help button home button
AJRCMB
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Perkowski, S.
Right arrow Articles by Albelda, S. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Perkowski, S.
Right arrow Articles by Albelda, S. M.
American Journal of Respiratory Cell and Molecular Biology. Vol. 28, pp. 682-696, 2003
© 2003 American Thoracic Society
DOI: 10.1165/rcmb.4692

Gene Expression Profiling of the Early Pulmonary Response to Hyperoxia in Mice

Sandra Perkowski, Jing Sun, Sunil Singhal, Jose Santiago, George D. Leikauf and Steven M. Albelda

Department of Clinical Studies-Philadelphia, School of Veterinary Medicine, and Department of Medicine, School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania; and Center for Environmental Genetics, University of Cincinnati, Cincinnati, Ohio

Address correspondence to: Sandra Perkowski, V.M.D., Ph.D., Dept. of Clinical Studies-Philadelphia, School of Veterinary Medicine, University of Pennsylvania, 3850 Spruce Street, Philadelphia, PA 19104-6010. E-mail: perksz{at}mail.vet.upenn.edu


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 References
 
To identify molecular events occurring during the early response to hyperoxia, we measured changes over time in total lung gene expression in C57BL/6 mice during prolonged exposure to > 95% O2. Specifically, differential gene expression of > 8,734 sequence-verified murine complementary DNAs was analyzed after 0, 8, 24, and 48 h of O2 exposure, with additional genes of interest analyzed at 24 h. Of the 385 genes differentially expressed, hyperoxia increased expression of 175 genes (2.0%) and decreased expression of 210 genes (2.3%). The majority of "classic" antioxidant enzymes, including catalase, MnSOD, and Cu-Zn SOD, showed no change in expression during hyperoxia, with a number of other antioxidant enzymes, including glutathione peroxidase, glutathione-S-Transferase (GST) {Pi}1, GST µ2, and heme oxygenase-1 showing relatively moderate increases. The exception was the heavy metal–binding protein metallothionein, which increased expression over 7-fold after 48 h of O2. We found no change in the expression of a number of known proinflammatory genes after 24 or 48 h of hyperoxia. A large increase in p21 expression was demonstrated, suggesting overall inhibition of cell cycle progression. Increases of the antiapoptotic gene Bcl-XL were counterbalanced by similar increases of the proapoptotic gene BAX. New findings included significant increases in expression of cysteine-rich protein 61(cyr61) at 48 h, suggesting a potential role for this factor in angiogenesis or remodeling of the extra cellular matrix during recovery from hyperoxia. In addition, downregulation of thrombomodulin expression occurred by 24 h and was further decreased at 48 h. Given the importance of thrombomodulin/thrombin interaction in regulating protein C activity, decreases in thrombomodulin may contribute to activation of the coagulation and inflammatory cascades and development of lung injury with hyperoxia.

Abbreviations: activator protein-1, AP-1 • activator protein C, APC • antioxidant response element, ARE • complementary DNA, cDNA • cysteine-rich protein 61, cyr61 • expressed sequence tags, ESTs • glyceraldehyde-3-phosphate dehydrogenase, GAPDH • glutathione peroxidase, GPx • glutathione reductase, GRed • oxidized glutathione, GSSG • glutathione-S-transferase, GST • heme oxygenase-1, HO-1 • intercellular adhesion molecule-1, ICAM-1 • interferon, IFN • interleukin, IL • IFN-{gamma}–inducible protein 10, IP-10 • ribosomal protein L32, L32 • lipopolysaccharide, LPS • macrophage inflammatory protein, MIP • messenger RNA, mRNA • metallothionein, MT • nitric oxide synthase, NOS • nuclear factor {kappa}B, NF-{kappa}B • plasminogen activator inhibitor-1, PAI-1 • platelet endothelial cell adhesion molecule-1, PECAM-1 • reactive oxygen species, ROS • RNAse protection assay, RPA • reverse transcriptase-polymerase chain reaction, RT-PCR • superoxide dismutase, SOD • transforming growth factor, TGF • thrombomodulin, TM • tumor necrosis factor, TNF


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 References
 
The pulmonary toxicity of prolonged exposure to high concentrations of oxygen is well recognized (1, 2). The pathogenesis of lung injury during oxygen exposure is thought to include direct endothelial and epithelial cell damage by an increase in reactive oxygen species (ROS), as well as release of products from activated leukocytes accumulating within the lung (3–5). The early stages of pulmonary oxygen toxicity begin with an initiation phase, occurring within the first 24 to 48 h of oxygen exposure, in which no morphologic injury is seen (1, 2). During this phase, however, many changes are occurring, including increases in the intracellular metabolism of oxygen and generation of reactive oxygen and nitrogen species. Changes in endothelial cell metabolism, crosslinking of membrane proteins, peroxidation of lipids, inhibition of cellular phosphatases, and DNA fragmentation have been found as a direct result of the increase in ROS (2, 5–7). The initiation phase of oxygen-induced lung injury, which culminates in morphologic evidence of endothelial injury, is rapidly followed by an inflammatory phase. This is associated with an influx of inflammatory cells, including neutrophils and platelets, into the airspace, release of inflammatory mediators, stimulation of alveolar macrophages, and rapid escalation of the morphologic signs of lung injury. Those factors that determine the extent of the inflammatory response and the ability to recover from oxygen exposure remain unknown.

Many of the early studies of hyperoxic lung injury focused on changes in antioxidant enzyme activity and its importance in protection from oxygen administration (6, 8). The majority of these early studies, however, were done using the rat as a model of oxygen toxicity, a species somewhat unique in its ability to upregulate antioxidant enzyme activity in the face of oxygen exposure. More recently, studies have focused on other cellular changes that may be important for oxygen toxicity, such as alterations in cytokine production, apoptotic activity, cell cycle proteins, or proteases (9–15).

Clearly, the development of lung injury during prolonged oxygen exposure is a complex process, associated with changes in expression of a number of genes important in the adaptive response to hyperoxia. Because it appears to be the balance between these factors, rather than any one factor, that modifies the development of lung injury during hyperoxia, we postulated that array technology would provide a powerful tool for studying the mechanisms behind this response. Accordingly, we first used a large-scale, custom array to study changes in the differential expression of > 8,734 genes over time, and then specifically looked at changes in the overall pattern of expression in genes associated with oxidative stress, inflammation, cell cycle progression, apoptosis, and extracellular matrix repair. We focused on changes occurring during the early stages of hyperoxic injury (8, 24, and 48 h of oxygen exposure) to dissect those changes occurring as a consequence of cellular activation rather than cellular injury. Second, because a number of known genes of interest were not included in the initial array analysis, a second set of arrays were examined using two commercially available murine kits (Clontech Mouse 1.2 and Stress arrays) (Clontech, Palo Alto, CA) encoding for 1,316 genes. Using this second set of arrays, we focused primarily on changes occurring at 24 h, a time when changes in gene expression and cellular trafficking are evident, but gross evidence of lung injury and emigration of inflammatory cells into the air space have not yet occurred. Third, we validated changes within specific sets of genes using semiquantitative reverse transcriptase–polymerase chain reaction (RT-PCR). Finally, we closely examined changes in genes likely to be expressed on endothelial cells as a more sensitive indication than current histopathologic techniques as to the onset of endothelial activation and/or injury during prolonged exposure to oxygen.


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 References
 
Animals and Exposure to Hyperoxia
C57BL/6 female mice (Taconic Farms Inc., Germantown, NY) were maintained at the animal facility a minimum of 2 wk before placement in the hyperoxic chamber to minimize stress from shipping. Mice (8–10 wk old) were placed in a Plexiglas chamber (30 x 38 x 76 cm3) through which 100% oxygen flowed at a continuous rate of 10 liters/min. The concentration of oxygen was maintained at > 95%, as measured using an oxygen analyzer (Model 600; ESD, Newark, DE), for a period of 8, 24, or 48 h. Mice were offered food and water ad libitum. At the designated time point, mice were anesthetized with ketamine and xylazine, and killed by cervical dislocation before harvesting of the lungs en bloc. All experimental protocols were approved by the Animal Care and Use Committee, University of Pennsylvania, Philadelphia, PA.

RNA Preparation
After mice were killed, lungs were immediately placed in 4 M guanidine isothiocyanate, 0.5% N-laurylsarcosine, 25 mM sodium citrate, and 0.1 M ß-mercaptoethanol solution and homogenized. Total lung RNA was isolated using a modified one-step method of acid guanidinium-thiocyanate phenol-chloroform extraction (16), followed by removal of contaminating genomic DNA by DNase I treatment (Roche Molecular Biochemicals, Indianapolis, IN). Only RNA with a 260/280 ratio of > 1.7 was used. To check for genomic DNA contamination, 2 µg of total RNA was used as a template in a PCR reaction with the primers for intronic sequences of the mouse PECAM-1 gene. No visible PCR product in total RNA sample was detected after 35 cycles, together with a positive control using as low as 500 pg of genomic DNA as a template in the PCR reaction.

