Recent evidence suggests that alveolar epithelial cells (AECs) may contribute to the development, propagation, and resolution of acute lung injury (ALI)/acute respiratory distress syndrome (ARDS). Proinflammatory cytokines, pathogen products, and injurious mechanical ventilation are important contributors of excessive inflammatory responses in the lung. In the present study, we used cDNA microarrays to define the gene expression patterns of A549 cells (an AEC line) in the early stages of three models of pulmonary parenchymal cell activation: cells treated with tumor necrosis factor-α (TNFα) (20 ng/ml), lipopolysaccharide (LPS, 1 μg/ml), or cyclic stretch (20% elongation) for either 1 h or 4 h. Differential gene expression profiles were determined by gene array analysis. TNFα induced an inflammatory response pattern, including induction of genes for chemokines, inflammatory mediators, and cell surface membrane proteins. TNFα also increased genes related to pro- and anti-apoptotic proteins, signal transduction proteins, and transcriptional factors. TNFα further induced a group of genes that may form a negative feedback loop to silence the NFκB pathway. Stimulation of AECs with mechanical stretch changed cell morphology and activated Src protein tyrosine kinase. The combination of TNFα plus stretch enhanced or attenuated expression of multiple genes. LPS decreased microfilament polymerization but had less impact on NFκB translocation and gene expression. Results from this study indicate that AECs can tailor their response to different stimuli or/and combination of stimuli and subsequently play an important role in acute inflammatory responses in the lung.
- acute respiratory distress syndrome
- acute lung injury
- ventilator-induced lung injury
- significant analysis of microarray
acute respiratory distress syndrome (ARDS) and acute lung injury (ALI) are part of a spectrum of acute inflammatory responses of the lung, incurring an unacceptably high mortality (40–50%) (21). Although mechanical ventilation is a life-saving strategy, in these patients, survival may be negatively influenced by the effects of injurious mechanical stretch (5). The clinical presentation of ventilator-induced lung injury (VILI) is indistinguishable to that of other causes of ALI. The functional injury is characterized by diffuse alveolar damage, structural changes in the alveolocapillary unit and loss of integrity of the alveolocapillary membrane. Injurious ventilation, pathogen products [such as lipopolysaccharide (LPS)], and proinflammatory cytokines [such as tumor necrosis factor-α (TNFα)] may all cause ALI and have been implicated in the pathophysiology of ARDS (21). However, it is not clear to what extent the early response of the lung to different proinflammatory or injurious signals is distinct and/or specific.
Most studies examining ALI have focused on the role of inflammatory cells, such as alveolar macrophages, neutrophils, monocytes, and lymphocytes. Until recently, there was a paucity of data in the literature exploring the role of pulmonary epithelial cells in the development and propagation of ALI. Over the past few years, it has become apparent that alveolar epithelial cells (AECs) are not merely innocent targets for inflammatory mediators but actively participate in the inflammatory response (17, 34). A number of studies have demonstrated that AECs can produce a host of inflammatory mediators such as monocytes chemoattractant protein 1 (MCP-1), interleukin-8 (IL-8), TNFα, and IL-6 in response to inflammatory stimuli (17). A recent work by Poynter and coworkers (28) has demonstrated that selective inhibition of the transcriptional factor nuclear factor κB (NFκB) in airway epithelial cells significantly reduced LPS-induced ALI. Tremblay and colleagues (41) demonstrated that an injurious ventilatory strategy induced production of various cytokines by AECs. Consequently, lung epithelial cells may play a prominent role in orchestrating innate immunity and inflammatory responses in the lung. This is of particular significance since one of the main challenges in developing therapeutic strategies for ALI/ARDS has been the inability to arrest injurious inflammatory responses while preserving the immune system intact.
To enhance our understanding of the role of AECs in ARDS/VILI, we undertook a comparative examination of the gene expression profile of these cells in response to injurious stimuli that are thought to be important in the pathogenesis of ARDS/VILI: cyclic stretch, TNFα, and LPS. Microarray technique allows us to compare the expression of huge numbers of genes (29). However, its application to lung tissues is limited by the dramatic change in cell population at the site of inflammation. It is often difficult to tease out whether the observed changes in various genes are associated with the underlying parenchymal cells or secondary to infiltrating cells. The involvement of multiple inflammatory mediators and cytokine networks also makes the interpretation of such results difficult. For this reason, cultured cells are thought to provide a “cleaner” background to study gene expression in specific cell types in the lungs.
