American Journal of Respiratory and Critical Care Medicine

Rationale: Chronic obstructive lung disease (COPD) is a common and disabling lung disease for which there are few therapeutic options.

Objectives: We reasoned that gene expression profiling of COPD lungs could reveal previously unidentified disease pathways.

Methods: Forty-eight human lung samples were obtained from tissue resected from five nonsmokers, 21 GOLD (Global Initiative for Chronic Obstructive Lung Disease) stage 0, 9 GOLD stage 1, 10 GOLD stage 2, and 3 GOLD stage 3 patients. mRNA from the specimens was profiled using Agilent's Functional ID v2.0 array (Agilent, Santa Clara, CA) containing 23,720 sequences.

Measurements and Main Results: The gene expression pattern was influenced by the percentage of the sample made up of parenchyma. Gene expression was related to forced expiratory flow between 25 and 75% of forced expiratory volume (FEF25–75% % predicted) revealing a signature gene set of 203 transcripts. Genes involved in extracellular matrix synthesis/degradation and apoptosis were among the up-regulated genes, whereas genes that participate in antiinflammatory responses were down-regulated. Immunohistochemistry confirmed expression of urokinase plasminogen activator (PLAU), urokinase plasminogen activator receptor (PLAUR), and thrombospondin (THBS1) by alveolar macrophages and airway epithelial cells. Genes in this pathway have been shown to be involved in the activation of transforming growth factor (TGF)-β1 and matrix metalloproteinases and are subject to inhibition by SERPINE2. Interestingly, both TGF-β1 and SERPINE2 have been identified as candidate genes in COPD genetic linkage and association studies.

Conclusions: The results provide evidence that genes involved in tissue remodeling and repair are differentially regulated in the lungs of obstructed smokers and suggest that they are potential therapeutic targets.

Data deposited in GEO at

Scientific Knowledge on the Subject

There have been few genomewide interrogations of gene expression in the lungs of smokers with and without chronic obstructive pulmonary disease. The separation of genes that are induced by smoking versus those that are induced in smokers who develop airflow obstruction is unclear.

What This Study Adds to the Field

The genes involved in tissue remodeling and repair are differentially regulated in the lungs of obstructed smokers and suggest that they are potential therapeutic targets.

Chronic obstructive lung disease (COPD) is characterized by airflow limitation that is not fully reversible: the airflow limitation is usually progressive and associated with an abnormal inflammatory response of the lungs to noxious particles and gases (1, 2). This abnormal chronic inflammatory immune response affects both airways and parenchyma and is characterized by thickening, fibrosis, and narrowing of the small conducting airways, and emphysematous destruction of the lungs' elastic properties (3).

Although it is generally agreed that inflammation plays a role in both the parenchymal and airway components of the disorder, traditional antiinflammatory therapy has proved unsuccessful in modifying the progression of the disease (4). New treatments for COPD under development include antagonists for leukotriene B4, 5-lipooxygenase, IL-8, tumor necrosis factor (TNF)-α, and inducible nitric oxide synthase. In addition, a panel of new antiinflammatory reagents, such as inhibitors of phosphodiesterase 4, nuclear factor-κB, p38, phosphoinositide 3 kinase (PI3K), and neutrophil elastase are being actively pursued as potential therapeutics (5). Although several of these have been tested in clinical trials, none of them has emerged as an effective drug.

The lack of success of traditional therapies suggests the need for a new paradigm and novel approaches to treatment. One method for generating new ideas about the pathogenesis of complex genetic disease is to perform genomewide transcriptional analysis on diseased tissue for comparison with normal tissue. This unbiased approach allows hypothesis generation unrestrained by existing schemas for pathogenesis but generates large volumes of data that are challenging to analyze and interpret. In addition, careful selection of subjects and samples is required to obtain reproducible results. A number of investigators have used the technologic advances in genomewide expression platforms to compare the lungs of patients who have COPD with those of nondiseased subjects (68).

In the present study, we took advantage of a well-characterized group of patients in whom detailed lung function, lung morphometry, and frozen lung tissue were available to examine whole lung tissue gene expression patterns using the Agilent microarray technology (Functional ID, ver 2.0; Agilent, Santa Clara, CA). By using quantitative morphometric analysis to determine the tissue-type content of the actual samples, we found a subset of genes that was related to the percentage of the sample made up of parenchyma versus airways and blood vessels. We also derived a list of 203 genes whose level of expression was related to lung function (forced expiratory flow between 25 and 75% of forced expiratory volume [FEF25–75% % predicted]) after correction for age, sex, and smoking. Among other genes, this approach identified urokinase plasminogen activator (PLAU), urokinase plasminogen activator receptor (PLAUR), and thrombospondin 1 (THBS1) as potential targets in COPD. Immunostaining of the same tissue samples showed expression of PLAU, PLAUR, and THBS1 in alveolar macrophages, alveolar epithelial cells, and the epithelium of the small airways. The small airway epithelium of smokers with airflow obstruction showed more extensive staining for PLAUR than the epithelium of the nonobstructed smokers.

Comparison with previously published studies of genomewide transcriptional analysis (68) showed little overlap for differentially expressed genes. This lack of reproducibility suggests that factors such as disease severity, sample acquisition, tissue type, and expression platform influence the results. A standardization of such factors will be necessary for meaningful comparisons across studies.

Some of the results of this study have been presented previously in abstract form at the American Thoracic Society annual meeting (9).

