutations is growing progressively, but it is already clear that these mutations can cause interstitial lung disease in children and adults. Several pathologic forms of IPF, including NSIP, DIP, and UIP, have been reported related to mutations in SFTPC. The clinical course of some individual patients with ILD due to SFTPC mutations appears to be quite long, sometimes even for decades. Anecdotal evidence suggests that common viral respiratory infections may trigger the clinical presentation. Further specific information about the clinical manifestations of SFTPC mutations is awaited with great interest, and surely will be developed in the next few years. Summary of Familial IPF Clinical findings, histopathology, and clinical course are indistinguishable between familial IPF and sporadic cases (6). Patients with familial IPF may be younger at diagnosis, but are otherwise indistinguishable from nonfamilial cases (6). Our experience in which we have collected 76 families with IPF at a single medical center suggests that IPF in families may be far more common than the 2% estimate (6) that is currently reported, perhaps even 10-fold higher. Nineteen percent of patients who received lung transplantation for IPF at our institution have a positive family history. It is possible that the incidence of familial IPF in our referral population is not a representative sample. The subsequent development of new cases in other members in families of patients formerly believed to have sporadic IPF suggests that some sporadic patients may have a genetic basis. The identification of the genetic basis of IPF in families may be the most effective method to identify the central pathogenetic features of IPF. Acknowledgments This work was supported by an Intramural Discovery Grant from Vanderbilt University on familial idiopathic pulmonary fibrosis. Footnotes This section was written by James E. Loyd (Vanderbilt University Medical School, Nashville, Tennessee). References
Gene Profiling and Kinase Screening in Asbestos-Exposed Epithelial Cells and LungsPulmonary fibrosis is a progressive and chronic inflammatory lung disease characterized by epithelial cell injury, mesenchymal cell (fibroblast, myofibroblast) proliferation, and extensive remodeling of the lung parenchyma (1, 2) . Lung remodeling may involve epithelial and inflammatory cell interactions with mesenchymal cells that lead to excessive accumulation of extracellular matrix (ECM), lung dysfunction, and fatality (1–3). Although the pathogenesis of pulmary fibrosis, which may exhibit a number of pathologies, is poorly understood, a variety of cytokines, chemokines, and regulators of apoptosis have been implicated in its development and progression (1–6). The failure of anti-inflammatory drugs to effectively treat this disease (7), as well as data that show inflammation per se is not intrinsic to the development of fibrosis (8, 9), suggest that other mechanisms such as epithelial cell injury and repair are important.Epithelial cell responses may be key to initiation of inflammation as well as regulating homeostasis of the ECM (10). It is becoming increasingly clear that maintenance of ECM is a dynamic process in which the synthesis of proteins such as fibrillar collagens, fibronectin, and proteoglycans is normally balanced by similar rates of proteolysis. Protein turnover in the ECM is mediated mainly by a class of proteases known as matrix metalloproteinases (MMPs) and their inhibitors (TIMPs) (11). The role of cell signaling pathways in eliciting cell injury, apoptosis, proliferation, and inflammation has been studied in a number of organs and cell types, but little is known about epithelial cell signaling and its relationship to the development of pulmonary fibrosis. In studies here, we used a murine nontransformed type II epithelial cell line, C10 (12), to characterize gene and protein expression by oligonucleotide microarray analysis (Affymetrix, Santa Clara, CA) and kinase profiling after exposure to the carcinogenic and fibrogenic mineral, crocidolite asbestos, for 8 and 24 h. In addition, we examined alterations in gene expression in whole lung homogenates of C57/BL6 mice at 3 d after inhalation of crocidolite asbestos at concentrations inducing proliferation of bronchiolar and alveolar epithelial cells (13). These acute periods of exposure were selected to identify genes and signaling proteins that may be involved in the initiation of epithelial cell injury and proliferation by asbestos. Materials and Methods
Cell Culture and Reagents
Kinase Screening
Microarray Assays on C10 Cells
Real Time RT-PCR
Microarrays on Mouse Lungs C57/BL6 mice (12 wk old) were exposed in inhalation chambers to clean air (shams) or NIEHS crocidolite asbestos (2 mg/m3 air, 6 h/d for 3 d) as described previously (13). Inflammation and focal interstitial fibrosis subsequently occur in this murine model of asbestosis. After preparation of RNA from lung samples (50 µg/sample) (13), RNA from two sham and two asbestos-exposed mice were examined for hybridization to Affymetrix mouse chips, and results analyzed using Affymetrix GeneChip software as described above. Results
Kinase Protein Screening Indicates that Multiple Pathways Are Involved in Asbestos- and EGF-Induced Effects on Pulmonary Epithelial Cell Proliferation
Exposure to EGF caused increases in many of the same signaling proteins that were observed with asbestos (Figure 1C). The observation that many of the same protein expression patterns were seen with EGF at early time points (30 min and 4 h), and asbestos at later time periods (8 and 24 h) may reflect differences in solubility and dose effects by these stimuli. EGF is soluble and affects all cells immediately after its addition to cultures, whereas asbestos fibers are insoluble and require time to precipitate upon cells. Thus their effects are more protracted and localized to cells in areas of deposition of fibers.
