Reduced gene expression levels after chronic exposure to high concentrations of air pollutants

Mutation Research 780 (2015) 60–70

Contents lists available at ScienceDirect

Mutation Research/Fundamental and Molecular Mechanisms of Mutagenesis journal homepage: www.elsevier.com/locate/molmut Community address: www.elsevier.com/locate/mutres

Reduced gene expression levels after chronic exposure to high concentrations of air pollutants Pavel Rossner Jr. a,∗ , Elena Tulupova a , Andrea Rossnerova a , Helena Libalova a , Katerina Honkova a , Hans Gmuender b , Anna Pastorkova a , Vlasta Svecova a , Jan Topinka a , Radim J. Sram a a b

Department of Genetic Ecotoxicology, Institute of Experimental Medicine, Prague, Czech Republic Genedata AG, Basel, Switzerland

a r t i c l e

i n f o

Article history: Received 3 April 2015 Received in revised form 6 August 2015 Accepted 6 August 2015 Available online 11 August 2015 Keywords: Chronic exposure Air pollution Gene expression profiles Human health Particulate matter Polycyclic aromatic hydrocarbons

a b s t r a c t We analyzed the ability of particulate matter (PM) and chemicals adsorbed onto it to induce diverse gene expression profiles in subjects living in two regions of the Czech Republic differing in levels and sources of the air pollution. A total of 312 samples from polluted Ostrava region and 154 control samples from Prague were collected in winter 2009, summer 2009 and winter 2010. The highest concentrations of air pollutants were detected in winter 2010 when the subjects were exposed to: PM of aerodynamic diameter <2.5 ␮m (PM2.5) (70 vs. 44.9 ␮g/m3 ); benzo[a]pyrene (9.02 vs. 2.56 ng/m3 ) and benzene (10.2 vs. 5.5 ␮g/m3 ) in Ostrava and Prague, respectively. Global gene expression analysis of total RNA extracted from leukocytes was performed using Illumina Expression BeadChips microarrays. The expression of selected genes was verified by quantitative real-time PCR (qRT-PCR). Gene expression profiles differed by locations and seasons. Despite lower concentrations of air pollutants a higher number of differentially expressed genes and affected KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways was found in subjects from Prague. In both locations immune response pathways were affected, in Prague also neurodegenerative diseases-related pathways. Over-representation of the latter pathways was associated with the exposure to PM2.5. The qRT-PCR analysis showed a significant decrease in expression of APEX, ATM, FAS, GSTM1, IL1B and RAD21 in subjects from Ostrava, in a comparison of winter 2010 and summer 2009. In Prague, an increase in gene expression was observed for GADD45A and PTGS2. In conclusion, high concentrations of pollutants in Ostrava were not associated with higher number of differentially expressed genes, affected KEGG pathways and expression levels of selected genes. This observation suggests that chronic exposure to air pollution may result in reduced gene expression response with possible negative health consequences. © 2015 Elsevier B.V. All rights reserved.

1. Introduction

Abbreviations: 8-oxodG, 8-oxo-7,8-dihydro-2 -deoxyguanosine; APEX1, apurinic/apyrimidinic endonuclease 1; ATM, ataxia-telangiectasia mutated; B[a]P, benzo[a]pyrene; c-PAHs, carcinogenic polycyclic aromatic hydrocarbons; DMAP1, DNA methyltransferase 1 associated protein 1; FAS, Fas cell surface death receptor; FG /100, genomic frequency of translocations/100 cells; GADD45A, growth arrest and DNA-damage-inducible, alpha; GSTM1, glutathione S-transferase mu 1; IL1B, interleukin 1 beta; 15-F2t -IsoP, 15-F2t-isoprostane; KEGG, Kyoto Encyclopedia of Genes and Genomes; MN/1000 BNC, micronuclei/1000 binucleated cells; PAHs, polycyclic aromatic hydrocarbons; PCA, principal component analysis; PM, particulate matter; PM2.5, particulate matter of aerodynamic diameter <2.5 ␮m; PTGS2, prostaglandin-endoperoxide synthase 2; qRT-PCR, quantitative real-time PCR; RAD21, RAD21 homolog (S. pombe). ∗ Corresponding author at: Videnska 1083, 14220 Prague, Czech Republic. Fax: +420 241062785. E-mail address: [email protected] (P. Rossner Jr.). http://dx.doi.org/10.1016/j.mrfmmm.2015.08.001 0027-5107/© 2015 Elsevier B.V. All rights reserved.

A significant number of inhabitants of both developed and developing countries are exposed to high concentrations of air pollutants. This has serious health consequences including increased incidence of cardiovascular and respiratory diseases, neurodegenerative disorders and cancer [1]. Air pollutants, which are often by-products of incomplete combustion of organic material (from traffic, local heating and industrial production), are adsorbed onto particulate matter (PM). The effect of PM on human health is determined by the size and chemical composition of the PM. Coarse particles of aerodynamic diameter <10 ␮m are deposited in the thoracic region of the lungs [2]. Fine particles of aerodynamic diameter <2.5 ␮m (PM2.5) can penetrate the lung alveoli and cause inflammation [3]. Ultrafine particles (<0.1 ␮m) can even enter cells and

P. Rossner Jr. et al. / Mutation Research 780 (2015) 60–70

cause direct damage to macromolecules [4]. Chemicals adsorbed on the surface of fine PM that impact human health include e.g. polycyclic aromatic hydrocarbons (PAHs), transition metals and volatile organic compounds (VOCs). These chemicals are released from PM in the lungs and enter the organism via the bloodstream. Exposure to PAHs is particularly deleterious to human health. Some PAHs are classified as probably or possibly carcinogenic to humans (c-PAHs) [5] and benzo[a]pyrene (B[a]P), arguably the best known and most studied c-PAH, is a group 1 carcinogen [6]. A study on a model system indicates that in the lungs B[a]P is released from PM in two steps. First, a smaller fraction (about 30%) is rapidly released into the circulation and metabolized. The remaining B[a]P is desorbed slowly and 5.6 months after inhalation about one third of the compound is still present on the particles with reduced bioavailability [7]. Biomarkers of exposure, effect and susceptibility have been used to evaluate the effects of air pollution on human health for many years [discussed in [8]]. Although many studies have successfully demonstrated the suitability of biomarkers for monitoring the health effects, both in environmentally and occupationally exposed populations, some reports have failed to show changes in their levels [reviewed examples in [9–12]]. The lack of an effect of air pollutants on biomarkers may be partly explained by the complex interactions between the organism and the environment, which cannot be captured by the measurement of a single (or several) biomarker(s). Technological developments over the last decade have opened up new avenues in biomarker research and many researchers shifted their focus to studies of mechanisms of action of air pollutants using analyses of gene expression. These studies, conducted mostly on cell or animal models, showed a crucial role of PM in changes of expression of genes participating in inflammatory response, oxidative stress, vascular dysfunction or progression of atherosclerosis in various body parts including the brain and cardiovascular system [13–16]. These results are consistent with the proposed pathophysiological effects of air pollution exposure mentioned above. Currently, analyses of expression of single or several pre-selected genes are commonly replaced by genome-wide approaches based on microarrays and next-generation sequencing that allow obtaining a comprehensive assessment of the gene expression signatures associated with exposures [17]. Global gene expression analysis has become affordable for many laboratories and has recently resulted in the publication of transcriptomics data [18]. Many of the deregulated pathways identified in the studies on model systems included the genes detected in gene-specific experiments; these pathways were associated with e.g. antioxidant response, xenobiotic metabolism, inflammatory signaling, endothelial dysfunction or immune response [19–22]. Although studies that report results of gene expression profiling in human subjects exposed to polluted air have recently been published, they are usually small and deal with exposure to tobacco smoke or inhalation exposure to air pollutants, rather than real exposure to environmental PM [23–27]. Thus, a larger study that involves human subjects from a highly-polluted environment is still required [28]. Recently, we have conducted a biomarker study on subjects living and working in a heavily polluted industrial region of Ostrava, located in the northeastern part of the Czech Republic [29]. Repeat samples were collected during the winter and summer of 2009 and the winter of 2010. The biological samples were analyzed for a panel of traditional biomarkers, which included levels of bulky DNA adducts, oxidative stress markers [8oxo-7,8-dihydro-2 -deoxyguanosine (8-oxodG), protein carbonyls and 15-F2t -isoprostane (15-F2t -IsoP)] and cytogenetic parameters (frequency of stable and unstable chromosomal aberrations). Unexpectedly, levels of most of these biomarkers did not reflect exposure

61

to PM2.5 and B[a]P, although the concentration of PM2.5 reached 119 ␮g/m3 and B[a]P reached 74.2 ng/m3 . The levels of biomarkers were lower or similar to the levels measured in the control subjects from Prague [29,30]. There were some exceptions: levels of B[a]Plike DNA adducts and protein carbonyls were higher in Ostrava than in Prague, but only in the winter of 2009. 15-F2t -IsoP was the only parameter that corresponded to differences in air pollution in all seasons, although the difference between locations in the winter of 2010 was not significant. The results of biomarkers analyses in subjects enrolled in the present study are presented in Table 1 . Here we aimed to explain these results by analysis of global gene expression in leukocytes from 466 blood samples. Our goal was to identify differentially expressed genes and signaling pathway differences between the regions. We further selected nine differentially expressed genes and verified their expression using quantitative real-time PCR (qRT-PCR). We hypothesized that analysis of global gene expression changes will help us to explain our previous data describing unexpected and minimal changes of biomarker levels in population exposed to high concentration of air pollutants in comparison to less exposed population. We further expected to observe gene expression changes related to DNA repair processes, cell cycle regulation, metabolism of xenobiotics and inflammatory response of the organism.