In the present study, we used cDNA microarrays to define gene expression patterns of A549 cells (an AEC line) in the early stages of three models of pulmonary parenchymal cell activation. We show that AECs can act as “effector” cells by tailoring their responses to different injurious agents and altering the expression of genes pertinent to the inflammatory responses. TNFα resulted in the most dramatic change in gene expression with a unique pattern of responses. AECs also responded to mechanical stretch and LPS; however, the alteration of gene expression profiles was less dramatic. Based on functional expression analysis, these experiments highlight the unique role TNFα plays in ALI at the cellular level.
MATERIALS AND METHODS
Cell Culture and Stimulation
Immortalized human pulmonary epithelial cells (A549 cells) [American Type Culture Collection (ATCC), Manassas, VA] were seeded at a density of 0.5 × 106 cells/well on BioFlex 6-well plates coated with collagen I (Flex I; Flexcell International, McKeesport, PA). Cells were grown overnight in Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum. For each experiment there were five groups: 1) control (no treatment), 2) mechanical stretch (18 kPa, 30 cycles/min), 3) LPS [1 μg/ml, E. coli (026:B6, 20 million U/mg); Sigma-Aldrich, St. Louis, MO], 4) TNFα [20 ng/ml (bioactivity ED50 range = 0.02–0.05 ng/ml); BioSource, Camarillo, CA], and 5) mechanical stretch plus TNFα. Cells were treated for either 1 h or 4 h. The Flexcell Strain Unit is designed to provide uniform radial and circumferential strain to a membrane surface. This is achieved by applying vacuum pressure to the stretchable membrane. Deformation of the elastic membrane leads to elongation of cells growing attached on the surface of the membrane (see the Flexcell International web site, at http://www.flexcellint.com/). For each time point, experiments were repeated twice. Totally 20 samples were used for microarray analysis.
Additional cell cultures were performed for confirmation studies. To further determine the effects of TNFα, LPS, and mechanical stretch on gene expression from A549 cells, similar experiments were also conducted at Mayo Clinic. The dose of TNFα, amplitude of stretch used were the same as described above, except the stretch apparatus was constructed locally, which could provide uniform and reproducible mechanical strain on cells cultured on flexible membrane (37). Human bronchial small airway epithelial cells (BEAS-2B, ATCC) were used for comparison. Experiments using BEAS-2B were performed in the same fashion as the experiments described for A549 cells, except cultured on 6-well plastic plates.
Total RNA was isolated for microarray, as well as for real-time quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR) studies. Once the treatment was completed, medium was removed, and 1 ml of TRIzol reagent (Life Technologies, Rockville, MD) was added to each well. Cells were placed at 4°C, and total RNA extraction was performed (RNeasy kit; Qiagen, Mississauga, Ontario, Canada). RNA quality was ensured by spectrophotometric analysis (OD260/280) and gel visualization (11, 22). All samples demonstrated high-quality characteristics using Test Probe Array (Affymetrix, Santa Clara, CA), prior to the final DNA GeneChip hybridization.
The Affymetrix GeneChip Human Genome Focus Array were used for transcript profiling, which represents over 8,500 verified human sequences from the National Center for Biotechnology Information (NCBI) RefSeq database (for specific information see http://www.affymetrix.com/products/arrays/specific/focus.affx). The data for 8,973 genes were generated from 20 hybridizations. Raw data was submitted to NCBI Gene Expression Omnibus (GEO) database (GEO accession number GSE1541). The data were scaled from each array to a target intensity value of 150 with Affymetrix Microarray Suite 5.0 software (MAS 5.0). After scaling, the data were imported into a statistical program, JMP (Statistical Discovery Software; SAS Campus Drive, Cary, NC). Normalization was performed according to the strategy described in Tusher et al. (44). Briefly, a reference data set was generated by averaging the expression of each gene over all 20 hybridizations. The data for each hybridization were compared with the reference data set in a cube root scatter plot. Subsequently, a linear least square fit to the cube root scatter plot was used to calibrate each hybridization. Normalized data were imported into Excel (Microsoft, Seattle, WA). Significant changes of genes in any of the groups were identified using the multiclass response parameter of the Significance Analysis of Microarrays(SAM) algorithm with 100 permutations (http://www-stat.stanford.edu/∼tibs/SAM/index.html; Ref. 45). Affymetrix MAS 5.0 was used to assess significant genes by comparing control samples to experimental samples. Comparison log ratios were used to determine fold change and their P values, in the expression of SAM significant genes.