Microarray Data Show a Common Subset of Genes Altered in Asbestos-Exposed Pulmonary Epithelial Cells and Lungs after Inhalation of Asbestos
Genes Linked to Cell Signaling and Fibrogenesis Are Upregulated in Pulmonary Epithelial Cells and Lungs Exposed to Asbestos In Tables 2–5, we present the designations and ontologies of genes upregulated in C10 cells and/or in lungs after exposure to asbestos. These are categorized for presentation here as genes involved in cell signaling, transcription factors, mitogenesis- or growth-related genes, and those regulating ECM homeostasis and lung remodeling.
A number of cell signaling genes were upregulated (Table 2), suggesting activation of multiple signaling pathways by asbestos and reinforcing the results of kinase screening assays. These included genes involved in G protein–coupled receptor and mitogen-activated protein kinase (MAPK) signaling, integrin-mediated signaling pathways, and cyclin-dependent kinases and their inhibitors. Other genes were linked to survival pathways (rsk, phosphatidylinositol 3-kinases) and cell surface receptor–cytokine pathways such as the urokinase plasminogen activator (uPA) receptor (uPAR), fibroblast growth factor, and transforming growth factor (TGF)-ß pathways. Several genes encoding transcription factors linked to MAPK signaling pathways, most notably the activator protein 1 (AP-1) components, FBJ osteosarcoma oncogene (Fos), Fos-like antigen 1, and JunB, were also upregulated by asbestos (Table 3). In addition, elevated expression of genes encoding transcription factors associated with altered proliferation and cell cycle control, e.g., E2F transcription factor 3, were observed.
Gene analysis also identified several growth factor–related genes encoding proteins that have been classically linked to the fibrogenic effects of asbestos or silica (6) and other models of fibrosis including connective tissue growth factor (15), insulin-like growth factor 1 (IGF-1) (16, 17), and TGF-ß1 (16, 18). (Table 4). A number of other factors associated in other organs and cell types with angiogenesis, cell cycle regulation, and proliferation, were also revealed.
Table 5 shows induction by asbestos of several genes linked to cytoskeletal organization/biogenesis, ECM regulation, and lung remodeling. Many of these, e.g., metalloproteinases (19), integrins (20), and the plasminogen-related group of genes (21–24), have been linked to the processes of fibrogenesis and repair, whereas other factors such as protein tyrosine phosphatases, serine proteases, plectin, and hydrolases need further characterization in lung. Table 6 shows genes upregulated by asbestos in C10 cells that are linked to the prevention or development of apoptosis (bcl, caspase 14), inflammation (GRO1 oncogene, phospholipase A2), and antioxidant responses (CuZn-SOD). These changes are consistent with previously published observations showing that asbestos induces apoptosis through mitochondrial pathways (25) and oxidative stress (26). Moreover, asbestos-induced inflammation and fibrosis in a rodent inhalation model can be ameliorated by administration of antioxidants (27).
Real-Time Q-PCR and Microarray Analysis Show Similar Trends of Expression for Selected Genes in Asbestos-Treated C10 Cells Table 7 presents the fold increases in expression of three genes (junB, fra-1, fos) associated with AP-1 activation, a known regulator of cell proliferation and responses to toxicants in the lung (28), and cd44, a gene recently linked to the resolution of inflammation after exposures to bleomycin (29), after comparative analysis using microarrays and Real-Time Q-PCR in C10 cells. The Real Time Q-PCR results validate the trends found after analysis of microarray data.