2. Materials and methods 2.1. Study subjects and exposure assessment In this study the number of samples analyzed during individual sampling seasons (winter 2009, summer 2009 and winter 2010) varied from 47 to 58 (control subjects from Prague) and 84-133 (exposed subjects from the Ostrava region). Most of the subjects (85%) were sampled repeatedly either in all (94 subjects, corresponding to 282 individual samples) or in two sampling seasons (57 subjects, corresponding to 114 individual samples); however, 70 subjects (15%), particularly those from the Ostrava region, provided samples only in summer 2009 or winter 2010. For all analyses, the repeatedly sampled subjects were considered individual samples. The samples represented a subset of the population analyzed and described previously [29]. All study participants were healthy, nonsmoking men. Smoking status was confirmed by analysis of urinary cotinine levels. Anyone who had undergone medical treatment, vaccination or radiography, up to three months before sampling, was excluded from the study to avoid potential effects on the parameters being analyzed. Prague volunteers were city policemen; subjects recruited in the Ostrava region were policemen who originated from Havirov and Karvina, smaller cities located southeast and north-east of Ostrava city, and office workers working in Ostrava city. In winter 2010 volunteers from Ostrava-Bartovice, a suburb of Ostrava located to the east of the biggest steelworks in the Ostrava region which is one of major sources of air pollution in the region also participated. The Ostrava region is characterized by a high concentration of heavy industry, particularly steelworks, coke ovens and coal mines. The region is infamous for being the most polluted part of the Czech Republic and, arguably, of the entire European Union [further details provided in [29,31]. The study participants provided informed consent and could cancel their participation at any time, in accordance with the Helsinki II declaration. The study was approved by the Ethical Committee of the Institute of Experimental Medicine AS CR, in Prague. Blood samples were collected by venipuncture, into vacuettes containing ethylenediamine tetra acetic acid (EDTA). Leukocytes were isolated and stabilized using the LeukoLOCKTM Total RNA Stabilization and Isolation System (LeukoLOCK; Ambion, Austin, TX,

62

P. Rossner Jr. et al. / Mutation Research 780 (2015) 60–70

Table 1 Basic characteristics of the study population. Sampling

Parameter

Prague: median (min, max)N = 47

Ostrava: median (min, max)N = 84

p

Winter 2009

Age (years) B[a]P (ng/m3 ) PM2.5 (␮g/m3 ) Benzene (␮g/m3 ) B[a]P-like DNA adducts/108 nucleotides FG /100 MN/1000 BNC 8-oxodG (nmol/mmol creatinine) 15-F2t -IsoP (pg/ml plasma) Protein carbonyl (nmol/ml plasma)

37.0 (25.0, 61.0) 0.68 (0.10, 3.37) 15.00 (7.60, 18.70) 3.45 (2.48, 76.20) 0.20 (0.12, 0.37) 1.12 (0.00, 3.73) 7.00 (2.67, 12.00) 4.91 (0.020, 14.36) 158.4 (108.7, 321.2) 21.3 (16.4, 36.4)

34.5 (22.0, 63.0) 1.90 (0.70, 12.6) 39.0 (15.0, 76.7) 6.59 (2.48, 28.4) 0.27 (0.16, 0.47) 1.12 (0.0, 6.72) 3.67 (0.33, 7.33) 5.53 (0.04, 11.8) 178.3 (101.3, 2410.3) 23.2 (16.2, 34.9)

=0.702 <0.001 <0.001 <0.001 <0.001 0.858 <0.001 0.209 <0.001 <0.05

Sampling

Parameter

Prague: median (min, max)N = 58

Ostrava: median (min, max)N = 95

p

Summer 2009

Age (years) B[a]P (ng/m3 ) PM2.5 (␮g/m3 ) Benzene (␮g/m3 ) B[a]P-like DNA adducts/108 nucleotides FG /100 MN/1000 BNC 8-oxodG (nmol/mmol creatinine) 15-F2t -IsoP (pg/ml plasma) Protein carbonyl (nmol/ml plasma)

37.0 (25.0, 61.0) 0.08 (0.04, 1.00) 13.2 (8.80, 21.0) 2.96 (1.21, 15.9) 0.23 (0.12, 0.66) 1.12 (0.00, 4.85) 9.33 (3.67, 20.0) 5.20 (0.10, 20.4) 116.3 (64.8, 376.5) 22.4 (16.7, 27.8)

35.0 (23.0, 60.0) 0.37 (0.08, 1.24) 13.0 (6.00, 20.0) 5.18 (1.52, 111.0) 0.11 (0.04, 0.29) 1.12 (0.0, 6.72) 7.00 (1.67, 16.3) 5.05 (0.09, 18.4) 127.8 (53.0, 262.1) 19.9 (13.7, 32.7)

=0.309 <0.001 =0.417 <0.001 <0.001 =0.523 <0.001 =0.835 <0.05 <0.001

Sampling

Parameter

Prague: median (min, max)N = 49

Ostrava: median (min, max)N = 133

p

Winter 2010

Age (years) B[a]P (ng/m3 ) PM2.5 (␮g/m3 ) Benzene (␮g/m3 ) B[a]P-like DNA adducts/108 nucleotides FG /100 MN/1000 BNC 8-oxodG (nmol/mmol creatinine) 15-F2t -IsoP (pg/ml plasma) Protein carbonyl (nmol/ml plasma)

38.0 (26.0, 62.0) 2.56 (0.28, 11.5) 44.9 (15.0, 55.1) 5.47 (2.28, 10.8) 0.23 (0.07, 0.62) 1.31 (0.0, 4.11) 7.0 (3.0, 14.0) 5.02 (0.52, 7.50) 236.9 (119.6, 647.8) 23.2 (10.0, 39.1)

37.0 (23.0, 61.0) 9.02 (2.19, 74.2) 70.0 (30.0, 119.0) 10.2 (5.34, 59.6) 0.15 (0.06, 0.40) 1.12 (0.0, 6.35) 6.33 (2.33, 15.0) 4.92 (0.19, 10.4) 266.4 (90.6, 814.6) 21.3 (11.2, 43.3)

=0.864 <0.001 <0.001 <0.001 <0.001 =0.083 =0.197 =0.881 =0.291 =0.085

B[a]P, benzo[a]pyrene; PM2.5, particulate matter of aerodynamic diameter <2.5 ␮m; FG /100, genomic frequency of translocations/100 cells; MN/1000 BNC, micronuclei/1000 binucleated cells, 8-oxodG, 8-oxo-7,8-dihydro-2 -deoxyguanosine; 15-F2t -IsoP – 15-F2t -isoprostane.

USA), within 12 h of blood collection. Stabilized leukocytes were stored at -80 ◦ C until analysis. Exposure to B[a]P bound to PM2.5 was monitored using personal samplers worn by the study participants for 48 h before collection of biological samples. The samplers were equipped with filters which collected PM2.5 [32]. Benzene was collected on Radiello® radial diffusive samplers (Fondazione Salvatore Maugeri, Padova, Italy) worn by the study subjects for 24 h. It was adsorbed on graphitized charcoal and recovered by thermal desorption. The analysis was performed by capillary gas chromatography with flame ionization detection in the certified laboratory ALS Czech Republic, Prague. Information on the PM2.5 concentrations was obtained from stationary monitors that were operated by the Czech Hydrometeorological Institute (CHMI, www.chmi.cz).

transcription using the Illumina TotalPrep RNA Amplification Kit (Ambion, Austin, TX, USA). Two of these samples were excluded from statistical analysis as described in section 2.5. Thus, the final sample size was 466. The resulting biotinylated cRNA (750 ng) was then hybridized to Illumina Human-HT12 v3 Expression BeadChips (Illumina, San Diego, CA, USA). The hybridization process and subsequent washing, staining and drying of the beadchips followed the standard instructions from Illumina. The hybridized beadchips were then scanned on the Illumina BeadStation 500GX. Bead level data were summarized by the Illumina BeadStudio Software v3.3.7. Each array on the Human-HT-12 Expression BeadChip targets more than 25,000 annotated genes, with more than 48,000 probes derived from the National Center for Biotechnology Information Reference Sequence (NCBI) RefSeq (Build 36.2, Rel 22) and the UniGene (Build 99) databases.