Array Processing: Incyte Array
Microarray analysis. Differential gene expression of 8,734 cDNAs were assessed using whole lung mRNA. The microarray was fabricated by the Genomic and Microarray Laboratory, Center for Environmental Genetics, University of Cincinnati, (http://microarray.uc.edu/). Briefly, clones from the Incyte Genomics mouse GEM1 Library (Incycle Pharmaceuticals Inc., Palo Alto, CA) were amplified by PCR and printed onto glass slides (Omnigrid Microarrayer; GeneMachines, San Carlos, CA). This set of cDNAs contains 3,205 clones of known genes, 2,045 RIKEN cDNAs, 2,103 nonannotated expressed sequenced tags (ESTs), 1,066 annotated ESTs, and 315 DNA segments or hypothetical proteins. RNA from five mice at each time point (0, 8, 24, and 48 h of oxygen exposure) was collected. RNA from three mice at each time point were combined and 20 µg of the pooled RNA was run on each of three arrays. RNA from the additional two mice at each time point were combined and 20 µg of the pooled RNA was run on each of two arrays. Each RNA sample was labeled with 5' Cy5-labeled, random 9-mers during reverse transcription (Operon Technologies, Inc., Alameda, CA) as detailed previously (17). To account for technical variation, pooled samples were run in duplicate or triplicate and the average of the intensities of each gene was calculated for each group, with a total of five microarrays compared at each time (n = 5 slides; 3 mice/slide [n = 3], 2 mice/slide [n = 2]). In each group, RNA was reciprocally tagged randomly (i.e., the Cy3 and Cy5 dye switched, e.g., Cy3 for exposed and Cy5 for control). Cy3 and Cy5 samples were co-hybridized with the printed cDNAs. Following hybridization, slides were washed and scanned at 635 (Cy5) and 532 (Cy3) nm (GenePix 4000B; Axon Instruments, Inc., Union City, CA). To adjust for differences in probe labeling efficiency, balance coefficients (ratio of total Cy3 to total Cy5 fluorescence signal) were derived for each array. The average balance coefficient was 0.85 ± 0.04. This balance coefficient was then multiplied to each Cy3 value (minus background) and results expressed as a ratio. For inclusion in analysis, a cDNA covered its grid location on a slide by a minimum of 50% and a maximum of 400%. The average number of cDNAs passing these criteria per slide exceeded 8,750/slide. A mean and standard error was calculated for each balanced differential expression value using GeneSpring (Silicon Genetics, Redwood City, CA).

Data normalization and statistical analysis. Data normalization was performed in three steps for each microarray separately. First, channel-specific local background intensities were subtracted from median intensity of each channel (Cy3 and Cy5). Let X1 and X2 denote the Cy5 and Cy3 intensities after subtracting local backgrounds, respectively. Second, background-adjusted intensities were log-transformed and the differences and averages of log-transformed X1 and X2 were calculated: R = log2(X1) - log2(X2) and A = [log2(X1) + log2(X2)]/2. Third, data centering was performed by fitting the local regression model of R as a function of A (18). The difference between the observed log-ratio and the corresponding fitted value represented the normalized log-transformed gene expression ratio. Normalized intensities for the two channels were then calculated by adding a half of the normalized ratio to A for the Cy5 channel and subtracting a half of the normalized ratio from A for the Cy3 channel. The statistical analysis was performed by fitting the following analysis of variance (ANOVA) for each gene separately: Yijk = m + Ai + Tj + Ck+ eijk, where the Yijkn corresponds to the normalized log-transformed expression levels on the ith array (i = 1,...,16), kth channel (k = 1 for Cy5 and k = 2 for Cy3), and for jth experimental treatment (j = 1 for the control and j = 2, 3, 4 for 0, 8, 24, 48 h of hyperoxia, respectively), m is the overall mean intensity, Ai is the effect of the ith array, Tj is the jth experimental treatment, and Ck is the effect of the kth channel. Assumptions about model parameters were the same as previously described (19). This model was fitted per gene and statistical significance of the differential expression for each postexposure time compared with the pre-exposure control was assessed by calculating P values of linear contrasts. Finally, multiple hypothesis testing adjustment was applied by calculating False Discovery Rates (FDR) (20) for each (21). Each value is presented as a mean, standard error of the mean, P values, estimated differential expressions after adjusting for the dye effects, and corresponding FDR.

Clontech Array
To examine changes in known genes of interest not included in the initial array analysis, we next examined changes at each time point using a commercially available kit (Atlas Mouse Stress Array, Clontech) containing an additional 140 genes of interest. A total of two arrays were examined at each time point, with each array containing pooled samples from three mice (for a total n = 6/time point). To further examine changes at 24 h, and to account for biological and technical variation, additional RNA from seven control and six hyperoxic mouse lungs was studied and compared. A quantity of 2.5 µg total RNA was reverse transcribed to cDNA and labeled with 32P following the manufacturer's instructions (Clontech, Palo Alto, CA). The 32P-labeled cDNA probes from each mouse were then hybridized to two different nylon array filters for a total of 1,316 genes (Atlas Mouse 1.2 Array and Atlas Mouse Stress Array; Clontech). After washing, hybridized membranes were exposed to a phosphorimager screen for 5 d and scanned. The results were analyzed by Array Vision software v5.1 (Imaging Research Inc., St. Catherine's, ON, Canada). Arrays were used only once to minimize variability due to stripping procedures.

The intensity of each spot was corrected by subtracting the background counts obtained from four regions surrounding each spot. Values were then normalized by dividing the background-corrected spot intensities by the mean intensity of all the background-corrected spots on the array. Each array contained at least six "negative control" DNA spots (i.e., bacterial, lambda phage, or genomic DNA). The average value of the negative control spots on the Mouse 1.2 arrays was 0.16, with a high value of 0.25. Thus, all genes with expression levels of < 0.25 were considered too close to background and were not analyzed further. The average value of the negative control spots on the Mouse Stress arrays was 0.26, with a high value of 0.35. All genes on this array with expression levels < 0.35 were not analyzed further. The normalized mean intensity of each gene was calculated for each group and an unpaired t test used for statistical analysis.

To assess variability, RNA from three different lung samples (two controls and one hyperoxic animal) was run on duplicate arrays performed at different times. When analyzed using a best fit regression model, the average value of r2 was 0.93. However, in the three comparisons, 36% of genes had more than 2-fold differences. Because of this variability, a relatively large number of animals (seven control and six hyperoxic mice) were studied and artificial "cutoffs" were not used to define differences between the groups. Instead, t tests were performed comparing the normalized expression levels of each gene for each treatment, and P values computed.

Data Analysis: Array
Temporal clustering. Data were sorted according to the criteria that a significant (P < 0.05) change in expression level occurred in at least one time point during exposure to hyperoxia. Data were then clustered using self-organizing maps (SOMs), which are useful for time-related data (22) With this technique, the number of centroids or patterns examined may be chosen (www.genome.wi.mit.edu/MPR). Using the available data from the Incyte and Mouse Stress arrays, the data was organized into the 25 most common patterns, and a closer examination made of the four patterns with the largest changes in expression in at least one time point.

All genes that were significantly changed (P < 0.05) and at least 1.5-fold different under hyperoxic conditions were more closely examined as to functional relationships within specific subsets of genes (see below).

Semiquantitative RT-PCR: Confirmation of Selected Genes
A semiquantitative analysis of mRNA expression was performed to confirm and/or evaluate the differential expression of 44 selected genes at 24 h of hyperoxia, with 25 genes of these genes evaluated at all of the time points. Two micrograms of total RNA were reverse transcribed to cDNA using Oligo(dT)15 primer (Promega, Madison, WI) and powerscript reverse transcriptase (Clontech). Synthesized cDNA was then submitted to real-time PCR using either the LightCycler System (Roche Molecular Biochemicals) as previously described (23) or the SmartCycler System (Cephied, Sunnyvale, CA). The amount of cDNA was normalized using ß-actin levels. A minimum of three samples from control lungs and each hyperoxic time point were pooled and analyzed in quadruplicate. The relative expression level based on cycle number was compared between groups.

Northern Blot Analysis
cDNA probes for ß-actin, thrombomodulin (TM), and cysteine-rich protein 61 (cyr61) were synthesized as described above. A quantity of 0.5 µg of cDNA probe was labeled with Psoralen-Biotin, following the protocol provided by the BrightStar Psoralen-Biotin nonisotopic labeling kit (Ambion Inc., Austin, TX). A quantity of 4.4 µg of mRNA was denatured in formamide, separated on a 1% agarose gel containing formaldehyde and transferred to a positively charged nylon membrane (BrightStar-Plus; Ambion). After hybridization with the Psoralen-Biotin–labeled probe, bands were detected using the BrightStar BioDetect Kit (Ambion). Intensity of the bands was analyzed by densitometric scanning of exposed film and quantitated by using Quantity 1 (PDI Inc., Huntingdon Station, NY). The expression levels of TM and cyr61 mRNA in each lung sample were standardized using ß-actin mRNA as an internal marker.

