Given the difficulties in obtaining sufficient primary human lung epithelial cells to study the complex response to asbestos, many studies have instead employed the A549 human adenocarcinoma cell line. In doing so one must keep in mind that the transformed cell line may not be entirely applicable to normal human lung cells. However, all processes noted below are recapitulated in primary human lung epithelial cells, suggesting that A549 cells represent a valid model system. Importantly, use of a human cell line avoids the inherent differences that are seen between individual patients. Alveolar epithelial type II cells are key participants in inflammation, fibrogenesis, and carcinogenesis 23 and have been described as the targets of asbestos-associated lung carcinomas 24.

The conditions used for obtaining the gene expression profile for crocidolite treated A549 cells were chosen to mimic conditions where a variety of biochemical observations have previously been made in our laboratories. Exposure of A549 cells to 6 μg/cm² crocidolite for 24 hours had previously been shown to result in the α_vβ₅ integrin receptor-mediated endocytosis of asbestos fibers 25, mobilization of iron from the fibers within the cell 26, upregulation of ferritin protein to combat the iron overload 27, production of reactive oxygen intermediates 28, efflux of reduced glutathione 9, DNA damage 7, PARP cleavage and activation of initiator caspases 29. Thus, at 6 μg/cm² of crocidolite the microarray results could be related to an extensive body of biochemical information already available in A549 cells. In addition, this cell line has been used by other investigators for several studies to explore the mechanisms of DNA damage 30, apoptosis 10, and invasiveness 31. Use of the A549 human lung epithelial cell line complements studies done in mesothelial cells 32 and by the less carcinogenic chrysotile in bronchial epithelial cells 33. Our goal was to extend our results beyond lists of genes and ontological classifications to discover new pathways based on functional interactions. Rather than diffusing statistical power by examining differences among multiple time points and asbestos concentrations, this experiment focused on the use of chemically defined crocidolite asbestos, a single concentration of crocidolite versus a control condition, and a single exposure time. This focus allowed for three replicates of mRNA from control and crocidolite-treated cells to be analyzed using the Affymetrix HGU133 Plus 2.0 GeneChip to provide the most comprehensive whole genome expression profile.

Graphical checks of the gene expression data revealed a high-quality data set, with no spatial artifacts in the chip images, and a high degree of reproducibility within both the control and treated replicates (Additional File 1). Using tools described in the Methods Section and a false discovery rate 34 of 0.05, 2,546 genes with q-values (FDR-adjusted p-values) less than 0.05 were called statistically significant. These results are represented in a volcano plot (Figure 1).

Among the 54,120 probe sets on the GeneChip, 1808 were significantly up-regulated (492 of which were two-fold or greater), and 738 were significantly down-regulated (27 of which were two-fold or greater) when A549 cells were exposed to 6 μg/cm² crocidolite asbestos for 24 h. Of the probe-sets that changed two-fold or greater, 234 correspond to known genes for the up-regulated probe sets, and 16 of the down-regulated probe sets correspond to known genes (some genes were represented more than once on the chip). Genes that increased in expression five-fold or greater or decreased two-fold or greater and were associated with q-values less than 0.05 are shown in Table 1. A complete table of all significant genes can be found as Additional File 2. The observation that most of the expression changes are upregulated is contradictory to a previous microarray study in A549 cells 22 which used a smaller dose of crocidolite (2 μg/cm²). A direct comparison of the two data sets was difficult due to differences in experimental design. Namely, replicates at a single time point were used in the current study versus single chips over a time course, and a lower crocidolite concentration, which is a likely cause for some differences. Cells exposed to asbestos may demonstrate a hierarchical oxidative stress response 35. Additionally, small amounts of asbestos have been shown to result in proliferation 24. This may be attributed to a transient response to an increase in iron, which is limiting in cells in culture. However, comparison of individual expression changes in our data set to other known experimental results (discussed below) demonstrated that our data set is consistent with the literature.

