Exposure to Metal-Rich Particulate Matter Modifies the Expression of Candidate MicroRNAs in Peripheral Blood Leukocytes

Background Altered patterns of gene expression mediate the effects of particulate matter (PM) on human health, but mechanisms through which PM modifies gene expression are largely undetermined. MicroRNAs (miRNAs) are highly conserved, noncoding small RNAs that regulate the expression of broad gene networks at the posttranscriptional level. Objectives We evaluated the effects of exposure to PM and PM metal components on candidate miRNAs (miR-222, miR-21, and miR-146a) related with oxidative stress and inflammatory processes in 63 workers at an electric-furnace steel plant. Methods We measured miR-222, miR-21, and miR-146a expression in blood leukocyte RNA on the first day of a workweek (baseline) and after 3 days of work (postexposure). Relative expression of miRNAs was measured by real-time polymerase chain reaction. We measured blood oxidative stress (8-hydroxyguanine) and estimated individual exposures to PM1 (< 1 μm in aerodynamic diameter), PM10 (< 10 μm in aerodynamic diameter), coarse PM (PM10 minus PM1), and PM metal components (chromium, lead, cadmium, arsenic, nickel, manganese) between the baseline and postexposure measurements. Results Expression of miR-222 and miR-21 (using the 2−ΔΔCT method) was significantly increased in postexposure samples (miR-222: baseline = 0.68 ± 3.41, postexposure = 2.16 ± 2.25, p = 0.002; miR-21: baseline = 4.10 ± 3.04, postexposure = 4.66 ± 2.63, p = 0.05). In postexposure samples, miR-222 expression was positively correlated with lead exposure (β = 0.41, p = 0.02), whereas miR-21 expression was associated with blood 8-hydroxyguanine (β = 0.11, p = 0.03) but not with individual PM size fractions or metal components. Postexposure expression of miR-146a was not significantly different from baseline (baseline = 0.61 ± 2.42, postexposure = 1.90 ± 3.94, p = 0.19) but was negatively correlated with exposure to lead (β = −0.51, p = 0.011) and cadmium (β = −0.42, p = 0.04). Conclusions Changes in miRNA expression may represent a novel mechanism mediating responses to PM and its metal components.


Research
Exposure to ambient particulate matter (PM) has been associated with increased morbidity and mortality from cardiovascular and respi ratory diseases (Baccarelli et al. 2008;Brook et al. 2004;Ciocco and Thompson 1961). Epidemiologic and in vivo studies suggest that the transition metal components of PM may be responsible for such effects (Brook et al. 2004;Chang et al. 2005;Corey et al. 2006). Foundry work has also been associated with adverse health outcomes, including cardio vascular disease (Kuo et al. 1999), potentially linked with PM exposure. Although prior studies have associated inhalation of ambi ent or occupational PM with systemic activa tion of inflammatory pathways, production of reactive oxygen species (ROS), and enhanced coagulation (Baccarelli et al. , 2008Chahine et al. 2007;Gurgueira et al. 2002;Li et al. 2006), the under lying mechanisms linking PM exposure with adverse health outcomes still need to be clarified (Nel et al. 2006). Inhaled PM pollutants have been shown to produce systemic changes in gene expres sion, which can be detected in peripheral blood of exposed individuals . Expression of human genes is con trolled by several genetic and epigenetic mechanisms, including microRNA (miRNA) regulation (He and Hannon 2004). MiRNAs are small, endogenous, singlestranded non coding RNAs of 20-22 nucleo tides (Baccarelli and Bollati 2009). They post transcriptionally regulate gene expression by hybridization to messenger RNA (mRNA), leading to trans la tional repression or degradation of the target mRNA (He and Hannon 2004). One single miRNA can regulate hundreds of mRNAs in inter related gene pathways, and a single mRNA can be targeted by several different miRNAs (Lewis et al. 2005).