RNase Protection Assays
At designated time points from 2–72 h, whole lungs were harvested for analysis of chemokine mRNA expression using commercially available multicytokine RNase protection assays (RPA; PharMingen, San Diego, CA). The intensity of each band was measured using a computer-linked PhorphorImager with ImageQuant software (Molecular Dynamics). Each intensity score was standardized as a relative ratio to L32 levels to correct for differences in RNA loading. The RPA was performed using a riboprobe template for mCK-5, which includes the C chemokine lymphotactin, the CC chemokines RANTES, eotaxin, MIP-1ß, MIP-1{alpha}, MCP-1, and TCA3, the CXC chemokines MIP-2 and interferon (IFN)-{gamma}–inducible protein 10 (IP-10), and the housekeeping genes L32 and GAPDH. A second RPA was performed using a riboprobe template for mCK-3b, which includes tumor necrosis factor (TNF)-{alpha}, TNF-ß, leukotriene-ß, interleukin (IL)-6, IFN-{gamma}, IFN-ß, transforming growth factor (TGF)-ß1, TGF-ß2, TGF-ß3, and macrophage migration inhibitory factor.

Protein Extraction and Western Blot Analysis
After killing of mice, mouse lung tissue was homogenized in ice-cold lysis buffer (10 mM Tris HCl, pH 7.4, 0.5% Triton X-100, 0.5% SDS) and protease inhibitor cocktail (complete protease inhibitor; Roche Molecular Biochemicals). The lysates were centrifuged at 12,000 rpm at 4°C in a microcentifuge for 15 min, and protein extracts were assayed for protein concentration, using the BCA Protein Assay kit (Pierce, Rockford, IL.). Five micrograms of extracted protein were separated by NuPAGE 10% Bis-Tris Gels (Invitrogen, Carlsbad, CA.) under either reducing or nonreducing conditions and electroblotted on to PolyScreen PVDF Transfer Membranes (NEN Life Science Products Inc., Boston, MA). Immunoblotting was accomplished using an anti–PECAM-1 polyclonal antibody produced against the cytoplasmic domain (24), anti-TM (a gift from Dr. Steve Kennel), and anti–ß-actin (Sigma, St. Louis, MO) antibodies. Protein bands were visualized by enhanced chemiluminescence with the ECL Kit (NEN Life Science Products Inc), and the intensity of bands was analyzed by densitometric scanning of exposed film and quantitated using Quantity 1 (pdi Inc.). The levels of PECAM-1 and TM proteins in each lung sample were standardized with that of ß-actin protein as a loading control.

Functional Thrombomodulin Activity
Whole lungs were harvested at 24 and 48 h for determination of thrombomodulin activity using a functional assay measuring the rate of hydrolysis of the chromogenic substrate S-2366 (Glu-Pro-Arg-p-nitroanilide, Chromogenix) at 405 nm (25). Briefly, lungs were removed and weighed and placed in 1 ml of extraction buffer (0.15 M NaCl, 20 mM Tris-HCl pH 7.5, 1% Triton X-100, 0.02% sodium azide, 5 mM iodoacetamide, 1 mM PMSF, 2 U/ml aprotinin) for every 20 mg of tissue. Tissues were then sonicated and placed at a ratio of 1:10 in dilution buffer buffer (0.15 M NaCl, 20 mM Tris-HCl pH 7.5, 5 mM CaCl2, 1 mg/ml gelatin). Ten microliters of diluted sample were then placed in 70 µl of reaction buffer containing 0.82 µM human protein C, 3.14 nM bovine thrombin, and 5 mM CaCl2. The reaction was allowed to proceed for 30 min at 37°C, then terminated by the addition of 2.98 µM recombinant hirudin. The activated protein C formed was diluted in 1:10 dilution buffer, and 200 µl of each sample was placed in duplicated wells of a 96-well plate. The chromogenic substrate spectrozyme Pca (0.2 mM) was then added to each well and the samples assayed at 405 nm at 1 min intervals for 15 min on a microplate reader. The results were compared with a freshly prepared standard curve.

Statistical Analysis
Analysis of the RT-PCR data was done using either an unpaired t test or one-way ANOVA with Fisher's exact test to compare differences between control and hyperoxic groups. RPA and protein data were analyzed by one-way ANOVA with Student-Neuman-Keuls (RPA) or Fisher's exact test. A P value <= 0.05 was accepted as indicating statistical significance.


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
References
 
Overall Gene Expression Changes
Evaluation of the mean intensity values of the 8,854 genes included in the array templates resulted in 8,031 genes (90.7%) with sufficient intensity above background levels in either the normal or hyperoxic group (see MATERIALS AND METHODS) for further analysis. Of the 8,031 genes further analyzed, 2,660 genes (33.1%) had significant changes (P < 0.05) from control values at any one time point. Over time, the number of genes with significant changes increased (8 h, 909 genes; 24 h, 1,052 genes; 48 h, 1,512 genes). A list of these genes, as well as a complete list of all gene expression changes, is available on the Web (www.uphs.upenn.edu/lungctr/academic_programs/pulmonary/research/labs/albelda) or by e-mailing the authors.

To evaluate temporal patterns of gene expression in response to hyperoxia, we grouped all significantly (P < 0.05) changed genes using self-organizing maps according to peak expression at 8, 24, and 48 h. From 25 patterns, we chose four as most interesting: early, transiently increased genes (Figure 1 , top left), genes increased consistently (Figure 1, top right), delayed, increased genes (Figure 1, bottom left) or decreased genes (Figure 1, bottom right). Genes sorted within each of these patterns are listed within each panel.



View larger version (56K):
[in this window]
[in a new window]
  		Figure 1. Self-organizing maps. All genes showing significant (P < 0.05) changes from control values were temporally categorized according to peak expression at 8, 24, and 48 h as early, transiently increased (A), increased consistently (B), delayed increased (C) or decreased (D) genes. Genes contained within each panel are listed accordingly.

 
We next evaluated genes according to functional relationships. In addition to using a P value <= 0.05 from control, only genes that were also differentially expressed (1.5-fold change from control), resulting in a total of 385 genes, were included in the analysis. A complete list is available on the Web (see website above) or by e-mailing the authors. However, no clear pattern of change was evident using this analysis. Of the 385 genes differentially expressed, hyperoxia increased expression of 175 genes (2.0%) and decreased expression of 210 genes (2.3%), supporting the emerging view that moderate oxidative stress specifically downregulates the expression of a number of genes (26). Table 1 lists the 15 genes (with P values <= 0.05) with the highest levels of up- and downregulation at 24 h. Genes with baseline expression levels approaching background and decreasing to levels below background are not included in the list of downregulated genes. A wide range of genes is represented here, including many genes without clearly understood functions in the lung. However, it is of interest that 3 of the 15 genes that were upregulated are involved in glutathione cycling. The cysteine-rich proteins cyr61 and metallothionein may also impact on glutathione levels by influencing cysteine reserves.


View this table:
[in this window]
[in a new window]
  		TABLE 1 Genes showing the greatest fold change by array analysis at 24 h (P <= 0.05)

 
Validation of Selected Findings using RT-PCR
Despite the use of large numbers of samples and statistical comparisons among genes, the multiple sources of variability inherent within array technique make validation of the results important. Therefore, for a subset of genes showing interesting changes in expression over time, results were validated using semi-quantitative RT-PCR and results compared with those obtained with the arrays (Table 2). Reassuringly, of the 14 genes evaluated in this fashion over time, where significant changes were noted, virtually every change seen on the arrays was similar in direction and magnitude to those seen using RT-PCR.


View this table:
[in this window]
[in a new window]
  		TABLE 2 Comparison of selected gene expression ratios over time: data generated by array versus RT-PCR

 
Analysis of Specific Subsets of Genes
Based on the known pathogenesis of hyperoxic lung injury, we more carefully analyzed changes in gene expression of (i) antioxidant enzymes that did or did not contain the antioxidant response promoter element, (ii) inflammatory mediators, (iii) cell cycle progression regulators, (iv) apoptosis and anti-apoptosis factors, (v) genes likely to be expressed on endothelial cells, and (vi) genes involved in extracellular matrix repair.

