|Home | About | Journals | Submit | Contact Us | Français|
Nuclear factor erythroid 2–related factor 2 (Nrf2) is an oxidant-responsive transcription factor known to induce detoxifying and antioxidant genes. Cigarette smoke, with its large oxidant content, is a major stress on the cells of small airway epithelium, which are vulnerable to oxidant damage. We assessed the role of cigarette smoke in activation of Nrf2 in the human small airway epithelium in vivo. Fiberoptic bronchoscopy was used to sample the small airway epithelium in healthy-nonsmoker and healthy-smoker, and gene expression was assessed using microarrays. Relative to nonsmokers, Nrf2 protein in the small airway epithelium of smokers was activated and localized in the nucleus. The human homologs of 201 known murine Nrf2-modulated genes were identified, and 13 highly smoking-responsive Nrf2-modulated genes were identified. Construction of an Nrf2 index to assess the expression levels of these 13 genes in the airway epithelium of smokers showed coordinate control, an observation confirmed by quantitative PCR. This coordinate level of expression of the 13 Nrf2-modulated genes was independent of smoking history or demographic parameters. The Nrf2 index was used to identify two novel Nrf2-modulated, smoking-responsive genes, pirin (PIR) and UDP glucuronosyltransferase 1-family polypeptide A4 (UGT1A4). Both genes were demonstrated to contain functional antioxidant response elements in the promoter region. These observations suggest that Nrf2 plays an important role in regulating cellular defenses against smoking in the highly vulnerable small airway epithelium cells, and that there is variability within the human population in the Nrf2 responsiveness to oxidant burden.
In cigarette smokers, phase II detoxifying and antioxidant enzymes are an important mechanism in the protection of the small airway epithelium from the more than 1014 oxidants present in each puff of cigarette smoke (1–4). Defense against oxidants is a critical cellular function, and oxidant-mediated somatic mutations play an important role in the mechanisms of malignant transformation of the airway epithelium in association with cigarette smoking (2,5–7). Central to the cellular oxidant defense mechanism is the nuclear factor erythroid 2–related factor 2 (NFE2L2, Nrf2), a member of the “Cap ‘n’ Collar” family of the basic leucine transcription factors, known to coordinate the induction of phase II detoxifying and antioxidant enzymes (8–11). In the resting state, Nrf2 is bound to the cytoplasmic inhibitor Kelch-like epichlorohydrin-associated protein 1 (KEAP1) (12,13). Upon activation, Nrf2 dissociates from KEAP1, translocates into the nucleus, and binds to antioxidant response element sequences, resulting in transcriptional activation of genes that help to protect the cell from oxidants (8,12,13).
Understanding of the role of Nrf2 in the lung has been advanced by studies of Nrf2 knockout mice exposed to a variety of mediators of lung injury, including bleomycin, elastase, mechanical ventilation, ovalbumin sensitization, lipopolysaccaride, and environmental stresses including hyperoxia, butylated hydroxytoluene, benz(a)pyrene, diesel exhaust particles, and cigarette smoke (14–26). Microarray analyses of lungs from Nrf2 knockout mice and wild-type controls have helped to identify Nrf2-modulated genes, for example, genes that help protect mice from the development of cigarette smoke–induced lung injuries (23).
On the basis of the knowledge that Nrf2-modulated genes respond to cigarette smoke and other toxic reagents, and in the context that cigarette smoking places a major oxidant stress on the lung, with the initial damage in the small airway epithelium (1–4,20,23,27–30), we asked: does cigarette smoking induce activation of Nrf2 in the human small airway epithelium with coordinate control of genes with antioxidant response elements, and are these genes the same or different from those observed in mice? To address these questions, we sampled the small airway epithelium of 83 individuals (38 healthy nonsmokers and 45 healthy smokers) to determine if Nrf2 was activated in healthy smokers, and if so, to identify which of the known murine Nrf2-modulated genes are up-regulated by smoking in this cell population. The data demonstrate that Nrf2 is activated in the small airway epithelium of healthy smokers. Using the known murine Nrf2-modulated genes in the lung as a starting point (14,16–26), and Affymetrix Human Genome U133 Plus 2.0 microarrays to assess genome-wide gene expression, we observed that 13 human homologues of 201 known murine Nrf2-modulated genes were highly responsive to cigarette smoke in the small airway epithelium of healthy smokers. To identify additional genes that may also be regulated by Nrf2 in the human small airway epithelium, an index was created to quantify the extent of Nrf2-responsive gene expression. We then searched among all genes expressed in the small airway epithelium of healthy individuals for genes whose expression, with TaqMan quantitative confirmation, correlated with the Nrf2 index. Using this strategy, we identified two genes not previously recognized to be controlled by Nrf2, PIR and UGT1A4, both of which contained several antioxidant response elements 5′ to the gene, and both of which responded in vitro to Nrf2.
Healthy nonsmokers (n = 38) and healthy current cigarette smokers (n = 45) were recruited through advertisements in local newspapers and on electronic bulletin boards, and through an ongoing program of free spirometry screening in the Department of Genetic Medicine and Division of Pulmonary and Critical Care Medicine. The evaluation of all individuals was performed at the Weill Cornell NIH Clinical and Translational Science Center and Department of Genetic Medicine Clinical Research Facility, using institutional review board–approved clinical protocols. Nonsmokers and smokers were determined to be healthy on the basis of standard history, physical exam, complete blood count, coagulation studies, liver function tests, HIV-1 serology, urine studies, chest x-ray, electrocardiogram, and pulmonary function tests. Current smoking status was evaluated on the basis of history (pack-year), venous carboxyhemoglobin levels, and urine analysis for nicotine metabolites. The inclusion criteria for healthy nonsmokers were “never smoking history” and normal physical exam, lung function, and chest x-ray, with smoking-related blood and urine within the nonsmoker range. The criteria for healthy smokers were current smoking history, normal physical exam, lung function, chest x-ray, and smoking-related urine and blood parameters consistent with that of a current smoker (31,32).