Changes in the expression of several miRNAs, including miR-222, miR-21, and miR-146a, have been implicated in disease mechanisms that may be related to PM exposure, such as oxidative stress (Babar et al. 2008) and regulation of inflammation (Xiao and Rajewsky 2009). In particular, miR-222 over expression indirectly reduces the expres sion of the endothelial nitric oxide syntha sein Dicer small interfering RNA-transfected cells (Suarez et al. 2007), an inflammation related hallmark of athero sclerosis and isch emic cardio myopathy (Zeiher 1996), and has been associated with altered redox signaling (Sen et al. 2009). miR-21 has been shown to respond to hydrogen peroxide stimula tion and participates in coordinated protec tive responses to oxidative stress (Cheng et al. 2009), as well as in inflammatory responses as suggested by animal models of allergic airway (Lu et al. 2009) and lipolysaccharideinduced inflammation (Moschos et al. 2007). Changes in miR-146a expression have been implicated in the negative regulation of inflammation induced via the innate immune response, which is also activated by PM (Shoenfelt et al. 2009). miR-146a expression during inflam mation is under the control of the transcrip tion factor NFκB (nuclear factorkappa B), a central mediator for PMrelated inflammation and health effects ).
In the present study, we investigated the effects of PM exposure on miR222, miR21, and miR146a measured in blood RNA from foundry workers with wellcharacterized expo sure to a wide range of PM levels. To clarify the mechanisms activated by PM exposure, we evaluated whether the expression of miR-222, miR-21, and miR-146a was associated with oxidative stress levels, as reflected in 8hydroxy guanine (8OHdG) measured in blood DNA.
Background: Altered patterns of gene expression mediate the effects of particulate matter (PM) on human health, but mechanisms through which PM modifies gene expression are largely undeter mined. MicroRNAs (miRNAs) are highly conserved, noncoding small RNAs that regulate the expression of broad gene networks at the posttranscriptional level. oBjectives: We evaluated the effects of exposure to PM and PM metal components on candidate miRNAs (miR-222, miR-21, and miR-146a) related with oxidative stress and inflammatory processes in 63 workers at an electricfurnace steel plant. Methods: We measured miR-222, miR-21, and miR-146a expression in blood leukocyte RNA on the first day of a workweek (baseline) and after 3 days of work (postexposure). Relative expression of miRNAs was measured by realtime polymerase chain reaction. We meas ured blood oxidative stress (8hydroxyguanine) and estimated individual exposures to PM 1 (< 1 µm in aero dynamic diameter), PM 10 (< 10 µm in aero dynamic diameter), coarse PM (PM 10 minus PM 1 ), and PM metal com ponents (chromium, lead, cadmium, arsenic, nickel, manganese) between the baseline and post exposure measurements. results: Expression of miR-222 and miR-21 (using the 2 -ΔΔC T method) was significantly increased in postexposure samples (miR222: baseline = 0.68 ± 3.41, postexposure = 2.16 ± 2.25, p = 0.002; miR21: baseline = 4.10 ± 3.04, postexposure = 4.66 ± 2.63, p = 0.05). In post exposure samples, miR-222 expression was positively correlated with lead exposure (β = 0.41, p = 0.02), whereas miR-21 expression was associated with blood 8hydroxyguanine (β = 0.11, p = 0.03) but not with individual PM size fractions or metal components. Postexposure expression of miR-146a was not significantly different from baseline (baseline = 0.61 ± 2.42, postexposure = 1.90 ± 3.94, p = 0.19) but was nega tively correlated with exposure to lead (β = -0.51, p = 0.011) and cadmium (β = -0.42, p = 0.04). conclusions: Changes in miRNA expression may represent a novel mechanism mediating responses to PM and its metal components.

Materials and Methods
Study subjects. We recruited 63 healthy male workers (mean age, 44 years; range, 27-55 years), in a steel production plant in Brescia, Northern Italy. These workers were free of cancer, cardio vascular disease, and pulmo nary disease, and all of them had been working in their current job position for at least 1 year. To discriminate short and longterm effects of PM, we obtained blood samples for DNA methylation analysis at two different times: a) a baseline sample collected in the morning of the first day of a workweek (after 2 days off work) before the beginning of any work activity (time 1); and b) a postexposure sample col lected at the same time on the fourth day of work, after 3 consecutive days of work (time 2). Written informed consent and approval from the local institutional review board IRCCS Fondazione Ca' Granda Ospedale Maggiore Policlinico were obtained before the study.