2.2. RNA isolation and quality control 2.4. Quantitative RT-PCR verification The RNA was extracted from frozen leukocytes using the LeukoLOCK system. Quantification was performed using a Nanodrop ND-1000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). The integrity of the RNA was assessed with an Agilent 2100 Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA, USA) and isolated RNA was stored at −80 ◦ C until further processing. To minimize a possible batch effect the RNA samples were processed together after all blood samples were collected. 2.3. Gene expression profiling and data analysis Biotinylated cRNA was prepared from 468 samples by reverse transcription of RNA (200 ng) to cDNA and subsequent in vitro

The cDNA was synthesized from 1000 ng of RNA using the High Fidelity cDNA Synthesis Kit (Roche, Mannheim, Germany). The 20 ␮l reaction mixture contained 2.5 ␮M oligo (dT) and 10 ␮M random hexanucleotides. The cDNA synthesis was performed under the following conditions: 30 min at 55 ◦ C and 5 min at 85 ◦ C. Quantitative RT-PCR measurements were performed using the 7900HT Fast Real-Time PCR System (Applied Biosystems, Carlsbad, CA, USA). Each qRT-PCR reaction was carried out in a final volume of 14 ␮l, which contained 3.5 ␮l of diluted cDNA, 2.8 ␮l of water and 7 ␮l of master mix (Primerdesign, Southampton, UK). To determine the level of each target gene, 0.7 ␮l of custom designed primers (PerfectProbe, Primerdesign) were added

P. Rossner Jr. et al. / Mutation Research 780 (2015) 60–70

to the reaction mixture (list of primers and their sequences are shown in Supplemental Table I). Cycling conditions were: 10 min at 95 ◦ C, followed by 40 cycles of amplification (15 s at 95 ◦ C, 30 s at 50 ◦ C and 15 s at 72 ◦ C). The SDS Relative Quantification Software version 2.3 (Applied Biosystems, Carlsbad, CA, USA) was used to analyze the raw data and to assign the baseline and threshold for Ct determination. The Ct values were used to calculate the fold change of gene expression using the delta delta Ct method. The expression levels of the target genes were normalized to the expression levels of the reference gene, YWHAZ. Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.mrfmmm.2015.08.001

2.5. Statistical analysis The expression values were imported into Genedata Expressionist from the Illumina Inc. BeadStudio file, as average signals. The detection scores were used as quality parameters and expression values were transformed into log values for analysis. The use of an Illumina quality parameter score of >0.95 (p-value < 0.05) resulted in approximately 24–40 % values below the selected pvalue threshold. In two samples only 10% and 22% values below the selected p-value threshold were obtained. Both samples showed different distributions of expression values in a box plot and histogram; the reason for these results was not identified. These two samples were excluded from further analysis. The values of the transcript intensities were shifted in a way that they have for each experiment the same median, namely 100. Pearson’s correlation analyses were used to compare gene expression values with other variables; for comparison of gene expression between two groups (locations or seasons) T-tests were applied; ANOVA was employed when more than two groups were compared. All the tests were performed with the log transformed and median normalized data using the following conditions: All available transcripts were used. Correction for multiple comparisons was performed using the Benjamini-Hochberg (BH) [33] method. Only transcripts with a BH q-value of <0.05 were considered to be significant. In addition, the ratios of the group medians (fold changes) were determined and the threshold for significance was set to a fold change of >1.5 (up-or down-regulated). Results of correlation analyses were considered significant if correlation coefficients were at least ±0.5. The transcripts that showed significant changes had to satisfy the condition that each group contained at least 50 % valid values. To determine the associations between gene expression and exposure to air pollutants (B[a]P, PM2.5 and benzene) in individual locations multivariate linear regression analyses were conducted. The linear models were adjusted to known life-style variables (plasma levels of vitamin A, C, E, HDL-, LDL- and total cholesterol levels, cotinine), age, BMI [29], and laboratory-analysis related parameters (RNA integrity number, microarray batch). All covariates were log-transformed. Principal component analysis (PCA) was applied to detect potential regional and seasonal effects. PCA was performed with all normalized transcription intensities and all 466 samples and colored according to the sampling season to visualize directions of maximum variance in the data using the covariance matrix. For PCA imputation was performed using the row means. 2D hierarchical clustering or divisive hierarchical clustering were performed with Manhattan (L1) as distance measure and Ward as linkage. Clustering was performed under the condition that for each transcript at least 20% valid values for 2D hierarchical clustering and at least 50% valid values for divisive clustering were available.

63

Differentially regulated biological processes in Gene Ontology categories were analyzed using the Functional classification analysis in the DAVID v6.7 tool [34]. Transcript overrepresentation was performed with respect to the total number of transcripts assayed and annotated within each group on the background of Illumina Human-HT12 v3 Expression BeadChips. Of the 48,803 probes on the arrays, 8380 could be annotated with KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway information. Fisher’s Exact tests on the significant transcripts, determined by the t-tests, ANOVA and fold change values were applied to identify over-represented KEGG pathways. Only pathways with p-values < 0.05 were considered to be significant. To account for redundant gene expression changes the Functional Annotation Clustering module in the DAVID v6.7 tool was used [34]. Fold change values obtained by qRT-PCR were analyzed by SPSS 20.0 software. Descriptive statistics were calculated and the groups were compared using either the Mann–Whitney U-test, for variables that did not follow a normal distribution, or the t-test for data that was normally distributed. 3. Results 3.1. Characteristics of the study population The basic characteristics of the study population are presented in Table 1. The data represent a subset (N = 466) of samples analyzed in our previous studies [29,30]. Concentrations of air pollutants (B[a]P, PM2.5 and benzene) were significantly higher in the Ostrava region, during all seasons, with the exception of PM2.5 in the summer of 2009. However, the higher air pollution in Ostrava was not reflected in the levels of biomarkers of damage to macromolecules described in our previous studies [29,30] and summarized in Table 1.

3.2. Principal component and hierarchical clustering analysis Principal component analysis (PCA) of microarray data showed that samples collected in winter 2010 are different from samples taken during the other two sampling seasons. As shown in Fig. 1, these samples form the most distinct cluster. In winter 2010, the highest concentrations of pollutants were noted in both regions. Although levels of pollutants were generally significantly higher in the Ostrava region, we did not observe clear clustering by the sampling season and location. To check for a possible batch effect we conducted a Z-normalization of the data which is based on the normalization of each microarray batch so that the transformed values have zero mean and unit variance. As this procedure did not change the results (data not shown), we concluded that the sampling season was the main driver of the separation in the PCA. Hierarchical clustering analysis of all transcripts and samples separated by locations and sampling seasons showed some clustering only for samples collected in winter 2010, but none for other sampling seasons (Supplemental Fig. 1). After restricting the clustering analysis to significantly differentially expressed transcripts the region-dependent clustering became more pronounced (Supplemental Fig. 2; a detailed description of selected transcripts is provided in Section 3.3). Further restriction of samples to those collected in both locations in winter 2010 showed that expression levels of differentially expressed transcripts were higher in Prague than in the Ostrava region (Fig. 2). Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.mrfmmm.2015.08.001 We further performed hierarchical clustering analysis of upand downregulated transcripts for samples from both locations

64

P. Rossner Jr. et al. / Mutation Research 780 (2015) 60–70

tion of downregulated transcripts was higher than those that were upregulated. Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.mrfmmm.2015.08.001

3.3. Identification of differentially expressed genes in individual locations and seasons

Fig. 1. Principal component analysis of gene expression data from microarrays for individual sampling seasons. PCA was used to identify potential regional and seasonal differences. Individual samples were colored according to the region and sampling season to visualize directions of maximum variance in the data using the covariance matrix. The data indicate a separation of winter 2010 samples from both locations along the axis 1.