Reverse Transcription of RNA and Real-Time PCR
Synthesis for single-strand cDNA was carried out using 1 μl of oligo-dT that was annealed to 10 μl of RNA (10 μg) at 70°C for 10 min and chilled on ice. Then, 9 μl of reaction mix [4 μl of 5× first-strand buffer, 2 μl of 0.1 M DTT, 1 μl of 10 mM deoxy-NTPs mixture, 2 μl of SuperScript II (Invitrogen, Burlington, Canada)] was added and incubated at 42°C for 60 min (11, 22).
Conditions for PCR included 50°C for 2 min, 95°C for 15 min, and 40 cycles of 95°C for 15 s, 60°C for 60 s, and an extension time of 72°C for 30 s (ABI Prism 7900HT; Applied Biosystems, Foster City, CA). Each assay included a standard curve of five serial dilutions of a known concentration of genomic DNA and a nontemplate control. Each well was filled with 4 μl of cDNA (100 ng), 1 μl of PCR 10× buffer, 0.6 μl of 25 mM of MgCl2, 0.2 μl of deoxy-NTPs, 0.2 μl of primer mix, 0.1 μl of HotStar Taq DNA Polymerase (Qiagen), 3.4 μl of sterile distilled water, 0.3 μl of SYBR Green I (Molecular Probes, Eugene, OR), and 0.2 μl of ROX Internal Reference Dye (Invitrogen). All assays were performed in triplicate. Primers were designed using the Primer Express 1.5 software (Applied Biosystems) and purchased from ACGT (Toronto, Ontario, Canada). Primer sequences are available upon request.
Cells were cultured on collagen I-coated BioFlex plates and exposed to mechanical stretch, or cultured on glass coverslips for LPS and TNFα treatment, and then subjected to fluorescence staining at room temperature (11, 22). Cells were fixed in 3.7% formaldehyde for 10 min, and permeabilized with 0.25% Triton X-100 in PBS for 5 min. After washing, cells were incubated with indicated primary antibody for 1 h, followed by incubation with Alexa-594-conjugated goat anti-rabbit IgG (Molecular Probes, Eugene, OR) at 1:1,000 dilution for 1.5 h in the dark. Primary antibodies used were anti-Src pTyr416 antibody (BioSource International) at 1:200 dilution, and anti-NFκB p65 antibody (Santa Cruz Biotech, Santa Cruz, CA) at 1:100 dilution. To determine the specificity of staining, the primary antibody was replaced with nonspecific rabbit IgG (Sigma, St. Louis, MO). For cytoskeleton staining, cells were fixed and permeabilized as above and then stained with Oregon green-labeled phalloidin (1:1,000) in dark for 30 min. After staining, the elastic membranes were fixed between two coverslips as a “sandwich” and then removed from the culture plates. Fluorescent images were examined and photographed under a fluorescent microscope (Nikon Canada, Toronto, Canada).
Immunoblotting was performed as previously described (18). Briefly, cells lysates were collected (lysis buffer: 50 mM Tris·HCl, pH 7.5, 150 mM NaCl, 2 mM EGTA, 2 mM EDTA, 1% Triton X-100, containing aprotinin 10 μg/ml, 10 μg/ml leupeptin, 1 mM PMSF, 1 mM Na3VO4, and 10 mM NaF). Lysates containing equal amounts of total protein were boiled with SDS sample buffer [0.06 M Tris, pH 8.0, 10% (vol/vol) glycerol, 2% (wt/vol) SDS, 5% (vol/vol) β-mercaptoethanol, and 0.0025% (wt/vol) bromophenol blue] and subjected to SDS-PAGE. Proteins were transferred onto nitrocellulose membranes and probed with anti-Src pTyr416 antibody (BioSource). Blots were visualized with an enhanced chemiluminescence detection kit (Amersham/Pharmacia Biotech, Baie d’Urfe, Quebec, Canada). The membrane was stripped and reblotted with anti-Src antibody (GD11) (Upstate Biotechnology, Lake Placid, NY).
