Research Article

Received: 30 July 2008, Revised: 3 October 2008, Accepted: 9 October 2008, Published online 19 November 2008 in Wiley Interscience

(www.interscience.wiley.com) DOI 10.1002/jat.1400

J. Appl. Toxicol. 2009; 29: 223–232 Copyright © 2008 John Wiley & Sons, Ltd.

223

John Wiley & Sons, Ltd.In vitro estrogenicity of ambient particulate matter: contribution of hydroxylated polycyclic aromatic hydrocarbonsDaniela Wenger,a,c Andreas C. Gerecke,a,* Norbert V. Heeb,a Peter Schmid,a Christoph Hueglin,a Hanspeter Naegelib and Renato Zenobic

ABSTRACT: Atmospheric particulate matter (PM1) was collected at an urban and a rural site in Switzerland during a hibernalhigh air pollution episode and was investigated for estrogenicity using an estrogen-sensitive reporter gene assay (ER-CALUX).All samples that were tested induced estrogen receptor-mediated gene expression in T47D human breast adenocarcinomacells. Observed estrogenic activities corresponded to 17b-estradiol (E2) CALUX equivalent concentrations ranging from 2 to23 ng E2-CEQ per gram of PM1 (particulate matter of £ 1 mm aerodynamic diameter) and from 0.07 to 1.25 pg E2-CEQ per m3

of sampled air. There was a strong correlation between the PM1 estrogenicity of the urban and rural sites (r = 0.92). Fivehydroxylated polycyclic aromatic hydrocarbons (hydroxy-PAHs), which show structural similarities to E2, were assessed fortheir estrogenic activity. The following order of estrogenic potency was found: 2-hydroxychrysene > 2-hydroxyphenanthrene> 1-hydroxypyrene > 2-hydroxynaphthalene > 1-hydroxynaphthalene. Three of these hydroxy-PAHs, namely 2-hydroxy-phenanthrene, 2-hydroxynaphthalene and 1-hydroxynaphthalene, were detected in all PM1 extracts. However, they con-tributed only 0.01–0.2% to the overall estrogenic activity. Hence, mainly other estrogenic compounds not yet identified bychemical analysis must be responsible for the observed activity. The temporal trend of PM1 estrogenicity at the urban andrural site, respectively, was compared with the time course of several air pollutants (NO2, NO, SO2, O3, CO) and meteorologicalparameters (temperature, humidity, air pressure, solar irradiation, wind velocity). However, specific emission sources andformation processes of atmospheric xenoestrogens could not be elucidated. This study showed that ambient particulatematter contains compounds that are able to interact with estrogen receptors in vitro and potentially also interfere with estrogen-regulated pathways in vivo. Copyright © 2008 John Wiley & Sons, Ltd.

Supporting information may be found in the online version of this article.

Keywords: endocrine disruption; air pollution; particulate matter; PM1; estrogenic activity; estrogen receptor; ER-CALUX;hydroxylated polycyclic aromatic hydrocarbons; hydroxy-PAHs

Introduction

Long-term exposure to high levels of respirable atmosphericparticles is considered to pose a health threat to humans andanimals. Several epidemiological studies have indicated thatparticulate levels in air influence the incidence of humanrespiratory and cardiovascular diseases and their mortality (Popeet al., 2004; Delfino et al., 2005). Health effects that may berelated to the endocrine system also have been associated withexposure to particles and other ambient air contaminants, forexample, adverse fetal development (Šrám et al., 2005) anddecreased fertility in men (De Rosa et al., 2003). The exact mech-anistic pathways linking exposure and observed effects haveso far not been established. It has been hypothesized that chem-icals bound to atmospheric particles play an important role(Sioutas et al., 2005). Some compounds found in vehicle exhaust,waste combustion emissions, and ambient air have been shownto act as endocrine disruptors (Clemons et al., 1998; Watanabeand Kurita, 2001; Wang et al., 2003, 2004; Sidhu et al., 2005; Kleinet al., 2006; Wenger et al., 2008a; 2008b). Endocrine disruptorsinterfere with the normal functioning of the endocrine systemby mimicking or antagonizing naturally occurring hormones and/or by modulating hormonal responses (Damstra et al., 2002). In

this way, endocrine disruptors can lead to adverse health effectsin humans and animals. Chemicals capable of disrupting theaction of gonadal steroids, for example estrogens, have receivedmost attention. Because of endocrine-disrupting effects observedin laboratory animals, a major concern is that these chemicals maycause reproductive disorders and hormone-sensitive cancers(Damstra et al., 2002).