Discussion Although the pathogenesis of pulmonary fibrosis is complex and incompletely understood, evidence suggests at least two critical routes influence its development, an inflammation-linked pathway and an epithelial cell pathway involving cross-talk with inflammatory and mesenchymal cells (3). Both routes trigger a number of chemokines/growth factors that induce fibroblast migration/proliferation, phenotypic changes to myofibroblasts, and subsequent accumulation of ECM. In this study, we provide new information on early epithelial responses to the fibrogenic mineral, asbestos, using transcriptional profiling and kinase screening on murine epithelial cells in culture (C10). We also analyzed global gene expression in mouse lungs after acute inhalation of asbestos fibers. These complementary approaches provide hypothetical models of cell signaling events that can be tested using transfection techniques in vitro and transgenic mouse models using constructs to disrupt these pathways and target them to lung epithelium. Figure 3 shows a diagram of asbestos-induced signaling events that is compiled from our past knowledge regarding the importance of the EGFR and ERK1/2 pathways in asbestos-associated cell injury and proliferation (14, 30–34) and new data from transcriptional and kinase profiling studies in C10 cells. The fact that asbestos fibers can activate the EGFR (30, 32, 35) and integrin-associated receptors (26), as well as calcium-mediated signaling pathways (36, 37) and PKCs (38–40), all of which activate the ERK1/2 cascade, suggests that blocking these events downstream is critical to modulating epithelial cell proliferation. This hypothesis is presently being tested in our laboratory using CC10-dnMEK1 transgenic mice and transgene-negative littermates exposed to asbestos fibers.
The fact that asbestos induces many cell-signaling pathways, transcription factors, and growth-related genes in epithelium and lung points to the complexity of these factors in the induction of fibrogenesis. Crosstalk between growth factor pathways may also be critical in determining fibrogenic versus repair responses. For example, like EGF, TGF-  also binds to the EGFR, and mice overexpressing TGF- targeted to lung epithelium using the CC10 promoter develop pulmonary fibrosis (41). If TGF- induces proliferation, like EGF, in lung epithelial cells, this observation is incongruous with the view that epithelial cell regeneration is thought to be a vital repair mechanism in fibrosis (42).
Tumor necrosis factor (TNF)- TGF-ß1 also induces expression of IGF-1 and CTGF, both of which were increased in C10 cells after exposure to asbestos (15). In addition to its role as a potent mitogenic polypeptide, IGF-1 is antiapoptotic to fibroblasts, thus inducing fibroblast proliferation and transcription of collagen and laminin genes (45). Constitutive and TGF-ß1–induced expression of IGF-1 is higher in fibroblasts from fibrotic lungs, and levels of IGF-1 are higher in patients with pulmonary fibrosis. The fact that increased fibrosis can be inhibited by antibodies to IGF-1 is encouraging (16). Our studies also reveal the potential importance of CTGF production, whose overexpression has been confirmed as important in fibrosis (15), by lung epithelial cells. In fibroblasts, CTGF expression is induced by TGF-ß. TGF-ß1 and CTGF are coordinately overexpressed during wound repair in a rat model of wound healing, a process that requires both fibroblast proliferation and ECM deposition (45) Abnormal matrix deposition and lung remodeling are fundamental features of fibrosis, and the many genes increased in expression by asbestos (Table 5) illustrate the complexity of these processes. Increased expression of many of these gene products such as fibronectin and integrins governing activation of TGF-ß1, laminin, and ligation of fibrin and fibrinogens have been related to fibrosis (20). Other proteins, such as CD44, play an important role in resolving lung inflammation and removal of ECM breakdown products (29). A rapidly evolving field has elucidated mechanisms by which fibrin turnover is altered in lung injury, fibrosis, and neoplasia. Tissue factor expression is increased in the lung, initiating coagulation and movement of coagulation substrates into the injured alveolar space, and potentiating thrombin generation and fibrin formation. Locally depressed fibrinolysis attributable to inhibition of plasminogen activator (PAI-1) often occurs concomitantly with increased procoagulant activity and promotes local fibrin deposition (46). Mice with deletion of the PAI-1 gene develop less fibrosis, and those constitutively overexpressing a PAI-1 transgene develop more fibrosis after exposure to bleomycin (47). We have shown previously that asbestos induces complex changes in the fibrinolytic cascade, including induction of urokinase plasminogen activator (uPA) and its receptor (uPAR) (48). Mice deficient in uPA or tPA also demonstrate increased pulmonary fibrosis after bleomycin-induced lung injury (46). These studies suggest complex interrelationships between these pathways and mitogenesis of fibroblasts versus epithelial cells. Gene expression profiles in mice exposed to asbestos were far more complicated than those observed with type II cells in vitro. We expect that this outcome is a result of the complex series of events that