Differential Response of Mono Mac 6, BEAS-2B, and Jurkat Cells to Indoor Dust

Background Airway toxicity of indoor dust is not sufficiently understood. Objectives Our goal in this study was to describe the effects of indoor dust on human monocyte, epithelial, and lymphocyte cell lines. We aimed to a) obtain a comprehensive and intelligible outline of the transcriptional response; b) correlate differential transcription with cellular protein secretion; c) identify cell line–specific features; and d) search for indoor dust–specific responses. Methods Settled dust was sampled in 42 German households, and various contaminants were characterized. We exposed Mono Mac 6, BEAS-2B, and Jurkat cells to 500 μg/mL indoor dust for 6 hr. Outcome parameters included the transcriptional profile of an oligonucleotide microarray covering 1,232 genes. Significantly enriched Gene Ontology themes were calculated. Supernatant protein levels of 24 inflammatory response proteins served to confirm transcriptional results. Results An intraclass correlation coefficient of 0.8 indicated reasonable microarray reproducibility. The transcriptional profile was characterized by enhancement of detoxification and a danger and defense response. Differential gene regulation correlated with protein secretion (Goodman and Kruskal’s gamma coefficient: 0.72; p < 0.01). Mono Mac 6 cells revealed the highest fraction of differentially expressed genes, dominated by up-regulation of various cytokines and chemokines. BEAS-2B cells revealed weaker changes in a limited set of inflammatory response proteins. No significant changes were observed in Jurkat cells. Conclusions Monocytes are particularly responsive to indoor dust. We observed a classical T-helper 1-dominated immune response, which suggested that bioorganic contaminants are relevant effectors in indoor dust.

Indoor dust contributes substantially to individual particle exposure (Butte and Heinzow 2002). Compounds tracked in from outdoors are a relevant source of indoor dust contaminants. Indoor penetration from outdoors has been calculated as 70% for trace elements including various transition metals, 50% for particles, and 35% for fungal spores (Butte and Heinzow 2002;Wallace et al. 2003). Indoor dust sources include lead from paint; pyrethroids from carpets and textiles; phenols, including pentachlorophenol and bisphenol A, from wood preservatives and pesticides; organochlorines and organophosphates from pesticides; polycyclic aromatic hydrocarbons (PAHs) from heating, smoking, cooking, and parquet floor glues; phthalates from softeners used in various home products; polychlorinated biphenyls (PCBs) from plastics and sealing materials; and polybrominated diphenyl esters from flame retardants (Butte and Heinzow 2002;Wallace et al. 2003). Moreover, higher organic carbohydrate and endotoxin levels have been reported in indoor rather than in outdoor particles (Long et al. 2001; Thorne et al. 2005), and the counts of viable bacteria are apparently higher in indoor air than in outdoor air (Gorny and Dutkiewicz 2002).
Few studies have addressed effects of indoor dust on airway mucosa cells. In one study, indoor dust from two buildings was found to be a particularly strong inducer of interleukin-6 (IL-6) and IL-8 release in airway epithelial cells (Saraf et al. 1999). Similarly, Long et al. (2001) reported that indoor particles induced significantly higher tumor necrosis factor (TNF) production than outdoor particles in alveolar macrophages, even when corrected for higher indoor endotoxin concentrations. In the present study, we characterized an indoor dust sample representative for German households and investigated its effect on human airway cells. As a surrogate for cells occurring in airway mucosa, we used the monocyte cell line Mono Mac 6 (MM6), the epithelial cell line BEAS-2B (B2B), and the T-cell line Jurkat (JKT). The cellular response to indoor dust was assessed with an oligonucleotide cDNA microarray (Bammler et al. 2005). Its 1,232 genes were selected based on differential transcription of human airway mucosa specimens in whole genome arrays under various experimental conditions. On the protein level, we studied the release of various cytokines using a microsphere-based flow cytometric assay. We questioned a) whether an intelligible outline of cell responses to indoor dust can be obtained using gene transcription profiling; b) if cell line-specific differences of the transcriptional and secretory response to indoor dust can be identified; c) if differential gene regulation in response to indoor dust corre-sponds with cellular protein secretion; and d) whether indoor dust specific responses can be observed.