PTX3 and IL-8 released from A549 cells were determined by using ELISA kits from Alexis (San Diego, CA) and BioSource (Camarillo, CA), respectively, following the manufacturers’ instructions.
Data Normalization and Selection of Significantly Regulated Genes
Figure 1, A and B, illustrates the salient features of the data set before and after normalization. Normalization improved the resolution of genes expressed at lower levels. Linear least fit square plots of the scatter plot of control 1 h and control 4 h experiments are shown as an example to demonstrate how normalization improved the correlation between data sets (Fig. 1, C and D). The normalized data set was processed using SAM. For a Δ set at 0.01745, 47 genes were significantly expressed, and the false discovery rate was ∼16%, i.e., 7.46 genes (Fig. 1E). MAS 5.0 was subsequently used to determine the fold change and P value for each of the significant genes. Of the 47 genes identified by the SAM analysis, 2 were excluded because their raw intensity values were below 20, which is in the range of the background noise of the microarray, and 5 were excluded because the fold changes were inconsistent between replicates. Consequently, 40 genes were selected, the expression levels of which are shown in Figs. 2–4. Interestingly, among these 40 genes, all were regulated by TNFα, whereas LPS or mechanical stretch alone had little independent effect.
TNFα-Induced Specific Gene Expression
Inflammatory response-related genes.
For comparison, in Figs. 2–4, the effects of TNFα alone and the combined effects of TNFα plus cyclic stretch are presented. A major effect of TNFα treatment was a marked upregulation of genes encoding chemokines (Fig. 2A). Two CC chemokines, CCL2/MCP-1 and CCL20/MIP-3α (macrophage inflammatory protein 3α), and four CXC chemokines, CXCL1/GRO1, CXCL2/GRO2, CXCL3/GRO3, and CXCL8/IL-8, were upregulated by TNFα. Note that the effects of TNFα alone and TNFα plus stretch are very similar. Genes related to inflammatory responses upregulated by TNFα include IL-6, cyclooxygenase (COX2), urokinase plasminogen activator (PLAU), and pentaxin 3 (PTX3) (Fig. 2B). Membrane-bound proteins regulated by TNFα included interferon-γ receptor 2 (IFNGR2), intercellular adhesion molecule-1 (ICAM-1), ephrin A1 (EFNA1), IL-15 receptor-α (IL15RA), and TNF superfamily member receptor 9 (TNFSFR9) (Fig. 2C).
TNFα-induced negative feedback on genes involved in NFκB pathway activation.
TNFα-induced acute inflammatory responses are mainly mediated through the NFκB pathway (20). Figure 3A shows the upregulation of several genes implicated in a negative feedback loop to turn off TNFα-induced NFκB activation. Binding of TNFα to its receptor leads to recruitment of the adaptor protein TRAF2 (20). Both TNF-associated inhibitor protein 3 (TNFAIP3/A20) (9) and IAP2/BIRC3 (31) have been shown to bind to TRAF2 and inhibit NFκB activation. In A549 cells, TNFα rapidly induced TNFAIP3/A20 gene expression and upregulated IAP2/BIRC3 in a time-dependent fashion. Translocation of NFκB from cytosol to nucleus occurs after its dissociation from its cytoplasmic inhibitor Iκβα (NFKBIA) (16). The p105 subunit of NFκB (NFKB1) has dual function. It is involved in cytoplasmic retention of attached NFκB proteins, thus inhibiting the activation of NFκB. It is also a precursor for the generation of p50 subunit of NFκB by a cotranslational processing (25). Both Iκα and p105 were upregulated by TNFα in A549 cells. Figure 3B shows a schematic representation of this auto-feedback loop.