Antioxidant enzymes. Given that generation of reactive oxygen species, including superoxide anion and hydrogen peroxide (H2O2), has been demonstrated during exposure to high concentrations of oxygen (5), upregulation of antioxidant enzymes (such as superoxide dismutase [SOD], glutathione peroxidase [GPx], and catalase) represent a first line of defense against oxidative stress (Figure 2) . An increase in lung SOD activity, more specifically the mitochondrial fraction MnSOD, has been correlated with the development of oxygen tolerance in rats (6, 8, 27). Expression of these enzymes was thus examined in detail.



View larger version (36K):
[in this window]
[in a new window]
  		Figure 2. Changes in antioxidant enzyme gene expression at 24 h of hyperoxia. Changes in antioxidant enzyme gene expression are expressed using the GenMapp format. Genes that had a ratio (hyperoxic/normoxic) greater than 1.5 and that had a significance value of P <= 0.05 are labeled red. Those genes that had a ratio less than 0.66 (equivalent to a 1.5 fold change from baseline) and that a significance value of P <= 0.05 are labeled blue. Those genes whose P value was > 0.05 are labeled yellow, and those genes that were not on the array or had values below our thresholds and were not examined further using RT-PCR are labeled white. The actual ratios are shown to the right of each box. CuZnSOD, copper zinc-containing SOD; MnSOD, manganese-containing SOD; ECSOD, extracellular SOD; GPX, glutathione peroxidase; GSH, reduced glutathione; GSSG, oxidized glutathione; G6PD, glucose 6-phosphate dehydrogenase; 6PGD, 6-phosphogluconate dehydrogenase; R5P, ribulose-5-phosphate; GST, glutathione-S-transferase.

 
As shown in Figure 2 and Table 3, the antioxidant enzymes showed no change or were only moderately upregulated by exposure to hyperoxia. Although both MnSOD and catalase were included in the array, expression levels were too low for further analysis, so changes in gene expression were examined using RT-PCR. Using this technique, we found no evidence for increased expression of MnSOD or CuZn SOD in response to hyperoxia. Catalase, already at a low level in the lung (28), did not change at 24 h, and actually appeared to be downregulated by 48 h of exposure. A decrease in lung catalase protein and activity in response to hyperoxia in mice has been noted by others (29). In contrast, plasma GPx (GPx3) and GPx2 expression increased significantly in response to hyperoxia (although the changes were rather small: 1.54- and 1.77-fold over control), while the cellular isoform GPx1, as measured by RT-PCR, did not change (Figure 2, Table 3). Interestingly, GPx2, previously identified as a cellular isoform in the gastrointestinal tract of rodents, was significantly elevated in the lung as early as 8 h after exposure to hyperoxia. More recently, Cho and coworkers found a similar increase in this isoform after 48 h in the lungs of mice exposed to hyperoxia (30). In contrast, glutathione reductase (GRed) was transiently decreased early in response to hyperoxia (8 h), but showed no change at the later time points. A change in the activity of GPx relative to GRed could result in a change in the balance between GSH and oxidized glutathione (GSSG) and overall redox status of the lung.


View this table:
[in this window]
[in a new window]
  		TABLE 3 Changes in gene expression ratios measured by RT-PCR over time

 
As a comparison to the above, Waxman and coworkers (15) measured changes in the Mn-SOD, CuZn-SOD, catalase, GRed, and GPx enzyme activities in C57BL/6 mice exposed to 24 or 72 h of hyperoxia. 100% oxygen caused a modest increase in whole lung MnSOD mRNA expression by 24 h, although MnSOD, CuZnSOD, catalase, and GPx enzyme activities did not change. Glutathione reductase activity was decreased. Folz and coworkers (31) found an increase in total lung glutathione content, with a significant increase in the percentage of GSSG in mice exposed to 48 h of hyperoxia. An increase in GPx activity in bronchoalveolar lavage fluid was also noted. Therefore, although upregulation in endogenous antioxidant enzymes has been repeatedly demonstrated in rats in response to hyperoxia, this does not appear to readily occur in adult mice.

It has been recently recognized that, in addition to increased expression of "classic" antioxidant enzymes, the cellular response to oxidative stress includes increased expression of a number of phase II xenobiotic detoxification enzymes, due to activation of an antioxidant-response element (ARE) first discovered in 1991 (32). The ARE is a unique cis-acting regulatory sequence (TGAG/CTACT/C) in the 5' regulatory region of a number of genes important in antioxidant function and detoxification. In addition to their role in detoxification, these enzymes also serve as indirect antioxidants by detoxifying reactive electrophilic metabolites such as organic peroxides, lipid peroxides, epoxides, and quinones.

The transcription factor that binds to the ARE element has recently been identified as Nrf-2 (33). Nrf-2 belongs to the basic region/leucine zipper transcription family related to AP-1. A number of enzymes reported to be induced by Nrf-2 were included in the array or examined using RT-PCR (Table 3). These include a group of glutathione-S-transferases, NAD(P)H: Quinone oxidoreductase (NQO1), and other antioxidants including UDP glycosyl transferase as well as catalase and SOD, heme-oxygenase (HO)-1, GPx2. and GRed. As may be seen from the table, evidence for activation of the ARE during the first 48 h of response to hyperoxia was mixed. For example, transcription of glutathione-S-transferase (GST) Pi2 was increased by 24 h and remained increased at 48 h. Increased transcription of this enzyme has been described in conjunction with GSH depletion and/ or increased H2O2 (34), due to activation of the ARE. Similarly, GST Pi1 and mu2 were increased by 24 h, although GST Ya and Yc, which together represent the class {alpha} GST, did not change. Alterations in GST expression by oxidative stress may form part of the antioxidant cellular defense system in response to hyperoxia.

HO-1 is induced by a variety of stimuli, including ultraviolet radiation and hydrogen peroxide, and may also play a role in protection against oxidant injury. A moderate increase (1.9-fold) in HO-1 expression has been demonstrated in rat lungs after 24 h of hyperoxia, and these changes have been correlated with increases in protein expression and enzyme activity (35). In the present experiment, HO-1 was significantly increased at 24 h, although changes were moderate and we failed to find an increase in expression at later time points. Changes in protein expression and enzyme activity were not examined.

In comparison, Kleeberger and coworkers have demonstrated upregulation in many Nrf-2 regulated enzymes in mice after 48–72 h of oxygen exposure (30). For example, GPx2, HO-1, and NQO1 were significantly increased from control values after 48 h, whereas GST Ya and UDP glycosyl transferase were increased after 72 h. It is likely that similar changes may have been seen in our mice at later time points. Interestingly, mice genetically deficient in Nrf-2 are markedly more susceptible to hyperoxia, with a more rapid development of lung injury (30).

In contrast to the moderate changes notes in other antioxidant enzymes, the cysteine-rich protein metallothionein (MT) was among those genes most highly upregulated in response to hyperoxia, increasing expression over 7-fold by 48 h of hyperoxia (Table 3). Similar large increases in MT have been demonstrated over time in a model of nickel-induced lung injury (17). The MT-1 promotor contains an ARE (36) and elevations in MT-1 during various forms of oxidative stress, including ultraviolet radiation, paraquat toxicity, hydrogen peroxide, and hyperoxia (36, 37), may constitute part of a second line of defense. Upregulation of MT in response to cadmium has been shown to induce cross-tolerance to ozone and hyperoxia (38). Similarly, overexpression of MT decreases sensitivity of pulmonary endothelial cells to oxidant injury (39), and MT has been shown to quench a wide range of reactive oxygen and nitrogen species in vitro (36).

In summary, during the first 48 h of hyperoxia in the mouse lung, upregulation of antioxidant enzymes was generally limited. Increases in GPx activity relative to GRed activity, in concert with an increase in GST activity, however, could result in some change in intracellular redox potential over time, leading to changes within a number of key signaling pathways within the lung. Increases in metallothionein could form part of the antioxidant cellular defense system in response to hyperoxia. This lack of an early robust, protective antioxidant response suggests that approaches aimed at delivering antioxidant enzymes or inducing a more rapid upregulation of these enzymes might be useful in treatment of oxygen toxicity.

Inflammatory mediators. In addition to direct endothelial and epithelial cell damage caused by increased generation of ROS, the pathogenesis of pulmonary oxygen toxicity is thought to include release of products from activated leukocytes accumulating within the lung (3, 4). The influx of activated inflammatory cells into the airspace occurs after 48–72 h of oxygen exposure (2, 3), although increased retention of intrapulmonary neutrophils, suggesting a change in neutrophil–endothelial interaction, has been reported before emigration into the airspace (40). Preliminary findings from our laboratory showed that sequestration of leukocytes within the pulmonary microvasculature occurs by 8–12 h of hyperoxia, although infiltration into alveoli and significant lung injury occurs later (41). However, the mechanisms leading to the initial sequestration of leukocytes in the pulmonary vasculature remain unclear.