occur by 3 d of exposure in mice; by this time epithelial cell injury, immune responses, and even compensatory proliferation are evident at focal sites. No doubt many cell types participate in these processes, and substantial effort will be required to describe the functional consequences of changes in gene expression in specific cell types, and how these influence the events that culminate in fibrosis. For example, induction of the antiproliferative cyclin-dependent kinase inhibitor p21 in response to asbestos is not likely to occur in the same cells that express cyclin E1, which is linked directly to cell proliferation (Table 4). In conclusion, transcriptional and protein profiling, especially to elucidate gene products that are post-transcriptionally regulated, provide sensitive tools for revealing new candidates for modulation of epithelial cell participation in fibrosis. These approaches and data confirmation using Real Time –Q-PCR, Western analyses, and kinase activity assays, provide information on pathways to be targeted for prevention of epithelial cell injury and fibrosis.> Acknowledgments This section was supported by grants ES/HL09213 (BTM) and PO1 HL67004 (BTM) from the National Institutes of Health. Footnotes This section was written by Maria E. Ramos-Nino, Nicolas Heintz, Luca Scapoli, Marcella Martinelli, Susan Land, Norma Nowak, Astrid Haegens, Brian Manning, Nicole Manning, Maximilian MacPherson, Maria Stern, and Brooke Mossman (Department of Pathology, University of Vermont College of Medicine, Burlington, Vermont; Applied Genomics Technology Center, Wayne State University, Detroit, Michigan; and Microarray and Genomics Facility, SUNY Buffalo School of Medicine and Biomedical Sciences and Roswell Park Cancer Institute, Buffalo, New York). References
The Importance of Sarcoidosis Genotype to Lung PhenotypeA genetic predisposition to sarcoidosis is becoming more evident. A recent family study in the United Kingdom has identified a risk ratio ( s) value for siblings of familial sarcoidosis of 36–73, indicating significant familial clustering of the disease, and the biggest case-control study on sarcoidosis from the United States (A Case Control Etiologic Study of Sarcoidosis) showed an odds ratio of 5.8 in siblings, 5.7 in avuncular relationships, 5.2 in grandparents, and 3.8 in parents (1, 2).
Genetic studies of sarcoidosis are becoming more prevalent. Two broad categories of studies have been reported: those involving the major histocompatibility complex (MHC), and those targeting non-MHC regions. MHC genes so far studied include the human leukocyte antigens (HLA) Class I + II genes, transporter associated with antigen-processing (TAP) genes, and tumor necrosis factor (TNF) (3–5). These studies, and a genome–wide search for predisposing genes in German sarcoidosis families (6), have provided inescapable evidence for a susceptibility locus for sarcoidosis somewhere in the MHC region of chromosome 6. Moreover, a number of non-MHC–related gene studies has provided evidence for other genes contributing to sarcoidosis susceptibility, including chemokine receptor (CCR) 2, CCR5, interleukin (IL)-1 HLA-DR and -DQ in Sarcoidosis
As the initiation of sarcoidosis is thought to begin with the presentation of an as yet unknown antigen by dendritic cells or macrophages, the HLA Class II genes are of particular interest for the pathogenesis of sarcoidosis. HLA molecules bind antigenic peptides within a groove formed by two Although some studies have reported associations between sarcoidosis and HLA Class I, Class II associations have been reported most frequently. The majority of these associations are found with HLA-DR and -DQ. Importantly, these reported associations tend to be with either disease susceptibility/chronicity or with a milder disease phenotype. Susceptibility/chronicity-related DR associations have been found in Japanese patients with sarcoidosis (HLA-DR5 [DRB1*11], -DR6 [DRB1*14], and -DR8 [DRB1*08]), in German patients (HLA-DR5), and in patients from Sweden (HLA-DR14 and DR15) (14–17). DQ susceptibility associations have been found in Japan (DQB1*0601), Sweden (DQB1*0201/0202), and Germany (DQB1*0603 and *0604) (17–19). DR associations with milder disease mainly involve DR3, and are reported for patients with a Swedish background (HLA-DR3 [17]), and also Polish patients (HLA-DRB1*03 and DRB3*0101) (17, 20). DR Profiling One of the earlier studies to highlight the dichotomy between alleles that confer likely chronicity or resolution was reported by Berlin and colleagues (17). They showed strong DR associations with two disease phenotypes: a susceptibility/chronicity phenotype and a milder disease phenotype. They investigated 122 Scandinavian patients with sarcoidosis and 250 healthy volunteers from the same ethnic background. All were typed genomically for HLA-DR, -DQA1, and -DQB1 alleles. The results showed that 33% of patients with nonchronic sarcoidosis, i.e., those who recovered within 2 yr, were HLA-DR3 (17)-positive by comparison with 17% of control subjects, whereas patients with chronic sarcoidosis had significantly increased frequencies of DR14 (18% compared with 6% in control subjects) and DR15 (60% compared with 30% in control subjects) (17). Recently, Foley and coworkers determined HLA-DRB1 alleles in three cohorts of white patients with