regardless sampling season and detected two distinct clusters of upregulated genes, one for each location (Supplemental Fig. 3). Further identification of the samples showed that they originate from the winter 2010 sampling (data not shown). Interestingly, levels of transcription were higher in samples from the Ostrava region. It is also interesting to note, that intensity of transcrip-

Gene expression changes were analyzed using t-tests for each location separately considering the sampling season as a variable. Out of 4270 differentially expressed genes in Prague population and 3360 genes in Ostrava population, surprisingly considerable amount of differentially expressed genes (2193) were overlapping in both locations. 2077 and 1167 differentially expressed genes were unique only for Prague population and Ostrava population, respectively (for a list of those regionspecific transcripts see Supplemental Table II). In Prague subjects, differentially expressed specific transcripts included e.g. those participating in cell cycle regulation [ABL proto-oncogene-1 (ABL1), cyclin A1 (CCNA1), RAD homologs (S. pombe) (RAD1, RAD17, RAD21)], DNA repair [alkylation repair homolog (E. coli) (ALKBH), excision repair cross-complementing rodent repair deficiency (ERCC1, ERCC2)], apoptosis [B-cell CLL/lymphoma 10 (BCL10), caspase 2 (CASP2)], processes associated with oxidative stress [catalase (CAT), myeloperoxidase (MPO), superoxide dismutase 2 (SOD2), 8-oxoguanine DNA glycosylase (OGG1)], and immune functions [chemokine (C–C motif) receptor 7 (CCR7), chemokine (C-X-C motif) ligand 10 (CXCL10), interleukin 1 receptor, type II (IL1R2)]. In subjects from the Ostrava region we detected e.g. the transcripts responsible for response to DNA damage [ataxia telangiectasia mutated (ATM)], apoptosis [caspase 6 and 8 (CASP6, CASP8)], immune functions [interferon, gamma (IFNG), interleukin 1, beta (IL1B), chemokine (C-X-C motif) ligand 6 (CXCL6), chemokine (C–C motif) ligand 3-like 1 (CCL3L1), tumor necro-

Fig. 2. Hierarchical clustering analysis of significantly differentially expressed transcripts in individual regions restricted to winter 2010 sampling. Expression levels of differentially expressed transcripts are higher in Prague than in the Ostrava region. Columns represent individual samples, intensities of individual transcripts are in rows.

P. Rossner Jr. et al. / Mutation Research 780 (2015) 60–70

sis factor receptor superfamily, member 12A (TNFRSF12A), tumor necrosis factor receptor superfamily, member 1A (TNFRSF1A), toll interacting protein (TOLLIP)] and metabolism of xenobiotics [glutathione S-transferase M1 (GSTM1), M2 (GSTM2) and T1 (GSTT1)]. The list of selected transcripts along with qvalues and fold changes is provided in Supplemental Table III. Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.mrfmmm.2015.08.001 Season specific analyses in Prague and the Ostrava region identified differences in the number of differentially expressed transcripts. Although the greatest differences in the concentrations of air pollutants were observed between summer 2009 and winter 2010 samplings, the biggest difference in the number of differentially expressed genes was detected for a comparison of winter 2009 and winter 2010 (1577 and 706 transcripts, for Prague and Ostrava region subjects, respectively). Interestingly, in both regions we detected substantially more downregulated than upregulated transcripts (1495 vs. 82 and 443 vs. 263 for downregulated vs. upregulated transcripts in Prague and Ostrava region subjects, respectively). We further compared gene expression changes between Prague and the Ostrava region in winter 2010 when the concentrations of air pollutants were the highest. We detected 165 differentially expressed transcripts, 85 of them were upregulated and 80 were downregulated in subjects from Prague. The list of the transcripts along with q-values and fold changes is presented in Supplemental Table IV. Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.mrfmmm.2015.08.001 To investigate possible associations between gene expression and traditional biomarkers, we also calculated correlations between gene expression levels and bulky DNA adduct levels, frequency of chromosomal aberrations and oxidative stress markers but no significant results were observed (data not shown). 3.4. Analysis of biological processes and KEGG pathways Functional classification analysis revealed that in both regions differentially regulated transcripts were involved in processes associated with transcription, translation, replication of DNA and cell division. In Prague subjects we further detected processes associated with mitochondria functions (mitochondrial membrane, oxidative phosphorylation, respiratory chain) (Table 2).

65

Table 2 Differentially regulated biological processes in individual locations. Prague-specific transcripts

Ostrava-specific transcripts

Nucleolus Nucleolus Cytosolic ribosome Transcription initiation mRNA processing Transcription Protein catabolic process RNA processing ATP-binding RNA splicing Protein kinase activity Ribosome Mitotic cell cycle Translation ATP-dependent helicase activity Endoplasmic reticulum Zinc ion binding Mitotic cell cycle Helicase activity Nucleosome Regulation of actin cytoskeleton organization Ubiquitin-dependent protein catabolic process GTP-binding ATP-binding Zinc ion binding Mitochondrial membrane Respiratory chain Oxidative phosphorylation Differentially regulated biological processes were analyzed using the DAVID v6.7 tool. Cutoff value of statistical significance was p < 0.05.

In a comparison between winter 2009 and winter 2010 sampling in Prague we observed upregulation of processes involved in negative regulation of transcription, translation, and macromolecule metabolic processes. In agreement with this observation we also found downregulation of processes associated with transcription, translation, replication or oxidative phosphorylation. In the Ostrava region, we detected upregulation of processes associated with translation and protein modification and downregulation of processes involved in gene regulation and signal transduction (Supplemental Table V). Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.mrfmmm.2015.08.001 A KEGG pathway analysis, with 2077 annotated Prague-specific transcripts and 1167 transcripts specific for Ostrava, identified a higher number of significantly (p-value < 0.05) over-represented pathways in Prague, compared with Ostrava (Table 3). For Prague subjects, significantly over-represented pathways included e.g. those related to neurodegenerative disorders (Alzheimer’s, Parkinson’s and Huntington’s disease), oxidative phosphorylation, asthma, cancer and B and T cell receptor signaling pathway. In the Ostrava region, only Natural killer cell mediated cytotoxicity pathway was significantly over-represented.

Table 3 Differentially expressed KEGG pathways specific for individual locations. Prague-specific transcripts

Ostrava-specific transcripts

KEGG pathway

N; %

Alzheimer’s disease Oxidative phosphorylation Parkinson’s disease Huntington’s disease Asthma Viral myocarditis Intestinal immune network for IgA production Epithelial cell signaling in Helicobacter pylori infection NOD-like receptor signaling pathway Endometrial cancer T cell receptor signaling pathway B cell receptor signaling pathway Acute myeloid leukemia Non-small cell lung cancer Prostate cancer Neurotrophin signaling pathway Chronic myeloid leukemia

36; 2.03 29; 1.64 26; 1.47 33; 1.86 9; 0.51 15; 0.85 11; 0.62 15; 0.85 13; 0.73 14; 0.79 22; 1.24 17; 0.96 14; 0.79 13; 0.73 18; 1.02 22; 1.24 15; 0.85

p-Value <0.001 <0.001 0.004 0.005 0.012 0.024 0.042 0.017 0.039 0.004 0.008 0.008 0.010 0.014 0.018 0.033 0.036

KEGG pathway

N; %

p-Value

Natural killer cell mediated cytotoxicity

14; 1.37

0.027

Differentially expressed KEGG pathways were analyzed using the DAVID v6.7 tool. Cutoff value of statistical significance was p < 0.05. N; % = a number of transcripts associated with the differentially expressed pathway and a percentage of the differential expressed transcripts in the given pathway.

66

P. Rossner Jr. et al. / Mutation Research 780 (2015) 60–70

Table 4 Biological processes associated with exposure to B[a]P and PM2.5 differentially regulated in individual locations. B[a]P: Prague-specific transcripts

B[a]P: Ostrava-specific transcripts

Nucleolus RNA processing RNA binding Transcription Transcription factor activity Negative regulation of transcription Positive regulation of transcription Metal ion binding Phosphatase activity

Nucleolus Regulation of transcription Positive regulation of transcription Translation Protein transport Microtubule cytoskeleton Mitosis Nucleosome assembly ATP binding Metal ion binding Protein tyrosine phosphatase activity

PM2.5: Prague-specific transcripts

PM2.5: Ostrava-specific transcripts

Nucleolus Ribosome Spliceosome Regulation of transcription Positive regulation of gene expression Helicase activity ATP biosynthetic process Mitochondrial inner membrane Respiratory chain Oxidative phosphorylation

Ribosome RNA splicing Transcription factor binding Regulation of transcription Protein folding Extracellular region Mitosis Protein serine/threonine kinase activity Microtubule cytoskeleton GTP binding

Differentially regulated biological processes were analyzed using the DAVID v6.7 tool. Cutoff value of statistical significance was p < 0.05.