Reactive oxygen species have been shown to initiate a number of signaling events that lead to endothelial cell "activation" and upregulation of cell adhesion molecules and chemoattractants. More specifically, ROS have been shown to increase production of cytokines such as TNF-{alpha} (42), chemokines such as IL-8 (43), and adhesion molecules such as intercellular adhesion molecule (ICAM)-1 (44) in vitro. Published reports of hyperoxic lung injury in vivo have focused on mediators released after 24 h of exposure, in association with the later migration of neutrophils into the alveolar space.

Of the limited number of inflammatory mediators included within the array, the majority had expression levels too low for further analysis. Therefore, in separate experiments, changes in pulmonary mRNA expression of a number of inflammatory mediators were measured by a complementary technique of RNAase Protection assay (RPA). The C chemokine lymphotactin, which is chemotactic for T-lymphocytes, was significantly increased at 24 h, from a ratio of 2.96 ± 0.7 at baseline to 19.94 ± 9.77 relative to L32, with values returning toward baseline by 48 h. No other changes were found in the expression of any of the measured chemokines after 24 h of hyperoxia. Interestingly, the expression of the CXC chemokines MIP-2 and IP-10 was increased at later time points. The mRNA values for MIP-2, which acts as a neutrophil chemoattractant, were significantly increased from control values after 48 and 72 h of hyperoxia (1.6-fold and 2.5-fold, respectively) while that of IP-10, which acts as a monocyte and T-lymphocyte chemoattractant, was increased 2.3-fold at 72 h. The relative balance between these two chemokines may also play a role in regulating other processes, such as angiogensis, during the later stages of hyperoxia. Increased pulmonary expression of TNF-{alpha}, IL-1ß, and IL-6 occurs after 72 h of hyperoxia (45, 46), although the temporal relationship between these cytokines and neutrophil recruitment into the airspace suggest that other factors are involved in mediating neutrophil emigration in this model.

The limited number of adhesion molecules included in the array, including P-selectin and ICAM-1, showed no significant changes in expression after 24 h of hyperoxia. PECAM-1 RNA expression was downregulated after 24 h (Table 4), a finding confirmed using RT-PCR (see below). Significant increases in the pulmonary expression of adhesion molecules, including ICAM-1 (47), PECAM-1 (48), and P-selectin (49), have been reported to occur after 48 h of hyperoxic exposure, in association with increased neutrophil influx into the alveoli.


View this table:
[in this window]
[in a new window]
  		TABLE 4 Gene expression array ratios at 24 h: endothelial-related genes

 
Galectin-3 (previously known as Mac-2) was significantly elevated after 24 h of hyperoxia and continued to increase at 48 h (see Table 2). Galectin-3 is a ß-galactosidase–binding protein which has been implicated in a number of processes, including cell adhesion, cell growth regulation, inflammation, and tumor metastasis (50). Increases in galectin-3 may be associated with activation of incoming macrophages, and a similar increase has been shown in a nickel-induced model of inflammatory lung injury (17). Galectin-3 has also recently been shown to play an anti-apoptotic role in some cell lines and may stimulate cell proliferation (51).

NF-{kappa}B activation by intracellular generation of ROS is widely accepted (52, 53). Generation of ROS by oxidative stress causes phosphorylation and degradation of the I{kappa}B protein, with rapid translocation of the NF-{kappa}B dimer to the nucleus. NF-{kappa}B regulates transcription of a number of genes, including those regulating inflammation and immune responses, cell–cell interactions, apoptosis, and cellular proliferation. The importance of NF-{kappa}B activation during hyperoxia in vivo remains controversial. NF-{kappa}B activation, with upregulation of p50, p65, and c-rel subunits, has been demonstrated in intrapulmonary neutrophils, monocytes, and lymphocytes harvested from mice after 24 h of hyperoxic exposure (54, 55). Other studies, however, have found no change in NF-{kappa}B activity until 72 h of hyperoxia (56).

In the present study, proinflammatory downstream products of NF-{kappa}B activation, including TNF-{alpha}, IL-1ß, IL-6, MIP-2, iNOS, and ICAM-1 evaluated either by RPA, RT-PCR, or by array analysis, were unchanged (Table 3) at 24 h. Thus, our data do not show clear evidence of increased NF-{kappa}B activity after 24 h of hyperoxia, although increases in pulmonary expression of inflammatory mediators, such as TNF, ICAM-1, and MIP-2 after 48–72 h of oxygen exposure suggest that NF-{kappa}B activation may occur at these later time points.

In summary, upregulation of key inflammatory mediators, including cytokines and adhesion molecules, was not apparent after 24 h of hyperoxia, despite the fact that pulmonary sequestration of leukocytes can be demonstrated at this time. Based on data from other studies, it is likely that these genes become upregulated later in the course of hyperoxic damage.

Cell cycle progression. It is well established that injuries inducing DNA damage, such as irradiation or carcinogens, cause cell cycle arrest. Although increases in oxygen pressure or ROS can also result in mutations and DNA damage (57), much less is known about the effect of oxidant injury on cell cycle kinetics. Examination of cell cycle and proliferation genes present on the array revealed decreases in cyclin D1, E1, and cdk4 at 24 h, whereas cyclin A1 increased. Expression level of cyclin D2, D3, and the key transcriptional regulator E2F-1 did not change. Consistent with previous reports (12), there was a large increase in p21 gene expression at 24 h, with further increase at 48 h confirmed using RT-PCR (Table 2). p27 levels did not change, whereas expression levels of the inhibitors p57 and p18 were decreased. The levels of proliferating cell nuclear antigen did not change. Changes in TGF-ß were variable. Although a moderate increase was seen with RT-PCR at 24 h (Table 2), levels had returned to baseline by 48 h. Therefore, the role of TGF-ß in this response remains unclear, although Corroyer and coworkers, using an anti–TGF-ß antibody in cultured alveolar Type II epithelial cells in response to oxidant stress, have suggested that it may be involved in regulating this process (58).

Consistent with our findings and this in vitro work, induction of the growth inhibitor p21Cip1/WAF1 in the lungs of mice exposed to hyperoxia has been noted by two groups (12, 59). Marked increases in p21 message and protein was seen primarily 48 h and later after exposure to oxygen. In contrast to the upregulation of p21, McGrath noted no differences in expression of p27Kip1 or p53 mRNA. Recently, p21-deficient mice have been shown to develop greater lung injury and have decreased survival times in response to hyperoxia than wild-type mice (60), suggesting that the increase in p21 may be part of a protective response.

In summary, our data suggest that an overall inhibition of cell cycle progression was induced by 24 h of hyperoxia and may be secondary to increased expression of p21.

Apoptosis. Exposure to hyperoxia eventually leads to endothelial and epithelial cell destruction. The mechanisms for this injury are still not known completely, but almost certainly involve direct oxidant-induced DNA damage, as well as damage to proteins and lipids. The mode of cell death also remains unclear. Reactive oxidant species induce apoptosis in vitro in some systems (61, 62), whereas in other models, including a lung epithelial cell-like line (A549 cells), exposed to hyperoxia, cell death occurred without apoptosis (61). Hyperoxic lung injury in animal models is characterized by a cell death response that has features of both necrosis and apoptosis (12, 61, 63).

Figure 3 shows the relative expression levels of recognized pro- and antiapoptotic genes that were significantly expressed on the array. Relatively few of these genes showed significant changes in expression. It was interesting to compare expression levels of the pro- versus antiapoptotic genes of the Bcl-2 family. Although Bcl-2 levels were not altered, there was a large upregulation of the antiapoptotic gene Bcl-XL at 24 h (confirmed by PCR), which continued to increase over time. Of the proapoptotic members of the family, BAD and BAK did not change, but BAX was significantly increased. These data are strikingly similar to that reported in a study of O'Reilly and coworkers (13) studying Bcl-2 family gene expression in mouse lung during hyperoxia. Although later time points were studied (48–88 h), mice were reported to show increased mRNA levels of Bax and Bcl-2 with no changes in other family members. Interestingly, protein analysis showed increased levels of Bcl-XL, but not Bax. Because the ratio of these family members is key, it appears that, overall, the lung is slightly antiapoptotic at this early time point.



View larger version (34K):
[in this window]
[in a new window]
  		Figure 3. Changes in apoptosis-related gene expression at 24 h of hyperoxia. Changes in apoptosis gene expression are expressed using the GenMapp format. Genes that had a ratio (hyperoxic/normoxic) greater than 1.5 and that had a significance value of P <= 0.05 are labeled red. Those genes that had a ratio less than 0.66 (equivalent to a 1.5 fold change from baseline) and that a significance value of P <= 0.05 are labeled blue. Those genes whose P value was > 0.05 are labeled yellow, and those genes that were not on the array or had values below our thresholds and were not evaluated using RT-PCR are labeled white. The actual ratios are shown to the right of each box.