Cell cultures and protein assay. The human monocyte cell line MM6, the human T-cell line JKT (German Resource Centre for Biological Material, Braunschweig, Germany), and the human bronchial epithelial cell line B2B (Cell Concepts, Umkirch, Germany) were adjusted to 5 × 10 5 cells/mL and grown to 80% confluence (B2B) or for a period of 24 hr (MM6 and JKT) in RPMI 1640 media supplemented with 10% fetal calf serum, 2 mM L-glutamine, 1% nonessential amino acids, and 1 mM sodium pyruvate, all from Promo Cell (Heidelberg, Germany); and 50 µg/mL penicillin and 50 µg/mL streptomycin both from Biochrom (Berlin, Germany). For each cell line, three cultures served as controls and three cultures were exposed to 500 µg/mL indoor dust for 6 hr. Cell viability was assessed before and after exposure with trypan blue dye exclusion. For gene expression analysis, RNA was extracted using the RNeasy Mini Kit (Qiagen, Cologne, Germany), and exposed cultures were compared with the corresponding control. In the supernatants, we used a microsphere-based flow cytometric assay (Luminex System; Microbionix, Munich, Germany) to detect 24 cytokines. The limits of detection (LOD) of this assay ranged between 6.9 and 14.8 pg/mL. Microarray spotting and hybridization. Ultra Gaps II coated slides (Corning, Schiphol-Rijk, The Netherlands) were spotted using an OmniGrid 100 spotter (Genemachines, San Carlos, CA, USA). We applied 70mer oligonucleotides (all oligonucleotides from Operon Biotechnologies, Inc., Cologne, Germany) and controls in four repetitive spots. On each slide, we included 1,232 human genes and 10 different extrahuman spiking controls from Arabidopsis and Sinorhizobium genes. Spotting buffer [3 saline sodium citrate (3SSC ):3 M NaCl, plus 0.3 M Na citrate and 1.5 M betaine (Sigma, Deisenhofen, Germany)] and randomized negative controls (i.e., oligonucleotides that do not bind human mRNA) served as controls in 300 and 12 spot quadruples, respectively. For immobilization, oligonucleotides were incubated for 10 min at 80°C followed by ultraviolet cross-linking (Stratalinker 2400; Stratagene, La Jolla, CA, USA) with 120 mJ/cm 2 . Spotting accuracy was checked with random 9mere (Operon Biotechnologies, Inc.).
For hybridization of spiking controls, synthetic mRNA-oligonucleotides of 10 different Arabidopsis and Sinorhizobium genes were synthesized on an ABI 394 synthesizer (Applied Biosystems, Foster City, CA, USA; PURIMEX, Staufenberg, Germany). Of this spiking mRNA, 200-10,000 fg was added to 5 µg total RNA of control and exposure cultures resulting in ratios of 1:2 to 1:10. Then, 5 µg total RNA from control cultures was reverse transcribed with an oligo dT primer carrying a cDNA capture sequence for the fluorescent dye cyanine 3 (Cy3), and 5 µg total mRNA from dustexposed cultures with a capture sequence for cyanine 5 (3 DNA Array 350 Kit; Genisphere, Hatfield, PA, USA). The resulting cDNA was further purified and concentrated with Millipore Microcon YM-30 Centrifugal Filter Device (Millipore, Billerica, MA, USA) and mounted on the spotted slides prehybridized with bovine serum albumin for 1 hr. cDNA was allowed to hybridize with the spotted oligonucleotides at 57°C overnight. cDNA hybridized with the spotted oligonucleotides was then incubated for 3 hr at 59°C with the 3DNA capture reagent.