3.5. The effects of exposure to air pollutants on gene expression profiles, biological processes and KEGG pathways In order to analyze specific effects of B[a]P, PM2.5 and benzene exposure on gene expression profiles we conducted region specific multivariate-adjusted linear regression analyses. For B[a]P, we detected 464 Prague-specific and 1001 Ostrava-specific transcripts. We also found 293 transcripts common for subjects from both locations. Exposure to PM2.5 was associated with a greater number of differentially regulated transcripts: 683 Prague-specific and 1205 Ostrava-specific; 240 transcripts were common for subjects from both locations. We further detected 1 Prague-specific and 3 Ostrava-specific transcripts associated with exposure to benzene. Due to the small number of differentially-expressed transcripts, processes and pathways affected by benzene exposure were not analyzed. The complete list of region specific differentially regulated transcripts is provided in Supplemental Table VI. Functional classification analysis (Table 4) revealed that exposure to B[a]P affected processes associated with regulation of transcription in subjects from both regions. In the Ostrava subjects the processes involved in translation and cell division were also detected. In general, exposure to PM2.5 affected similar processes as exposure to B[a]P; in the Prague subjects we further found processes involved in oxidative phosphorylation, respiratory chain and mitochondrial inner membrane. Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.mrfmmm.2015.08.001 The KEGG pathway analysis of Prague- and Ostrava-specific transcripts showed that exposure to B[a]P resulted in overrepresentation of immune response-associated pathways in subjects in both regions. In the Prague subjects we also observed over-representation of Neurotrophin signaling pathway, in the Ostrava region Purine metabolism was over-represented. Similarly, exposure to PM2.5 affected immune system-related pathways in subjects from both regions. In the Prague subjects it was also associated with oxidative phosphorylation and neurodegenerative diseases-related pathways; in the subjects from the Ostrava region Chronic myeloid leukemia pathway was affected (Table 5).

A summary of the findings from the analysis of KEGG pathways is presented in Fig. 3. 3.6. Verification of microarray data by qRT-PCR and comparison of expression levels of selected genes Nine genes were selected for qRT-PCR verification. They were chosen because they showed differential expression in the microarray analysis and could be relevant to the traditional biomarkers analyzed in our previous studies [29,30]. These genes were apurinic/apyrimidinic endonuclease 1 (APEX1), ataxia-telangiectasia mutated (ATM), DNA methyltransferase 1 associated protein 1 (DMAP1), Fas cell surface death receptor (FAS), growth arrest and DNA-damage-inducible alpha (GADD45A), glutathione S-transferase mu 1 (GSTM1), interleukin 1 beta (IL1B), prostaglandin-endoperoxide synthase 2 (PTGS2) and RAD21 homolog (S. pombe) (RAD21). The mean correlation across all the transcripts between the qRT-PCR and microarray data was significant (Pearson R = 0.99, p < 0.001). This correlation was based on a comparison of fold change values obtained for individual transcripts in both locations using the respective method. The individual correlation coefficients and p-values ranged from R = 0.02, p = 0.667 to R = 0.83, p < 0.001 (Supplemental Table VII and Supplemental Fig. 4). Among nine verified transcripts we detected five with statistically significant associations (DMAP1, GADD45A, GSTM1, IL1B and PTGS2). Supplementry material related to this article found, in the online version, at http://dx.doi.org/10.1016/j.mrfmmm.2015.08.001 Due to large differences in concentrations of air pollutants in summer 2009 and winter 2010, we focused on the comparison of gene expression in these two seasons separately in Prague and the Ostrava region. The results presented in Table 6 indicated that there were significant differences in expression levels between these seasons in twice as many genes in Ostrava than in Prague (six vs. three genes). In addition, with the exception of APEX1, the differentially expressed genes did not overlap between the locations. In the Ostrava region, higher exposure to air pollutants in winter 2010 was accompanied by down-regulation of all significantly differentially expressed genes (APEX1, ATM, FAS, GSTM1, IL1B and RAD21). In Prague, a similar trend was found for APEX1 only. An increase was observed in the expression of two other significantly differentially expressed genes (GADD45A and PTGS2) in the subjects from Prague in the winter season. 4. Discussion Here, we report the results of whole genome expression profiling of two populations exposed to different levels of air pollution, over three seasons. Winter 2010 in the Ostrava region was distinct due to high levels of pollution. The fact that concentrations of B[a]P recorded in the Ostrava region are probably one of the highest in the entire European Union makes our work unique. The study was prompted by our recent data, which implied that there was no effect of air pollutants on traditional biomarkers such as B[a]P-like DNA adducts, cytogenetic parameters and oxidative stress markers, particularly in a heavily polluted region of Ostrava [29,30]. Although several studies of the effects of air pollutants on the transcriptome have been published [reviewed in [17], they have mostly dealt with occupational exposure or tobacco smoking. Studies of the effects of environmental pollutants on gene expression profiles are scarce [35–37]. In two such studies, higher exposure to air pollutants, which included c-PAHs, was associated with an increased number of differentially expressed genes [35,36]. Although none of these studies gave detailed information from personal monitoring on exposure to environmental pollutants, the

P. Rossner Jr. et al. / Mutation Research 780 (2015) 60–70

67

Table 5 Differentially expressed KEGG pathways associated with exposure to B[a]P and PM2.5 in individual locations. B[a]P: Prague-specific transcripts

B[a]P: Ostrava-specific transcripts

KEGG pathway

N; %

p-Value

KEGG pathway

N; %

p-Value

T cell receptor signaling pathway Natural killer cell mediated cytotoxicity Neurotrophin signaling pathway

14; 2.07 12; 1.78 10; 1.48

<0.001 0.012 0.046

T cell receptor signaling pathway B cell receptor signaling pathway Purine metabolism

15; 1.30 12; 1.04 16; 1.39

0.004 0.004 0.033

PM2.5: Prague-specific transcripts

PM2.5: Ostrava-specific transcripts

KEGG pathway

N; %

p-Value

KEGG pathway

N; %

p-Value

Oxidative phosphorylation Parkinson’s disease Alzheimer’s disease Huntington’s disease Pyrimidine metabolism Purine metabolism T cell receptor signaling pathway Natural killer cell mediated cytotoxicity

14; 1.62 13; 1.51 14; 1.62 15; 1.74 11; 1.28 14; 1.62 16; 1.85 14; 1.62

0.008 0.018 0.046 0.047 0.014 0.029 <0.001 0.010

Chronic myeloid leukemia B cell receptor signaling pathway

11; 0.85 11; 0.85

0.025 0.025

Differentially expressed KEGG pathways were analyzed using the DAVID v6.7 tool. Cutoff value of statistical significance was p < 0.05. N; % = a number of transcripts associated with the differentially expressed pathway and a percentage of the differential expressed transcripts in the given pathway.

Fig. 3. An overview of KEGG pathways associated with environmental pollution in subjects from Prague and the Ostrava region. Region-specific: KEGG pathways overrepresented in individual locations not considering concentrations of air pollutants; B[a]P, PM2.5: over-represented KEGG pathways associated with exposure to air pollutants in individual locations. For a complete list of KEGG pathways see Tables 3 and 5.