We collected small airway epithelium brushes by using fiberoptic bronchoscopy, as previously described (33,34). Briefly, after the study participant had received mild sedation with meperidine and midazolam and routine anesthesia of the vocal cords and bronchial airways with topical lidocaine, a fiberoptic bronchoscope (Pentax, EB-1530T3) was directed to proximal to the opening of a desired lobar bronchus. Small airway epithelium cells were collected from the 10th- to 12th-order bronchi of the right lower lobe. A 2.0-mm disposable brush was advanced 7 to 10 cm further distally from the third-order bronchial branching, and the distal end of the brush was wedged into the 10th- to 12th-generation branching of the right lower lobe. Small airway epithelium was collected by gently gliding the brush back and forth 5 to 10 times in 8 to 10 different locations in the same general area. Cells were detached from the brush by flicking and were immediately transferred into aliquots of ice-cold LHC8 medium (Gibco, Grand Island, NY, USA). Total cell number was counted on a hemocytometer, and cell viability was estimated by Trypan Blue exclusion and expressed as a percentage of the total cells recovered. About 1–2 × 104 cells were processed immediately for RNA extraction, and 2–5 × 106 cells were frozen at −80°C in cell-freezing medium (Sigma-Aldrich, St. Louis, MO, USA) for further processing for Western analysis and electrophoretic mobility shift assays. Additionally, 2 × 104 cells per slide were used for a differential cell count and for immunohistochemistry. Slides for a differential cell count and for immunohistochemistry were prepared by centrifugation (Cytospin 11; Shandon Instruments, Pittsburgh, PA, USA), and stored at 4°C until further processing. For differential cell count, cells were stained with DiffQuik (Baxter Healthcare, Miami, FL, USA). All samples were confirmed as bonafide small airway epithelial samples by expression of genes encoding surfactant proteins and Clara secretory protein (33).
Analyses were performed by using Affymetrix (Santa Clara, CA, USA) Human Genome U133 Plus 2.0 microarrays (54,675 probe sets representing approximately 47,000 full-length human gene transcripts) and associated protocols. Total RNA was extracted by using a modified version of the TRIzol method (Invitrogen, Carlsbad, CA, USA), followed by RNeasy (Qiagen, Valencia, CA, USA) to remove residual DNA. RNA samples were stored in RNA Secure (Ambion, Austin, TX, USA) at −80°C until further processing. An aliquot of each RNA sample was run on an Agilent Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA) to visualize and quantify the degree of RNA integrity. The concentration was determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA).
Double-stranded cDNA was synthesized from 1–2 μg of total RNA using the GeneChip One-Cycle cDNA Synthesis Kit, followed by cleanup with a GeneChip Sample Cleanup Module, in vitro transcription reaction performed by using the GeneChip IVT Labeling Kit, and clean-up and quantification of the biotin-labeled cRNA yield by spectrophotometric analysis. All kits were from Affymetrix. Hybridizations to test chips and to the microarrays were performed according to Affymetrix protocols, and microarrays were processed by the Affymetrix fluidics station and scanned with the Affymetrix GeneChip Array Scanner 3000 7G. Overall microarray quality was verified by the following criteria: (a) RNA integrity number ≥7.0; (b) 3′/5′ ratio for GAPDH ≤3; and (c) scaling factor ≤10.0. Captured images were processed by using the MAS 5.0 algorithm (Affymetrix Microarray Suite Version 5.0 software), which takes into account the perfect match and mismatch probes. The data were normalized by using GeneSpring version 7.3 software (Agilent Technologies) per array, by dividing the raw data by the 50th percentile of all measurements and by gene.
On the basis of the knowledge that Nrf2 translocates to the nucleus to protect cells from oxidative stress, such as that caused by smoking (11,12,20,23,30, 35), we applied immunohistochemical and Western analysis of Nrf2 to assess the activation status of Nrf2 in small airway epithelial cells obtained from healthy nonsmokers and healthy smokers.
For immunohistochemistry, slides with cytospins of the small airway epithelium were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS), pH 7.4, at 23°C for 20 min. To enhance staining, an antigen retrieval step was carried out by incubating the slide at 90°C for 20 min in a water steamer (Black & Decker, Hunt Valley, MD, USA), in citrate solution (BD Biosciences Pharmingen, San Diego, CA, USA), followed by cooling at 23°C for 20 min. The slides were blocked in 5% donkey serum for 45 min to reduce background staining, then incubated overnight at 4°C with an affinity-purified rabbit polyclonal antibody, mapped to the C-terminus of human Nrf2 (Santa Cruz Biotechnology, Santa Cruz, CA, USA), diluted 1:1000 in PBS with 5% donkey serum. To ensure the specificity of the Nrf2 antibody, blocking Nrf2 peptide (Santa Cruz Biotechnology) was used as a control. The Vectastain Elite ABC kit (Vector Laboratories, Burlingame, CA, USA) and the 3-amino-9-ethyl carbazole substrate kit (Vector Laboratories) were used to detect antibody binding, and the slides were counterstained with hematoxylin (Sigma-Aldrich) and mounted using GVA mounting medium (Zymed, San Francisco, CA, USA). Brightfield microscopy was performed using a Nikon Microphot microscope equipped with a Plan 40 × N.A. 0.70 objective lens. Images were captured with an Olympus DP70 CCD camera.