Hierarchical clustering analysis (Figure 2A) of our data set showed that the majority of the up-regulated probe sets clustered together in four different hierarchical clusters. The down-regulated probe sets were mainly in one hierarchical cluster, but had several individual genes spread throughout the dendrogram. Within the four up-regulated clusters, 24 genes that were upregulated 5-fold could be mapped by functional String analysis shown in Fig 2B. Most of the gene products are connected; i.e., could be linked functionally by way of known molecular interactions. Three major pathways (MAPK, JNK/SAPK and cytokine-cytokine receptor interactions) are represented in the cluster. This result is consistent with what is currently known about the effect of asbestos in epithelial and mesothelial cells. When the down-regulated genes were functionally clustered, only two (fyr and uvo) out of 16 had a connection (data not shown).

Validation of the microarray data was confirmed by quantitative RT-PCR on four of the genes (Egr1, ATF3, c-Jun, and JunB) and was further corroborated with previously noted results from experiments using human cells/cell lines (Table 2). Quantitative RT-PCR also demonstrated that increases in Egr-1, ATF3, c-Jun, and JunB message occurred after 6 hours and continued to increase through 24 hours (data not shown). Increased mRNA expression from ATF3, c-Jun, and JunB was also observed by quantitative RT-PCR in primary human epithelial cells from small airways (SAEC) treated with crocidolite (data not shown). Increased protein expression in A549 cells exposed to crocidolite was observed for ATF3, c-Jun, and JunB (data not shown). Finally, previous results using asbestos-treated A549 cells have demonstrated increased mRNA and/or protein levels for IL8, TP53, HMOX1, CDKN1A, SOD2 and c-myc (see Table 2 for additional references). The reproducibility of the microarray replicates, the significance level of the expression changes, and the validation by quantitative RT-PCR suggest that a highly useful and validated data set was obtained in this study. Furthermore, the large number of significant genes allowed us to perform a comprehensive analysis of the crocidolite-induced transcriptome in human lung epithelial cells.

Although it is clearly established that asbestos induces DNA damage in both primary cells and cell culture lines, the literature illustrates that the molecular mechanism underlying this process and how the fate of the cell is dictated are multifaceted. Although ROS are intimately tied to the mechanism of asbestos-induced fibrosis and carcinogenesis, ROS alone do not offer a complete understanding of the asbestos response. In order to identify genes for which expression level changes are specific for crocidolite compared to oxidative stress, we compared our data set to expression changes that occur in A549 cells when exposed to hydrogen peroxide. In a previous study that documented changes in the A549 cell transcriptome, Cotgreave and co-workers 36 showed that treatment of A549 cells with hydrogen peroxide results in DNA damage and apoptosis. Both hydrogen peroxide and crocidolite asbestos induced the upregulation of TNFRS10B, PPP1R15A, GADD34, CDKN1A, BTG2, DUSP1, DUSP5, DUSP14, SDC4, GDF15, IL8, ADM, FST, IER3, FOS, HMOX1, ATF3, and ZFP36. Therefore, expression changes in these genes may represent a response to an oxidative stress. Several unique genes that are differentially regulated in the crocidolite data set are noted in Table 1 and include some genes which are associated with carcinogenesis such as STC1, IL6, FN1, BRE, and PTGS2. Differences between the hydrogen peroxide-treated and crocidolite-treated transcriptomes may be due to the additional iron released from fibers, present in cells treated with crocidolite, or reactive nitrogen species, changes in glutathione content and/or the involvement of processes initiated at the cell surface by the fiber. It will be very interesting to investigate if these apparent crocidolite-specific gene regulations play a role in crocidolite-induced cytobiological endpoints.

In order to identify themes in the global expression pattern of crocidolite-treated A549 cells, we used Gene Ontology (GO) classification. The differentially expressed genes (up-regulated/down-regulated ± 2-fold with p-values less than 0.2) were analyzed using Onto-Express http://vortex.cs.wayne.edu/. The Biological Process tree was expanded to the fourth tier in the hierarchy as a balance between specificity and coverage. Processes affected by crocidolite are shown in Figure 3A. Within the Cellular Metabolism classification, processes involved in transcription and phosphorus metabolism were highly represented. Expansion of Molecular Functions to the third tier revealed a large number of genes with functions related to transcription (over-represented in Protein binding (Figure 3B)). This observation was expanded upon by performing a global test 373839 to identify specific gene ontologies related to transcription, where the expression levels of all the genes in the ontology give useful information for predicting the clinical outcome (or in this case, a cellular outcome of crocidolite exposure). Ontologies of particular interest related to transcription factor activity (GO:0003700) are summarized in Figure 3C, with nodes colored to correspond to global test p-values; the lighter nodes have lower p-values. To reduce visual clutter, only the last four digits of each ontology's identifier are reported in Figure 3C. The ontological analysis illustrates that transcriptional corepressor activity and repressor activity are highly represented in the data set. Overall the transcriptome of crocidolite-treated human lung epithelial (A549) cells is heavily populated by gene products associated with transcription.