TNFα-induced apoptosis is an important component in acute inflammatory responses. SAM analysis revealed simultaneous increases in the expression of genes involved in both death and anti-death pathways (Fig. 4A). BID (BH3 interacting domain death agonist) encodes a death agonist that heterodimerizes with either agonists or antagonists (19). BCL2A1 is an anti-apoptotic Bcl-2 family member that forms a heterodimer with BID (6). The human intracellular serine protease inhibitor 9 (SERPIN B9) is a potent inhibitor of caspase-1-induced (1) and of granzyme-B-induced (38) apoptosis. Interferon regulatory factor 1 (IRF-1) may directly mediate the IFNγ-induced apoptosis via the activation of caspase-1 gene expression (13). In addition, IRF-1 plays a pivotal role in the IFNγ-mediated enhancement of Fas/CD95-mediated apoptosis, through regulation of caspase-7 (CASP7) (39). IRF-1 gene expression was significantly increased by TNFα at both 1 h and 4 h, and expression of caspase-7 was significantly increased at 4 h (Fig. 4A). Death-associated protein kinase 1 (DAPk1) is a Ca2+/calmodulin-regulated Ser/Thr death kinase, and it has been implicated as a positive mediator of IFNγ-induced programmed cell death (4). Its expression is downregulated by TNF-α treatment (Fig. 4A).
TNFα also increased mRNA levels of other genes related to intracellular signal transduction and transcriptional factors (Fig. 4B) and genes with other functions (Fig. 4C). There are differences in the temporal pattern of TNFα-induced gene expression. Six genes were upregulated only at 1 h, including three transcription factors (ZFP36, EGR1, and COPEB), two chemokines (CXCL2 and CXCL3), and IL-6. Ten genes were upregulated at both 1 h and 4 h (CCL20, CXCL1, IL-8, A20, NFKBIA, IRF-1, COX2, EFNA1, PLAU, and IEGR3). All other genes were only upregulated after 4 h of treatment with TNFα (Figs. 2–4).
Confirming Microarray Results with Multiple Approaches
Confirmation of microarray data with qRT-PCR.
To confirm the microarray results, we first chose 20 genes and measured their expression with real-time qRT-PCR with the same sample used for microarray studies. Eight of these genes had moderate to high expression levels. Nine had very low expression levels and three were commonly used housekeeping genes. All of them showed very high concordance between microarray and PCR data. Examples of three genes, IL-8, PTX3, and EFNA1, are presented in Fig. 5. Figure 5A shows normalized raw intensity values from microarray. For comparison, Fig. 5B shows real-time qRT-PCR data expressed as ratio to a housekeeping gene [hydroxymethylbilane synthase (HMBS)].
Confirmation with independent experiments.
To further determine whether results from our microarray studies are reproducible, three different experimental approaches were used: 1) repeat experiments using A549 cells; 2) independent experiments with A549 cells in Dr. Rolf Hubmayr’s laboratory at the Mayo Clinic in Rochester, MN; and 3) experiments in BEAS-2B cells, a bronchoepithelial airway cell line. The doses of TNFα and LPS and the amplitude of mechanical stretch were the same as used for microarray studies. All treatments were 4 h. We examined the expression of a total of 18 genes with qRT-PCR. Examples of selected results are shown in Fig. 6. Figure 6A show results from A549 cells treated with TNFα or mechanical stretch (from Mayo Clinic). Figure 6B show results from BEAS-2B cells treated with TNFα and LPS. Although the absolute fold of change in response to TNFα, mechanical stretch, or LPS is not the same between these two cell lines and between different experimental conditions (microarray vs. qRT-PCR, cells from Toronto vs. from Mayo), the general trend of the results is very similar. That is, cells responded to TNFα significantly, but only weakly to mechanical stretch and LPS. In addition to focusing on a few key genes known to be regulated by TNFα that we perceived as essential for quality control (e.g., IL-8), we purposefully selected genes that had either low levels of expression or where the data from the microarray experiments were unclear (data not shown). We also included genes whose expression is not regulated by TNFα in our microarray studies. Similar results were confirmed by qRT-PCR (see Fig. 6, where HSP90 is used as an example). Expression for all genes selected showed excellent concordance with our microarray data.
Confirmation of protein levels in A549 cells.
It is known that several cytokines including IL-8 can be regulated by TNFα at both gene and protein levels. We chose two of them, IL-8 and PTX3, as examples and examined their protein levels. The absolute message level and fold change for IL-8 was markedly elevated in our microarray and qRT-PCR experiments, whereas for PTX3, although the fold change was markedly increased, the absolute values for this gene were lower. The protein levels of these two molecules were very similar to their mRNA levels. Furthermore, both of them were markedly upregulated by TNFα but not by LPS in A549 cells (Fig. 7).