Exposure assessment. Measures of airborne PM mass and PM metal components obtained in each of the 11 work areas of the steel pro duction facility were used to estimate individ ual exposures. Measures of PM mass included levels of PM with aerodynamic diameters ≤ 10 µm (PM 10 ) and ≤ 1 µm (PM 1 ) measured using a GRIMM 1100 lightscattering dust analyzer (Grimm Technologies, Inc., Douglasville, GA, USA). Concentrations of coarse particles were calculated from these meas ures as the differ ence between PM 10 and PM 1 . We measured PM metal components on the PM 10 fraction of PM mass through multi elemental analysis by means of inductively coupled plasma mass spectrometer (ELAN DRC II; PerkinElmer, Waltham, MA, USA). We meas ured arsenic, cadmium, lead, manganese, and nickel con centrations using the Total Quant method, and chromium concentrations using the Dynamic Reaction Cell (DRC) method with ammonia (De Palma et al. 2008).
During the 3 working days between times 1 and 2, each of the study subjects recorded in a personal log the time that he spent in each work area. Individual exposures were calculated as the timeweighted averages of area concentrations (Tarantini et al. 2009). miRNA analysis. Buffy coat samples were separated within 30 min of blood draw, immediately snapfrozen, and stored at -80°C. RNA was extracted from the buffy coats using the Ribopure Kit (Ambion, Inc., Austin, TX, USA), modified for miRNA extraction [see Supplemental Material (doi:10.1289/ehp.0901300)]. We used spe cific TaqMan MicroRNA Assays (Applied Biosystems, Uppsala, Sweden) to detect and quantify mature miRNAs as recom mended by the manufacturer, using ABI Prism 7900HT sequence detection sys tems (Applied Biosystems). In the reverse transcription (RT) step, 10 ng total RNA was employed in RT reactions (16°C for 30 min, 42°C for 30 min, 85°C for 5 min, and then to 4°C) using reagents from the TaqMan MicroRNA Reverse Transcription kit (Applied Biosystems) and specific miRNA primers provided with the TaqMan MicroRNA Assays. Realtime polymerase chain reaction (PCR) was performed using TaqMan MicroRNA Assays together with TaqMan Universal PCR Master Mix on an Applied Biosystems 7900 Sequence Detection System (95°C for 1 min and 40 cycles of 95°C for 15 sec and 60°C for 30 sec) (Chen 2005). Normalization was performed with RNU6B (RNA, U6 small nuclear 2) endogenous con trol (Applied Biosystems, Foster City, CA, USA). Realtime PCR was performed in trip licate, including notemplate controls. The threshold cycle (Ct) was defined as the frac tional cycle number at which the fluorescence passes the fixed threshold. The relative gene expression was calculated via a 2 −ΔΔCt method (Livak and Schmittgen 2001). Data are pre sented as relative quantity of target miRNA, normalized to RNU6B endogenous control and a calibrator built as a pool of 50 random samples. All laboratory analyses were per formed in the same batch using samples that remained frozen until the analyses.

Determination of 8-OH-dG content in mitochondrial DNA (mtDNA).
We meas ured the content of 8OHdG in mtDNA using quantitative realtime PCR to amplify dif ferent fragments of mtDNA on each sample with and without hOGG1 (human 8oxogua nine DNA glycosylase 1) pre treatment, as previously described by Lin et al. (2008). All experimental DNA samples were assayed in triplicates, and the mean value of ΔCt was used in the statistical analyses.
Target prediction and pathway mining. We used the miRNA target prediction soft ware miRanda (http://www.microrna.org/ microrna/home.do) to predict the target genes of miR-222, miR-21, and miR-146a. A total of 1,056 target genes were annotated for miR-222, 1,065 for miR-21, and 1,038 for miR-146a. Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.genome. jp/kegg) pathway searching was performed by mapping the predicted target genes from Values shown are mean ± SD or n (%).   To find signal transduction pathways related to miR-222, miR-21, and miR-146a, we used LitInspector (http://www.litinspector.org), a liter ature search tool providing gene and signal trans duction pathway mining within the National Center for Biotechnology Information's PubMed database (Frisch et al. 2009). Two LitInspector pathways were annotated for miR-222, seven for miR-21, and six for miR-146a.