Evidence of estrogen-like compounds present in atmosphericparticulate matter was first reported by Clemons et al. (1998)and subsequently confirmed by two other in vitro studies (Wanget al., 2004; Klein et al., 2006). Estrogens, of which 17β-estradiol

* Correspondence to: A. Gerecke, Empa, Swiss Federal Laboratories for MaterialsTesting and Research, Ueberlandstrasse 129, CH-8600 Duebendorf, Switzerland.E-mail: andreas.gerecke@empa.ch

a Empa, Swiss Federal Laboratories for Materials Testing and Research, Ueber-landstrasse 129, CH-8600 Duebendorf, Switzerland.

b Institute of Pharmacology and Toxicology, University of Zurich-Vetsuisse,Winterthurerstrasse 260, CH-8057 Zurich, Switzerland.

c Department of Chemistry and Applied Biosciences, ETH Zurich, CH-8093Zurich, Switzerland.

D. Wenger et al.

www.interscience.wiley.com/journal/jat Copyright © 2008 John Wiley & Sons, Ltd. J. Appl. Toxicol. 2009; 29: 223–232

224

(E2) is the major one in humans, regulate the growth, differentia-tion, and functioning of diverse target tissues, including thoseinvolved in reproduction (Katzenellenbogen et al., 2000; Nilssonet al., 2001). Most of the estrogenic effects appear to be mediatedvia estrogen receptors, of which two subtypes (ERα, ERβ ) havebeen found in humans. ERs are ligand-activated transcriptionfactors and belong to the nuclear receptor superfamily. Besidesthe physiological ligands, both natural and man-made exogen-ous compounds (i.e. xenoestrogens) are able to bind to ERs.Through activation or blockage of ERs, these exogenous ligandsinterfere with estrogen-regulated gene expression. Alterationsin gene expression are thought to represent an early but critical stepin the regulation of biological processes (Damstra et al., 2002).

A class of anthropogenic chemicals that have structural simi-larities to steroid hormones such as E2 are hydroxylated poly-cyclic aromatic hydrocarbons (PAHs) (Fig. 1). Several hydroxy-PAHssuch as hydroxylated derivatives of benzo[a]pyrene and chrysenehave been shown to possess estrogenic activity, as indicatedby in vitro test systems based on genetically modified yeast ormammalian cell lines (Charles et al., 2000; Fertuck et al., 2001;Kamiya et al., 2005; van Lipzig et al., 2005; Hayakawa et al., 2007).Therefore, we hypothesized that hydroxy-PAHs contribute in someextent to estrogenic activity previously observed with ambientparticulate matter (Clemons et al., 1998; Wang et al., 2004; Kleinet al., 2006). In the atmosphere, hydroxy-PAHs are formed by sec-ondary oxidation of PAHs in the gas-phase, involving hydroxyradicals (HO•) during the day and nitrate radicals (•NO3) duringnight-time (Vione et al., 2004). During nighttime, hydroxy-PAHscan further react to nitro-oxy-PAHs. Hydroxy-PAHs are generallyless volatile than the parent PAHs and tend to be more particle-associated (Vione et al., 2004). In addition to atmospheric pro-cesses, hydroxy-PAHs and their precursors are emitted directlyfrom combustion processes and are found, for example, in dieselexhaust (Sidhu et al., 2005; Heeb et al., 2008). Furthermore,hydroxylated PAHs can be formed in organisms, after the intakeof parental PAHs, through biotransformation catalyzed by cyto-chrome P450 (CYP) enzymes (Charles et al., 2000; Fertuck et al.,2001; van Lipzig et al., 2005). In vitro estrogenicity observed withcertain PAHs is supposed to be predominantly due to their

hydroxylated metabolites formed in the exposed cells (Charleset al., 2000; Fertuck et al., 2001; van Lipzig et al., 2005).

In this study, estrogenic compounds were investigated thatare bound to atmospheric particles with an aerodynamic diameterof less than 1 μm (PM1). PM1 was collected at an urban and arural site in Switzerland. Based on our findings that diesel exhaustcontains considerable amounts of estrogenic compounds andpotential precursors (Heeb et al., 2008; Wenger et al., 2008a), wehypothesized that ambient air in traffic-exposed areas mustalso have some estrogenic activity. On the other hand, airbornexenoestrogens formed via transformation of precursors mightresult in increased estrogenic activity of aged air masses evenin rural areas far from direct emission sources. In contrast to pre-vious in vitro investigations, we focused on PM1 because fine andultrafine particles are thought to be of particular importance inthe context of adverse health effects induced by particulate airpollution. Compared with larger particles, fine and ultrafine parti-cles have a higher number concentration and surface area, and largeramounts of sorbed chemicals per unit mass (Sioutas et al., 2005).Daily PM1 samples were collected at the urban and the rural siteduring a 15-day measuring campaign during a winter smog periodin 2006. We applied an estrogen-sensitive in vitro reporter geneassay (estrogen-responsive chemically activated luciferase geneexpression, ER-CALUX) to detect and quantify PM1-associatedchemicals that induce ER-controlled gene expression (Legler et al.,1999). So far, the contribution of hydroxylated PAHs to the estro-genicity of PM has not been assessed. In this study, concentrationsof five hydroxy-PAHs were determined, using gas chromatography/high resolution mass spectrometry (GC/HRMS). Relative potency(REP) factors were deduced for these compounds, using the ER-CALUX assay. Finally, the contribution of these hydroxy-PAHs tothe overall estrogenic activity was estimated for ambient particles.