Effect of Mechanical Stretch and LPS
Figure 8 shows the effect of mechanical stretch and LPS on A549 cells. Although in general, stretch alone and LPS alone had little effect on gene expression profiles compared with the TNFα-treated and TNFα-plus-stretch groups, we noted that stretch had an additive effect on the expression of 16 genes at 1 h and/or 4 h. The expression of nine genes was reduced and that of seven genes was enhanced by stretch in the TNFα-plus-stretch group. For example, CCL20 was enhanced by stretch at both 1 h and 4 h (Fig. 8A). In contrast, the combined effect of these two stimuli resulted in an initial downregulation of ICAM-1 at 1 h (Fig. 8A).
Because the effects of stretch alone on gene expression were relatively small, we examined whether there was a response of A549 cells to mechanical stretch alone under the experimental conditions used in the present study. Cell morphology was examined with phase-contrast microscopy. After 4 h of mechanical stretch, cells appeared smaller and rounder (Fig. 8B). Activation of Src protein tyrosine kinase is an important event in mechanotransduction (18). Phosphorylation of Tyr416 of c-Src is a measure of Src activation, which was increased by mechanical stretch, as determined by immunoblotting and immunofluorescent straining (Fig. 8C).
Similarly, we examined whether cells responded to LPS under the experimental conditions employed in the present study. Cells were stained for F-actin. After 1 h of LPS treatment, polymerization of actin fibers was reduced compared with control cells (Fig. 8D). Stimulation of A549 cells with TNFα for 1 h caused rapid translocation of the p65 subunit of NFκB from cytoplasm into the nuclei, as shown by immunofluorescent staining. In contrast, NFκB translocation was less evident in LPS-treated cells (Fig. 8E).
Taken together, these data indicate that the experimental conditions of mechanical stretch and LPS used in the present study induced cellular responses. However, because we purposely used relatively low doses of LPS and less injurious mechanical stretch conditions, their effects are much more subtle than those of TNFα. These changes may be masked by the dramatic effects of TNFα on gene expression. Importantly, however, cyclic stretch may act to modulate the response of cells to proinflammatory molecules such as TNF-α. This may have clinical implications.
Traditionally, alveolar macrophages and leukocytes have been considered as the sources of cytokines and inflammatory mediators, whereas AECs and other lung parenchymal cells were regarded as innocent targets. In the present study, we have shown that production of multiple chemokines and inflammatory mediators from AECs can be induced by TNFα, as well as genes related to membrane protein, apoptosis, and intracellular signal transduction. Identifying the pattern of acute inflammatory responses in AECs may lead to the development of therapeutic strategies that selectively block inflammation and tissue injury in lung epithelial cells, while preserving the normal functions of itinerant immune cells for host defense.
Despite its obvious advantages, microarray analysis also has a number of limitations. The number of genes is large (frequently more than thousands), thus inflating the likelihood of false-positive results (type I error). The number of replicates usually is small, leading to noisy point estimates (43). Consequently, conventional statistical approaches yield overly conservative results (type II error). The advantage of SAM is that it provides an estimate of the distribution of the data for each single gene, and therefore no particular distribution (such as a normal distribution) has to be assumed (44). Another important feature of SAM is that it does not calculate conventional type I errors, but instead far less conservative false discovery rates, which can be selected by setting the cutoff parameter, Δ. This provides a statistically robust means of performing multiple hypotheses testing of microarray data. To focus on the transcriptional regulation on gene expression, we treated cells only up to 4 h. The consequence of longer exposure to these factors is not addressed in the present study. We have two separate time points (1 h and 4 h) for each condition measured in duplicate. The concordance in gene expression levels between the two time points in the control groups confirms the reproducibility of the gene array analysis (27). Consistency between the genes upregulated by TNFα-plus-stretch and TNFα-alone groups further corroborated the effects of TNFα on A549 cells. To minimize misinterpretations, we focused only on the genes that were most highly upregulated. The genes picked up by SAM analysis responded to TNFα in a reproducible and consistent fashion, confirming the usefulness of the SAM procedure.