To quantify the nuclear accumulation of Nrf2 in small airway epithelium in response to smoking, cytoplasmic and nuclear extracts from small airway epithelium cells of healthy nonsmokers and healthy smokers were assessed by Western analysis. Cytoplasmic and nuclear extracts were isolated stepwise by using a commercially available nuclear and cytoplasmic extraction kit (Pierce Biotechnology, Rockford, IL, USA). Briefly, after thawing and one washing with PBS (pH 7.4), small airway epithelium cells were lysed in 200 μL of ice cold cytoplasmic extraction reagent 1 for 10 min, and 11 μL of ice-cold cytoplasmic extraction reagent 2 for 1 min, according to the manufacturer’s protocol. After centrifugation at 16,000g for 5 min, the supernatant (cytoplasmic extract) was stored on ice, while the insoluble pellet fraction was resuspended in 25 μL of nuclear-extraction reagent. After a 40-min incubation period, with vortexing every 10 min, and another centrifugation at 16,000g for 10 min, the supernatant (nuclear extract) was collected. Protein concentration was assessed in nuclear and cytoplasmic extracts by using a BCA protein concentration kit (Pierce). Samples were adjusted to equal amounts of protein (50 μg protein/lane for cytoplasmic protein and 30 μg protein/lane for nuclear protein). Samples were mixed with Nupage LDS sample buffer and reducing agent (Invitrogen) and loaded onto a Novex Tris-Glycine 4%–12% gel (Invitrogen). Proteins were transferred (25 V, 90 min, 23ºC) to a 0.45-μm pore size polyvinylidene fluoride membrane (BioRad) in Tris-glycine transfer buffer (Invitrogen). After transfer, the membrane was blocked in PBS containing 5% milk powder (Bio-Rad) for 1 h, then the membrane was incubated overnight at 4ºC with a 1:1000 dilution of antihuman Nrf2 monoclonal IgG (Abnova, Taipei, Taiwan). Detection was performed by using horse-radish peroxidase–conjugated secondary antimouse antibody (Santa Cruz), at a 1:2000 dilution, and the enhanced chemiluminescence reagent system (GE Healthcare, Pittsburgh, PA, USA), using Kodak BioMax light film. To evaluate the levels of nuclear protein in the nuclear extracts, the blots were stripped in Western-blot stripping buffer (Pierce) for 15 min, at 23ºC. After being blocked in PBS containing 5% milk powder for 1 h at 23ºC, the blots with nuclear extracts were incubated with a polyclonal goat-antihuman lamin B antibody (Santa Cruz Biotechnology) at a 1:300 dilution, and with a secondary antigoat horseradish peroxidase-conjugated antibody (Santa Cruz Biotechnology) at 1:2000, for 90 min each at 23ºC, with detection as described above. The blots with cytoplasmic extracts were incubated with a polyclonal rabbit anti–β-tubulin antibody (Imgenex, San Diego, CA, USA) at a 1:1000 dilution, and with a secondary antirabbit horse-radish peroxidase–conjugated antibody (Santa Cruz Biotechnology) at 1:2000, for 90 min each at 23ºC. To quantitatively assess the amount of Nrf2 in the nucleus and cytoplasm, the films were digitally imaged, maintaining exposure within the linear range of detection. The contrast was inverted, the pixel intensity of each band determined, and the background pixel intensity for a negative area of film of identical size was subtracted by using MetaMorph image analysis software (Universal Imaging, Downingtown, PA, USA). Bands on the lamin B and β-tubulin Western blots were scanned and analyzed by densitometry to ensure equal protein loading.
We compiled a list of known murine Nrf2-modulated genes consisting of all murine genes reported to be responsive to different oxidative stress conditions in the lung, as identified in a total of 12 reports by several groups using Nrf2 knockout mice (14,16–26). This list included 201 genes that were upregulated in wild-type mice compared with Nrf2 knockout mice (Supplemental Data, Table 1). Among these 201 genes, 187 human homologs were identified by using the NCBI (National Center for Bioinformatics) HomoloGene database (http://www.ncbi.nlm.nih.gov/homologene) and the Information Hyperlinked Over Proteins database (http://www.ihop-net.org/UniPub/iHOP/). The impact of smoking on these 187 human genes was assessed by using the gene expression micorarray data from the 45 smokers and 38 nonsmokers with following criteria: (a) Affymetrix detection call of Present (“P call”) in > 20 % of samples, (b) magnitude of fold-change in average expression value for healthy smokers versus nonsmokers > 1.5, (c) significance level in expression value for healthy smokers versus nonsmokers P < 0.01. A final list was established consisting of putative, highly smoking-responsive, Nrf2-modulated human genes.
If Nrf2-modulated genes are coordinately controlled in the small airway epithelium of smokers, then all of the Nrf2 human homologs that are significantly modulated by smoking should be modulated to a similar extent (for example, low, medium, or high expression level) in each individual. To assess this thesis, an Nrf2 index was created that ranked healthy smokers based on their expression levels of the 13 Nrf2-modulated genes. For each gene, the individuals were divided into quartiles based on the level of gene expression; each individual was assigned a quartile designation of 1, 2, 3, or 4 for each of the putative Nrf2-responsive genes. To avoid biasing the Nrf2 index toward genes that were represented by multiple probe sets, the quartile designations were averaged and rounded to the nearest single-decimal unit for genes represented by more than 1 probe set. To calculate the Nrf2 index for each individual, the quartile designations of that individual were averaged across all Nrf2-modulated genes and reported as mean ± SD. A z score Nrf-2 index was used to validate the quartile-based index. This validation was accomplished by determining for each subject the z score for each expression value and summing the z scores for the 13 genes in the initial gene list for each of the 45 smokers.
To visualize whether Nrf2-modulated genes were concordantly regulated among the healthy smokers, Nrf2 genes were ordered alphabetically, and all healthy smokers were ranked according to the Nrf2 index. Gene clusters were visualized based on the level of gene expression, with a different color assigned to each quartile. The distribution of the Nrf2 index was compared with that of indices generated in the same way using control genes. Two other gene lists were compiled as controls, both compromised of randomly chosen genes (“P call” in >20% of samples) generated by using Excel 2003 software (Microsoft Corporation, Redmond, WA, USA): (a) genes randomly chosen from all probe sets corresponding to unique genes and (b) genes randomly chosen from a total of 476 smoking-responsive genes identified from our dataset (defined as > 1.5-fold, P < 0.01 after Benjamini-Hochberg correction, for healthy smokers versus nonsmokers; Supplemental Data, Table 2). To rule out the possibility that the Nrf2 index might reflect differing demographic characteristic or differing exposure to smoking, we compared the demographics and smoking history of individuals from each quartile. In addition, the overall coordination of the expression levels of the Nf2-related genes was determined by averaging the r2 values for all pairs of genes from all healthy smokers and comparing that mean r2 value to the mean value from the set of random genes and set of random smoking-related genes (Supplementary Table 2).
TaqMan real-time reverse-transcription–polymerase chain reaction (RT-PCR) was performed on RNA samples from the small airway epithelium of healthy smokers with low and high Nrf2 index values, and the results compared with those from the Human Genome U133 Plus 2.0 microarray. cDNA was synthesized from 2 μg RNA in a 100 μL reaction volume, using the TaqMan Reverse Transcriptase Reaction Kit (Applied Biosystems, Foster City, CA, USA), with random hexamers as primers. Two dilutions, 1:10 and 1:100, were made from each sample, and duplicate wells were run for each dilution. TaqMan PCR reactions were carried out using pre-made gene-expression kits from Applied Biosystems for Nrf2-modulated genes, and 2 μL of cDNA was used in each 25 μL reaction. The endogenous control was human β-actin (Applied Biosystems). Relative expression levels were determined using the Δ Δ Ct method, with the average value of the low-ranked smokers as the calibrator. The PCR reactions were run in an Applied Biosystems Sequence Detection System 7500, and the relative quantity was determined using the algorithm provided by the manufacturer.