Table 6 Differences in expression of selected genes between summer 2009 and winter 2010 sampling seasons, assessed by qRT-PCR. Fold changes represent increase/decrease of gene expression levels in winter 2010 relative to summer 2009. Gene

Prague

Ostrava

Name

Abbreviation

Fold change

p-Value

q-Value

Fold change

p-Value

q-Value

APEX nuclease (multifunctional DNA repair enzyme) 1 Ataxia telangiectasia mutated DNA methyltransferase 1 associated protein 1 Fas (TNF receptor superfamily, member 6) Growth arrest and DNA-damage-inducible, alpha Glutathione S-transferase M1 Interleukin 1, beta Prostaglandin-endoperoxide synthase 2 RAD21 homolog (S. pombe)

APEX1 ATM DMAP FAS GADD45A GSTM1 IL1B PTGS2 RAD21

0.87 1.22 1.11 1.01 1.21 0.91 1.08 1.38 1.38

0.035 0.089 0.187 0.432 0.005 0.764 0.696 0.008 0.220

0.048 0.091 0.151 0.253 0.016 0.349 0.349 0.016 0.151

0.76 0.80 1.01 0.80 1.00 0.80 0.67 0.90 0.71

<0.001 0.005 0.393 0.019 0.794 <0.001 0.005 0.848 0.005

<0.001 0.007 0.386 0.022 0.648 <0.001 0.007 0.648 0.007

68

P. Rossner Jr. et al. / Mutation Research 780 (2015) 60–70

subjects were likely exposed to lower concentrations of pollutants than the subjects from the Ostrava region. In addition, when compared with our analysis, these studies were small and included a maximum of 71 subjects from both genders [38]. In this study we found a large difference in the number of differentially expressed genes between locations. There was approximately 44% fewer transcripts that showed specific changes in the Ostrava region than observed in the subjects from Prague (1167 and 2077 unique differentially expressed genes for Ostrava and Prague, respectively). To elucidate biological meaning of these differences we performed analyses of biological processes and pathways. Gene functional classification identified a significant number of common processes in both regions, most of them associated with transcription, translation and cell cycle. However, it should be noted that although many processes overlapped, the transcripts involved in them differed suggesting that the same cellular function is achieved by different sets of genes. On the other hand, there was only one differentially expressed KEGG pathway in the Ostrava subjects, and 17 KEGG pathways in the subjects from Prague. Interestingly, in the subjects from Prague we found differentially expressed processes and pathways involved in mitochondrial functions and neurodegenerative disorders; none of them was detected in the subjects from the Ostrava region. There is a causal link between mitochondrial activity, reactive oxygen species formation (ROS) and subsequent oxidative stress [39]. Also, it is believed that oxidative stress is involved in etiology of neurodegenerative disorders (Alzheimer’s and Parkinson’s disease) [40], in which inhalation of PM may play a significant role [41]. Those diseases share similar pathogenesis related to the disruption of endoplasmic reticulum (ER) functioning and accumulation of misfolded proteins [42]. Endoplasmic reticulum was a process significantly differentially regulated in the Prague subjects. Thus, environmental factors may affect processes in the ER compartment and trigger ER stress, involving changes in intraluminal Ca2+ , redox status or energy deprivation. It consequently leads to chronic unfolded protein response activation and induction of apoptotic response. ER stress has been associated with increased ROS production, leakage of Ca2+ from the ER lumen and stimulation of mitochondrial ROS production during oxidative phosphorylation [42]. Oxidative phosphorylation pathway was significantly over-represented in the Prague subjects and may indicate the link between ER stress-associated ROS generation and neurodegenerative diseases. We should further mention differential regulation of pathways associated with immune response and inflammation, most of them detected in the Prague subjects. These observations are in agreement with studies reporting the modifying effect of particulate matter pollution on immune response of organism [43,44]. Finally, five affected pathways detected in the Prague subjects are involved in various cancers (chronic and acute myeloid leukemia, prostate, lung and endometrial cancer). It has been repeatedly shown that exposure to air pollution affects cancer mortality [reviewed e.g. in [45]]. Recently, outdoor air pollution has been classified by International Agency for Research on Cancer as carcinogenic to humans [46]. The results of linear regression analyses suggest that exposure to PM2.5, but not to B[a]P, is responsible for differential regulation of processes and pathways associated with neurodegenerative disorders and mitochondrial functions in the Prague subjects. This observation is supported by the ability of PM2.5 to cause inflammation [3], resulting in the induction of oxidative stress which may contribute to the development of neurodegenerative disorders. No cancer-related pathway was affected by the exposure to either pollutant in this group. In summary, our results indicate that, in general, subjects in the Ostrava region respond to adverse effects of the environment by changes of expression of a lower number of genes, processes and

pathways. Our data may be an indication that subjects in Prague are at a greater risk of development of neurodegenerative disorders and cancer, despite the fact that the Ostrava region subjects are exposed to substantially higher concentrations of air pollutants. Nine differentially expressed genes were selected for verification of the microarray data. The genes included those that are relevant for cell cycle regulation and DNA repair (APEX1, ATM, DMAP1, GADD45A and RAD21), inflammation (IL1B, PTGS2), apoptosis (FAS) and metabolism of xenobiotics (GSTM1). For further comparison, we used qRT-PCR data from samples collected in summer 2009 and winter 2010 seasons. The analysis showed a distinct difference between the regions. Despite a significant increase in concentration of air pollutants, the levels of expression of six genes (APEX1, ATM, RAD21, IL1B, FAS and GSTM1), measured in the Ostrava subjects in winter 2010, were significantly lower than in the summer season. In the substantially less polluted city of Prague, expression of two genes (GADD45A and PTGS2) was upregulated. Down-regulation was observed for APEX1 only. With the exception of APEX1, the differentially expressed genes differed between the Ostrava region and Prague. Down-regulation of gene expression in winter 2010 in the Ostrava region, was an unexpected finding. The APEX1 gene encodes an essential protein that participates in base excision repair [47], which removes, among others, oxidatively damaged bases from DNA. Protein products of the ATM and RAD21 genes are active in DNA double-strand break repair [48,49] and ATM protein kinase is one of the key activators of DNA damage response [reviewed in [48]]. The FAS receptor is a pro-apoptotic factor that is induced by genotoxic compounds that cause DNA adducts and double-strand DNA breaks [50]. Interleukin 1 beta plays an important role in inflammation and response of the organism to pathogens [51]. The GSTM1 gene encodes a key enzyme of the phase II biotransformation of xenobiotics, such as PAHs [52]. It can be argued that lower expression of these genes compromises the effective response of the organism to the insult of xenobiotics. Their up-regulation would seem to be a logical response. But this was observed only in Prague subjects for GADD45A, a gene reported to be induced by environmental stresses that are responsible for cell cycle arrest [53], and PTGS2, which is induced by inflammatory stimuli. In the comparison between regions, the levels of expression of the afore-mentioned genes tended to be higher in the more polluted Ostrava region than in Prague, in all seasons but winter 2010 (data not shown). This indicated that, in the polluted region, although the numbers of differentially expressed genes changed less than in the control location, there may be higher levels of expression of the genes that are differentially expressed. This has the potential to provide more efficient protection against environmental pollutants for subjects living in the Ostrava region. This hypothesis is in agreement with our previous study [54], in which we observed upregulation of expression of XRCC5 in a subset of subjects living in the Ostrava region. However, to confirm these hypotheses, further analyses are required on the level of regulation and effectiveness of translation.

5. Limitations of the study A possible limitation of our study stems from the fact that repeatedly sampled subjects were treated as individual samples in statistical analyses. We performed linear regression analyses that included the subjects as pairing variables. However, due to the fact that only 94 subjects (282 individual samples; 60.5% of all samples) participated in all sampling seasons, this approach resulted in greatly reduced statistical power and consequently in a low number of significantly deregulated transcripts (data not shown). Therefore we opted for the statistical approach described in Section 2.5.

P. Rossner Jr. et al. / Mutation Research 780 (2015) 60–70

Our data showed seasonal effects on gene expression profiles. This was particularly apparent for the third sampling period (winter 2010), which formed a distinct cluster in PCA. Interestingly, despite significant differences in concentrations of pollutants in this season between the locations, no clear clustering by region was observed in the PCA. This indicated that, although levels of B[a]P, benzene and PM2.5 affect gene expression profiles, there is likely to be other, unidentified interfering variables. There may be several sources of variability potentially affecting our results. First, our study was conducted on mRNA extracted from leukocytes that contain neutrophils and lymphocytes as the most abundant fractions. It has been reported that variation in the proportions of these blood cell subsets affects gene expression patterns [55]. As we did not investigate blood composition we cannot rule out differences between populations and/or individuals that would affect our results. Second, seasonal changes in gene expression in healthy individuals has been reported recently [56]. The authors observed circannual patterns in three co-expression modules when studying whole blood transcriptomics data of 233 subjects. However, as these changes stem primarily from red blood cells and thrombocytes that were not present in our samples, it is not likely that they would impact our data. Third, we concentrated on effects of long-term, chronic exposure to environmental pollutants. The neutrophils as well as mRNAs are short-lived and one may argue that this makes both materials not suitable for monitoring of long-term effects. However, we believe that chronic exposure that occurs over a long period of time affects all generations of neutrophils in a similar way resulting in expression of a comparable set of genes and in synthesis of a comparable set of mRNAs. Forth, as it is not possible to directly include a confounding factor (i.e. a microarray batch) in the PCA, we cannot completely exclude the role of the batch effect. As mentioned before, we used the Z-normalization instead to confirm that the sampling season was the most probable driver of separation in the PCA. This is further supported by the fact that the data on environmental pollution exposure correlated with individual sampling seasons (data not shown). Also, to avoid any experimental variables that would contribute to the possible batch effect, we strictly applied standardized protocols in all experimental steps.