One potential limitation of this study is the use of A549 cells, a cell line derived from lung carcinoma. These cells do not form tight junctions in culture, and in some instances their responses are different from that of primary cultured cells. Several groups have established methods to isolate primary cultured human lung epithelial cells. However, the purity of these cell preparations is difficult to control, and the cellular responses vary from batch to batch. In addition, primary human AECs failed to attach in the silicon elastic membrane coated with collagen I for mechanotransduction studies (data not shown). A549 cells are a widely used tool for lung cell biology studies providing a large database against which new findings can be compared. Consequently, we used A549 cells to generate new hypotheses. Using qRT-PCR, we confirmed that our results from microarray analysis were reliable. This was not only proved by analyzing the same sample with both techniques but also by using samples from separate experiments. Moreover, since A549 cells in different laboratories may develop different phenotypes, it was important to establish that the responses were well reproducible with cells from different institutes treated by different personals. Notably, the stretch apparatuses used in these two laboratories are also different. Alveolar and airway epithelial cells may respond differently to inflammatory stimuli. Our experiments carried out in BEAS-2B cells indicate that TNFα induced similar responses in small airway epithelial cells and AECs, at least with the genes we determined. Furthermore, using ELISA, we demonstrated that the changes in cytokine gene expression were associated with similar changes at the protein levels. Therefore, results from the present study may provide useful and valid clues for further investigation of the responsiveness of human lung epithelial cells to various inflammatory stimuli.
The most significant finding of the present study is the TNFα-induced gene expression pattern in human AECs, which includes: 1) increased chemokine genes and genes related to the generation of inflammatory mediators, 2) cellular membrane proteins related to inflammation, 3) genes related to cell death and survival, and 4) genes related to signal transduction. These results share some degree of similarity with the activation of inflammatory cells; however, they highlight the fact that TNFα stimulation likely plays an important and nonredundant role in AECs during pulmonary inflammation in vivo.
The CXC chemokines play a key role in ALI (30). It is well known that TNFα can induce IL-8 gene expression and protein release from A549 cells (35). The genes for GRO1, GRO2, GRO3, and IL-8 are located closely together on human chromosome 4q13–21. The fact that their expression is increased simultaneously suggests a common transcriptional regulatory mechanism. TNFα also induced a strong upregulation of CCL2/MCP-1 and CCL20/MIP-3α. CCL2 displays chemotactic activity for monocytes and basophils. It has been implicated in the pathogenesis of diseases characterized by monocytic infiltrates (30). CCL20 has been shown to have antimicrobial properties and participates in pulmonary innate immunity (49). IL-6 was the only cytokine gene that was upregulated by TNFα in the present study. Although the GeneChips contain probes for over 100 cytokines/chemokines and their receptors and many other genes related to inflammatory reactions, this selected group of chemokines could be a “signature” of TNFα stimulation of AECs.
In the present study we detected three genes that are less known in ALI: PTX3, EFNA1, and TNFRSF9. PTX3 has been shown to function as a secreted pattern-recognition receptor that has a nonredundant role in resistance to selected microbial agents, specifically Aspergillus fumigatus in the lung (8). PTX3 is elevated in the serum of critically ill patients, with a gradient from systemic inflammatory response syndrome to septic shock (24). Therefore, PTX3 may be a novel inflammatory mediator for ARDS/ALI. EFNA1 encodes a member of the ephrin family. Among other functions, it is known to inhibit chemokine-induced actin polymerization of T cells, thereby blocking migration of these cells (33). TNFRSF9 (also called CD137 or 4–1BB) is a costimulatory member of the TNFR family expressed on activated T cells. Its ligand is expressed on activated antigen-presenting cells. Soluble CD137 ligand is released following T cell activation. Signals relayed through CD137 inhibit activation-induced cell death. Although CD137 has been shown to be expressed in pulmonary parenchymal cells (2), its function in these cells remains unknown. The roles of these genes in ALI/ARDS merit further investigation.