To identify unknown genes that may also be modulated by Nrf2 in the small airway epithelium, the Nrf2 index was correlated with the expression values of all genes expressed in the lung of healthy smokers (“P call” in >20% of samples), using the r2 value to identify known and putative new Nrf2-modulated genes. For the 25 most highly correlated genes, the presence of the primary core sequence of the antioxidant response element (RTGAYNNNGCR) (23,36) in the promoter region of each gene (5′ untranslated region and 10,000-bp upstream of the transcriptional start site) was determined by searching with Genamics Expression 1.1 Pattern Finder Tool Software (Genamics, Hamilton, New Zealand).
Newly identified Nrf2-modulated genes were validated with electrophoretic mobility shift assay (37,38). We incubated 10 μg of nuclear extract from small airway epithelium with a 32P-labeled antioxidant response element oligonucleotide from NAD(P)H dehydrogenase, quinone 1 (NQO1), a known Nrf2-modulated gene (10,37). A fifty-fold excess of unlabeled NQO1 oligonucleotide was used as a specific competitor to determine the specificity of binding. A fiftyfold excess of nonspecific oligonucleotide (a random sequence from the von Wille-brand factor gene) was used as a negative control for the competition. To demonstrate that putative antioxidant response elements from the newly identified Nrf2-modulated genes bind to Nrf2 in a similar manner as does the NQO1-antioxidant response element, unlabeled antioxidant response element oligonucleotides from each newly identified Nrf2-modulated gene were used as specific competitors, in a fifty-fold excess, against the labeled NQO1 oligonucleotide. To confirm the results, one of the antioxidant response element oligonucleotides from each gene proven to be functional was radioactively labeled and competed with an unlabeled oligonucleotide as a specific competitor, and with a nonspecific oligonucleotide as a negative control. The sequences of the oligonucleotides (Operon Biotechnologies, Huntsville, AL, USA) are listed in Supplemental Data Table 3. The oligonucleotides were annealed by heating to 95ºC in an annealing buffer of 10 mmol/L Tris, pH 8.0, 50 mmol/L NaCl, and 1 mmol/L EDTA (ethylenediaminetetraacetic acid). These oligonucleotides (200 ng) were end-labeled with 300 μCi 32P ATP by a T4 polynucleotide kinase (New England Biolab, Ipswich, MA, USA), and afterward purified, according to the protocol, using the QIAquick Nucleotide Removal kit (Qiagen, Valencia, CA, USA). Approximately 3 ng (300,000 cpm) of labeled probe was used in each reaction. The probes were incubated with nuclear protein, extracted as described above, for 30 min at 23ºC in a binding buffer of 0.02 mol/L Hepes pH 8.0, 0.05 mol/L KCl, 0.05 μmol/L EDTA, 1 μmol/L MgCl2, and 5% glycerol, and with a fifty-fold excess of an unrelated, noncompetitive oligonucleotide to reduce nonspecific binding. For super shift analysis (39), labeled oligonucleotides were incubated with 1.65 μg of an affinity purified antihuman Nrf2 rabbit polyclonal IgG (Atlas Antibodies, Stockholm, Sweden) for 30 min. Pure rabbit IgG (Jackson Immunoresearch, West Grove, PA, USA) was used as a control at the same concentration. The mixture was separated on a 6% native polyacrylamide gel (Invitrogen). The gels were subsequently dried, and exposed to a BioMax X-Ray film for 5 to 24 h.
Human Genome U133 Plus 2.0 microarrays were analyzed using GeneSpring software. Average expression values in small airway samples were calculated from normalized expression levels for healthy nonsmokers and healthy smokers. Statistical comparisons between nominal variables were calculated using an unpaired, two-tailed t test with unequal variance. The comparison of SDs in the index of Nrf2-modulated genes against indices of random genes and random smoking-responsive genes was calculated with the nonparametric Wilcoxon rank test. A chi-square test was used for statistical comparison for categorical variables. To rule out the possibility that the Nrf2 index might reflect differing demographic characteristics or differing exposure to smoking, individuals of each quartile were compared by ANOVA. Correlation analyses were all performed using the Spearman Correlation. All tests other than the two-tailed t test were calculated using StatView version 5.0 (SAS Institute). Data are presented as mean ± standard error. A P value < 0.05 was considered to be significant.
Small airway epithelial samples were collected from 83 individuals comprising 38 healthy nonsmokers and 45 healthy smokers (Table 1). The nonsmoker and smoker groups were similar with regard to age (P > 0.4), sex (P > 0.8), and ethnicity (P > 0.2). Healthy smokers had a mean smoking history of 27 ± 2 pack-years. Compared with healthy nonsmokers, healthy smokers demonstrated a mildly higher total number of cells recovered in the brushings (P < 0.04, Table 1). The differentials of the epithelium cells mostly did not differ between healthy nonsmokers and smokers (P > 0.07), except that undifferentiated cells were more frequently found in smokers than in nonsmokers (P < 0.01).
To confirm airway epithelium expression, and on the basis of knowledge that Nrf2 translocates to the nucleus to protect cells from oxidative stresses such as smoking (11,12,20,23), we carried out immunohistochemical analysis by using small airway epithelial cells obtained by brushing. With an antibody dilution of 1:1000, we stained about 20%–30% of brushed cells for Nrf2 (Figure 1A–B, D–F), which were specifically blocked by using a blocking peptide as a negative control (Figure 1C, F). The intensity of Nrf2 staining was variable among cells. However, for nonsmokers there was minimal Nrf2 detected in the nucleus (Figure 1A, B). In contrast, the epithelial cells of smokers showed faint cytoplasmic Nrf2 staining with prominent strong areas of stain in the nucleus (Figure 1D, E). There were typically 2 or 3 spots of Nrf2 staining per cell consistent with location at the nucleoli (arrows).
To quantify the nuclear accumulation of Nrf2 in small airway epithelium in response to smoking, nuclear and cytoplasmic extracts from healthy nonsmokers and smokers were assessed by Western analysis. Using multiple brushes from the airway, we obtained approximately 30 μg of nuclear and 50 μg of cytoplasmic protein from each individual. Equal protein loading among the samples was verified with antibodies that identify the nucleus (lamin B) and cytoplasm (β-tubulin). Western analysis demonstrated increased Nrf2 protein concentration in the cytoplasm of healthy nonsmokers as compared with healthy smokers (Figure 1G). In contrast, the nuclear extract showed a higher amount of Nrf2 protein in smokers compared with nonsmokers (Figure 1H). Quantification of single bands by densitometry revealed a 3.6-fold increase in levels of Nrf2 in the cytoplasm of healthy nonsmokers compared with healthy smokers (P < 0.02; Figure 1I), and a 3.7-fold increase in levels of Nrf2 in the nucleus of healthy smokers compared with healthy nonsmokers (P < 0.05, Figure 1J). Because of limitations in biological materials from human subjects, it was not possible to correlate gene expression at the mRNA level with protein levels measured by Western analysis in the same subject.