Subject Selection

The patients for gene profiling were selected from a larger group of patients who required pneumonectomy or lobectomy for small peripheral lung nodules and who are part of an ongoing investigation of lung structure and function (10, 11). Criteria for inclusion were as follows: that the total volume of the lung lesion, as calculated from radiographs and/or computed tomographic scans, was less than 120 ml, and that the lesions did not obstruct subsegmental or larger airways. Patients who had radiographic or pathologic evidence of obstructive pneumonitis were excluded. Approximately 731 subjects fulfilled these criteria and gave informed consent to have their clinical and lung function data, together with their resected tissue, examined in research studies using methods approved by the Providence Health Care (Vancouver, BC, Canada) clinical ethics review board. Of these subjects, 184 had their resected lung tissue frozen and archived in a manner suitable for gene expression studies. From these 184 subjects, we selected 74 subjects to represent a range of smoking and lung function. We randomly selected 6 of the 12 nonsmokers in the biobank and randomly sampled approximately one-third of those in the lower GOLD (Global Initiative for Chronic Obstructive Lung Disease) categories (GOLD 0–1), approximately one-half of those in GOLD 2, and all three of the subjects in GOLD 3. Of these, 48 were found to have mRNA of suitable quality for gene expression profiling (Figure 1).

The 26 subjects whose RNA did not pass quality-control criteria did not differ from those included in the study with respect to age, male/female ratio, FEV1% predicted, or pack-years smoked (63.4 ± 2 yr vs. 62.8 ± 2 yr, 11/15 vs. 17/31 male/female ratio, 86.8 ± 4 vs. 86.5 ± 3% predicted, and 34.4 ± 6.3 vs. 39.1 ± 3.2 pack-years, respectively; P > 0.2 for all comparisons). In addition, there was no difference in the time from tissue collection to RNA extraction for those samples that yielded high- or lesser quality RNA.

Before surgery, subdivisions of lung volume, spirometry, and single-breath diffusing capacity were measured as previously described and according to American Thoracic Society (ATS) standards. (10) A modified ATS questionnaire was applied to gather demographic and clinical information. A detailed smoking exposure was determined and expressed as pack-years. On the basis of smoking history and lung function, subjects were classified as lifetime nonsmokers or into the GOLD categories of COPD severity (3). We expressed FEF25–75% as a percentage of the predicted value (FEF25–75% % predicted) (12) to use as a continuous variable for comparison with gene expression data. Table 1 shows the number of patients in each GOLD category and their mean FEF25–75% % predicted.



Age (yr)



FEV1% predicted

FEV1/FVC (%)

FEF25–75% % predicted

Nonsmokers548 ± 83/2O92 ± 382 ± 1100 ± 155 Carcinoid
GOLD stage
 02165 ± 213/836 ± 499 ± 376 ± 190 ± 46 Squamous
10 Adeno
4 Large cell
1 Small cell
 1963 ± 46/351 ± 590 ± 365 ± 150 ± 34 Squamous
3 Adeno
1 Large cell
1 Small cell
 21064 ± 37/351 ± 766 ± 258 ± 128 ± 21 Squamous
5 Adeno
1 Large cell
2 Poorly differentiated
1 Amyloid
 3367 ± 52/149 ± 746 ± 0.153 ± 519 ± 11 Adeno (BAC)
1 Squamous

1 Carcinoid

Definition of abbreviations: Adeno = adenocarcinoma; BAC = bronchoalveolar cell carcinoma; FEF25–75% % predicted = forced expiratory flow between 25 and 75% of forced expiratory volume; GOLD = Global Initiative for Chronic Obstructive Lung Disease.

Values are means ± SE. An analysis of variance showed no significant difference in age by GOLD category but significant differences for pack-years, FEV1% predicted, FEV1/FVC%, and FEF25–75% % predicted (P < 0.0001 for all comparisons). Among smokers, there was no significant difference in pack-years by GOLD category. GOLD 0, 14/21 current smokers; GOLD 1, 6/9 current smokers; GOLD 2, 9/10 current smokers; and GOLD 3, 2 current smokers and 1 status unknown. Current smoker = within 6 mo of surgical resection.

Tissue Processing

Immediately after resection, the lung or lobe was obtained from the operating room, and after the clinical specimens of the lesion, lymph nodes, and the resection margin were obtained, the lobes and lungs were inflated using a 50% mixture of Cryomatrix (Shandon, Pittsburgh, PA) and saline and frozen in liquid nitrogen fumes. The frozen lungs and lobes were then cut into 7 to 15 2-cm-thick slices using a band saw (Sears, Hoffman Estates, IL) and multiple randomly stratified blocks were acquired (1–3/slice) using a power-driven hole saw (Hobart, Troy, OH) fitted with a 1.5-cm diameter bit (11). The frozen “cores” were stored at −80°C for later cryosectioning and RNA extraction (see the online supplement for details).

Frozen sections were obtained from the surface of the half core immediately adjacent to the portion used for RNA extraction. The 10-μm sections were stained with hematoxylin and eosin and a digital image of the entire sample was captured. The fraction of the tissue area occupied by air and tissue (% parenchyma) was determined and the tissue content was further divided into alveolar tissue, membranous airway, cartilaginous airway, and blood vessel tissue using a standard point-counting method (13).

Microarray Study Design and Methodology

RNA from 8 of the 21 GOLD 0 subjects (nonobstructed smokers) was pooled to form the reference RNA (the 8 subjects were chosen because enough RNA was available from their lung sample). The reverse-transcribed cDNA from this reference was used for competitive hybridization against all of the other samples, including all of the GOLD 0 samples that made up the reference pool. Microarray profiling was done for 23,720 sequences as previously described (14) and as outlined in the online supplement.


Frozen human lung specimens were cut at 10-μm thickness onto slides and stored at −70°C until used. For PLAUR and THBS1 staining, the APAAP (alkaline phosphatase–antialkaline phosphatase) method was used, whereas for PLAU the ABC (avidin-biotin-peroxidase complex) method as described by Boenisch (52) was used. Details are included in the online supplement.

Image Analysis

For PLAU and PLAUR, all images were captured using a Nikon Eclipse E600 microscope (Nikon Corp., Tokyo, Japan) fitted with a SPOT camera. The image analysis was performed using Image-Pro Plus 4.0 (Media Cybernetics, Silver Spring, MD). Alveolar macrophages positive for PLAU and PLAUR were expressed as a percentage of total macrophages, and the length of the membranous bronchiolar epithelium that stained positive for these proteins was expressed as a percentage of the total epithelial length. Only selected sections were stained for thrombospondin. Details are included in the online supplement.