sarcoidosis from the United Kingdom (n = 189), Poland (n = 87), and Czech Republic (n = 69), and confirmed the associations found between HLA-DRB1*14(DR14) and -DRB1*15(DR15) and sarcoidosis susceptibility in all populations (3). In addition, HLA-DRB1*3(DR3) carrier frequency was found to be increased in Czech patients with sarcoidosis, and -DRB1*12(DR12) in British patients. Furthermore, they showed that another HLA-DRB1 allele, DRB1*01(DR1), was consistently protective for sarcoidosis in all three populations, and also consistent with results from previously published Scandinavian, Italian, and Japanese case–control studies. Furthermore, the carrier frequency of HLA-DRB1*04(DR4) was found to be significantly reduced in British patients, suggesting an additional protective allele for sarcoidosis, which is also consistent with results previously published for Scandinavian, Italian, and Japanese sarcoidosis. From these studies, covering six populations, therefore, a DR profile emerges; i.e., DR1 and DR4 can be classified as "protective alleles," and DR3, DR12, DR14, and DR15 as "susceptible" alleles. The finding that certain HLA-DR alleles act as protective markers for sarcoidosis in a number of populations from different background suggests importantly that this protective effect is independent of ethnic or geographic background. When HLA-DRB1 allele protein sequences were compared to identify shared residues forming pockets within the peptide-binding groove of the HLA-DR complex, remarkably, all alleles identified as being protective for sarcoidosis, encoding DR1 and DR4 antigens, were found to share small hydrophobic residues at position 11. By contrast, the susceptibility alleles did not share these hydrophobic residues, but instead had a hydrophilic residue at this position. The residue at position 11 is within a pocket of the HLA-DR complex antigen-binding groove (designated P6), where it is the only variable amino acid and may determine the peptide binding preference of this pocket (3). The findings imply that specific peptide binding is determined by amino acid sequences in the binding groove, and that variability in this binding may be important in initiating sarcoidosis. The basis for this protective effect of hydrophobic residues at HLA-DRB1 position 11 in sarcoidosis is, however, unknown at this time, but the alterations in the number of H2O molecules in P6 might be of relevance, as these might influence the strength of the MHC binding of the potentially sarcoid-triggering peptide(s). Other important data have emerged from the multicenter ACCESS study, which has recently reported the findings on HLA typing at recent international meetings. High-resolution HLA typing was performed for HLA-DPB1, -DQB1, -DRB1, and -DRB3, and low-resolution typing for HLA-DRB4 and -DRB5 to explore possible HLA associations with sarcoidosis. This study comprised the biggest sarcoidosis cohort typed for HLA to date, including 474 patients and matched control subjects. The results showed that HLA-DRB1*1101 was significantly associated with sarcoidosis in both African Americans and in whites, again confirming a genetic predisposition to sarcoidosis, and again identifying the HLA-DRB1 locus as a major locus for disease susceptibility in this disease. Importantly, this study confirmed HLA-DRB1*1501(DR15) and *1201(DR12) as being susceptibility alleles, and HLA-DRB1*04(DR4) as a protective allele. A summary of the susceptibility and protective alleles is shown in Table 1.
Genome-Wide Screen for Candidate Genes
Schürmann and colleagues have performed a genome-wide search for predisposing genes in sarcoidosis. They used microsatellite linkage analysis to identify chromosomal regions that contribute to the risk of sarcoidosis. The investigators chose 225 microsatellite markers, covering the whole of the genome, but with an average spacing as high as 19.6 cM (
The most prominent peak at the MHC has more recently been shown to be closer to the class II region, in keeping with numerous reports of associations between gene variants in this region and sarcoidosis. Interestingly, the peak at chromosome 3 is in a region (although at 8 cM distance) containing the chemokine receptor genes CCR2 and CCR5, of which polymorphisms have been associated with sarcoidosis in two populations (Czech and Japanese) (7, 8). One of two minor peaks at chromosome 7 (D7S3070) attracts attention because it is close to the T cell receptor (TCR) B gene cluster. Further, the gene encoding transforming growth factor-ß receptor 1 (TGFBR1) is located in the area of the peak at chromosome 9. Although an interesting candidate gene in sarcoidosis, however, no significant association studies of TGFBR1 have been reported so far. Finally, the gene encoding interleukin-2 receptor (IL-2R) Despite the limitations of this study, especially the low-density screen, likely causing insufficient resolution to identify all markers of interest, it provides very useful information for those involved in sarcoidosis genetic research and should encourage further fine mapping of the areas identified. Sarcoidosis: Concepts of Severity In clinical practice, it is clearly recognized that there are large differences in presentation of sarcoidosis and prognosis, and that some clinical characteristics strongly relate to disease severity. In this regard, erythema