Microarray data analysis. Slides were scanned on a dual-laser microarray scanner (GenePix 4000 B; Axon Instruments, Foster City, CA) and analyzed with Gene Pix pro 4.1 software (Axon Instruments, Foster City, CA, USA). We used a nonparametric algorithm without background subtraction to assess differential gene expression. In a recent report (Bammler et al. 2005), background subtraction was identified as a significant source of data variability. Therefore, instead of background subtraction, we disregarded spots with a background above the 99th percentile. From the remaining spots per gene, we calculated the median intensity if at least three of four repetitive spots were available. Following lowess-and block-normalization, the dual logarithm of the Cy5/Cy3 intensity ratio was formed for all genes and controls. To identify differentially regulated genes, we calculated the upper and lower quartiles of the log 2 intensity ratios of the 300 control spot quadruples (spotting buffer only). Following the method of Tukey and Aase (1977), we calculated outer fences three interquartile ranges above or below the hinges of the control spots. Values outside the outer fences of the log 2  Tukey and Aase (1977). Oligonucleotides of 10 Arabidopsis and Shinorizobium genes were synthesized and added to the probes in concentration of 200-10,000 fg (spiking controls). Spiking controls were reproducibly separated from buffer spots, whereas randomized negative controls (i.e., oligonucleotides that do not bind human RNA) were always within the calculated fences (not visible due to high spot density). Genes (rhombi) outside the fences were regarded as differentially transcribed. The y-axis is the dual logarithm of mRNA spot intensity after indoor dust exposure divided by spot intensity after control exposure; the x-axis is the geometric mean of spot intensity (arbitrary scale). Intensity ratios of the 300 control spots were defined as differentially transcribed. With this algorithm, control spots could be reproducibly separated from spiking controls ( Figure 1). For each cell line, only genes up-or down-regulated in at least two of three arrays were considered differentially transcribed. All calculations were performed with Systat 10.2 (Systat Software Inc., Richmond, CA, USA). Gene-annotation enrichment analysis. Genes were named according to the Human Genome Organization and grouped into categories defined by the Gene Ontology (GO) Consortium (Bammler et al. 2005;GO Consortium 2006) based on their molecular function and the involved biological process. The number of observed versus expected differentially transcribed genes per GO category were calculated using the web-based platform GOTree Machine (Zhang et al. 2004). In this context, the expected number of differentially transcribed genes equals the number per GO category, if all categories on the particular array are equally affected by up-or down-regulation. The expected number thus represents the fraction of differentially transcribed genes on the whole array multiplied by the number of genes in a particular GO-category on this array, and is not directly associated with control exposure. The enrichment factor equals the odds ratio of expected and actually observed differentially transcribed genes per GO category. We present only significantly enriched GO categories containing a disproportionate amount of differentially transcribed genes (Fisher's exact p < 0.01). These categories provide detailed, cell line-specific information about transcriptional responses to indoor dust. To provide an intelligible overview of biological processes, we used 14 GO slim terms defined by the GO Consortium (Harris et al. 2004). GO slims are a reduced version of the GO ontologies representing a high-level summary of molecular functions and biological processes of differentially transcribed genes. They are presented for all three cell lines, whether enriched or not, and thus allow a direct comparison of the responses of the three cell lines.
Microarray reproducibility. The three arrays per cell line yielded reasonably consistent results. The intraclass correlation coefficients of the log 2 ratios between the three arrays per cell type were 0.87 (0.86-0.88, 95% confidence interval) for MM6, and 0.79 (0.78-0.80) for both B2B and JKT. In none of the nine arrays were the 12 randomized negative controls categorized as differentially transcribed. In contrast, all 10 spiking controls were grouped as differentially transcribed in all arrays. The log 2 intensity ratios of the spiking controls compared well among the nine arrays, with an average coefficient of variation of 0.32. MM6 cells revealed the highest number of up-regulated genes (n = 90), followed by B2B (n = 28) and JKT cells (n = 30, Figure 2). In MM6 cells, 91 genes were down-regulated, whereas only 1 gene was down-regulated in B2B cells and 2 genes in JKT cells.
Significantly enriched GO categories. Eleven genes were up-regulated in all three cell lines following exposure to house dust. Significantly enriched GO categories in all three cell lines included transition metal ion binding (observed, 5; expected, 1.21; enrichment factor, 4.13; p = 0.004) and UDPglycosyltransferase activity (observed, 2; expected, 0.04; enrichment factor, 50; p < 0.001).
In MM6 cells, up-regulated genes were significantly overrepresented in the GO categories of cadmium and copper ion binding, chemokine and cytokine activity, and response to chemical stimulus. Moreover, the GO category apoptosis was enriched, mainly with antiapoptotic genes. Down-regulated genes were overrepresented in the GO categories alcohol and cholesterol metabolism, unfolded protein binding, cell aging, and male sex differentiation (Figure 3).