Statistical Analysis

To identify genes correlated to lung function we used a continuous variable, the post-bronchodilator FEF25–75% % predicted. This measure of lung function provided the largest range of values, and as is shown in Table 1, related well to GOLD stage. Among demographic and clinical variables that were potentially predictive of FEF25–75% % predicted (age, sex, smoking duration), we found that only exposure to cigarette smoke, expressed as pack-years, was significantly negatively related (r = −0.5, P < 0.0005) as shown in Figure 2. Thus, in selecting genes related to FEF25–75% % predicted, we corrected for any relationship with pack-years by fitting the data to a multivariate linear model:

Where Y is gene expression for each gene, β0 is the intercept, β1 and β2 are slope coefficients for FEF25–75% % predicted and pack-years, respectively, and ε is random noise with zero mean and standard deviation σ. We were primarily interested in the genes that showed a significant relationship to FEF25–75% % predicted adjusted for pack-years. The significance of this relationship is given by the P value for the slope of FEF25–75% % predicted as well as overall P value of the model. Because microarrays do not provide absolute readouts of gene abundance, the model intercept is strongly related to the type of hybridization pool used and is therefore of no specific interest. Because there was a significant influence of the tissue content (% parenchyma) on gene expression profile, we also used a model in which % lung parenchyma is included as well as FEF25–75% % predicted and pack-years.

The extent of PLAU and PLAUR staining was compared with FEF25–75% % predicted and pack-years using linear regression analysis and, in addition, the mean % positive macrophages and % positive airway epithelial length in subjects who had a FEF25–75% % predicted of less than 60% predicted were compared with those who had a FEF25–75% % predicted greater than 60% predicted using Student's t test.

TaqMan Validation

A number of the genes whose level of expression proved to be related to the level of lung function (COPD signature genes) were tested using TaqMan real-time polymerase chain reaction. Details of the TaqMan protocol are available in the online supplement.

Study Subjects

Forty-two of the subjects had lung resection for bronchial carcinoma, five subjects had carcinoid tumors, and one subject had resection of an amyloid lesion. Table 1 shows demographic and clinical data for the 48 subjects on whom mRNA profiling was done. The nonsmokers were younger than the smokers, but there was no difference in age between GOLD categories. Most of the subjects were in the GOLD 0, 1, and 2 categories. This distribution is consistent with our archived samples and reflects the fact that the majority of heavy smokers do not develop severe airflow limitation and that severely obstructed subjects are not candidates for lung resection.

Tissue Samples

All of the tissue samples showed good expansion, allowing accurate quantification of tissue compartments. The tissue-type content of the samples varied. Some were pure parenchyma, whereas others showed substantial percentages of airways and blood vessels. The mean content of “air” was 69 ± 13% (range, 18–87%) and the remainder was made up of tissue that was subdivided into parenchymal tissue, membranous bronchi, cartilaginous bronchi, and blood vessels larger than capillaries. The parenchyma made up 75 ± 19% (range, 7–95%); membranous bronchioles, 2.5 ± 2.9% (range, 0–11%); cartilaginous bronchi, 1.7 ± 4% (range, 0–19%); and blood vessels, 19 ± 15% (range, 0–68%), of the total tissue area. Of the 48 samples, 30, 12, and 47 contained some membranous bronchiolar, cartilaginous bronchial, and/or blood vessel tissue, respectively.


All 48 samples passed Rosetta's quality control (see the online supplement for details). Microarray hybridizations comparing all 48 lung samples to the control pool of 8 GOLD 0 nonobstructed smokers showed 3,222 sequences with a significant difference in expression at a fold-change difference of 1.5 or greater (P ⩽ 0.01 in more than 5 patients using the Rosetta error model) (15). Two-dimensional cluster analysis, however, did not result in a coclustering of samples derived from patients of the same GOLD stage. Approximately half (22/48) of the samples contained a signature that was highly enriched with tracheal genes, obtained from monkey and human body gene expression atlases (data not shown).

The effect of the percentage of the tissue sample made up of lung parenchyma on gene expression was assessed for all transcripts by correlating individual transcripts to the % parenchyma. A total of 5,159 genes correlated with % parenchyma with a correlation coefficient r ⩾ +0.3 or ⩽ −0.3 (see the online supplement). The significance of this correlation, P ⩽ 0.05, was evaluated using a Monte Carlo permutation (14). Eighty-seven percent of the genes in the “tracheal signature” subset were represented in this category. These data indicate that the gene expression profile of individual tissue samples was influenced by their tissue content; samples with a low parenchymal content, and therefore a higher content of airways and blood vessels, contained a unique subset of “tracheally” expressed transcripts.

Lung Function–related Gene Expression Patterns

Two hundred three genes whose level of expression was related to lung function were selected on the following basis: (1) a level of expression average across all experiments with log10 intensity greater than −1.5, abs(log ratio) greater than 0.2, (2) Rosetta error model P value of less than 0.05 in four or more experiments, and (3) a P value less than 0.02 for the independent relationship of expression level with FEF25–75% % predicted as well as a P value less than 0.02 in the model that incorporated FEF25–75% % predicted and pack-years. Figure 3 shows a heat map of these “COPD signature” transcripts and their relationship to FEF25–75% % predicted and GOLD stage. A list of the 203 COPD signature genes in rank order of their relationship with FEF25–75% % predicted is available in the online supplement. A sample relationship between FEF25–75% % predicted and the level of expression of PLAUR, an example of one of the COPD signature genes, is shown in Figure 4.

To examine if tissue content confounded the relationship between gene expression and lung function, we also analyzed a model that included % parenchyma as a variable. The inclusion of % parenchyma did not substantially impact the slopes and P values for the relationships between the level of expression of the 203 transcripts and FEF25–75% % predicted (see the online supplement).