TNFα stimulation of A549 cells induced expression of genes potentially involved in a putative negative feedback loop aimed at silencing TNFα-induced NFκB activation (47). Genes encoding four inhibitory proteins on TNFα-induced NFκB activation were upregulated. In addition to genes illustrated in Fig. 3, we also noticed upregulation of IL-15RA by TNFα (Fig. 4A). IL-15 is a potent inhibitor of TNFα-mediated apoptosis (3). IL-15 blocks TNFR1-induced apoptosis via IL-15RA chain signaling. The intracellular tail of IL-15RA shows sequence homologies to the TRAF2 binding motifs of CD30 and CD40. Binding of IL-15 to IL-15RA successfully competes with the TNFR1 complex for TRAF2 binding, which may impede assembly of key adaptor proteins to the TNFR1 complex. The interaction between IL-15 and TNFα has not been studied in the ALI setting.
Using cell morphology, Src activation, and cytoskeleton organization as readout, we demonstrated that the mechanical stretch condition and the dose and time of LPS stimulation indeed induced cellular responses in the present study. However, the increase in gene expression levels in response to mechanical stretch or LPS were less than that to TNFα and thus excluded by SAM. Real-time PCR confirmation studies revealed that both mechanical stretch and LPS increased the expression of IL-8 by twofold. We also attempted to analyze the effects of stretch or LPS alone on gene profiling in A549 cells, using MAS 5.0 software. Groups of genes were found to be regulated by either stretch or LPS in each experiment at either 1 h or 4 h. However, compared with the effects of TNFα, these changes were relatively small and less consistent (data not shown). The effects of these stimuli on gene expression therefore need to be interpreted with caution and further investigated.
The lack of dramatic changes in gene expression to mechanical stretch in AECs is not completely surprising. We have previously shown that mechanical stretch increases secretion of MIP-2, but not gene expression, from fetal rat lung cells (23). Similarly, when A549 cells were stretched the release of IL-8 protein was significantly increased, but the increase in IL-8 mRNA was much less compared with the effect of TNFα (46). Although there are several reports that injurious mechanical ventilation induces cytokine gene expression in the lung (40) and in alveolar epithelia type II cells in particular (45), we found that mechanical stretch alone does not induce dramatic changes in gene expression of A549 cells. However, we found that stretch enhanced or attenuated the TNFα-induced expression of multiple genes. It is known that mechanical ventilation of patients with normal lung function, such as during anesthesia, usually does not induce significant lung injury, whereas patients with underlying pathology, such as trauma or sepsis, are much more susceptible to mechanical ventilation (48). Thus, in the presence of other proinflammatory cytokines and/or pathogen products, mechanical stretch may influence the gene expression profiles induced by these mediators at transcriptional or posttranscriptional levels.
Our data on the ability of A549 cells to respond to LPS are consistent with previous reports that these cells are hyporesponsive to LPS (12, 35, 42), possibly because expression of major LPS receptors (Toll-like receptor 4 and CD14) is very low in A549 cells (32, 42). LPS can induce cytokine release from A549 cells when used at very high doses (50–200 μg/ml) (7, 14) or in the presence of soluble CD14 (15, 32). However, to avoid potential nonspecific effect of LPS on gene expression in AECs, we used a relatively lower dose of LPS. Under this experimental condition, LPS did not induce NFκB activation and translocation. In line with this, the effect on gene expression was not very impressive, even with thousands of candidate genes analyzed. Nonetheless, LPS treatment led to a small increase in the expression of IL-8 as well as some other genes. Similar to what happens in primary cultured rat pneumocytes (10, 11), LPS induced depolymerization of microfilaments. The mechanisms and functions of these LPS-induced cellular responses need to be further studied.
In summary, our studies using A549 cells demonstrate that AECs respond differently to specific injurious agents. Moreover, these studies highlight the unique role TNFα plays at the cellular level. Other cytokines, such as IL-1β (26), are also known to participate in the development and evolution of ALI/VILI; the roles of these molecules in human lung epithelial cells as well as in other cell types in the lung also merit further investigations with microarray and other bioinformatics approaches.
This work was funded by Canadian Institutes of Health Research Grants MOP-13270 and MOP-42546. C. C. dos Santos and B. Han are CIHR Fellows, and M. Liu is a recipient of Premier’s Research Excellence Award.
We thank Dr. Sandy Der for assistance with developing real-time PCR protocol.
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
Address for reprint requests and other correspondence: M. Liu, Professor of Surgery, Univ. of Toronto, University Health Network Toronto General Hospital, 200 Elizabeth St., Toronto, Ontario, M5G 2C4 Canada (E-mail:).
- Copyright © 2004 the American Physiological Society