A list of murine Nrf2-modulated genes was compiled based on reports of genes responsive to different oxidative stress conditions in the lung of Nrf2 knockout mice (14, 16–26). In total, 201 murine oxidative stress responsive genes in the lung were identified, and 187 human homologs were determined to be expressed in the small airway epithelium (Supplemental Data, Table 1). The impact of smoking on these 187 human genes was assessed by using the gene expression micorarray data from the 45 smokers and 38 nonsmokers described above. From these 187 genes, we established a final list consisting of 13 putative human, Nrf2-modulated, highly smoking-responsive genes (Table 2). All 13 genes were significantly upregulated in smokers compared with nonsmokers (all comparisons, P < 10−4; Figure 2), with a mean increase in expression level of 4.5- ± 1.7-fold in response to smoking. Among the other known murine Nrf2-modulated genes, an additional 16 genes were found to be significantly modulated by smoking, with smoker-to-nonsmoker P values of 0.05 to 10−4. However, because the fold-change of smoker to nonsmoker was <1.5 in all cases, these genes were not considered further (Supplemental Data, Table 4).
If the human Nrf2-modulated, smoking responsive genes have been correctly identified, then it follows that there should be coordinate regulation of the 13 genes among the smoking subjects (that is, a subject who has a high expression level of 1 of the 13 genes should also have high expression of the other 12). To assess the average small airway epithelium gene expression of the Nrf2-modulated smoking responsive genes, an Nrf2 index was created that ranks healthy smokers on the basis of their overall expression levels of the 13 Nrf2-modulated genes. For each gene, individuals were divided into quartiles on the basis of the level of gene expression, so each individual was assigned a quartile designation of 1 to 4 for each of the 13 genes. This analysis demonstrated that among smokers, there were clearly individuals who had expression levels for most or all genes in the upper quartile, whereas other subjects had expression levels for most or all genes in the lower quartile. The Nrf2 index for each individual is the average of the 13 quartile designations (Figure 3A). The use of a discontinuous score for determining the Nrf2 index was validated by calculating the index by using an alternate strategy in which the sum of the z scores for each gene was calculated for each subject. This method gave an index that was strongly correlated (r2 = 0.95) with the quartile-based score (see Supplementary Figure S1).
The healthy smokers showed a wide range of Nrf2 index values, with a 25th-to-75th percentile range of 1.7 to 3.2, and individuals were almost evenly distributed among the index groups. Of all 45 smokers, 13 smokers ranked in the Nrf2 index between 1 and 2, 17 smokers between 2 and 3, and 15 smokers between 3 and 4 (Figure 3B). To demonstrate that this result represents coordinate control, the distribution of the smokers according to the 13 Nrf2-modulated gene index was compared with that of an index generated in the same way by using 13 random genes (Supplemental Table 2). This comparison gave a narrower range with a 25th to 75th percentile range of 2.2 to 2.7. Similarly, for 13 smoking-responsive genes randomly chosen from a total of 476 smoking-responsive genes identified from our data set (corrected P < 0.05, fold-change > 1.5), the 25th and 75th perctentile of the index was 2.2 to 2.8, narrower than the range of the putative Nrf2-modulated, smoking-responsive genes. For both control gene indices, smokers were predominately distributed between the index values two and three, suggesting a random distribution of these control genes, without coordinate regulation (Figure 3B). Of the 45 smokers 37 (82%) were ranked between the index values two and three in the random gene index, and 33 of 45 smokers (73%) were ranked between two and three in the random smoking-responsive gene index (P < 0.0001 versus Nrf2-modulated genes, chi-square test). Other evidence of the coordinate regulation of the Nrf2 genes is the comparison of SDs of the Nrf2 index for each individual, which should be small if all genes are coordinately regulated. For Nrf2-modulated genes, the mean ± SD of all smokers was 0.67 ± 0.03, a value significantly lower than for random genes (1.08 ± 0.02, P < 10−4) or that of random smoking-responsive genes (1.06 ± 0.03, P < 10−4, Figure 3C).
To confirm the relatedness of the expression profile of these genes, we used pairwise Pearson correlations between the 13 genes across all subjects and then compared the average of these correlations to correlations derived from random sets of genes on the array or smoking-related genes. The mean r2 for the Nrf2-modulated genes was 0.36 ± 0.02 compared with 0.10 ± 0.01 for random genes and 0.05 ± 0.01 for random smoking-regulated genes.
To assess whether the Nrf2 index reflected differing demographic characteristics or differing exposure to smoking among the subjects examined, the healthy smokers were divided into three ranges (1–2, low expression of Nrf2-modulated genes; 2–3, moderate expression; 3–4, high expression). The demographic characteristics (sex, age, and ethnicity) were not different among the three groups (P > 0.05 by chi-square test for sex and ethnicity, ANOVA for age, Table 3). Similarly, the smoking exposure as assessed by history in pack-years, urine nicotine, urine cotinine, and serum carboxyhemoglobin was not different among the three groups (all comparisons P > 0.05 by ANOVA, Table 3). Thus, the Nrf2 index is not influenced by known demographic characteristics. Similarly, with analysis using the z score method, the P value for correlation of Nrf2 index with smoking history (P = 0.10), urine nicotine (P = 0.50), urine cotinine (P = 0.19), and carboxyhemoglobin (P = 0.24) were all insignificant.
