Gene expression profiles in asbestos-exposed epithelial and mesothelial lung cell lines

Asbestos causes DNA double strand breaks [1], chromosomal aberrations, and abnormal chromosome segregation [2]. Asbestos fibre-induced genotoxicity has been proposed to be caused by both a direct interaction with the genetic material and also indirect effects via production of reactive oxygen species (ROS) [3]. The respiratory diseases linked to asbestos exposure include fibrotic lung disease, mesothelioma, and lung cancer. The pathogenesis and histopathology of asbestosis resemble that of idiopathic pulmonary fibrosis. Some specific genes contributing to the development of asbestosis and asbestos-related cancer have been described, as reviewed in [4]. However, the exact molecular mechanisms behind asbestos-associated carcinogenesis and fibrosis/asbestosis are thought to be very complex and involve several parallel pathways [4, 5] that remain to be clarified.

Asbestos exposure has been reported to alter the expression of several genes involved in integrin-mediated signaling pathways, MAPK pathways, and NFKB/IKB pathways [6]. A recent study in a model for asbestos-induced oncogenesis demonstrated that the tumor necrosis factor, TNF-α, triggered by macrophages, induced activation of the NFKB cascade and thereby blocked apoptosis [7].

Studies from this laboratory as well as from others, have examined the gene copy number and expression changes in lung cancer patients with asbestos exposure and in mesothelioma. These have revealed a very complex pattern of chromosomal aberrations and altered gene expression, i.e. evidence for specific asbestos-associated aberrations and altered expression profile in these cancer genomes [8–10].

Given the evidence of association between fibrosis and cancer, it is clearly important to identify the specific genes involved in the inflammatory, fibrotic, and carcinogenetic processes in the lung following asbestos exposure, as reviewed in [4]. Chromosomal aberrations, together with more accurate information of gene expression alterations caused by asbestos, could provide valuable clues in the search for novel therapeutic targets.

We used three human lung cell lines to map the gene expression following crocidolite exposure. In addition to epithelial cells, mesothelial cells with features of both mesenchymal and epithelial cells [11], were used in the identification of asbestos-affected genetic pathways. Crocidolite was selected as it is probably the most pathogenic amphibole fibre, possibly due to its high iron content [4]. We report for the first time specific time-dependent genome-wide expression profiles in asbestos exposed human lung adenocarcinoma cells (A549), non-tumorigenic SV40-immortalized bronchial (Beas-2B), and pleural mesothelial cells (MeT5A).

We exposed transformed lung epithelial and mesothelial cells to asbestos and hybridized the samples from 3 to 6 time points to Affymetrix oligonucleotide arrays. The expression results were analyzed using three different methods to profile the expression pattern evoked by asbestos exposure (Fig. 1). The aim of this study was to profile gene expression at a genome-wide level to enable also other researchers to choose possible candidates for further investigation. Therefore we do not intend to discuss all the new potentially asbestos-related genes, but rather to select some examples. We are well aware that single genes are of little interest in profiling experiments when thousands of genes are evaluated. The genes we discuss here are some of those highlighted in three very diverse analyses. For complete results, see Additional files 1, 2, 3, 4, 5: GOanalysis, Cluster_analysis, CCA, A549clusters and Beas2Bclusters.

Many of the measured expression changes are probably due to the triggering of various universal cellular responses to foreign toxic substances, e.g., apoptosis or cell death. We anticipated that the use of three different cell lines would help us to pinpoint the specific asbestos-related effects. I.e. by comparing and identifying changes common to all cell lines, we expected to be able to neglect the expression changes associated with the malignancy or cell type and to concentrate on specific asbestos-related changes common to all cell types. Indeed, the number of GO terms shared by the cell lines increased with time, implying that response to the exposure occurs through the same pathways. Furthermore, the number of downregulated biological processes in each cell line increased with time, while the number of upregulated processes decreased (Fig. 1, bottom charts). This could be a consequence of apoptosis or cell death related functions. However, the failure of apoptotic functions to trigger cell death has been proposed to precede carcinogenic changes in a cell [16, 17], making also the pathways and genes involved in the apoptotic processes worth investigating.