TaqMan Validation

To validate the results of the microarray, 8 genes selected from among the 203 COPD signature genes were subjected to TaqMan real-time quantitative polymerase chain reaction in three to seven subjects using the same RNA sample that was used in profiling (CXCL1, IL8, THBS1, PLAU, PLAUR, CCL2, BMPR2, and MMP9). Although the level of expression relative to the pooled reference sample was systematically higher using TaqMan, the relationship was highly concordant. The equation relating the two methods was as follows: microarry value = 0.69 TaqMan value + 0.48 (R2 = 0.63) (for details, see Table E1 and Figure E1 in the online supplement).


To select promising candidate genes for further analysis, we developed a scoring scheme that included the following factors: (1) the fold difference in mRNA levels between obstructed and unobstructed smokers, (2) the analysis of variance P values, (3) the abundance of the mRNA, (4) location of the gene in a locus linked to a COPD phenotype, (5) altered expression in more than one candidate in a pathway, (6) the biologic plausibility that the candidate could contribute to the pathogenesis of a COPD, (7) a reported COPD-like phenotype in a mouse model if overexpressed or knocked out, (8) altered expression in one or more inflammatory disorder other than COPD, and (9) attractiveness as a therapeutic target. On the basis of this schema, we selected PLAU, PLAUR, and THBS1 for validation at the protein level using immunohistochemistry.

Immunohistochemical staining for PLAU and PLAUR was performed on sections from each of the 48 lung samples that were profiled. Staining for thrombospondin was performed on a few selected specimens. Figure 5 shows representative sections illustrating that both PLAU and its receptor were strongly expressed in alveolar macrophages and in bronchiolar epithelial cells. Staining was also detected, to a lesser extent, in alveolar surface epithelial cells (not shown) and circulating inflammatory cells within the capillaries (Figure 6). The majority of the macrophages expressed both PLAU (68 ± 32%) and its receptor (67 ± 25%). Similarly, most of the length of the bronchiolar epithelium stained positively for both PLAU (83 ± 25%) and PLAUR (76 ± 29%). The extent of staining for PLAU and PLAUR was correlated in both macrophages (Pearson correlation coefficient = 0.59 for % positive) and epithelial cells (Pearson correlation coefficient = 0.40 for % positive). There was no significant correlation between FEF25–75% % predicted or pack-years and the number, or percentage, of alveolar macrophages positive for PLAU or PLAUR. Similarly, there were no significant correlations between the percentage of the length of the bronchiolar epithelium that was stained for PLAU and PLAUR and the measurements of lung function or smoking. However, when subjects were divided into those whose FEF25–75% % predicted was less than and more than 60%, the obstructed smokers showed greater bronchiolar epithelial staining for PLAUR (Table 2).


Bronchiolar Epithelial Staining

Alveolar Macrophage Staining

PLAU (% positive)
PLAUR (% positive)
PLAU (% positive)
PLAUR (% positive)
FEF25–75% % predicted < 60%89 ± 2890 ± 1668 ± 3369 ± 23
FEF25–75% % predicted > 60%80 ± 2269 ± 3268 ± 3265 ± 26
P value

Definition of abbreviations: FEF25–75% % predicted = forced expiratory flow between 25 and 75% of forced expiratory volume; PLAU = urokinase plasminogen activator; PLAUR = urokinase plasminogen activator receptor.

Values are mean ± SD.

The 60% predicted cutoff is arbitrary, but the rationale was that an FEF25–75% % predicted of less than 60 is abnormal. The 95% confidence lower limit in the normal population is approximately 70% predicted. In addition, this divided the group into two with a reasonable number in each group: 21 versus 27.

Selected sections were stained for thrombospondin and the pattern of staining was similar to that for PLAU and PLAUR. Alveolar macrophages and membranous bronchiolar epithelium were positive with no staining of other cell types (data not shown).

Network Analysis of COPD Signature Genes

The 203 genes that were related to lung function were uploaded into the Ingenuity Pathways Analysis tool (Ingenuity Systems, for gene network search. Figure 7 shows a portion of the resulting network. This analysis showed a number of pathways containing two or more differentially expressed genes that are biologically plausible candidates for the pathogenesis of COPD. For example, PLAU, PLAUR, and THBS1, three genes involved in the activation of TGF-β1 and matrix metalloproteinases (MMPs) (16, 17), were significantly increased in individuals who had lower values for FEF25–75% % predicted. PLAUR resides on chromosome 19 in a locus linked to early-onset human COPD in a recent study (18). Epidermal growth factor receptor (EGFR), a receptor tyrosine kinase linked to COPD (19), was down-regulated as was the antiinflammatory gene SCGB1A1/uteroglobin. Many genes within this network, including EDG5, HIPK2, FOXO3A, GAS, BDNF, FAS, PLSCR1, and BAX, are associated with apoptosis, a mechanism implicated in the pathogenesis of COPD (20). Two cytoskeleton assembly–related genes, YWHAH and YWHAZ, and a calcium modulating gene (CAMLG) involved in T-cell activation, were up-regulated. Finally, PTGS2/COX-2, a well-characterized inflammatory mediator gene, was also increased.

The study of differential gene expression in normal and diseased tissue has been proposed as a method to identify novel pathogenetic pathways in complex disease as well as to confirm suspected disease processes. The rationale is that gene pathways involved in the processes that lead to tissue injury and/or repair will be activated or suppressed in the tissue of subjects who have disease. This approach can be pursued in a hypothesis-driven format by selecting known genes or pathways for analysis or by using a genomewide, unbiased approach designed to identify genes and pathways involved in a disease that were not previously implicated in its pathogenesis. This approach has been made possible by advances in genomewide transcriptional analysis platforms and has been successful in identifying potential new asthma targets by profiling in the ovalbumin-sensitized animal model (21).