We also noted that DNA repair terms were not highly represented in the data set. Although this result was initially surprising, we noted that the transcriptome of A549 cells exposed to hydrogen peroxide, which also induces DNA damage, was also underrepresented in DNA repair terms 36. In order to gain a better understanding of how the transcriptome of crocidolite-treated A549 cells is reorganized in response to the resulting DNA damage, the results of the global test discussed above were also applied to ontologies related to DNA damage/repair. Although this analysis (see Additional File 3) confirms the overall observation that many of the processes related to DNA repair/damage are not highly represented, the analysis also highlights specific nodes in the GO tree related to DNA damage-induced phosphorylation (GO:6975), DNA dealkylation (GO:6307), DNA methylation (GO:6306), and DNA alkylation (GO:6305). In light of the several studies directed at linking promoter methylation status and carcinogenesis (reviewed in 40), it is interesting to point out that in as little as 24 hours of crocidolite exposure, the A549 transcriptome may be poised to affect the DNA methylation status that is associated with lung cancer pathogenesis.

Even though this study was initiated to understand the broad changes that occur in the lung transcriptome upon crocidolite exposure, we also sought to identify novel genes that had no previous association with crocidolite and to identify genes that may be downstream targets of well-known crocidolite-related players. Given the complex effects that asbestos has on the lung system, we expected that a systems biology approach may provide novel avenues to study. NR4A1, NR4A2, and NR4A3 belong to the steroid nuclear hormone receptor superfamily of immediate early genes that are induced by serum, growth factors and receptor engagement and are thus implicated in cell mitogenic responses. Although previously characterized as pro-survival, studies have also suggested an important role for these receptors in cell transformation and tumorigenicity via their anti-apoptotic and pro-apoptotic functions 41. Thus, depending on cellular context, these gene products may serve as switches in determining cell fate. All three members of this family show increased expression in our data set (Additional File 2) but have not previously been implicated in crocidolite-induced pulmonary toxicity.

Modulation of apoptosis can also be affected by BRE (brain and reproductive organ-expressed protein), a stress-modulating protein also known as TNFRSF1A modulator. BRE expression was upregulated >7-fold in crocidolite-treated A549 cells. Exogenous overexpression of BRE can attenuate intrinsic apoptosis and promote growth of the transfected Lewis lung carcinoma line in mice 42 which is consistent with the recent finding that BRE protein is overexpressed in human hepatocellular carcinomas 43. Given the ability of BRE to interact with both Fas 44 and the TNF receptor 1 45, and the observation that TNFα can attenuate asbestos cytotoxicity in mesothelial cells 17, it will be very interesting to investigate possible roles of BRE in crocidolite-treated human lung cells. Also noteworthy is the upregulation of several "early response" NF-κB targets 46 in our dataset (Additional File 2) including TNFAIP3, IL8, IL6, CXCL1, CXCL2, CXCL3, PTGS2, and PLAU. Activation of the NF-κB pathway is thought to play a critical role in cell survival in asbestos-treated cells 17, but only a few of the downstream targets of NF-κB have been identified.

The relationship between asbestos and calcium has received little attention in recent years, but the initial studies suggest that calcium may have an important role to play (reviewed in 47). Perhaps most compelling is the ability of the calcium-chelator Quin-2 to prevent crocidolite-induced DNA breaks 48. Additionally, several players in the asbestos response are regulated by calcium levels, e.g., protein kinase C 15. We were therefore interested in determining if the crocidolite-treated transcriptome demonstrated any clues regarding the regulation of calcium homeostasis. We observed that expression of MCTP1 and STC1 were both upregulated in our data set by 7.3- and 12.9-fold, respectively. MCTP1 is a transmembrane protein that binds calcium ions via C₂ domains. Unique properties of this protein suggest that these proteins function in Ca²⁺ signaling at the membrane 49. Stanniocalcin 1 (STC1) has long been studied as a regulator of both phosphate and calcium homeostasis in bony fish, but has recently received attention in mammalian systems. STC1 is a glycoprotein present in a variety of mammalian tissues where it can function as a regulator of gene expression and modulator of transendothelial cell migration 50, and can also affect cellular metabolism by perturbing mitochondrial electron transport chain and mitochondrial calcium transport 51. STC1 affects calcium homeostasis in the heart 5253 and the brain 54. Growing evidence also points to a correlation between STC1 expression and the development of human cancers 5556. Quantitative RT-PCR also demonstrated a 30.7 ± 5.1-fold increase in STC1 message in primary human SAEC exposed to crocidolite.