nodosum (EN) and chest radiographic stage I are known to be associated with a milder disease phenotype, which is often spontaneously remitting and has a favorable prognosis. In combination with joint symptoms, EN and stage I chest radiography are grouped into a syndrome, Löfgren's syndrome, that generally has an excellent prognosis (22). Other extrapulmonary disease manifestations such as uveitis, cardiac disease, central nervous system involvement, and stage II/III disease represent a more severe phenotype with a more likely chronic course (23). Interestingly, mild and severe disease phenotypes often do not evolve into each other, but appear to retain the same phenotype from first presentation. This concept of at least two clinically different severity phenotypes in sarcoidosis was recently given more genetic basis in a high-resolution HLA-DQB1 typing study by our group (24). We studied the relationship between 19 HLA-DQB1 alleles and sarcoidosis severity and progression in 133 white patients with sarcoidosis and 354 control subjects from the United Kingdom, and 102 patients and 214 control subjects from The Netherlands. Disease severity was evaluated by chest radiography, pulmonary function tests, and extrapulmonary disease, including uveitis and central nervous system disease, at presentation, and progression, measured by chest radiographic follow-up at 2 and 4 yr. We found that HLA-DQB1*0201 was strongly associated with milder disease manifestations (EN, Löfgren's syndrome, and stage I chest radiography at presentation, 2 and 4 yr), and protective against severe sarcoidosis (uveitis, chest radiograph stage II or greater, and diffusing capacity < 80% predicted). Furthermore, the DQB1*0201 allele was also strongly associated with a reduced risk of disease progression (improved or stable stage I chest radiograph, and no progression or persistent stage II/III disease) regardless of corticosteroid treatment. Remarkably, another HLA-DQB1 allele, *0602, showed susceptibility associations with more severe disease. Importantly, the above findings are consistent with previous HLA-DRB1 association studies. The HLA-DQB1*0201 allele is in tight linkage disequilibrium with HLA-DRB1*03 that has been associated with good prognosis in several sarcoidosis cohorts. In addition, the findings that the HLA-DQB1*0602 allele is associated with more severe disease are consistent with the Scandinavian studies described above as this allele is closely linked to the HLA-DRB1*15 allele. Therefore, taken together, there is now good evidence that there are at least two HLA haplotypes that are very influential on different sarcoidosis phenotypes: (i) HLA-DRB1*0301/DQB1*0201 determines mild-type disease with remitting course, and (ii) HLA-DRB1*15-DQB1*0602 determines more severe, nonremitting-type disease. However, there might be other relevant genotype–susceptibility/phenotype associations, either MHC-based or involving other, yet to be identified, loci. Sarcoidosis, a Polygenic Disease Although some clear HLA genotype–susceptibility/phenotype associations have been described, many patients have different HLA haplotypes. Therefore, we are presently extending our studies to a large MHC typing study, including the major HLA loci. At this time > 500 sarcoidosis patients are included in the study. Interim analysis shows some potentially important results. When all positive allelic HLA associations are reclassified as either "susceptible" or "protective" alleles, almost half of all patients carry either a protective or susceptible or both allele(s) (our unpublished data) (Figure 1) . Interestingly, the observed effects of some of these alleles appear to be age-dependent (our unpublished data). Although these HLA protective/susceptibility–sarcoidosis associations are highly significant, it also has to be recognized that roughly half the patients do not have evidence for an HLA contribution to the pathogenesis of their disease. This observation confirms the conceptualized polygenic nature of sarcoidosis, and highlights the importance of studying other candidate genes, either MHC-associated (e.g., TNF) or located in other chromosomal areas (e.g., the areas identified by Schürmann and colleagues). In this respect, we have recently studied a series of different TNF promoter polymorphisms, at position –1031, –863, –857, –307, and –237 in 96 British and 100 Dutch patients with sarcoidosis, and 354 British and 222 Dutch control subjects (5). The results showed a significant increase in the rarer TNF –857T allele in both sarcoidosis populations. In total, 25.5% of the patients with sarcoidosis carried the TNF –857T allele, compared with 14.1% of the control subjects (P = 0.0003, pc = 0.002; population attributable risk percentage 13.3%). Interestingly, subgroup analysis showed a significant increase of the rarer TNF –307A allele in patients presenting with Löfgren's syndrome, and a decrease of the TNF –857T allele in this subgroup. Haplotype construction of the investigated polymorphisms revealed six haplotypes, of which haplotype-2 contained the –307A allele and haplotype-4 the –857T allele. This study of genetic variants in the TNF promoter, therefore, provides further support that mild-type sarcoidosis is genetically distinguishable, i.e., associating with haplotype-2 and not -4 as occurs in more persistent disease.