In this study, we used FEF25–75% % predicted for comparison with gene expression data in patients whose lung tissue was resected for small peripheral lung lesions. FEF25–75% % predicted was used as the measure of airflow obstruction because it provided a wider range of values than FEV1% predicted. Although it is recognized that there is a larger between- and within-subject variation in FEF25–75% % predicted, it is also recognized to be a sensitive indicator of airflow obstruction. The subjects that we studied all had lung function that was considered adequate to undergo lung resection and thus represent a selected population that has relatively preserved lung function. The use of FEV1% predicted as the measure of lung function in our patient group has a smaller dynamic range. Nevertheless, we also used FEV1% predicted as the measure of lung function for correlation with gene expression and found substantial overlap of the genes that correlated with both measures of airflow obstruction (see the online supplement).

We used a model to select genes whose expression correlated with lung function independently of pack-years smoked. Although there is often a poor correlation between the level of exposure to cigarette smoke and lung dysfunction, the level expression of many genes is increased by exposure to cigarette smoke (2225). Correcting for the effect of smoking on gene expression reveals genes whose expression is related to the effects of smoking rather than the degree of smoking.

Because most of the subjects in this study had lung cancer as the reason for their lung resection, the results cannot be easily extrapolated to the broader COPD population. Spira and coworkers (26) have recently shown that there is a cancer-specific gene expression profile in cytologically normal large-airway epithelial cells from patients who have pulmonary carcinoma. Although it is likely that the genes whose expression correlates with lung function in smokers with cancer will also correlate with lung function in smokers without cancer, this remains to be tested.

Gene profiling of whole lung samples has the power to detect differential gene expression from any of the lung cell types, but the fact that more than 12 cell types contribute to the expression signature makes interpretation of the results challenging. In addition to resident cell types, the lung contains an abundance of migratory inflammatory cells, and the expression pattern of whole lung tissue represents an amalgam of expression by all of these cell types. In addition, the process of COPD development involves many years of chronic injury and repair, and the expression pattern from a gene profiling exercise is restricted to one cross-sectional look at complex processes that could have varying time courses. Additional complexity is related to nonuniformity of lung structure and the known spatial heterogeneity of the pathologic processes that underlie COPD.

COPD is a functional diagnosis and at least two processes, parenchymal destruction and inflammatory airway narrowing and scarring, combine to cause the abnormal function. Tissue samples subjected to gene profiling may contain varying amounts of alveolar tissue, blood vessels, and airways, and have varying pathologic involvement of these constituents. In addition, lung gene expression patterns could be altered over the short term by environmental factors unrelated to the processes causing tissue injury. These include exposure to cigarette smoke, microbial agents, changes in nutrition, and the use of pharmaceutical agents.

In this study, we attempted to minimize these confounding factors by studying a relatively large number of well-characterized subjects and by performing a precise quantitative characterization of the tissue-type content of the profiled samples. This was especially important because the samples were relatively small compared with those sampled in previous studies (68). An advantage of profiling RNA extracted from a whole lung or lobe is that regional differences in tissue content and disease process are averaged; however, a disadvantage is that relevant signals may be diluted or masked. The importance of quantitative morphometry was supported by the concordance between tissue content and the expression levels for genes from tracheal tissue derived from body atlases of gene expression. Most body atlases studies lump “lung” as one tissue type, but as our analysis shows, a lung sample can contain a variable amount of specific tissue elements and this correlates with the relative expression of tissue components (e.g., tracheal signature in samples with a higher content of airways and blood vessels). Although the tissue content did not substantially alter the relationship between lung function and gene expression in this study, the failure to account for tissue content could lead to spurious results if one component is overrepresented in some samples and/or one compartment is predominantly involved in a disease process.

In an ideal experimental design, samples would come from patients who are matched for age, sex, smoking, and medical history. In addition, samples should be collected from a similar region of lung with comparable proportions of structural components. To select for disease-related genes, clear endpoint differences, such as FEV1 or extent of emphysema, should be observed between the control and the diseased samples. When working with an archived tissue bank, few of these criteria can be met (Table 1). However, with the availability of body atlases and histologic information, we were able to ascertain that a major signature from these samples was related to the relative amount of nonparenchymal tissue that they contained.

The COPD signature gene set that we identified is relatively small, but included several genes in pathways known or suspected of being involved in the pathogenesis of COPD. The portion of the network analysis shown in Figure 7 reveals a potential role for the urokinase plasminogen activator (PLAU/uPA) system. This protease and its receptor have been suggested as candidate genes for COPD (27, 28). PLAUR anchors PLAU at the cell surface to participate in extracellular matrix degradation, cell migration, cell adhesion, and cell proliferation (29). This ligand–receptor pair plays a critical role in the development and persistence of inflammatory and immune responses (16). Active PLAU proteolytically converts inactive plasminogen to active plasmin, which then activates latent forms of MMPs and TGF-β1 (30). PLAU also interacts directly with THBS1, another up-regulated gene related to lung dysfunction, and modulates its activity (31). THBS1 has previously been shown to be a major activator of TGF-β1 in vivo (17) and is also an inducer of MMP-9 (32). It is conceivable that the net result of this up-regulated gene set, which includes PLAU, PLAUR, and THBS1, is a significant increase in the activities of both TGF-β1 and MMP-9.

A review of the current literature supports the concept that dysregulated expression of both TGF-β1 and MMP-9 could be involved in the pathogenesis of COPD (33, 34). In addition, PLAUR is located on chromosome 19 in a region that has been linked to early-onset COPD (18), and PLAU is subject to inhibition by SERPINE2, another candidate gene that is located within a COPD-linked locus on chromosome 2q. Finally, DeMeo and colleagues (35) have shown association of polymorphisms in SERPINE2 and COPD phenotypes. Taken together, these data and our results suggest that the PLAU–PLAUR system could represent a potential new area for therapeutic intervention in COPD.