Finally, in recent years a significant number of studies have been directed at understanding how the disruption of dynamic chromatin remodeling is linked to carcinogenesis. Mechanisms including the previously mentioned DNA methylation status and the use of histone modifications have led to the discovery of prognostic biomarkers 57 and the use of HDAC inhibitors as cancer therapeutics 58. It is of little surprise then to find several differentially regulated genes in crocidolite-treated A549 cells that could participate in chromatin remodeling including the Jumonji domain histone demethylases JMJD1C, JMJD3 and JMJD1A, all of which showed increased expression (Supplementary Table 1). Other genes of interest are discussed below.

In order to extract novel biological insight from the large number of genes upregulated/downregulated in our study, we employed a structured network knowledge-based approach to analyze genome-wide transcriptional responses in the context of known functional interrelationships among proteins, small molecules and phenotypes http://www.ingenuity.com. Networks are generated not only based on functional interactions but also statistical likelihood. Genes/gene products are represented as nodes and are color-coded to represent fold-change in expression level. Interactions between nodes are designated as edges, or lines, and represent physical, transcriptional, and enzymatic interactions.

Crocidolite was obtained from Dr. Richard Griesemer, NIEHS/NTP. The chemical formula for crocidolite is Na₂Fe^III₂(Fe^II, Mg)₃Si₈O₂₂(OH)₂. It has a mean length of 10 μm, a density of 3.2–3.3 g/cm³ and contains 27% iron by weight 69. A custom prepared Ham's F-12 cell culture medium free of iron salts and 0.5% trypsin with 0.2% EDTA were obtained from Invitrogen Inc. (Carlsbad, CA). Fetal bovine serum (FBS) was purchased from Hyclone (Logan, UT). Antibodies for Egr-1 (sc-189) and ATF3 (sc-188) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA) and the antibody for β-actin was purchased from Sigma (St. Louis, MO). Anti-rabbit and anti-mouse enzyme-conjugated secondary antibodies were purchased from Jackson Immunoresearch Laboratories Inc. (West Grove, PA). Primers were purchased from IdtDNA (Coralville, IA). All remaining reagents were purchased in the highest purity possible.

A549 cells were purchased from American Type Culture Collection and grown in F12 medium without iron, with 10% FBS and 50 μg/mL gentamicin (BioWhittaker, Walkersville, MD). Cells were plated for treatments at a concentration of 20,000 cells/cm² and incubated 24 h before treatment. For all treatments, medium was removed and replaced with either fresh medium (for control samples) or medium containing the appropriate stimulus (treated samples). For crocidolite treatments, fibers were suspended in sodium bicarbonate at a concentration of 1 mg/mL and immediately diluted to a final concentration of 6 μg/cm² in complete medium. Sodium bicarbonate is used specifically to avoid iron mobilization in the media before the fibers can by endocytosed and maintains the pH 26. These are the exact same conditions and methodologies used previously in several other studies 911252627.

mRNA was isolated from cells at passage 9 using TRIzol^® reagent from Invitrogen (Carlsbad, CA) following manufactures directions. This procedure yielded RNA of both high quantity and quality, as verified by both the A₂₆₀/A₂₈₀ ratio (> 1.8) and the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). The extracted RNA was processed by the Affymetrix core service at the Center for Integrated BioSystems, Utah State University (CIB-USU; Logan, UT). RNA was processed as per the manufacturer's instructions (Affymetrix Inc., Santa Clara, CA) for the first-strand cDNA synthesis and amplification, followed by cDNA synthesis and labeling. The cDNA was hybridized overnight to the HGU133-Plus2.0 Human Genome 2.0 genechips and post-hybridization washing, detection and data collection were performed as per manufacturer's instructions. Three biological replicates were performed. Each replicate consisted of a control plate and a crocidolite-treated plate. All experiments were done at passage 9.