Conclusion Evidence is accumulating that there are strong genotype–phenotype relationships in sarcoidosis. Of these, specific alleles are associated with disease susceptibility/chronicity (HLA-DRB1*03, *11, *12, *14, *15, and DQB1*0602 and TNF –857T), and others are clearly protective or associated with mild disease (HLA-DRB1*01, 03, 04, DQB1*0201, and TNF –307A). The DQB1*0201 allele is in linkage disequilibrium with DRB1*0301 and TNF –307A, and it is this haplotype that is almost invariably associated with mild-type disease, i.e., sarcoidosis presenting as Löfgren's syndrome or stage I disease only. The true protective locus on this extended MHC haplotype, however, still needs identification. This will need much larger cohorts of patients with sarcoidosis than those studied to date, or DRB1*0301-deficient cohorts, most likely to be found in populations from a different ethnic background. Fine mapping across the MHC region is also required, including genes such as the HLA-B-associated transcript (BAT) 1. Furthermore, we need fine mapping of important candidate chromosomal areas outside the MHC, and confirmation of previously described associations in genes including the chemokine receptors CCR 2 and 5. In conclusion, is sarcoidosis a single disease or not? There appear to be at least two clearly different disease phenotypes, one mild-type/remitting and one chronic/progressive, which split now can be given genetic backing, implying a different pathogenesis that might relate to differences in trigger factors and/or their processing by the mononuclear phagocyte system. Footnotes This section was written by Jan C. Grutters, Hiroe Sato, Kenneth I. Welsh, and Roland M. du Bois (Clinical Genomics Group, Imperial College of Science, Technology and Medicine, National Heart and Lung Institute & Royal Brompton Hospital, London, United Kingdom). References
Cytokine Phenotypes Serve as a Paradigm for Experimental Immune-Mediated Lung Diseases and RemodelingThe pathologic presentation of a number of chronic lung disease is often associated with an inflammatory response, with subsequent fibroproliferation and deposition of extracellular matrix. Many chronic lung disorders share common characteristics, including an unknown etiology, ill-defined mechanisms of disease progression, inability to effectively resolve, and end-stage fibrosis. Unfortunately, these progressive chronic lung diseases are usually refractory to treatment and are associated with substantial morbidity and mortality. The inability to identify efficacious therapeutic options to treat the actively progressing forms of these lung disorders likely reflects the scientific community's limited mechanistic understanding of these disorders. However, investigative inroads have been made that suggest that cytokine networks are operative in dictating the progression of these diseases. For example, it is known that various cytokines can promote and maintain the chronicity of inflammation by sustaining the recruitment of leukocyte subpopulations and inducing fibroblast activation, proliferation, and collagen deposition during the maintenance of chronic lung disease.
Clearly, the etiology of chronic lung disease, which possesses a fibrotic outcome, is diverse and includes such insults as radiation injury, cytotoxic drugs, particulates, and idiopathic events. However, we present a working model based on immune dependent cytokine phenotypes that appear to be important in dictating the progression of lung inflammation and end-stage outcome. The involvement of various cytokines in the initiation and maintenance of chronic immune-mediated lung disease, which eventually may mature to end-stage fibrosis, may be directed by a sequence of host cytokine responses that have gone awry. Under a normal host defense paradigm, it is likely that the initial cell-mediated reaction involves the expression of However, if the initiating antigen or pathogen is not cleared by this cell-mediated immune response, the host enters a transition phase, which is characterized by the appearance of either a hybrid cytokine phenotype or a totally different cytokine phenotype. The subsequent host response is represented by the expression of cytokines, which would fall under the rubric of a type 2 immune process with accompanying levels of the prototypic cytokines interleukin (IL)-4, IL-5, and IL-13. The significance of this new mix of mediators to the host defense lies in the fact that a different type of immune process is now available to aid in attacking and clearing the antigen or pathogen. The switch to a more sophisticated immune response allows the host to mount a continued response with renewed vigor of the host, as antibody switching occurs and an additional leukocyte, the eosinophil, is cast into the inflammatory mix. It is also recognized that this phenotype consists of cytokines that ultimately activate resident fibroblasts and cause these cells to proliferate and deposit collagen. Thus, if the antigen continues to persist and escape the grasp of the different cytokine-directed responses, the final cytokine phenotype targets and activates the resident fibroblasts to proliferate, deposit collagen ("lay down bricks and mortar"), and wall the inciting agent away from the host. With the above paradigm in mind, the end-stage