GO slim categories.
External stimulusrelated GO slim categories, including response to abiotic stimulus and response to stress, were significantly enriched in both MM6 and B2B cells. In MM6 cells, additional immune response-related GO slim categories were enriched. These included response to biotic stimulus (38 vs. 21 expected; p < 0.001), cell adhesion (up-regulated,12; expected, 5; p < 0.05) and cell-cell signaling (up-regulated, 18; expected, 9; p = 0.001). GO slim categories with enrichment of up-regulated genes were associated with underrepresentation of down-regulated genes, and vice versa in MM6 cells (Figure 4). GO slim categories related to external stimuli or immune responses were not significantly enriched in JKT cells.
The GO slim term "cell death" is mainly determined by programmed cell death and apoptosis. In MM6 cells, 19 cell death-related genes were up-regulated (vs. 13 expected; p < 0.05). However, consistent with regulation of apoptosis in all three cell lines, antiapoptotic genes were preferentially up-regulated (observed, 9; expected, 4; p = 0.015). Apoptosis-related genes were not significantly enriched in B2B or JKT cells.
The GO slim terms "cell cycle" and "proliferation" relate to genes involved in cell replication and multiplication. Of the MM6 genes belonging to the GO term cell cycle, 6 were up-regulated versus 10 expected. Of these 6 genes, 4 coded for proteins with negative regulation of cell cycle. Consistently, down-regulated genes were overrepresented in the "cell cycle" category in MM6 cells (down-regulated, 15; expected, 9; p = 0.02). Neither slim category was significantly enriched with up-or down-regulated genes in B2B or JKT cells.
The GO slim term "signal transduction" refers to the cascade of processes mediating a change in the functioning of the cell when a receptor is activated. In MM6 cells, up-regulated genes were significantly enriched (upregulated, 42; expected, 32; p = 0.009). "Transcription" in this context refers the regulation of the synthesis of RNA on a template of DNA. In neither cell line was the category "transcription" significantly enriched with upor down-regulated genes.
The GO slim term "cellular metabolism" covers the chemical reactions and pathways by which individual cells transform chemical substances. Down-regulated genes were significantly enriched in this GO slim category (down-regulated, 49; expected, 38; p = 0.005) in MM6 cells. Neither up-nor downregulated genes were significantly enriched in the GO slim categories development, cell organization and biogenesis, and transport. Although not significant, down-regulated genes revealed a comparatively high enrichment factor (1.9) in the category "transport" in MM6 cells.    Protein concentrations. Supernatant concentrations of 24 proteins were assessed in the three cell lines exposed to indoor dust or control medium for 6 hr. The means ± SEs of three cultures per cell line per exposure group were calculated (Table 2).
Under control and exposure conditions, MM6 cells released the highest amount of cytokines. Increased cytokine release was concordant with gene up-regulation for C-C chemokine ligand 3 (CCL3), colony-stimulating factor 2 (CSF2), C-X-C chemokine ligand 10 (CXCL10), IL-1β, IL-8, and TNF-α. Increased protein release without gene upregulation was observed for interferon gamma (IFN-γ), IL-3, and IL-6. Decreased protein concentration without gene down-regulation was observed for CCL, CCL11, IL-1α, and vascular endothelial growth factor (VEGF). No change in protein and mRNA expression was concordantly found in the remainder.
In B2B supernatants, fewer cytokines were detectable than in MM6 supernatants, and the concentrations were generally lower (Table 2). Higher cytokine concentrations after dust exposure were concordant with gene up-regulation for CCL2, IL-6, and IL-8. Higher protein concentrations without gene up-regulation were observed for IFN-γ, IL-1α, IL-3, and IL-7. Protein and mRNA expression for IL12p40 were discordant. No change in protein and mRNA expression was concordantly found for CCL5, CCL11, CXCL10, and VEGF. No differential gene regulation was observed in proteins undetectable in the B2B supernatants.