Statistical analysis. The Student's paired ttest was used to assess differences in miRNA expression between baseline (time 1) and post exposure (time 2). We evaluated the association of PM mass and PM metal component levels with miRNA expression measured in post exposure samples using simple linear regres sion models and multi variable models. The adjusting variables were selected a priori based on the assumption that potential confounders had to be related to inflammation or oxidative stress generation. Age, body mass index (BMI), smoking, and number of cigarettes per day are all potential determinants of inflamma tion, oxidative stress, or both. In addition, we adjusted for percent granulocytes to control for possible shifts in leukocyte differential count.
As a sensitivity analysis, we also fitted duration of smoking in the model described above. The results of the models including duration of smoking were very similar to those including number of cigarettes per day.
To compare the magnitude of the asso ciations of miRNA expression with different exposures, we calculated standardized regres sion coefficients. Outliers were excluded from regression analysis by dropping observations with studentized residuals that exceeded +3 or -3. We checked regression assumptions by performing diagnostic tests for each model, including the Shapiro-Wilk test to verify nor mality of residuals and the White test to verify the homo geneity of variance of the residuals. For miRNAs that did not show differences in their expression between time 1 and time 2, we also used mixed models that regressed PM and metal exposures against all measures of miRNA expression, regardless of whether they were measured at baseline or post exposure samples (Tarantini et al. 2009). A twosided pvalue < 0.05 was considered statistically sig nificant. All statistical analyses were performed using SAS (version 9.1.3; SAS Institute Inc., Cary, NC, USA).

Results
Subject characteristics and exposure. The char acteristics of the 63 study subjects are summa rized in Table 1 Table S2, shows the average levels and distributions of the individual exposures to PM and PM metal components during the 3 workdays between the baseline and post exposure miRNA measurements.
Target mapping and functional analysis: potential interactions of miR222, miR21, and miR146a with signal transduction pathways. To explore the functional significance of the miRNAs investigated, we applied KEGG (http://www.genome.jp/kegg), a pathway analysis database, to the target genes iden tified for miR-21, miR-222, and miR-146a using miRanda. The enriched pathways we identified appeared to be largely overlapping (i.e., many target genes were in more than one pathway). Among the topranked pathways, we found those related to general functions (e.g., purine metabolism, cell cycle) and others with more specific functions, including a high proportion of pathways related to oxidative stress and inflammation [Table 3; for a com plete pathway list, see Supplementary Material, Table S5 (doi:10.1289/ehp.0901300)]. Using LitInspector, we scanned PubMed for co occurrence of the user input gene (miR-222, miR-21, or miR-146a) and the general pathway key words in the same sentence. We found two pathways related to miR-222, seven related to miR-21, and six related to miR-146a (Table 4).

Discussion
In this study of foundry workers in an electric furnace steel plant, we evaluated the effect of exposure to PM and PM metal components on the expression of three candidate miRNAs that regulate genes in pathways related to oxida tive stress and inflammation. We found that miR-222 and miR-21 expression was increased in post exposure samples collected after 3 work days, compared with baseline samples. miR-222 expression in post exposure samples was posi tively associated with the mean lead exposure meas ured in the PM 10 fraction of the PM mass during the 3 workdays, whereas we found no significant association of miR-21 expression with the exposures levels we evaluated. In addi tion, although miR-146a expression did not differ in post exposure and baseline samples, we observed that individual levels of exposure to lead and cadmium in the PM 10 fraction were associated with miR-146a expression in post exposure samples, as well as in a pooled data analysis that included both baseline and post exposure samples.
Although many studies have focused on comparing miRNA expression between pathologic samples and normal tissues, very few studies have evaluated changes in miRNA expression in response to environ mental stimuli (Baccarelli and Bollati 2009). A recent in vitro study by Jardim et al. (2009) showed that miRNA expression profiles in human airway cells change in response to die sel exhaust particles. In that study, 197 of the 313 detectable miRNAs (62.9%) were either up or downregulated ≥ 1.5 times, including many miRNAs associated with responses in inflammatory pathways. To the best of our knowledge, ours is the first human study to show that PM modifies the in vivo expression of candidate miRNAs.