 
The majority of other studies examining apoptosis in hyperoxic lung injury have focused on later time points (48 or 72 h), making comparisons with our data difficult. For the most part, these studies have shown increased apoptosis as evidenced by TUNEL staining and increased immunostaining for p53 protein (11, 61, 63). Despite this increase, p53-null mice, Fas null mice, or mice treated with a caspase inhibitor (Z-VAD) and exposed to high levels of oxygen are not protected from lung injury, suggesting that either these particular apoptosis pathways, or apoptosis in general, many be only a minor contributor in the cell death induced by hyperoxia.

Endothelial cell–related changes. Although the toxic route of entry is inhalational rather than systemic during oxygen administration, many of the early signs of injury are associated with the endothelial, rather than the epithelial, side of the alveolar–capillary barrier, suggesting a prominent role for endothelial cell activation and injury in orchestrating further events. To determine how endothelial cell activation might contribute to the development of oxygen-induced lung injury, we examined a total of 16 relatively specific endothelial cell markers included in the array that had sufficient expression levels for further analysis (see Table 4). Six of the 16 genes evaluated had a greater than 1.5-fold change from control values at 24 h, although only 4 changes reached statistical significance.

Two of the more interesting changes were decreases in expression levels of PECAM-1 and thrombomodulin (TM). When re-examined using RT-PCR, the decrease in TM was significant (ratio = 0.63, P < 0.02). Changes in TM were further examined at 48 h using RT-PCR and revealed a further decrease in gene expression (ratio = 0.41, P < 0.006).

PECAM-1 is a highly expressed member of the immunoglobulin superfamily on the surface of endothelial cells that functions in both cell adhesion and signal transduction. It is most clearly recognized as a regulator of leukocyte transmigration (64). The decrease in the PECAM transcript was unexpected in light of the increases in PECAM-1 protein expression that have been previously reported in response to more prolonged exposures to oxygen (48). To determine whether changes in protein expression paralleled the changes seen in mRNA expression, immunoblot analysis was used. Although PECAM-1 mRNA was decreased, PECAM-1 protein expression was not changed from baseline values at the 24 or 48 h time points examined (Figure 4 , upper panel). This suggests that the PECAM protein may be relatively stable and that it lags behind the changes in RNA. It is also possible that PECAM-1 protein on the surface on infiltrating leukocytes or platelets may contribute to the total PECAM protein detected by immunoblot. Further studies using immunostaining, in situ hybridization, and intravascular marking with labeled antibodies are in progress to explore this finding.



View larger version (29K):
[in this window]
[in a new window]
  		Figure 4. Protein expression levels of PECAM-1 and thrombomodulin before and after hyperoxia. Whole lung protein extracts were prepared from four control animals (0 h) or four animals exposed to > 95% oxygen for 24 or 48 h. Equal amounts of protein were run on SDS-PAGE gels, transferred to nylon membranes, and immunoblotted with anti-PECAM mAb, anti–ß-actin mAb (as a loading control), or anti-Thrombomodulin mAb. Densitometry was performed and a relative density of expression (on an arbitrary scale) was calculated by normalizing for ß-actin expression. The upper panels show expression of PECAM-1 protein was unchanged at 24 and 48 h of hyperoxia. In contrast, the amount of thrombomodulin protein (lower panels) was significantly decreased at 24 h with even lower levels by 48 h.

 
Thrombomodulin is a membrane-bound glycoprotein found primarily on the endothelial cell that acts to regulate the coagulation pathway by modifying the action of thrombin (65). Thrombomodulin directly inhibits the procoagulant activities of thrombin, as well as acting as a cofactor in the protein C anticoagulant pathway. Activated protein C (APC) functions as an anticoagulant by proteolytic inactivation of the coagulation cofactors Va and VIIIa. In addition, a number of anti-inflammatory effects, including decreased leukocyte adhesion in vitro and in vivo, have been attributed to protein C activation (66–69). Finally, APC can also increase the fibrinolytic response by inhibiting plasminogen activator inhibitor 1 (PAI-1). Deposition of fibrin within the alveoli and an increase in PAI-1 expression has been demonstrated after 48–72 h of hyperoxia (9).

Given the potential importance of TM expression in endothelial cell activation and in modifying the coagulation, inflammatory and fibrinolytic cascades, further validation of changes in TM mRNA expression in total lung extracts was done. Northern analysis showed a 70% downregulation of message (Figure 5) . In addition, changes in mRNA expression were correlated to changes in protein expression by immunoblot analysis of total lung extracts. As shown in Figure 4 (lower panel), in contrast to PECAM-1, TM protein expression was significantly decreased by ~ 40% by 24 h, and continued to decrease to 80% of control values at 48 h.



View larger version (32K):
[in this window]
[in a new window]
  		Figure 5. Northern blot analysis of thrombomodulin mRNA expression after hyperoxia. Total RNA was extracted from lung tissues of three control mice and four mice exposed to 24 h of hyperoxia, respectively. Pooled samples were analyzed for thrombomodulin mRNA expression by Northern blot analysis. Densitometry was performed and a relative density of thrombomodulin mRNA expression was calculated by normalizing for ß actin mRNA expression (upper panels). As a comparison, changes in thrombomodulin mRNA expression, normalized to ß actin mRNA expression, as measured by RT-PCR at 24 h (n = 7/time point, mean ± SEM) are depicted in the lower panel.

 
Oxidation of TM in vitro has been shown to rapidly decrease 75–90% of TM activity (25). To determine whether decreases in TM activity paralleled the decreases in protein expression in vivo, we used an assay measuring TM activity by the ability of tissue homogenates to catalyze the conversion of a chromogenically labeled Protein C to APC in the presence of thrombin. Compared with control, TM activity was decreased to 80.7 + 8.9% (mean ± SEM) at 24 h and 53.8 ± 8.6% (P < 0.05) at 48 h of hyperoxia.

Although activation of the coagulation cascade and platelet accumulation within the lung in response to hyperoxia in vivo has been demonstrated by other investigators (3, 70), the significance of these events have not been explored in much detail. Given the key role of endothelial thrombomodulin in binding thrombin and activating Protein C, these data raise the intriguing possibility that decreases in thrombomodulin expression and/or activity and associated changes in the protein C pathway could contribute to not only to activation of the coagulation cascade, but also to inhibition of the fibrinolytic system and activation of the inflammatory response, and the development of lung injury in response to hyperoxia. Experiments to test this hypothesis are ongoing.

Extracellular matrix. Oxygen-induced lung injury resulted in increased expression of a number of genes involved in extracellular matrix repair (Table 2). For example, elastin was upregulated ~ 2-fold by 48 h and this increase was confirmed by RT-PCR. Another highly upregulated gene at 24 and 48 h after hyperoxia was Cysteine-rich protein 61 (cyr61). Cyr61 and its related family member, connective tissue growth factor (CTGF) are heparin-binding extracellular matrix–associated signaling proteins of the CCN protein family, a group of proteins involved in the regulation of cell growth and differentiation (71). Upon synthesis, both proteins are secreted and become associated with the cell surface and extracellular matrix, where they support cell adhesion, migration, and proliferation in fibroblasts. Both TGF-ß1–dependent and independent mechanisms for cyr61 upregulation have been described. Cyr61 in particular has been shown to regulate several genes important for wound healing, including those involved with angiogenesis (VEGF), inflammation (IL-1ß), extracellular remodeling (collagenase-1 [MMP-1], stromelysin-1 [MMP-3], TIMP1, uPa, and PAI-1), and cell–matrix interactions (integrins {alpha}3 and {alpha}5). Interestingly, increased expression of TIMP-1, PAI-1, uPA, and VEGF have all been demonstrated in response to hyperoxia (10, 14, 37).

Both RT-PCR (Table 2) and Northern blot analysis (Figure 6) confirmed a large increase in cyr61 expression at 48 h, suggesting that this protein may play an important role in the development of fibrosis after prolonged oxygen exposure. RT-PCR revealed that CTGF was similarly increased after 48 h.



View larger version (31K):
[in this window]
[in a new window]
  		Figure 6. Northern blot analysis of cyr61 mRNA expression after hyperoxia. Total RNA was extracted from lung tissues of three control mice and four mice exposed to 24 h of hyperoxia, respectively. Pooled samples were analyzed for cyr61 mRNA expression by Northern blot analysis. Densitometry was performed and a relative density of cyr61 mRNA expression was calculated by normalizing for ß actin mRNA expression (upper panels). As a comparison, changes in cyr61 mRNA expression, normalized to ß actin mRNA expression, as measured by RT-PCR at 8, 24, and 48 h (n = 3/time point, mean ± SEM) are depicted in the lower panel.