pathology observed in chronic fibroproliferative lung disease may be due to dysregulation of this final walling-off process. The persistence of the causative agent, coupled with the continued long-term expression of type 2 cytokines, drives fibroblasts to a highly active state and serves as the underpinning for end-stage disease. Thus, experimental models of chronic lung inflammation, which are characterized by either a type-1 or a type-2 response, would be useful in delineating the mechanisms that maintain and resolve chronic lung inflammation. These experimental systems will prove to be especially important to investigate, as the degree of inflammation and fibroblast activation during the pathogenesis of chronic pulmonary inflammation may be dependent upon a polarized expression of type-1 and type-2 cytokines during the evolution of the disease. The Type 1/Type 2 Paradigm of Tissue Remodeling
One of the most common worldwide diseases, which is dominated by type-2 cytokines and eventual end-stage fibrosis of target tissue, is schistosomiasis. This helminth parasitic infection induces a chronic cell-mediated inflammatory process characterized by high levels of IL-4, IL-5, IL-10, and IL-13, with corresponding low levels of IFN-
Interestingly, the treatment of murine schistosomiasis with exogenous IFN- Additional data support the concept that the chronicity of certain cell-mediated responses in vivo ultimately results in pathology due to a characteristic type-2 cytokine profile (6). Murine models of chronic graft-versus-host disease, as a result of experimental bone marrow transplant, have been characterized by hypergammaglobulinemia, high levels of IgE, immune complex deposition in tissues, and elevated concentrations of the type-2 cytokine, IL-4 (7). When mice with bone marrow transplant-graft-versus-host disease were treated with neutralizing IL-4 antibodies, IgE levels dropped, immune complex–induced lesions resolved, and splenomegaly was prevented. Interestingly, cyclosporine A, an agent known to suppress type-1 cytokine responses, caused an exacerbation of bone marrow transplant graft-versus-host disease in these models (8). Clinical studies assessing longitudinal alterations in cytokine levels and corresponding fibroproliferative changes in lung pathology are difficult to perform in human bone marrow transplants with subsequent graft-versus-host disease. However, it is known that fibrosis and associated cell proliferation associated with bronchiolitis obliterans may be a consequence of the transplant. One of the more compelling pieces of information, which may link the expression of type-2 cytokines to the evolution of chronic fibrosis, is the association of fibroblast activation and the presence of eosinophils (9). A number of studies have demonstrated that asthma and parasitic infections are associated with both Th2 cytokine expression (IL-4 and IL-5) and a profound eosinophilia, as IL-5 is both an eosinophilopoeitic and chemotactic factor for eosinophils. Although the mechanistic role of eosinophils and type-2 cytokines has been demonstrated in asthma and parasitic infections, the role of these cells and type-2 cytokines in other disease states is not as clear. However, in vitro experiments have shown that eosinophils are capable of a time-dependent release of factors that stimulate human lung fibroblasts to undergo replication and synthesize extracellular matrix. The interactions between fibroblasts and eosinophils appear to be rather complex, as fibroblast-conditioned media has also been shown to prolong the survival of eosinophils. Nonetheless, studies have identified an increase in eosinophils in association with fibrotic changes in pulmonary fibrosis. Thus, a potential fibrotic network, leading to end-stage pathology, may be established between the triumvirate of type-2 type cytokines, eosinophils, and fibroblasts. Using unique models of experimental lung granuloma formation defined by a characteristic cytokine phenotype, which may share characteristics of specific immune, antigen-driven responses, we have assessed many of the above processes that are outlined in Figure 1 .
Type 2 Cytokines Induce Profibrotic Cytokine Cascades Information regarding the role of IL-13 and TGF-ß as key mediators in the fibrotic process stems from recent studies using either a transgenic expression system for the overexpression of IL-13 or an adenovector-mediated gene transfer system for the overexpression of TGF-ß1 (10–12). The overexpression of an IL-13 transgene under the control of the Clara cell 10-kD protein promoter in the lungs of mice demonstrated a unique model system to assess the in vivo biology of this cytokine. The overexpression of IL-13 resulted in an eosinophilic inflammation, mucus hypersecretion, globlet cell hyperplasia, and subepithelial airway fibrosis. Because IL-4 and IL-13 are related cytokines, are both located on chromosome 5, and may have resulted from gene duplication during evolution |
. Several studies in both animals and humans have suggested that IFN-
1% in a Scandinavian population (
, focal adhesion kinase [FAK], etc.) of which have been confirmed by Western blots and kinase activity assays. An 8-h exposure to asbestos fibers caused increases in Raf-1, an upstream activator of MEK1/2 and the ERK1/2 pathway, PKC