6. Conclusions In the present study we observed an effect of environmental air pollution by PM2.5, B[a]P and benzene on gene expression profiles. However, we did not find a clear relationship between the concentrations of air pollutants and the number of differentially expressed genes. In the polluted Ostrava region, not only is the number of differentially expressed genes, processes and pathways lower than in the control region, but the levels of expression of some genes dropped during the smog episode, in winter 2010. Pathway analysis clearly showed negative effects of environmental pollution in Prague, particularly those pathways associated with neurodegenerative diseases. To the best of our knowledge, there is no study that would conclusively show the existence of adaptive response in humans, although there is some indication that radiation may have this effect [57]. However, our data suggest that chronic exposure to high levels of air pollutants may result in lower sensitivity of the organism to the deleterious effects of the environment that is manifested by limited changes in expression of some genes and a lack of response by traditional biomarkers. Whilst our study aimed to describe global gene expression changes in a population exposed to extremely high levels of air pollution, analysis of functional proteins and/or regulatory elements (microRNAs, DNA methylation) may shed additional light on underlying mechanisms of this observation.

69

Conflict of Interest statement The authors declare that there are no conflicts of interest.

Acknowledgements We thank Prof. William W. Au for his critical comments. This work was supported by the Czech Ministry of the Environment [SP/1b3/8/08], the Czech Ministry of Education [2B08005 and LO1508] and the Grant Agency of the Czech Republic [P301/13/13458S].

References [1] C.A. Pope 3rd, D.W. Dockery, Health effects of fine particulate air pollution: lines that connect, J. Air Waste Manage. Assoc. 56 (2006) 709–742. [2] B. Brunekreef, S.T. Holgate, Air pollution and health, Lancet 360 (2002) 1233–1242. [3] A. Valavanidis, K. Fiotakis, T. Vlachogianni, Airborne particulate matter and human health: toxicological assessment and importance of size and composition of particles for oxidative damage and carcinogenic mechanisms, J. Environ. Sci. Health C Environ. Carcinog. Ecotoxicol. Rev. 26 (2008) 339–362. [4] G. Oberdorster, E. Oberdorster, J. Oberdorster, Nanotoxicology: an emerging discipline evolving from studies of ultrafine particles, Environ. Health Perspect. 113 (2005) 823–839. [5] IARC, IARC monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, in: Monograph on Some Non-Heterocyclic Polycyclic Aromatic Hydrocarbons and Some Related Exposures, IARC, Publications, Lyon, France, 2010. [6] IARC, IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans. Chemical Agents and Related Occupations, IARC, Publications, Lyon, France, 2012. [7] P. Gerde, B.A. Muggenburg, M. Lundborg, A.R. Dahl, The rapid alveolar absorption of diesel soot-adsorbed benzo[a]pyrene: bioavailability, metabolism and dosimetry of an inhaled particle-borne carcinogen, Carcinogenesis 22 (2001) 741–749. [8] C.A. Demetriou, O. Raaschou-Nielsen, S. Loft, P. Moller, R. Vermeulen, D. Palli, M. Chadeau-Hyam, W.W. Xun, P. Vineis, Biomarkers of ambient air pollution and lung cancer: a systematic review, Occup. Environ. Med. 69 (2012) 619–627. [9] P. Moller, S. Loft, Oxidative damage to DNA and lipids as biomarkers of exposure to air pollution, Environ. Health Perspect. 118 (2010) 1126–1136. [10] G. Castano-Vinyals, A. D’Errico, N. Malats, M. Kogevinas, Biomarkers of exposure to polycyclic aromatic hydrocarbons from environmental air pollution, Occup. Environ. Med. 61 (2004) e12. [11] R. Mateuca, N. Lombaert, P.V. Aka, I. Decordier, M. Kirsch-Volders, Chromosomal changes: induction, detection methods and applicability in human biomonitoring, Biochimie 88 (2006) 1515–1531. [12] A. Rossnerova, M. Spatova, C. Schunck, R.J. Sram, Automated scoring of lymphocyte micronuclei by the MetaSystems Metafer image cytometry system and its application in studies of human mutagen sensitivity and biodosimetry of genotoxin exposure, Mutagenesis 26 (2011) 169–175. [13] M.E. Gerlofs-Nijland, D. van Berlo, F.R. Cassee, R.P. Schins, K. Wang, A. Campbell, Effect of prolonged exposure to diesel engine exhaust on proinflammatory markers in different regions of the rat brain, Part Fibre Toxicol. 7 (2010) 12. [14] N. Bai, T. Kido, T.J. Kavanagh, J.D. Kaufman, M.E. Rosenfeld, C. van Breemen, S.F. van Eeden, Exposure to diesel exhaust up-regulates iNOS expression in ApoE knockout mice, Toxicol. Appl. Pharmacol. 255 (2011) 184–192. [15] X. Xu, C. Liu, Z. Xu, K. Tzan, M. Zhong, A. Wang, M. Lippmann, L.C. Chen, S. Rajagopalan, Q. Sun, Long-term exposure to ambient fine particulate pollution induces insulin resistance and mitochondrial alteration in adipose tissue, Toxicol. Sci. 124 (2011) 88–98. [16] I. Bos, P. De Boever, J. Emmerechts, J. Buekers, J. Vanoirbeek, R. Meeusen, M. Van Poppel, B. Nemery, T. Nawrot, L.I. Panis, Changed gene expression in brains of mice exposed to traffic in a highway tunnel, Inhal. Toxicol. 24 (2012) 676–686. [17] C.P. Wild, A. Scalbert, Z. Herceg, Measuring the exposome: a powerful basis for evaluating environmental exposures and cancer risk, Environ. Mol. Mutagen. 54 (2013) 480–499. [18] B. Sen, B. Mahadevan, D.M. DeMarini, Transcriptional responses to complex mixtures: a review, Mutat. Res. 636 (2007) 144–177. [19] E.M. Thomson, D. Vladisavljevic, S. Mohottalage, P. Kumarathasan, R. Vincent, Mapping acute systemic effects of inhaled particulate matter and ozone: multiorgan gene expression and glucocorticoid activity, Toxicol. Sci. 135 (2013) 169–181. [20] S. Yokota, H. Hori, M. Umezawa, N. Kubota, R. Niki, S. Yanagita, K. Takeda, Gene expression changes in the olfactory bulb of mice induced by exposure to diesel exhaust are dependent on animal rearing environment, PLoS One 8 (2013) e70145.