If Nrf2-modulated genes have coordinate regulation in the human small airway epithelium, then other potential Nrf2-modulated genes might be identified by searching among all genes expressed in the lung of healthy smokers for genes whose expression correlates with the Nrf2 index. Of the top 25 probe sets whose small airway expression level was correlated with the Nrf2 index (Table 4), 24 were smoking responsive, with a mean increase of expression level of 2.9- ± 0.2-fold in response to smoking (P values all < 6 × 10−5). The 25 probe sets corresponded to 18 uniquely identified genes. Overall, these 25 probe sets were highly correlated with the Nrf2 index, (r2 > 0.51, P < 0.0001), suggesting that all these genes might have the same coordinate regulation and control. The gene expression that was most strongly associated with the Nrf2 index was that of transaldolase 1 (TALDO1, Figure 4A), a key enzyme of the pentose phosphate pathway, which protects cellular integrity from oxygen injury. Other strongly correlated genes included: ALDH3A1 (Figure 4B), an enzyme that plays an important role in the detoxification of alcohol-derived acetaldehydes; NQO1 (Figure 4C), an enzyme that prevents the reduction of quinones that results in the production of radical species; and TXNRD1 (Figure 4D), a protein that reduces thioredoxins and plays a role in selenium metabolism and protection from oxidative stress.
The coordinate regulation of these genes was confirmed by unsupervised clustering. The complete set of smoking-responsive genes was clustered for a set of healthy smokers, and the locations of the NRF2-modulated genes within the cluster was determined. A small subsection of the cluster containing 52 genes (Supplementary Figure S2) included 12 of the genes listed in Table 4, a finding that suggests a strong coordinate regulation. Other genes found in this part of the cluster, for example, carbonyl reductases 1 and 3 (CBR1, CBR3) may also be modulated by Nrf2.
To validate the results obtained from the microarrays, TaqMan real-time RT-PCR was carried out to assess several Nrf2-modulated genes by use of RNA samples from six healthy smokers classified as low expressers (Nrf2 index 1–2) and six healthy smokers categorized as high expressers (Nrf2 index 3–4). The TaqMan-PCR data were consistent with the microarray data for all three genes (Figure 5A). In the microarray analyses, TALDO1 (3.0-fold, P < 10−6), ME1 (4.3-fold, P < 10−5) and TXNRD1 (2.8-fold, P < 10−4) were significantly upregulated in the high expressers compared with the low expressers, in small airway epithelium. Use of TaqMan analysis on the same subjects confirmed that TALDO1, ME1, and TXNRD1, were significantly upregulated in the high expressers (2.4-fold, 4.3-fold and 2.8-fold, respectively; all P < 0.05; Figure 5B).
If correlation with the Nrf2 index predicts regulation of small airway gene expression level by Nrf2, then some of the genes identified by this method may have previously been reported to be Nrf2 modulated. Of the 18 genes that were identified as having a strong association to the Nrf2 index, 14 have previously been reported as Nrf2-modulated on the basis of a transcription induction by Nrf2 in studies using Nrf2 knockout mice and/or by gel-shift assays specific for Nrf2: TALDO1 (10,40); ALDH3A1 (23), NQO1 (10,16–21,23,41); TXNRD1 (18,19, 22,23,26); ME1 (10,19,23,26); aldo-keto reductase family 1, member C1 and member C2 (AKR1C1, AKR1C2) (42); glutathione peroxidase 2 (GPX2) (10,17,18,23,25,26); transketolase (TKT) (19); UGT1A6 (16–19, 43); glutamate-cysteine ligase, modifier subunit (GCLM) (10,19,23,25); sequestome 1 (SQSTM1) (23); ADH7 (19,23); and GSR (10,23,24). We identified a very similar (22 of 25) list of potential Nrf2-modulated genes by using the continuous z score method of index calculation.
However, some genes identified had not previously been reported as transcriptionally induced by Nrf2 but have a strong association to the Nrf2 index. These included PIR, a gene of the cubin family (Figure 6A); ATB binding cassette, sub-family B, member 6 (ABCB6), encoding a transporter protein of the superfamily of the ATP-binding cassette (Figure 6B); UDP gluconosyltransferase 1 family, polypeptide A4 (UGT1A4), part of the complex locus which encodes several UDP-glucuromosyltransferases (Figure 6C); and aldo-keto reductase family 1, member C3 (AKR1C3), a member of the aldo/ketoreductase superfamily (Figure 6D).
To evaluate whether PIR, ABCB6, UGT1A4, and AKRC3 are potential candidate genes modulated by Nrf2, the presence of the primary core sequence of the antioxidant response element (RTGAYNNNGCR) (10,37) was assessed by searching the promoter region of each gene with Genamics Expression 1.1 Pattern Finder Tool software. This analysis showed that PIR, ABCB6, and UGT1A4, but not AKR1C3, contain one or more antioxidant response elements in the genomic sequence upstream of their transcription start sites, suggesting these genes as novel Nrf2-modulated genes (Table 4).
If PIR, ABCB6, and UGT1A4 are Nrf2-modulated genes, then the DNA binding activity of Nrf2 to the antioxidant response element of these genes should be detectable by electrophoretic mobility-shift assay. A strong DNA-binding activity of nuclear protein from the small airway epithelium of healthy smokers for NQO1-antioxidant response element was observed (Figure 7A, lanes 2, 6, 10, 18, 23). Nrf2 binding was demonstrated by the disappearance of DNA–protein complexes by competition with excess unlabeled NQO1 oligonucleotides (lanes 3, 11, 19, 24), but not with nonspecific unlabeled oligonucleotides (lanes 4, 16, 21, 29). Specificity was further confirmed by the disappearance of the gel shift band in the presence of an anti-Nrf2 affinity-purified polyclonal antibody (lane 7), whereas no disappearance was seen with the use of a corresponding IgG control (lane 8). Instead of a supershift, the disappearance of the gel shift band suggests a competition for binding between the Nrf2 antibody and NQO1 oligonucleotide, similar to that reported for Nrf2 binding to an antioxidant response element of UGT1A1 with nuclear extracts from HepG2 cells after oxidative stress (38).
PIR contains four antioxidant response elements located 33 bp (lane 12) downstream of the transcription site and −3209 bp (lane 13), −3480 bp (lane 14), and −5566 bp (lane 15) upstream of the transcription start site. If these are functional Nrf2-regulated antioxidant response elements, then an excess of oligonucleotides consisting of these antioxidant response elements should compete specifically against the labeled NQO1 antioxidant response element. For PIR, functional anti-oxidant response elements were found at the locations of −3209 bp (lane 13) and −5566 bp (lane 15) upstream of the transcription start site, but not at 33 bp (lane 12) or −3480 bp (lane 14) from the transcription site.
For ABCB6, one antioxidant response element was identified at −7575 bp upstream of the transcription site. An excess of oligonucleotides consisting of this antioxidant response element did not compete against the labeled NQO1 anti-oxidant response (lane 20).