There are few reports concerning the expression and function of PLAU and PLAUR in the human lung. In infant respiratory distress syndrome, which is characterized by intraalveolar fibrin deposition, the ratio of plasminogen activator inhibitor (PAI)-1 to PLAU in tracheal aspirates was higher than in control subjects and immunohistochemical analysis demonstrated increased expression of both PAI-1 and PLAU, predominantly in alveolar epithelium (36). These results suggest a role for PLAU in intraalveolar fibinolysis, and are supported by studies in mice in which lung-specific overexpression of PLAU led to a reduction in the accumulation of collagen in the lung and reduced mortality after bleomycin-induced lung injury (37). The same transgenic mice were studied as a model of emphysema by inducing PLAU expression over 30 weeks. These animals showed increases in lung compliance, lung volume, and alveolar size, which are typical features of emphysema; however, because these same changes were documented in mice that expressed only the transgene that controls lung specific expression (the Clara cell secretory protein promoter controlling the reverse tetracycline transactivator gene), these emphysematous changes could not be solely attributed to the overexpression of PLAU (38). Murine studies also suggest a protective role for PLAU/PLAUR in bacterial infection of the lung (3941); bacterial infections frequently complicate COPD and may contribute to progressive disease (42).

TaqMan gene expression analysis was concordant with the microarray data for PLAU, PLAUR, and a number of additional genes. Immunostaining of sections of the same samples that were profiled showed widespread expression of these proteins, predominantly in alveolar macrophages and airway epithelium, although expression was also detected in the alveolar epithelium and circulating inflammatory cells. Quantification of PLAU and PLAUR expression by morphometry showed a moderate increase in expression of PLAUR in the airway epithelial cells of individuals who had more severe airflow limitation. The lack of tight correlation of mRNA and protein expression assessed by immunohistochemistry is not surprising. Immunohistochemisty is an inherently nonquantitative technique. It is excellent to detect the presence or absence of expression of a protein, but when there is relatively abundant expression, it fails as a quantitative measure.

Figure 7 also shows that a number of genes involved in apoptosis were differentially expressed. Several pathways in the network emanate from EGFR. That decreased expression of this growth factor receptor could result in increased expression of its direct downstream effectors is counterintuitive; however, a compensatory reaction by these targets is a possibility. The increased expression of the phospholipid scramblase 1, PLSCR1, which mediates migration of phosphatidylserine/ethanolamine to the outer leaflet of the plasma membrane (43), and of FAS support the concept that apoptosis may contribute to the emphysematous destruction of the lung parenchyma in COPD (reviewed in Reference 44). Similarly, two additional up-regulated genes (YWHAH and YWHAZ) are involved in apoptosis through interaction with a ubiqutin–protein ligase, parkin (45), and regulation of cytoskeletal architecture (46). On the other hand, the results show decreased expression of FOXO3A as airway obstruction progresses. The FOXO family of forkhead transcription factors is regulated by the PI3K/protein kinase B (PI3K/Akt) pathway and plays a key role in controlling cell cycle, apoptosis, oxidative stress, and immunomodulation (47). Down-regulation of FOXO3A would be expected to be proinflammatory, but antiapoptotic, which is consistent with its down-regulation by FAS and the proinflammatory phenotype of COPD (48).

Other genes of interest include the Coxsackie virus and adenovirus receptor (CXADR), which was found to be up-regulated in the lungs of obstructed subjects. In addition to serving as the receptor that mediates viral attachment to the cell surface and viral uptake, it also plays a role in cell adhesion, paracellular solute movement, cell proliferation, and cell signaling (49). Adenovirus, especially its E1A protein, has been implicated in the pathogenesis of COPD, and high levels of E1A DNA were located in the lungs of patients with COPD, and its level of expression increased with disease severity (50). Consistent with a previous report (51), COX-2 was up-regulated in COPD lungs and could potentially contribute to the inflammatory status of the disease.

Three COPD-related microarray studies have been recently published (68). The genes that we report to be correlated with lung function showed little overlap with the genes whose level of expression has been correlated with lung function in these reports. A comparison of the study designs and results is included in the online supplement. These differences could relate to differences in patient selection, tissue sampling, and gene expression platform.

The unique features of this study are as follows: (1) we profiled tissue from a relatively large group of individuals who were well characterized preoperatively and (2) we related the gene expression pattern to the tissue content of the samples that were profiled. Our results not only support the possible involvement of some previously reported genes and pathways in COPD pathogenesis but also suggest some novel potential pathways.