Using tools from the Bioconductor project 70, the data were pre-processed using the RMA algorithm 71, and a test for differential expression between control and treated (crocidolite-exposed) conditions was performed using the moderated t-statistic from the limma/eBayes approach 72. The raw p-values from this test were adjusted to control the false discovery rate 34 at 0.05. Dendrograms (Figure 2) were drawn by Hierarchical Clustering Explorer (HCE; http://www.cs.umd.edu/hcil/hce/hce3.html) software version 3.0 to summarize RMA-normalized data after averaging across replicates. The cutoff value for minimum levels of gene expression was based on the whole genome expression profiles generated by HCE. The median cutoff value was set at the software-calculated median of gene expression and was colored black. The data were further subdivided into gene functional classes that had been annotated by the Human Genome Database. Each class was then clustered separately and the determined cutoff values were applied to generate colored expression maps for each individual class for ease of visualization. In addition, a global test 373839 was used to identify gene ontologies of interest, where the expression levels of the genes in the ontology were statistically significant (p-value less than 0.05) for predicting the clinical outcome (crocidolite exposure). STRING analysis was performed at http://string.embl.de/ using default parameters.

mRNA was isolated using TRIzol^® reagent from Invitrogen (Carlsbad, CA) following manufactures directions. DNase from the RNAqueous^®-4PCR kit (Ambion Inc., Austin, TX) was added to eliminate genomic DNA and removed following the manufacture protocol. cDNA was transcribed using SuperScript™ II Reverse Transcriptase (Invitrogen Inc., Carlsbad, CA). Specificity was confirmed by performing a melting curve on the final amplicons and running the amplicons on a 2% agarose gel. All qRT-PCR samples were also run on samples without reverse transcriptase to confirm the product was from mRNA, not DNA, and all of these samples showed no amplicon. Three different dilutions of cDNA were used to confirm that the samples were in the linear range.

Cells were harvested with trypsin and then lysed using RIPA buffer (0.15 M NaCl, 50 mM Tris, pH 7.2, 1% deoxycholate, 1% Triton X-100, 0.1% SDS) containing protease inhibitors (30 μL/mL aprotinin, 4 μg/mL leupeptin, 4 μg/mL soybean trypsin inhibitor, 0.1 mM PMSF and 1 μM benzamidine). After incubation on ice for 30 min, cells were centrifuged for 10 min at 8200 × g to remove cellular debris. Protein concentrations were determined using the Bradford method (Bio-Rad Laboratories, Hercules, CA). Cellular protein (50 μg) was analyzed using SDS-PAGE (10–15%) and transferred to polyvinyl difluoride (PVDF) membranes. Primary antibodies were diluted 1:200 in 5% dried milk in TTBS and incubated for 1 h at room temperature. Secondary antibodies were diluted 1:5000 and incubated for 1 h at room temperature. Blots were visualized using ECL Plus reagent from GE Healthcare Life Sciences (Piscataway, NJ).

Data were analyzed through the use of Ingenuity Pathways (Ingenuity^® Systems, http://www.ingenuity.com). The asbestos data set with gene identifiers and corresponding expression values was uploaded into the application. Each gene identifier was mapped to its corresponding gene object in the Ingenuity Pathways Knowledge Base. A cutoff of p < 0.01 was set to identify genes whose expression was significantly differentially regulated by ± 2-fold. These genes, called focus genes, were overlaid onto a global network developed from information contained in the Ingenuity Pathways Knowledge Base. Networks of these focus genes were then algorithmically generated based on their connectivity. Biological functions associated with these focus genes were identified using the Functional Analysis using the Ingenuity Pathways Knowledge Base. Fischer's exact test was used to calculate a p-value determining the probability that each biological function assigned to the data set was due to chance alone. Networks are graphical representations of the molecular relationships between genes/gene products. Genes or gene products are represented as nodes, and the biological relationship between two nodes is represented as an edge (line). All edges are supported by at least one reference from the literature, a textbook, or from canonical pathway information stored in the Ingenuity Pathways Knowledge Base.