For UGT1A4, which contains 4 anti-oxidant response elements at −3043 bp (lane 25), −3159 bp (lane 26), −3885 bp (lane 27), and −5523 bp (lane 28) upstream of the transcriptional start site, two antioxidants response elements, at −3885 bp (lane 27) and −5523 bp (lane 28), competed against the NQO1-antioxidant response element—though not as strongly as either the NQO1 olignucleotide itself or as the functional pirin antioxidant response elements—suggesting that these two are probably functional antioxidant response elements.
To confirm these results, one of the antioxidant response element oligonucleotides of PIR and UGT1A4 that was proven to be functional was radioactively labeled and competed with an unlabeled oligonucleotide as a specific competitor, and with a nonspecific oligonucleotide as a negative control. A strong DNA binding activity of nuclear protein from small airway epithelium was observed for labeled PIR antioxidant response element −3209 bp (Figure 7B, lane 31, 35) and for labeled UGT1A4 antioxidant response element −3385 bp (Figure 7C, lane 39, 43). Nrf2 binding was demonstrated by the disappearance of DNA–protein complexes by competition with excess unlabeled PIR oligonucleotides (Figure 7B, lane 32) and UGT1A4 oligonucleotide (Figure 7C, lane 40), but not with nonspecific vWF unlabeled oligonucleotides (Figure 7B and 7C, lanes 33 and 41). Specificity was further confirmed by the disappearance of the gel shift band in the presence of an anti-Nrf2 affinity purified polyclonal antibody (Figure 7B and 7C, lanes 36 and 44), whereas no disappearance was seen using a corresponding IgG control (Figure 7B and 7C, lanes 37 and 45).
Nuclear factor erythroid 2–related factor 2 (NFE2L2, Nrf2), an oxidant responding transcription factor, is known to induce phase 2 detoxifying and anti-oxidant genes to protect cells from oxidative stress (8–11). Based on the knowledge that cigarette smoke with its large oxidant content is a major stressor of the small airway epithelium, with cells that are vulnerable to oxidant damage and its associated malignant transformation (2,5–7), we asked the questions: does cigarette smoking induce activation of Nrf2 in the human small airway epithelium, is there an associated coordinate control of Nrf2-modulated genes, and are the highly-responsive Nrf2-modulated genes the same or different from those observed in Nrf2 murine models? Fiber-optic bronchoscopy was used to sample pure populations of small airway epithelium in 38 healthy nonsmokers and 45 healthy smokers, and gene expression of human homologs of 201 known murine Nrf2-modulated genes were assessed using Affymetrix Human Genome U133 Plus 2.0 microarrays. In the human small airway epithelium of healthy normal smokers compared with healthy non-smokers, Nrf2 protein was significantly activated in the cells, and 13 Nrf2-modulated genes were identified as highly smoking responsive. Comparing different indices created based on the extent of gene expression of Nrf2-modulated genes and other control genes, we observed a significant concordant regulation for Nrf2-modulated genes. To identify new human Nrf2-modulated genes, the index of Nrf2-modulated genes was correlated to all genes expressed in the small airway epithelium. There was a concordant regulation of gene expression among some highly associated genes, which was confirmed with TaqMan quantitative PCR and was independent of smoking history. Two additional highly smoking-responsive Nrf2-modulated genes in the small airway epithelium were identified, PIR and UGT1A4, both of which contain several antioxidant response elements 5′ to the gene. Using electrophoretic mobility shift assay, we found that some antioxidant response elements of PIR and UGT1A4 responded in vitro to activated Nrf2. These observations are consistent with the concept that Nrf2 plays an important role in regulating cellular defenses against smoking in the highly vulnerable small airway epithelium cell population, and that there is variability among the population in the relative Nrf2 responsiveness to a similar oxidant burden.
Oxidants and related radicals are known inducers of phase II detoxification and antioxidant pathways (1–4). The ability of cells to activate these pathways is an important cellular defense system. Studies of the “Cap ‘n’ Collar” family of transcription factors identified Nrf2 as the master regulator of antioxidant protection (8–11). The cellular mechanism of Nrf2 activation is not fully understood and may include several pathways. It is known that phosphorylation of Nrf2 through protein kinases controls the binding of Nrf2 to Keap1 in the cytoplasm, where it rests in the inactivated form and is targeted for ubiquitin-dependent degradation in the absence of antioxidant stress (12,13). Keap1 constitutively targets Nrf2, and this complex serves as a biological sensor in the cytosol (12,13). During oxidative stress, oxidants react with thiol groups of Keap1, eliciting a conformational change that causes a dissociation of Nrf2, allowing Nrf2 to translocate into the nucleus (12). An alternative model proposes that electrophilic stresses impair the Keap1-modulated proteosomal degradation of Nrf2, which enables Nrf2 synthesized de novo after exposure to the stress to accumulate directly in the nucleus, bypassing the Keap1 gate (13). Once in the nucleus, Nrf2 forms a heterodimer with muscle aponeurotic fibrosarcoma protein, and the heterodimer binds with high affinity to antioxidant response elements upstream of the anti-oxidant genes (8,12,13). In addition, mitogen-activated protein kinases also regulate antioxidant response element binding activity of Nrf2 (8).
The role of Nrf2 in the lung has been significantly advanced by studies of Nrf2 knockout mice exposed to a variety of mediators of lung injury. After stress, lungs of Nrf2-deficient mice are characterized by inflammation, evidence of oxidative stress, and increased apoptosis compared with the wild-type littermates (14–26). Microarray analyses of gene expression in lungs from Nrf2 knockout mice and wild-type controls have helped to identify Nrf2-modulated genes, for example, genes that help protect mice from development of cigarette smoke–induced lung injuries (23). Comprehensive studies of changes in gene expression using microarrays have helped to identify novel Nrf2-modulated genes (19,22,23). For example, Rangasamy et al. (23) have shown that after exposure to cigarette smoke, Nrf2 induces expression of 45 genes in the lung, mainly related to the glutathione, thioredoxin, and NADPH regenerating systems. Similarly, oxidative lung injury upregulated 125 Nrf2-modulated genes, including many known antioxidant genes, but also novel genes, such as protein kinase C α(19), a gene linked to asthma and pulmonary adenocarcinoma (44). Aoki et al. (15) demonstrated that Nrf2 knockout mice bred with transgenic mice that are susceptible to in vivo mutations (guanine phosphoribosyltransferase delta mice) have enhanced spontaneous and benz(a)pyrene–induced mutation rates in the lungs, suggesting that Nrf2 also protects genomic DNA from certain types of cancer. These observations are consistent with the concept that Nrf2-modulated genes confer protection against carcinogenesis. For example, the NQO1 gene catalyzes the detoxification of quinones and stabilizes the tumor suppressor p53 protein (45,46). Additionally, NQO1-deficient individuals have a higher risk of developing leukemia (47,48) and other type of cancers (49,50). Another example is the family members of glutathione-S-transferases (GSTs) that catalyze the conjugation of reduced glutathione. Humans lacking GST have a increased risk of developing cancer in many organs, including breast, colorectal, thyroid, lung, stomach, prostate, and bladder (51–57). Some cancer chemopreventive agents, such as isothiocyanates, induce Nrf2-modulated genes and are protective against carcinogen-induced tumorgenesis (10,58,59). Some cancer cell lines are also characterized by high induction of Nrf2-modulated genes. For example, mutations in Keap1 of lung cancer cell lines lead to an increased constitutive expression of Nrf2-modulated genes, with associated increased resistance to chemotherapeutic drugs (60–62).