1. Pauwels RA, Buist AS, Calverley PM, Jenkins CR, Hurd SS; GOLD Scientific Committee. NHLBI/WHO Global Initiative for Chronic Obstructive Lung Disease (GOLD) workshop summary. Am J Respir Crit Care Med 2001;163:1256–1276.
2. Global Initiative for Chronic Obstructive Lung Disease (GOLD). Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease: NHLBI/WHO workshop report [Internet]. Bethesda (MD): National Heart, Lung, and Blood Institute; 2001 Apr [revised 2003; accessed 2006 Jan]. NIH Publication No. 2701. Available from:
3. Hogg JC, Chu F, Utokaparch S, Woods R, Elliot WM, Buzatu L, Otternack RM, Rogers RM, Sciuba IC, Coxson HO, et al. The nature of small-airway obstruction in chronic obstructive pulmonary disease. N Engl J Med 2004;350:2645–2653.
4. Alsaeedi A, Sin DD, McAlister FA. The effects of inhaled corticosteroids in chronic obstructive pulmonary disease: a systematic review of randomized placebo-controlled trials. Am J Med 2002;113:59–65.
5. Barnes P, Hansel T. Prospects for new drugs for chronic obstructive pulmonary disease. Lancet 2004;364:985–996.
6. Spira A, Beane J, Pinto-Plata V, Kadar A, Liu G, Shah V, Celli B, Brody JS. Gene expression profiling of human lung tissue from smokers with severe emphysema. Am J Respir Cell Mol Biol 2004;31:601–610.
7. Golpon HA, Coldren CD, Zamora MR, Gosgrov GP, Moore MD, Tuder RM, Geraci MV, Voekel NF. Emphysema lung tissue gene expression profiling. Am J Respir Cell Mol Biol 2004;31:595–600.
8. Ning W, Li CJ, Kaminski N, Feghali-Bostwick CA, Alber SM, Di YP, Otterbein SL, Song R, Hayashi S, Zhou Z, et al. Comprehensive gene expression profiles reveal pathways related to the pathogenesis of chronic obstructive pulmonary disease. Proc Natl Acad Sci USA 2004;101:14895–14900.
9. Wang I, Boie Y, Stepaniants S, Mortimer J, Kennedy B, Elliott M, Hayashi S, Coulter MS, Thorton M, Roberts C, et al. Gene expression profiling in the lungs of patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2005;117:A512.
10. Retamales I, Elliott WM, Meshi B, Coxson HO, Paré PD, Sciurba FC, Rogers RM, Hayashi S, Hogg JC. Amplification of inflammation in emphysema and its association with latent adenoviral infection. Am J Respir Crit Care Med 2001;164:469–473.
11. Ding L, Quinlan KF, Elliott WM, Hamodat M, Paré PD, Hogg JC, Hayashi S. A lung tissue bank for gene expression studies in chronic obstructive pulmonary disease. COPD 2004;1:191–204.
12. Crapo RO, Morris AH, Gardner RM. Reference spirometric values using techniques and equipment that meet ATS recommendations. Am Rev Respir Dis 1981;1123:659–664.
13. Coxson HO, Hogg JC, Mayo JR, Behzad H, Whittall KP, Schwartz DA. Quantification of idiopathic pulmonary fibrosis using computed tomography and histology. Am J Respir Crit Care Med 1997;155:1649–1656.
14. Lampe JW, Stepaniants SB, Mao M, Radich JP, Dai H, Linsley PS, Friend SH, Potter JD. Signatures of environmental exposures using peripheral leukocyte gene expression: tobacco smoke. Cancer Epidemiol Biomarkers Prev 2004;13:445–453.
15. Weng L, Dai H, Zhan Y, He Y, Stepaniants SB, Bassett DE. Rosetta error model for gene expression analysis. Bioinformatics 2006;22:1111–1121.
16. Blasi F, Carmeliet P. uPAR: a versatile signaling orchestrator. Natl Rev 2002;3:932–943.
17. Crawford SE, Stellmach V, Murphy-Ullrich JE, Ribeiro SM, Lawler J, Hynes RO, Boivin GP, Bouck N. Thrombospondin-1 is a major activator of TGF-β1 in vivo. Cell 1998;93:1159–1170.
18. Silverman EK, Palmer LJ, Mosley JD, Barth M, Senter JM, Brown A, Drazen JM, Kwiatkowski DJ, Chapman HA, Campbell EJ, et al. Genomewide linkage analysis of quantitative spirometric phenotypes in severe early-onset chronic obstructive pulmonary disease. Am J Hum Genet 2002;70:1229–1239.
19. Kranenburg AR, de Boer WI, van Krieken JHJM, Mooi WJ, Walters JE, Saxena PR, Sterk PJ, Sharma HS. Enhanced expression of fibroblast growth factors and receptor FGFR-1 during vascular remodeling in chronic obstructive pulmonary disease. Am J Respir Cell Mol Biol 2002;27:517–525.
20. de Souza PM, Lindsay MA. Apoptosis as a therapeutic target for the treatment of lung disease. Curr Opin Pharmacol 2005;5:232–237.
21. Rolph MS, Sisavanh M, Liu SM, Mackay CR. Clues to asthma pathogenesis from microarray expression studies. Pharmacol Ther 2006;109:284–294.
22. Heguy A, O'Connor TP, Luettich K, Worgall S, Cieciuch A, Harvey BG, Hackett NR, Crystal RG. Gene expression profiling of human alveolar macrophages of phenotypically normal smokers and nonsmokers reveals a previously unrecognized subset of genes modulated by cigarette smoking. J Mol Med 2006;84:318–328.
23. Harvey BG, Heguy A, Leopold PL, Carolan BJ, Ferris B, Crystal RG. Modification of gene expression of the small airway epithelium in response to cigarette smoking. J Mol Med 2007;85:39–53.
24. Pierrou S, Broberg P, O'Donnell RA, Pawlowski K, Virtala R, Lindqvist E, Richter A, Wilson SJ, Angco G, Moller S, et al. Expression of genes involved in oxidative stress responses in airway epithelial cells of smokers with chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2007;175:577–586.
25. Stevenson CS, Docx C, Webster R, Battram C, Hynx D, Giddings J, Cooper PR, Chakravarty P, Rahman I, Marwick JA, et al. Comprehensive gene expression profiling of rat lung reveals distinct acute and chronic responses to cigarette smoke inhalation. Am J Physiol Lung Cell Mol Physiol 2007;293:L1183–L1193.
26. Spira A, Beane JE, Shah V, Steiling K, Liu G, Schembri F, Gilman S, Dumas YM, Calner P, Sebastiani P, et al. Airway epithelial gene expression in the diagnostic evaluation of smokers with suspect lung cancer. Nat Med 2007;13:361–366.