70

P. Rossner Jr. et al. / Mutation Research 780 (2015) 60–70

[21] J.G. Maresh, M.J. Campen, M.D. Reed, A.L. Darrow, R.V. Shohet, Hypercholesterolemia potentiates aortic endothelial response to inhaled diesel exhaust, Inhal. Toxicol. 23 (1) (2011) 1–10. [22] R.R. Gottipolu, J.G. Wallenborn, E.D. Karoly, M.C. Schladweiler, A.D. Ledbetter, T. Krantz, W.P. Linak, A. Nyska, J.A. Johnson, R. Thomas, J.E. Richards, R.H. Jaskot, U.P. Kodavanti, One-month diesel exhaust inhalation produces hypertensive gene expression pattern in healthy rats, Environ. Health Perspect. 117 (2009) 38–46. [23] A.P. Pettit, A. Brooks, R. Laumbach, N. Fiedler, Q. Wang, P.O. Strickland, K. Madura, J. Zhang, H.M. Kipen, Alteration of peripheral blood monocyte gene expression in humans following diesel exhaust inhalation, Inhal. Toxicol. 24 (2012) 172–181. [24] Y.C. Huang, M. Schmitt, Z. Yang, L.G. Que, J.C. Stewart, M.W. Frampton, R.B. Devlin, Gene expression profile in circulating mononuclear cells after exposure to ultrafine carbon particles, Inhal. Toxicol. 22 (2010) 835–846. [25] Y. Strulovici-Barel, L. Omberg, M. O’Mahony, C. Gordon, C. Hollmann, A.E. Tilley, J. Salit, J. Mezey, B.G. Harvey, R.G. Crystal, Threshold of biologic responses of the small airway epithelium to low levels of tobacco smoke, Am. J. Respir. Crit. Care Med. 182 (2010) 1524–1532. [26] A. Peretz, E.C. Peck, T.K. Bammler, R.P. Beyer, J.H. Sullivan, C.A. Trenga, S. Srinouanprachnah, F.M. Farin, J.D. Kaufman, Diesel exhaust inhalation and assessment of peripheral blood mononuclear cell gene transcription effects: an exploratory study of healthy human volunteers, Inhal. Toxicol. 19 (2007) 1107–1119. [27] M. Lodovici, C. Luceri, C. De Filippo, C. Romualdi, F. Bambi, P. Dolara, Smokers and passive smokers gene expression profiles: correlation with the DNA oxidation damage, Free Radical Biol. Med. 43 (2007) 415–422. [28] J.W. Holloway, S. Savarimuthu Francis, K.M. Fong, I.A. Yang, Genomics and the respiratory effects of air pollution exposure, Respirology 17 (2012) 590–600. [29] P. Rossner Jr, V. Svecova, J. Schmuczerova, A. Milcova, N. Tabashidze, J. Topinka, A. Pastorkova, R.J. Sram, Analysis of biomarkers in a Czech population exposed to heavy air pollution. Part I: bulky DNA adducts, Mutagenesis 28 (2013) 89–95. [30] P. Rossner Jr, A. Rossnerova, M. Spatova, O. Beskid, K. Uhlirova, H. Libalova, I. Solansky, J. Topinka, R.J. Sram, Analysis of biomarkers in a Czech population exposed to heavy air pollution. Part II: chromosomal aberrations and oxidative stress, Mutagenesis 28 (2013) 97–106. [31] R.J. Sram, M. Dostal, H. Libalova, P. Rossner Jr, A. Rossnerova, V. Svecova, J. Topinka, A. Bartonova, The European Hot Spot of B[a]P and PM2. 5 Exposure—the Ostrava Region, Czech Republic: health research results, ISRN Public Health 2013 (2013) 1. [32] B. Binkova, J. Topinka, G. Mrackova, D. Gajdosova, P. Vidova, Z. Stavkova, V. Peterka, T. Pilcik, V. Rimar, L. Dobias, P.B. Farmer, R.J. Sram, Coke oven workers study: the effect of exposure and GSTM1 and NAT2 genotypes on DNA adduct levels in white blood cells and lymphocytes as determined by 32P-postlabelling, Mutat. Res. 416 (1998) 67–84. [33] Y. Benjamini, Y. Hochberg, Controlling the false discovery rate: a practival and powerful approach to multiple testing, J. R. Stat. Soc. Ser. B Methodol. 57 (1995) 289–300. [34] W. Huang da, B.T. Sherman, R.A. Lempicki, Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists, Nucleic Acids Res. 37 (2009) 1–13. [35] D.M. van Leeuwen, M.H. van Herwijnen, M. Pedersen, L.E. Knudsen, M. Kirsch-Volders, R.J. Sram, Y.C. Staal, E. Bajak, J.H. van Delft, J.C. Kleinjans, Genome-wide differential gene expression in children exposed to air pollution in the Czech Republic, Mutat. Res. 600 (2006) 12–22. [36] D.M. van Leeuwen, M. Pedersen, P.J. Hendriksen, A. Boorsma, M.H. van Herwijnen, R.W. Gottschalk, M. Kirsch-Volders, L.E. Knudsen, R.J. Sram, E. Bajak, J.H. van Delft, J.C. Kleinjans, Genomic analysis suggests higher susceptibility of children to air pollution, Carcinogenesis 29 (2008) 977–983. [37] S. De Coster, D.M. van Leeuwen, D.G. Jennen, G. Koppen, E. Den Hond, V. Nelen, G. Schoeters, W. Baeyens, J.H. van Delft, J.C. Kleinjans, N. van Larebeke, Gender-specific transcriptomic response to environmental exposure in Flemish adults, Environ. Mol. Mutagen. 54 (2013) 574–588.

[38] D.M. van Leeuwen, R.W. Gottschalk, G. Schoeters, N.A. van Larebeke, V. Nelen, W.F. Baeyens, J.C. Kleinjans, J.H. van Delft, Transcriptome analysis in peripheral blood of humans exposed to environmental carcinogens: a promising new biomarker in environmental health studies, Environ. Health Perspect. 116 (2008) 1519–1525. [39] A. Barzilai, K. Yamamoto, DNA damage responses to oxidative stress, DNA Rep. 3 (2004) 1109–1115 (Amst). [40] L. Migliore, F. Coppede, Environmental-induced oxidative stress in neurodegenerative disorders and aging, Mutat. Res. 674 (2009) 73–84. [41] L. Calderon-Garciduenas, A.C. Solt, C. Henriquez-Roldan, R. Torres-Jardon, B. Nuse, L. Herritt, R. Villarreal-Calderon, N. Osnaya, I. Stone, R. Garcia, D.M. Brooks, A. Gonzalez-Maciel, R. Reynoso-Robles, R. Delgado-Chavez, W. Reed, Long-term air pollution exposure is associated with neuroinflammation, an altered innate immune response, disruption of the blood-brain barrier, ultrafine particulate deposition, and accumulation of amyloid beta-42 and alpha-synuclein in children and young adults, Toxicol. Pathol. 36 (2008) 289–310. [42] J.D. Malhotra, R.J. Kaufman, Endoplasmic reticulum stress and oxidative stress: a vicious cycle or a double-edged sword? Antioxid. Redox. Signal. 9 (2007) 2277–2293. [43] J.A. Deiuliis, T. Kampfrath, J. Zhong, S. Oghumu, A. Maiseyeu, L.C. Chen, Q. Sun, A.R. Satoskar, S. Rajagopalan, Pulmonary T cell activation in response to chronic particulate air pollution, Am. J. Physiol. Lung Cell Mol. Physiol. 302 (2012) L399–L409. [44] L. Muller, C.V. Chehrazi, M.W. Henderson, T.L. Noah, I. Jaspers, Diesel exhaust particles modify natural killer cell function and cytokine release, Part Fibre Toxicol. 10 (2013) 16, 16. [45] C. Ren, S. Tong, Health effects of ambient air pollution–recent research development and contemporary methodological challenges, Environ. Health 7 (2008) 56. [46] D. Loomis, Y. Grosse, B. Lauby-Secretan, F. El Ghissassi, V. Bouvard, L. Benbrahim-Tallaa, N. Guha, R. Baan, H. Mattock, K. Straif, The carcinogenicity of outdoor air pollution, Lancet Oncol. 14 (2013) 1262–1263. [47] R. Abbotts, S. Madhusudan, Human AP endonuclease 1 (APE1): from mechanistic insights to druggable target in cancer, Cancer Treat. Rev. 36 (2010) 425–435. [48] Y. Shiloh, Y. Ziv, The ATM protein kinase: regulating the cellular response to genotoxic stress, and more, Nat. Rev. Mol. Cell Biol. 14 (2013) 197–210. [49] K.M. Feeney, C.W. Wasson, J.L. Parish, Cohesin: a regulator of genome integrity and gene expression, Biochem. J. 428 (2010) 147–161. [50] W.P. Roos, B. Kaina, DNA damage-induced cell death by apoptosis, Trends Mol. Med. 12 (2006) 440–450. [51] A. Kasza, IL-1 and EGF regulate expression of genes important in inflammation and cancer, Cytokine 62 (2013) 22–33. [52] H.M. Bolt, R. Thier, Relevance of the deletion polymorphisms of the glutathione S-transferases GSTT1 and GSTM1 in pharmacology and toxicology, Curr. Drug Metab. 7 (2006) 613–628. [53] A. Rosemary Siafakas, D.R. Richardson, Growth arrest and DNA damage-45 alpha (GADD45alpha), Int. J. Biochem. Cell Biol. 41 (2009) 986–989. [54] P. Rossner Jr, K. Uhlirova, O. Beskid, A. Rossnerova, V. Svecova, R.J. Sram, Expression of XRCC5 in peripheral blood lymphocytes is upregulated in subjects from a heavily polluted region in the Czech Republic, Mutat. Res. 713 (2011) 76–82. [55] A.R. Whitney, M. Diehn, S.J. Popper, A.A. Alizadeh, J.C. Boldrick, D.A. Relman, P.O. Brown, Individuality and variation in gene expression patterns in human blood, Proc. Natl. Acad. Sci. U. S. A. 100 (2003) 1896–1901. [56] S. De Jong, M. Neeleman, J.J. Luykx, M.J. ten Berg, E. Strengman, H.H. Den Breeijen, L.C. Stijvers, J.E. Buizer-Voskamp, S.C. Bakker, R.S. Kahn, S. Horvath, W.W. Van Solinge, R.A. Ophoff, Seasonal changes in gene expression represent cell-type composition in whole blood, Hum. Mol. Genet. 23 (2014) 2721–2728. [57] Z. Jaworowski, Radiation hormesis–a remedy for fear, Hum. Exp. Toxicol. 29 (2010) 263–270.