The asbestos associated genes NFKB2 and IKBKB [18, 19] were present in the significant clusters of all cell lines (Table 3) and the GO term, "positive regulation of I-kappaB kinase/NF-kappaB cascade", associated with these genes was common for all the cell lines in both the cluster analysis and the GO analysis (Table 2). In addition, the genes within the GO term "cytoplasmic sequestering of NF-kappa B" were downregulated after 48h of asbestos exposure in all cell lines (Table 1). NFKB is known to be upregulated by asbestos and indeed, both the pathways and the probe sets corresponding to this gene exhibited upregulation at the initial time points, but downregulation at the later time points. This might be due to the fact that cells treated with asbestos often undergo apoptosis or cell death, whereas NFKB activation is involved in inhibition of apoptosis and cell survival [7].

Thioredoxin (TXN) and thioredoxin reductase (TXNRD1) downregulate apoptotic processes. These genes have been found to be upregulated in malignant pleural mesothelioma [20, 21] and TXN is also known to be involved in the DNA-binding activity of NFKB [22, 23]. TXNDC is located in 14q22, found in the CCA results. This thioredoxin-related gene was also represented in a highly significant cluster in all three cell lines. It is also one of the genes contributing to differential expression of the GO term "positive regulation of transcription, DNA-dependent", which was downregulated in all cell lines after 48h of asbestos exposure (Table 1). This suggests that the activation of thioredoxin and thioredoxin-related products, and subsequently NFKB, is evoked by other asbestos-activated products, e.g., TNF-α, derived from cell types other than the epithelial cells.

BNIP3L is a potential tumor suppressor gene, associated with hypoxia-induced epithelial injury [24]. Similar to TXNDC, BNIP3L was represented in all analyses, i.e., in the region 8p21 respresented in the CCA results, in the significantly downregulated GO term "negative regulation of survival gene product activity" (Tables 1 and 4), and it was present in the highly significant clusters 1, 8, and 2 (A549, Beas-2B, and MeT5A, respectively). TXNDC and BNIP3L were the only genes represented with such high significance in all analysis results and they can be readily envisaged as being associated with asbestos-related cellular damage based on their functions and their relationships with previously known asbestos-associated genes.

Asbestos-associated protein kinase c, delta (PKC-δ)and p-adducin (ADD) [4, 25] were represented in significant clusters in all cell lines and ADD3 was represented in the most significant cluster of all cell lines (Fig 2A, D and 2G). ADD1 has also been found to be upregulated in asbestos-exposed lung cancer patients [10]. Furthermore, PKC-δ maps to a chromosomal area closely adjacent to 3p21, a region that has recently been found to be more frequently affected by loss of heterozygosity in asbestos-exposed than in non-exposed lung cancer patients [8, 26]. Crocidolite asbestos is believed to modulate the intracellular calcium levels through activation of PKC, pointing to a possible involvement of calcium related pathways and genes [4]. We found both calcium-related genes and GO terms in the results of all analyses.CAMK2D was represented in high-ranking clusters of all cell lines and the potential asbestos hotspot regions identified by CCA harbored CAMK1. In addition, many other genes belonging to calcium-related biological processes were represented in the CCA data (Additional file 5: CCA). The GO analysis revealed that the biological process "calcium ion transport" had been affected in both Beas-2B and MeT5A cells (Additional file 1: GOanalysis).