miR-222 expression has been related with nitric oxide (Suarez et al. 2007) and redox   Dai et al. 2008 signaling (Sen et al. 2009). An upregulation of miR-222, as measured in our study, may suggest an increased proliferation rate of blood leuko cytes in response to environmental stimuli that are able to induce inflammation responses. Expression of miR-21 is part of a response aimed at limiting injuries from ROS (Cheng et al. 2009). In the present study, we observed an upregulation of miR-21 after 3 days of work compared with the baseline measure ment. This may reflect a non specific response to ROS production in blood due to increased PMinduced oxidative stress (Brauner et al. 2007), as also suggested by the positive associa tion that we observed between miR-21 expres sion and 8OHdG. However, we found no significant associations between miR-21 expres sion and PM or PM metal components. Other unknown biological changes intervening during the 3 workdays-or other unmeas ured expo sures that are present in foundry facilities, such as heat, carbon monoxide, and non ionizing radiation (Tarantini et al. 2009)-might have modified miR-21 expression.
Studies in myeloid cells activated by bac terial and fungal components or after expo sure to the pro inflammatory cyto kines tumor necrosis factor α or interleukin1β have shown that miR-146a is involved in limit ing inflammatory responses triggered through the innate immune system Williams et al. 2008). The negative association of miR-146a expression with lead and cad mium levels that we observed in the present study may indicate that PM metal components enhance inflammatory processes initiated by organic PM antigens through the innate sys tem (Descotes 1992;Li et al. 2008).
Bioinformatic strategies, such as those implemented in miRanda software (Bentwich 2005), are now available to identify poten tial miRNA target sites in the 3´ untranslated region (UTR) of a proteincoding gene. The potential targets of miRNAs often include hundreds of genes because the reverse comple ment of some "seeds" (bases 2-8 of the mature miRNA) appears in multiple locations in many premRNA 3´ UTRs. With the under standing that recog nition of mRNA targets is speculative, we explored miR-222, miR-21, and miR-146a targets and examined whether these targeted genes are over represented in pathways annotated in the KEGG database. miRanda identified > 3,000 genes that are potential targets of the three miRNAs evalu ated in our study. Using the KEGG pathway database, we identified several pathways that are involved in immune/inflammatory and oxi dative stress. Transition metals are common components of ambient PM and have been shown to inter act with the immune system in antigen non specific fashion (Mishra 2009). MacNee and Donaldson (2000) proposed that the generation of oxidative stress, either directly by transition metal components of PM or indi rectly from the recruitment into the airspaces and activation of blood leukocytes, is a primary mechanism determining the inflammation related health outcomes of PM. Our results indicate novel pathways through which metals may elicit specific PMrelated responses and help explain previous studies that have spe cifically indicated a role of metals in promoting PM effects (Arisawa et al. 2001;Magari et al. 2002;Messner et al. 2009;Park et al. 2008).
Results of the present study show that asso ciations of metals with miR-222 and miR-146a expression were limited to nonsmokers. In addition, when we divided the study subjects in two sub groups according to age (25-45 years vs. 45-60 years), we observed stronger asso ciations among younger subjects. Identification of sets of individuals who have enhanced responses to PM may suggest possible mecha nisms of physio logic assault, and provide data that can be used for more detailed risk assess ment (Bateson and Schwartz 2004).
In the present study we investigated a popu lation with wellcharacterized expo sure to PM and PM metal components that allowed for contrasting subjects over a wide range of different exposure levels. Because of the limited number of study subjects, it is possible that the associations observed were due to chance. However, the occupational exposure and relatively controlled environ ment of a foundry provide a good setting for evaluating these mechanistic questions and reduce bias and chance findings. Our study was based on subjects working in several areas of the same factory but did not include a dif ferent population of subjects without expo sure to PM. Limiting our investigation to individuals who have all been working in the same facility avoided potential concerns related to the selection of external referents who might have differed from the exposed population in terms of socio economic factors and other chara cteristics determining hiring into the plant (Pearce et al. 2007). However, the differences in the individual levels of exposure within our study group were large, providing sufficient contrast for identifying exposurerelated changes in miRNA expres sion (Anselmi and Patelli 2006).
In summary, our findings suggest that air particles, particularly those rich in lead and cadmium, are able to modify miRNAs expres sion. Further studies are required to deter mine the role of such alterations along the pathways determining the effects of PM on human health.