Cigarette smoke can be divided into particulate matter and a gas phase, both rich sources of oxidants (1–4,63). The gas phase of cigarette smoke contains a large number of various species of oxidants and free radicals, including nitrogen-, carbon-, and oxygen-centered radicals, and produces esters and peroxyesters of nitrous and nitric acids (63,64). The radicals in tar include a mixture of semiquinones, hydroquinones, and quinones, which can penetrate mammalian cells and bind to and nick DNA (63,65). Other polyphenolic agents, such as catechol, also generate hydrogen peroxide and superoxide free radicals (63). It is estimated that more than 1014 free radicals are inhaled per puff in the respiratory tract, putting an enormous oxidative stress on the small airway epithelium (1,27–30). The epithelial surface of the respiratory tract has an antioxidant defense system that includes glutathione, vitamins C and E, β-carotene, uric acid, and the enzymatic antioxidants, such as superoxide dismutases, catalases, enzymes associated with glutathione metabolism, and other proteins such as peroxiredoxins, thioredoxins, glutaredoxins, heme oxygenases and reductases (2,4,11). The homeostasis of lung cells depends on these antioxidants, and lung injury results where oxidative stress exceeds antioxidant defenses. In smokers, cigarette smoking induces a broad spectrum of oxidant-related genes in the airway epithelium not seen in non-smokers (11,12,20,23,28–30,33–35,66,67). Interestingly, Spira et al. (67) have evaluated gene expression in the large airway epithelium of smokers with possible lung cancer and found that oxidant related defense genes were downregulated in smokers with proven lung cancers as compared with smokers without lung cancer.
Based on all reports of genes responsive to different oxidative stress conditions in the lungs of Nrf2 knockout mice (14,16–26), we compiled a list of 201 murine oxidative stress-responsive genes in the lung, and 187 human homologs were identified as expressed in small airway epithelium in humans. Using the gene expression data from 45 smokers and 38 nonsmokers, we identified a final list of 13 putative human, Nrf2-modulated, highly smoking-responsive genes. In agreement with the murine data (14,16–26), these genes showed a coordinate regulation in the small airway epithelium with certain smokers showing a consistently high level or consistently low level of Nrf2-modulated genes by microarray analysis with TaqMan-quantitative confirmation. Interestingly, the Nrf2 index was not dependent on current smoking, smoking history, or other demographic characteristics such as sex, age, and ethnicity, raising the possibility that the variability of Nrf2-modulated gene expression is, at least in part, genetically determined. This variability among smokers might lead to the hypothesis for future studies that healthy smokers with low Nrf2-modulated gene expression may represent a group of individuals at higher risk for smoking-induced lung injuries. A recent study identified a single polymorphism in the promoter of Nrf2 significantly associated with a higher risk of acute lung injury after major trauma in humans, and proposed Nrf2 as a candidate susceptibility gene for oxidative related illnesses (68). This is in agreement with a report published by Malhotra et al. (69), that the level of expression of Nrf2-modulated genes in the lung parenchyma were negatively correlated with severity of chronic obstructive lung disease.
The Nrf2 index, which is based on the expression level of the 13 putative human highly smoking-responsive genes, was used to identify further Nrf2-responsive genes expressed in human small airway epithelium. Most of the top genes which were identified as having a strong association with the Nrf2 index have been previously reported as Nrf2-modulated on the basis of transcription induction by Nrf2 in studies using Nrf2 knockout mice and/or by gel shift assays specific for Nrf2. Two additional highly smoking-responsive genes were identified, PIR and UGT1A4, both of which contain several functional antioxidant response elements. PIR, a gene of the cubin family, has been recently described to be highly upregulated by chronic cigarette smoking, and it was associated with bronchial epithelium cell apoptosis (70). Interestingly, the structure of pirin comprises two β-barrel domains, with an iron-binding cofactor, suggesting an enzymatic role in biological redox reactions (71). UGT1A4 is part of the complex locus encoding several UDP-glucuronosyltransferases, which are enzymes of the glucuronidation pathway that transforms small lipophilic molecules, such as steroids, bilirubin, hormones, and drugs, into water-soluble, excretable metabolites (43). Although other UDP-glucuronosyltransferases have already been connected with Nrf2 (16–19), UGT1A4 has not specifically been shown to have a functional antioxidant response element.
In conclusion, these observations demonstrate that Nrf2 plays an important role in regulating cellular defenses in highly vulnerable cell populations such as the small airway epithelium after cigarette smoking. Nrf2-modulated genes have been found to be upregulated in healthy cigarette smokers in a concordant pattern. The variability among the population in the relative Nrf2 responsiveness to a similar oxidant burden might suggest a genetic determination, which would have implications in the risk for lung disease related to cigarette smoking.
We thank the Fellows of the Division of Pulmonary and Critical Care Medicine for assistance with screening of subjects and bronchoscopies, T O’Connor and P Leopold for helpful discussions; M Teater, H Webb, T Raman, and D Dang for excellent technical assistance; H Lou for providing differential counts; and T Virgin-Bryan and N Mohamed for help in preparing this manuscript. These studies were supported, in part, by R01 HL074326, P50 HL084936 and UL1RR024996 (the WMC CTSC); and the Will Rogers Memorial Fund, Los Angeles, CA.
Online address: http://www.molmed.org
We declare that the authors have no competing interests as defined by Molecular Medicine, or other interests that might be perceived to influence the results and discussion reported in this paper.