Only a few cytokines were detectable in the supernatants of JKT-cells. IL-1α supernatant concentrations were higher after dust exposure without detectable differential gene regulation. CCL11, CXCL10, and VEGF were detectable, but neither supernatant concentrations nor mRNA expression differed between control and dust exposures.

Discussion
In this explorative investigation, we assessed short-term transcriptional and secretory responses of MM6 and B2B human cell lines to an indoor dust sample typical for German homes in vitro. These two cell lines frequently serve as a surrogate for airway mucosa cells. In addition, the undifferentiated human T-cell line JKT was exploratively included. Based on physiologic function of these cell lines, we hypothesized a different response pattern of the three cell lines employed. Moreover, we were interested in the correlation of the transcriptional and secretory response. The study was not designed and the results do not allow to infer on real life conditions; particularly chronic effects that are very important for human health in vivo were not assessed. Outcome parameters included cell viability, a transcriptional profile of 1,232 genes, and the supernatant concentrations of several cytokines and chemokines.
Microarray analysis and gene-annotation enrichment analysis. One question posed by this investigation was whether plausible and interpretable transcriptional profiles can be obtained using large-scale cDNA microarrays. Interarray correlation coefficients of approximately 0.8 indicate the sound reliability of the laboratory processes (Bammler et al. 2005). The accuracy in differentiating randomized negative controls from Arabidopsis and Sinorhizobium spiking controls suggests adequate validity. The fundamental framework for the biological interpretation of gene transcription data was provided by the GO Consortium (2006). Current computational tools, such as the GOTree Machine platform, allow identification of overrepresentation of up-or down-regulated genes within the various GO categories (Zhang et al. 2004). Using these resources, we analyzed differentially transcribed genes in a two-step approach. First, GO themes significantly enriched with up-or down-regulated genes were listed including all levels of the GO domains Biological Process and Molecular Function. Thus, a comprehensive and detailed view on the transcriptional response was obtained ( Figure 3). In addition, a concise transcriptional profile was obtained using a fixed set of GO slim categories (Harris et al. 2004), which allow direct comparisons of the transcriptional response of the three cell lines (Figure 4). Combining both approaches, a biologically sound profile of the cellular responses to indoor dust could be conceptualized. This transcriptional profile is characterized by enhancement of detoxification, a danger and defense response, securing of cell survival, and a cellular energy-saving program with reduced proliferative and metabolic activity.
Cell survival rather than apoptosis was promoted in all three cell lines. In MM6 cells, 19 cell death related genes were up-regulated compared with 13 expected (p = 0.04). Consistent with the viability assays, which did not show relevant cell death in response to indoor dust exposure, antiapoptotic genes were preferentially up-regulated. Various transcriptional responses resembled a cellular economy drive under stress conditions. Cell functions not involved in detoxification, defense, and survival were down-regulated. Genes grouped under the GO category "cell proliferation" preferentially transmitted negative regulation of cell proliferation (p = 0.04). Similarly, emissions of indoor dust and ambient particulates reduced epithelial cell proliferation in two recent studies (Baulig et al. 2003;Mathiesen et al. 2004). Consistently, genes promoting the progression through the cell cycle were significantly downregulated in MM6 cells. Down-regulated genes were also overrepresented in the GO slim categories "cellular metabolism" and "transport'" A term related to transport is "endocytosis," including phagocytosis, which is particularly interesting in dust-exposed cells. Up-and down-regulated genes were not significantly enriched in the category "endocytosis" or in any of its related terms. This observation is in line with a recent publication, which reported decreased phagocytic activity after particle exposure (Lundborg et al. 2006).

Cellular mRNA levels and protein secretion.
In large-scale studies, microarrays demonstrated reasonable validity to assess cellular mRNA levels when compared with various polymerase chain reaction (PCR) methods (Canales et al. 2006;Cortez et al. 2006;Wang et al. 2006). In the present study microarrays were calibrated with quantitatively added mRNA spiking controls, as described by Wang et al. (2003). Moreover, we assessed the differential transcription of functional groups of genes, rendering single gene products less relevant. Additional confirmation of differential gene expression with PCR-techniques was therefore considered dispensable. However, conclusions on the transcriptional response of single genes should be made with caution without PCR confirmation, particularly if the transcriptional and secretory response are inconsistent.