 
Limitations and Future Directions
Although the array provides an amazingly comprehensive overview of gene expression in the lung after hyperoxia, there are a number of caveats that must be kept in mind. First, array data provides information on gene expression levels. Important functional changes induced by post-translational modifications (such as phosphorylation) cannot be identified. It is also clear that changes gene expression levels may not necessarily reflect actual protein levels. For example, although PECAM mRNA expression fell, protein levels did not. Array data findings of potential pathophysiologic importance (i.e., changes in PECAM-1 or thrombomodulin mRNA levels) need to be studied at the protein or functional level. Second, in our experiments, we have examined whole lung tissue gene expression that includes epithelial cells, endothelial cells, interstitial cells, as well as circulating or adherent hematopoietic cells. Although some cell type "specific" genes can be identified (e.g., endothelial cell–related genes), most of the genes on the array are likely to be expressed by many cell types. The compact anatomy of the lung, especially in the alveolar spaces, will make unraveling specific cell type changes difficult, requiring techniques such as immunoelectron microscopy or laser capture microdissection. In addition, important changes in a subset of genes could easily be "diluted" out, markedly lowering the sensitivity of the analysis. Third, the large number of genes included in the initial array did not allow for careful analysis of every gene. Clearly, there were many additional genes with significant changes that would have been of interest to more fully examine. It would also have been exciting to further evaluate genes of unknown function (ESTs), that were up- or downregulated by hyperoxia; however, these experiments must await future studies.

Lastly, array data contains large amounts of variability, requiring confirmatory tests. We were encouraged to observe that, of those genes where array data was compared with semiquantitative RT-PCR, changes occurred in same direction and were of similar magnitude. Therefore, we feel the data from these arrays may be used to serve as guideposts to interesting changes.

Summary
In summary, most of the genes included in the array showed no significant change after 24 h of oxygen exposure. Of those genes that did change, the majority appeared to be downregulated rather than upregulated, supporting the emerging view that moderate oxidative stress specifically downregulates the expression of a number of genes. Relatively small changes in genes expressing lung antioxidant enzymes were seen, with only a few oxidant genes, including glutathione peroxidase, GST Pi1, and HO-1, showing moderate increases their expression in response to oxygen. The one exception was metallothionein, which increased dramatically in response to hyperoxia and may represent an important part of the antioxidant defense mechanism in mice.

We found no evidence for NF-{kappa}B activation after 24 h of oxygen exposure and relatively few changes in proinflammatory genes. A new finding was the downregulation of TM expression. Given that the TM/thrombin interaction is integral in regulating Protein C activity, it is interesting to speculate that decreases in TM may contribute not only to activation of the coagulation cascade, but also to activation of the inflammatory cascade and inhibition of fibrinolysis in response to hyperoxia. Large increases in the expression of cysteine-rich protein 61 suggest that this factor may also play an important role in changes within the extracellular matrix, angiogenesis, and the development of fibrosis in response to prolonged oxygen exposure. Studies are currently underway to more fully explore this possibility and the possible relationship to changes in the TM/thrombin/protein C cascade.


   Acknowledgments
 
The authors thank Dr. David Earle and Aron Fisher for their useful comments and suggestions, Eliza Windsor for her technical expertise and Mario Medvedovic for microarray statistical analysis. This work was partially supported by the SCOR in Acute Lung Injury ((NHLBI- P50-HL60290), SCOR in Hyperbaric Oxygen Therapy (AT-00428), NIH KO8 HL 67913-01A1, NHLBI-R01-HL65612, -HL656213, NIEHS-P30-ES06096 and –R01-10562.