The only upregulated biological process common to all three cell lines at the 48h time point in the GO analysis was "sensory perception of smell", which has not previously been proposed to be asbestos-associated. The majority of the genes involved in this process are G protein coupled olfactory receptors and taste receptors. G proteins have been suggested to be involved in the respiratory burst (release of ROS) caused by asbestos [27]. Furthermore, many G protein-associated biological processes, e.g., "G-protein signaling, coupled to IP3 second messenger" and "G-protein coupled receptor protein signaling pathway", were represented in our results (see Additional files 1 and 5: GOanalysis and CCA). This could indicate that the expression of G proteins is specifically altered by asbestos, possibly causing the release of ROS, which is known to contribute to carcinogenesis and progression to malignancy [28].

The most significant region according to the CCA was 2p22 (Table 4). It precedes a region (2p21-p16.3) that we have previously shown to be differentially altered in lung carcinomas of asbestos-exposed patients than in non-exposed patients' carcinomas [8]. A region homologous to the human 2p25-p21 has been reported to be amplified in radon-induced rat lung tumors [29], which could point to preferential damage that, for example, ROS production caused in this region of the genome. Additionally, a c-fos-like gene (FOSL2), which maps close to 2p21-p16.3, was present in significant clusters of all cell lines (Table 3). The early response genes c-fos and c-jun are closely linked to asbestos exposure [30]. Our results also revealed that the GO term "negative regulation of JNK activity", which regulates JUN kinase activity, was downregulated at the 48h time point in all cell lines (Table 1). Based on these results, the short arm of chromosome 2 could be an important potential target for DNA damage by asbestos, either directly or indirectly through ROS or other by-products. It is noteworthy that the probe sets located in 2p were substantially enriched in the highly significant clusters 9 of A549 (p = 0.00032) and 5 of Beas-2B (p = 0.000114) (data not shown).

Some of the crucial chromosomal aberrations in mesothelioma occur in 3, 4q, 5p, 6q, 8q22-q23, 9p, 14q12-q24 and 15q [9, 31–35]. These regions correspond to the CCA results, strengthening the hypothesis that some chromosomal regions could be more prone to DNA damage caused by asbestos. Additionally, CCA revealed two interesting integrin genes in 2q31, ITGA4 and ITGA6. ITGA4 was significantly underexpressed in asbestos-exposed lung cancer patients [10] and ITGA6 has been associated with asbestos [6, 36]. The long arm of chromosome 2 contains two fragile sites, FRA2G and FRA2H closely preceding and adjacent, respectively, to the region in the CCA results, suggesting that damage in these areas could cause specific changes in the expression of genes localized between the fragile sites. We have previously found fragile sites to be associated with asbestos-related copy number aberrations in lung cancer patients [8]. Furthermore, the ITGA genes are involved in calcium ion binding functions.

The significant clusters of the differentially expressed genes in the A549 cell line contained more genes located in regions corresponding to regions with differential copy number changes in asbestos-related lung cancer than was the case with the Beas-2B and MeT5A cell lines. This could possibly be attributable to the fact that A549 is a malignancy-derived cell line from a lung adenocarcinoma, whereas Beas-2B and MeT5A are SV40-transformed non-malignant cells that may exhibit the early effects of asbestos more clearly. This hypothesis is strengthened by the finding of enrichment of differentially expressed genes in regions corresponding to the CCA results in the clusters of the MeT5A cells, although this cell line was not included in the CCA.

This study describes the asbestos-related gene expression profiles in lung epithelial and mesothelial cells at different time points. Clearly, in vitro experiments can never precisely reflect the conditions and mechanisms in vivo; for example, interactions between inflammatory cells and with the extra-cellular matrix are missing. Furthermore, both Beas-2B and MeT5A are SV40-immortalized cells which may contribute to some of the gene expression changes. However, efficient use of controls and careful selection of bioinformatics methods should account for these limitations and minimize the false negatives and positives. We expect these expression profiles may provide a better understanding of the mechanisms behind asbestos-associated disease, especially when correlated with gene expression data and CGH array data from asbestos-exposed lung cancer and mesothelioma patients [6, 8–10, 36–39].