In the present study we focused on the concordance of mRNA and protein expression. The correlation of cellular mRNA concentrations and protein expression depends on various parameters, including intracellular mRNA amount and location, translational regulation, ribosome density and occupancy, cellular protein degradation, experimental conditions, and stochastic noise inherent in large-scale experiments. In comparisons of mRNA levels with limited protein expression data, fair correlations were found (Andrew et al. 2003;Cortez et al. 2006;Greenbaum et al. 2003). However, on a genomic scale, mRNA-protein correlations were poor (Greenbaum et al. 2003). In the present study, we compared changes of cellular mRNA levels of 24 inflammatory response proteins with the according change in supernatant protein concentrations. Differential expression of mRNA and protein expression agreed in 53 of 69 pairs assessed (Table 3), indicating a relevant association of the transcriptional response and protein secretion in this small set of functionally corresponding proteins. Interestingly, all genes that were up-regulated at least 2-fold also showed increased protein secretion. Moreover, a T H1 -skewed immune response profile was observed on both the transcriptional and the protein levels.
Cell line-specific differences in the response to indoor dust. Airway mucosa contains various cell types that may differ in their response to inhaled dust constituents. One question posed by this study was whether cell line-specific differences of the transcriptional and secretory response to indoor dust can be identified. MM6 cells revealed the highest fraction of differentially transcribed defense genes and the highest supernatant concentrations of cytokines and chemokines. This suggests a key role in the mucosal defense response. The high responsiveness of MM6 cells is consistent with the physiologic function of monocytes/ macrophages representing the first line defense and surveillance of the innate immune system. Moreover, these cells are able to present antigen to immune cells, which is associated with extensive transcriptional and secretory activity.
In contrast, the main physiologic function of epithelial cells is to form a dermal or mucosal barrier. Consistent with this function, less intense transcriptional and secretory response was expected. However, epithelial cells also contribute actively to innate immune functions (Bals and Hiemstra 2004;Baulig et al. 2003). Although the result of the present study suggests that epithelial immune functions are less intense than those of macrophages, the high number of epithelial cells in the airways may render the epithelial innate immune response highly effective.
Naive T cells in respiratory mucosa are generally activated by antigen on major histocompatibility class II (MHC-II) molecules (Del Prete et al. 1993). They are not able to phagocytize. In monoculture, JKT cells responded to a limited number of stimuli including phorbol ester, phytohemagglutinin, metals, formaldehyde, calcium ionophores, and pyrene, but not to hydrogen peroxide or mitomycin C ( Brundage et al. 2004;Saito et al. 2005;Verstraeten 2006). After exposure to indoor dust, JKT cells remained more or less immunologically dormant. This result suggests that without the help of antigen-presenting cells, indoor dust is not able to elicit a significant response in JKT cells. In coculture with MHC-matched antigen-presenting cells, JKT cells may be more intensively activated.

Conclusion
We obtained a biologically plausible transcriptional profile in response to coarse indoor dust as deposited in the upper airways using cDNA microarrays and gene-annotation enrichment analysis. Validity of this new type of bioinformatics analysis will probably be further substantiated within the next few years. MM6, B2B, and JKT cells responded similarly with detoxification, antiapoptosis, and reduced mitotic and metabolic activity to a representative indoor dust from German homes. The three cell lines differed particularly in the intensity of their defense response on both the transcriptional and secretory levels. Consistent with previous studies, the observed defense response to indoor dust in MM6 and B2B cells promoted a T H1 -skew and suggested weak oxidative stress and subtle DNA damage (Allermann et al. 2003;Long et al. 2001;Saraf et al. 1999). This points toward a major role of bioorganic dust constituents such as endotoxin, fungal, and bacterial contaminants. However, ambient air particles of smaller particle sizes with higher metal and lower bioorganic content induced a marked release of inflammatory mediators in association with a more pronounced oxidative stress response (Ghio 2004;Hetland et al. 2005). Also, allergens in indoor dust may alter cell functions because of their enzymatic and toxic activity (Martinez et al. 1999); however, in contrast with the in vivo situation, IgE mediated effects are considered less likely because mast cells were absent in the cell cultures.