Received in original form August 6, 2001

Received in final form November 22, 2002


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 References
 

  1. Clark, J. M., and C. J. Lambertsen. 1971. Pulmonary oxygen toxicity: a review. Pharmacol. Rev. 23:37–133.[Free Full Text]
  2. Crapo, J. D. 1986. Morphologic changes in pulmonary oxygen toxicity. Annu. Rev. Physiol. 48:721–731.[CrossRef][Medline]
  3. Barry, B. E., and J. D. Crapo. 1985. Patterns of accumulation of platelets and neutrophils in rat lungs during exposure to 100% and 85% oxygen. Am. Rev. Respir. Dis. 132:548–555.[Medline]
  4. Fox, R. B., J. R. Hoidal, D. M. Brown, and J. E. Repine. 1981. Pulmonary inflammation due to oxygen toxicity: involvement of chemotactic factors and polymorphonuclear leukocytes. Am. Rev. Respir. Dis. 123:521–523.[Medline]
  5. Freeman, B. A., and J. D. Crapo. 1981. Hyperoxia increases oxygen radical production in rat lungs and lung mitochondria. J. Biol. Chem. 256:10986–10992.[Free Full Text]
  6. Frank, L. 1991. Developmental aspects of experimental pulmonary oxygen toxicity. Free Radic. Biol. Med 11:463–494.[CrossRef][Medline]
  7. Smith, L. J. 1985. Hyperoxic lung injury: biochemical, cellular, and morphologic characterization in the mouse. J. Lab. Clin. Med. 106:269–278.[Medline]
  8. Crapo, J. D., B. E. Barry, H. A. Foscue, and J. Shelburne. 1980. Structural and biochemical changes in rat lungs occurring during exposures to lethal and adaptive doses of oxygen. Am. Rev. Respir. Dis. 122:123–143.[Medline]
  9. Barazzone, C., D. Belin, P.-F. Piguet, J.-D. Vassalli, and A.-P. Sappino. 1996. Plasminogen activator inhibitor-1 in acute hyperoxic mouse lung injury. J. Clin. Invest. 98:2666–2673.[Medline]
  10. Barazzone, C., Y. R. Donati, A. F. Rochat, C. Vesin, C. Kan, J. C. Pache, and P. F. Piguet. 1999. Keratinocyte growth factor protects alveolar epithelium and endothelium from oxygen-induced injury in mice. Am. J. Pathol. 154:1479–1487.[Abstract/Free Full Text]
  11. O'Reilly, M., R. J. Staversky, B. R. Stripp, and J. N. Finkelstein. 1998. Exposure to hyperoxia induces p53 expression in mouse lung epithelium. Am. J. Respir. Cell Mol. Biol. 18:43–50.[Abstract/Free Full Text]
  12. O'Reilly, M. A., R. J. Staversky, R. H. Watkins, and W. H. Maniscalco. 1998. Accumulation of p21CipI/WAFI during hyperoxic lung injury in mice. Am. J. Respir. Cell Mol. Biol. 19:777–785.[Abstract/Free Full Text]
  13. O'Reilly, M. A., R. J. Staversky, H. L. Huyck, R. H. Watkins, M. B. LoMonaco, C. T. D'Angio, R. B. Baggs, W. H. Maniscalco, and G. S. Pryhuber. 2000. Bcl-2 family gene expression during severe hyperoxia induced lung injury. Lab. Invest. 80:1845–1854.[Medline]
  14. Ward, N. S., A. B. Waxman, R. J. Homer, L. L. Mantell, O. Einarsson, Y. Du, and J. Elias. 2000. Interleukin-6-induced protection in hyperoxic acute lung injury. Am. J. Respir. Cell Mol. Biol. 22:535–542.[Abstract/Free Full Text]
  15. Waxman, A. B., O. Einarsson, T. Seres, R. G. Knickerlbein, J. B. Warshaw, R. Johnston, R. J. Homer, and J. A. Elias. 1998. Targeted lung expression of interleukin-11 enhances murine tolerance of 100% oxygen and diminishes hyperoxia-induced DNA fragmentation. J. Clin. Invest. 101:1970–1982.[Medline]
  16. Chomcyzynski, P., and N. Sacchi. 1987. Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction. Anal. Biochem. 162:156–159.[Medline]
  17. McDowell, S. A., K. Gammon, C. J. Bachurski, J. S. Wiest, J. E. Leikauf, D. R. Prows, and G. D. Leikauf. 2000. Differential gene expression in the initiation and progression of nickel-induced acute lung injury. Am. J. Respir. Cell Mol. Biol. 23:466–474.[Abstract/Free Full Text]
  18. Dudoit, S., Y. Yang, and T. P. Speed. 2002. Statistical methods for identifying differentially expressed genes in replicated CDNA microarray experiments. Statistica Sinica 12:111–139.
  19. Wolfinger, R. D., G. Gibson, E. D. Wolfinger, L. Bennett, H. Hamadeh, P. Bushel, C. Afshari, and R. S. Paules. 2001. Assessing Gene Significance From CDNA Microarray Expression Data Via Mixed Models. J. Comput. Biol. 8:625–637.[CrossRef][Medline]
  20. Benjamini, Y., and Y. Hochberg. 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. Royal Statistical Soc. B. 57:289–300.
  21. Tusher, V. G., R. Tibshirani, and G. Chu. 2001. Significance analysis of microarrays applied to the ionizing radiation response. Proc. Natl. Acad. Sci. USA 98:5116–5121.[Abstract/Free Full Text]
  22. Tamayo, P., D. Slonim, J. Mesirov, Q. Zhu, S. Kitareewan, E. Dmitrovsky, E. S. Lander, and T. R. Golub. 1999. Interpreting patterns of gene expression with self-organizing maps: methods and application to hematopoietic differentiation. Proc. Natl. Acad. Sci. USA 96:2907–2912.[Abstract/Free Full Text]
  23. Wittwer, C. T., K. M. Ririe, R. V. Andrew, D. A. David, R. A. Gundry, and U. J. Balis. 1997. The LightCycler: a microvolume multisample fluorimeter with rapid temperature control. Biotechniques 22:176–181.[Medline]
  24. Cao, M. Y., M. Huber, N. Beauchemin, J. Famiglietti, S. M. Albelda, and A. Veillette. 1998. Regulation of mouse PECAM-1 tyrosine phosphorylation by the Src and Csk families of protein-tyrosine kinases. J. Biol. Chem. 273:15765–15772.[Abstract/Free Full Text]
  25. Glaser, C. B., J. Morser, J. H. Clarke, E. Blasko, K. McLean, I. Kuhm, R.-J. Chang, J.-H. Lin, L. Vilander, W. H. Andrews, and D. R. Light. 1992. Oxidation of a specific methionine in thrombomodulin by activated neutrophil products blocks cofactor activity: a potential rapid mechanism for modulation of coagulation. J. Clin. Invest. 90: 2565–2573.
  26. Morel, Y. 1999. Repression of gene expression by oxidative stress. Biochem. J. 342:481–496.
  27. Lewis-Molock, Y., K. Suzuki, N. Taniguchi, D. H. Nguyen, R. J. Mason, and C. W. White. 1994. Lung manganese superoxide dismutase increases during cytokine-mediated protection against pulmonary oxygen toxicity in rats. Am. J. Respir. Cell Mol. Biol. 10:133–141.[Abstract]
  28. Chabot, F., J. A. Mitchell, J. M. C. Gutteridge, and T. W. Evans. 1998. Reactive oxygen species in acute lung injury. Eur. Respir. J. 11:745–757.[Abstract]
  29. Levy, M. A., Y.-H. Tsai, A. Reaume, and T. M. Bray. 2001. Cellular response of antioxidant metalloproteins in Cu/An SOD transgenic mice exposed to hyperoxia. Am. J. Physiol. Lung Cell. Mol. Physiol. 281:L172–L182.[Abstract/Free Full Text]
  30. Cho, H.-Y., A. E. Jedlicka, S. P. M. Reddy, T. W. Kensler, M. Yamamoto, L.-Y. Zhang, and S. R. Kleeberger. 2002. Role of NRF2 in protection against hyperoxic lung injury in mice. Am. J. Respir. Cell Mol. Biol. 26:175–182.[Abstract/Free Full Text]
  31. Folz, R. J., A. M. Abushamaa, and H. B. Suliman. 1999. Extracellular superoxide dismutase in the airways of transgenic mice reduces inflammation and attenuates lung toxicity following hyperoxia. J. Clin. Invest. 103:1055–1066.[Medline]
  32. Rushmore, T. H., M. R. Morton, and C. R. Pickett. 1991. The antioxidant responsive element:activation by oxidative stress and identification of the DNA consensus sequence required for functional activity. J. Biol. Chem. 266:11632–11639.[Abstract/Free Full Text]
  33. Venugopal, R., and A. K. Jaiswal. 1998. Nrf2 and Nrf1 in association with Jun proteins regulate antioxidant response element-mediated expression and coordinated induction of genes encoding detoxifying enzymes. Oncogene 17:3145–3156.[CrossRef][Medline]
  34. Forman, H. J., R.-M. Liu, and L. Tian. 1997. Glutathione cycling in oxidative stress. In Lung Biology in Health and Disease: Oxygen, Gene Expression, and Cellular Function. L. B. Clerch and D. J. Massaro, editors. Vol. 105. Marcel Dekker, New York. 99–121.
  35. Lee, P. J., J. Alam, S. L. Sylvester, N. Inamdar, L. Otterbein, and A. M. K. Choi. 1996. Regulation of heme oxygenase-1 expression in vivo and in vitro in hyperoxic lung injury. Am. J. Respir. Cell Mol. Biol. 14:556–568.[Abstract]
  36. Andrews, G. K. 2000. Regulation of metallothionein gene expression by oxidative stress and metal ions. Biochem. Pharmacol. 59:95–104.[CrossRef][Medline]
  37. Piedboeuf, B., C. J. Johnston, R. H. Watkins, B. B. Hudak, J. S. Lazo, M. G. Cherian, and S. Horowitz. 1994. Increased expression of tissue inhibitor of metalloproteinases (TIMP-1) and metallothionein in murine lungs after hyperoxic exposure. Am. J. Respir. Cell Mol. Biol. 10:123–132.[Abstract]
  38. Hart, B. S., G. W. Voss, M. A. Shatos, and J. Doherty. 1990. Cross tolerance to hyperoxia following cadmium aeroxol pretreatment. Toxicol. Appl. Pharmacol. 103:255–270.[CrossRef][Medline]
  39. Pitt, B. R., M. Schwarz, E. D. Woo, E. Yee, K. Wasserloos, S. Tran, W. Weng, R. J. Mannix, S. A. Watkins, Y. Y. Tyurina, V. A. Tyurin, V. E. Kagan, and J. S. Lazo. 1997. Overexpression of metallothionein decreases sensitivity of pulmonary endothelial cells to oxidant injury. Am. J. Physiol. (Lung Cell Mol. Physiol.)273:L856–L865.
  40. Rinaldo, J. E., D. English, J. Levine, R. Stiller, and J. Henson. 1988. Increased intrapulmonary retention of radiolabeled neutrophils in early oxygen toxicity. Am. Rev. Respir. Dis. 137:345–352.[Medline]
  41. Perkowski, S. Z., A. Scherpereel, G. Masiko, S. M. Albelda, and M. Christofidou-Solomidou. 2000. Pulmonary neutrophil accumulation and chemokine expression during early hyperoxia in mice. FASEB J. 14:A607. (Abstr.)
  42. VanOtteren, G. M., T. J. Standiford, S. L. Kunkel, J. M. Danforth, and R. M. Strieter. 1995. Alterations of ambient oxygen tension modulate the expression of tumor necrosis factor and macrophage inflammatory protein-1{alpha} from murine alveolar macrophages. Am. J. Respir. Cell Mol. Biol. 13:399–409.[Abstract]
  43. DeForge, L. E., A. M. Preston, E. Takeuchi, J. Kenney, L. A. Noxer, and D. G. Remick. 1993. Regulation of interleukin 8 gene expression by oxidant stress. J. Biol. Chem. 268:25568–25576.[Abstract/Free Full Text]
  44. Nishio, K., Y. Suzuki, T. Aoki, K. Suzuki, A. Miyata, N. Sato, K. Naoki, H. Kudo, H. Tsumura, H. Serizawa, S. Morooka, Y. Ishimura, M. Suematsu, and K. Yamaguchi. 1998. Differential contribution of various adhesion molecules to leukocyte kinetics in pulmonary microvessels of hyperoxia-exposed rat lungs. Am. J. Respir. Crit. Care Med. 157:599–609.
  45. Jensen, J. C., H. W. Pogrebniak, H. I. Pass, C. Buresh, M. J. Merino, D. Kauffman, D. Venzon, H. N. Langstein, and J. A. Norton. 1992. Role of tumor necrosis factor in oxygen toxicity. J. Appl. Physiol. 72:1902–1907.[Abstract/Free Full Text]
  46. Johnston, C. J., T. W. Wright, C. K. Reed, and J. N. Finkelstein. 1997. Comparison of adult and newborn pulmonary cytokine mRNA expression after hyperoxia. Exp. Lung Res. 23:537–552.[Medline]
  47. Welty, S. E., J. L. Rivera, J. F. Elliston, C. V. Smith, T. Zeb, C. M. Ballantyne, C. A. Montgomery, and T. N. Hansen. 1993. Increases in lung tissue expression of intercellular adhesion molecule-1 are associated with hyperoxic lung injury and inflammation in mice. Am. J. Respir. Cell Mol. Biol. 9:393–400.
  48. Piedboeuf, B., M. Gamache, J. Frenette, S. Horowitz, H. S. Baldwin, and P. Petrov. 1998. Increased endothelial cell expression of platelet-endothelial cell adhesion molecule-1 during hyperoxic lung injury. Am. J. Respir. Cell Mol. Biol. 19:543–553.[Abstract/Free Full Text]
  49. Zeb, T., F. B. Piedboeuf, M. Gamache, C. Langston, and S. E. Welty. 1996. P-selectin is upregulated early in the course of hyperoxic lung injury. Free Rad. Biol. Med. 21:567–574.[CrossRef][Medline]