27. Abboud RT, Ofulue AF, Sansores RH, Müller NL. Relationship of alveolar macrophage plasminogen activator and elastase activities to lung function and CT evidence of emphysema. Chest 1998;113:1257–1263.
28. Xiao W, Hsu YP, Ishizaka A, Kirikae T, Moss RB. Sputum cathelicidin, urokinase plasminogen activation system components, and cytokines discriminate cystic fibrosis, COPD, and asthma inflammation. Chest 2005;128:2316–2326.
29. Mondino A, Blasi F. uPA and uPAR in fibrinolysis, immunity and pathology. Trends Immunol 2004;25:450–455.
30. Carmeliet P, Moons L, Lijnen R, Baes M, Lemaitre V, Tipping P, Drew A, Eeckhout Y, Shapiro S, Lupu F, Collen D. Urokinase-generated plasmin activates matrix metalloproteinases during aneurysm formation. Nat Genet 1997;17:439–444.
31. Albo D, Tuszynski GP. Thrombospondin-1 up-regulates tumor cell invasion through the urokinase plasminogen activator receptor in head and neck cancer cells. J Surg Res 2004;120:21–26.
32. Qian X, Wang TN, Rothman VL, Nicosia RF, Tuszynski GP. Thrombospondin-1 modulates angiogenesis in vitro by up-regulation of matrix metalloproteinase-9 in endothelial cells. Exp Cell Res 1997;235:403–412.
33. Russell REK, Culpitt SV, DeMatos C, Donnelly L, Smith M, Wiggins J, Barnes PJ. Release and activity of matrix metalloproteinase-9 and tissue inhibitor of metalloproteinase-1 by alveolar macrophages from patients with chronic obstructive pulmonary disease. Am J Respir Cell Mol Biol 2002;26:602–609.
34. Takizawa H, Tanaka M, Takami K, Ohtoshi T, Ito K, Satoh M, Okada Y, Yamasawa F, Nakahara K, Umeda A. Increased expression of transforming growth factor-β1 in small airway epithelium from tobacco smokers and patients with chronic obstructive pulmonary disease (COPD). Am J Respir Crit Care Med 2001;163:1476–1483.
35. DeMeo DL, Mariani TJ, Lange C, Srisuma S, Litonjua AA, Celedon JC, Lake SL, Reilly JJ, Chapman HA, Mecham BH, et al. The SERPINE2 gene is associated with chronic obstructive pulmonary disease. Am J Hum Genet 2006;78:253–264.
36. Cederqvist K, Siren V, Petaja J, Vaheri A, Haglund C, Andersson S. High concentrations of plasminogen activator inhibitor-1 in lungs of preterm infants with respiratory distress syndrome. Pediatrics 2006;117:1226–1234.
37. Sisson TH, Hanson KE, Subbotina N, Patwardhan A, Hattori N, Simon RH. Inducible lung-specific urokinase expression reduces fibrosis and mortality after lung injury in mice. Am J Physiol Lung Cell Mol Physiol 2002;283:L1023–L1032.
38. Sisson HT, Hansen JM, Shah M, Hanson KE, Du M, Ling T, Simon RH, Christensen PJ. Expression of the reverse tetracycline-transactivator gene causes emphysema-like changes in mice. Am J Respir Cell Mol Biol 2006;34:552–560.
39. Gyetko MR, Sud S, Kendall T, Fuller JA, Newstead MW, Standiford TJ. Urokinase receptor-deficient mice have impaired neutrophil recruitment in response to pulmonary Pseudomonas aeruginosa infection. J Immunol 2000;165:1513.
40. Rijneveld AW, Levi M, Florquin S, Speelman P, Carmeliet P, van Der Poll T. Urokinase receptor is necessary for adequate host defense against pneumococcal pneumonia. J Immunol 2002;168:3507–3511.
41. Gyetko MR, Sud S, Sonstein J, Polak T, Sud A, Curtis JL. Antigen-driven lymphocyte recruitment to the lung is diminished in the absence of urokinase-type plasminogen activator (uPA) receptor, but is independent of uPA. J Immunol 2001;167:5539–5542.
42. Sethi S, Murphy TF. Bacterial infection in chronic obstructive pulmonary disease. Clin Microbiol Rev 2001;14:336–363.
43. Nanjundan M, Sun J, Zhao J, Zhou Q, Sims PJ, Wiedmer T. Plasma membrane phospholipid scramblase 1 promotes EGF-dependent activation of c-Src through the epidermal growth factor receptor. J Biol Chem 2003;278:37413–37418.
44. Tuder RM, Petrache I, Elias JA, Voelkel NF, Henson PM. Apoptosis and emphysema: the missing link. Am J Respir Cell Mol Biol 2003;28:551–554.
45. Sato S, Chiba T, Sakata E, Kato K, Mizuno Y, Hattori N, Tanaka K. 14-3-3η is a novel regulator of parkin ubiquitin ligase. EMBO J 2006;25:211–221.
46. Jin J, Smith FD, Stark C, Wells CD, Fawcett JP, Kulkarni S, Metalnikov P, O'Donnell P, Taylor P, Taylor L, et al. Proteomic, functional, and domain-based analysis of in vivo 14–3-3 binding proteins involved in cytoskeletal regulation and cellular organization. Curr Biol 2004;14:1436–1450.
47. Lin L, Hron JD, Peng SL. Regulation of NF-κB, Th activation, and autoinflammation by the forkhead transcription factor Foxo3a. Immunity 2004;21:203–213.
48. Turato G, Zuin R, Miniati M, Baraldo S, Rea F, Beghé B, Monti S, Formichi B, Boschetto P, Harari S, et al. Airway inflammation in severe chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2002;166:105–110.
49. Fok PT, Huang K-C, Holland PC, Nalbantoglu J. The Coxsackie and adenovirus receptor binds microtubules and plays a role in cell migration. J Biol Chem 2007;282:7512–7521.
50. Hayashi S, Hogg JC. Adenovirus infections and lung disease. Curr Opin Pharmacol 2007;7:237–243.
51. Xaubet A, Roca-Ferrer J, Pujols L, Ramirez J, Mullol J, Marin-Arguedas A, Torrego A, Gimferrer JM, Picado C. Cyclooxygenase-2 is up-regulated in lung parenchyma of chronic obstructive pulmonary disease and down-regulated in idiopathic pulmonary fibrosis. Sarcoidosis Vasc Diffuse Lung Dis 2004;21:35–42.
52. Boenisch T. Staining methods. In: Boenisch T, editor. Immunochemical staining methods handbook, 3rd ed. Carpinteria (CA): Dako Corporation; 2001. p. 26–31.
Correspondence and requests for reprints should be addressed to Peter D. Paré, M.D., McDonald Research Wing, Room 166, St. Paul's Hospital, 1081 Burrard Street, Vancouver, BC, Canada V6Z 1Y6. E-mail:


No related items
American Journal of Respiratory and Critical Care Medicine

Click to see any corrections or updates and to confirm this is the authentic version of record