Experimental Section

Site Locations and Ambient Particulate Matter Sampling

Sampling of PM1 was performed at measuring sites of the SwissNational Air Pollution Monitoring Network (NABEL). The NABEL

Figure 1. The molecular structures and relative potencies (REPEC20 E2) of 17β-estradiol (E2; 1) and five hydroxy-PAHs (2–6) exhibiting in vitro estrogenicity.The hydroxy-PAHs are 1-hydroxynaphthalene (2), 2-hydroxynaphthalene (3), 2-hydroxyphenanthrene (4), 2-hydroxychrysene (5) and 1-hydroxypyrene(6). To point out the similarities to E2, the molecular structures of the hydroxy-PAHs are not orientated according to the IUPAC convention.

In vitro estrogenicity of ambient particulate matter

J. Appl. Toxicol. 2009; 29: 223–232 Copyright © 2008 John Wiley & Sons, Ltd. www.interscience.wiley.com/journal/jat

225

comprises 16 sampling sites that represent typical and impor-tant air quality situations in Switzerland (Bruggisser et al., 2007).For the present study, we selected an urban and a rural site onthe Swiss plateau (supporting information, Fig. S1). The urbansite (46°57′04″N, 7°26′27″E; 546 m above sea level) is located inthe city center of Bern (128 000 inhabitants). The sampling site issituated near to the central railway station and next to a highlyfrequented road. During the sampling period, 15 000–30 000vehicles per day passed along the road, 5–12% of which wereheavy goods vehicles. The road is bordered by multi-storiedhouses, forming an urban street canyon. There is no industry inthe proximity, but many service companies and stores. The ruralsampling site (46°48′47″N, 6°56′40″E; 489 m a.s.l., typical altitudeof the Swiss plateau) is located approximately 1 km southeast ofPayerne, a small town (8000 inhabitants) in western Switzerland,and about 40 km southwest of Bern. The site is surrounded byan agrarian landscape. The nearest main road proceeds severalhundred meters southeast of the sampling site. From bothsampling sites, we analyzed PM1 collected between 26 Januaryand 9 February 2006. As shown in Fig. 2, this was a period withincreased atmospheric PM1 concentrations in the Swiss plateaucaused by a strong winter inversion (Lanz et al., 2008). PM1 filtersof every fourth day were not available.

Ambient particulate matter sampling was performed on quartzfiber filters (Whatman QMA, diameter 150 mm) using high-volumesamplers (Digitel DHA 80). The intake probes were placed 3.5 m

above ground level. Prior to sampling, filters were conditionedfor 48 h at 22 ± 0.5 °C and 50 ± 2% relative humidity and weighted.Filters were neither pre-washed with solvents nor pre-heated.The PM1 samples were taken during 24 h (i.e. from midnight tomidnight of the following day) at a flow-rate of 30 m3 h−1. Thisresulted in 720 m3 of sampled air per filter. After sampling, thefilters were re-conditioned for another 48 h and gravimetricallyanalyzed for total mass. The exposed filters contained 2–50 mgPM1. They were stored in a freezer at −18 °C. Further informationon sampling is given in the technical report of the NABEL(Bruggisser et al., 2007).

Two field blanks were collected in January 2006. They wereidentically handled as filters used for PM1 sampling. The fieldblanks were taken to a NABEL sampling site and placed in thefilter stock of the high-volume sampler, but they were not usedfor sampling. In addition, filters directly taken from the supplier’sbox were extracted and analyzed, yielding blank controls ofthe filter material and the extraction and sample preparationprocedures.

Filter Extraction and Sample Preparation

All solvents were suitable for residue analytical quality andwere purchased from Merck (Darmstadt, Germany) or Biosolve(Valkenswaard, The Netherlands). Glassware used for extractionwas baked out at 450 °C overnight and subsequently rinsed with

Figure 2. Average daily concentrations of PM1 (μg m−3) at the urban (Bern) and rural site (Payerne) in January and February 2006. The grey areamarks the sampling dates used to investigate PM1 estrogenicity. The insets indicate that PM1 levels are highly correlated between Bern and Payerneboth from January through February (upper inset A; linear regression, r2 = 0.87; empty circle, outlier) and during the sampling period studied (lowerinset B; r2 = 0.90).

D. Wenger et al.

www.interscience.wiley.com/journal/jat Copyright © 2008 John Wiley & Sons, Ltd. J. Appl. Toxicol. 2009; 29: 223–232

226

dichloromethane. Filters were extracted with dichloromethanefor 24 h in a Soxhlet apparatus. Extracts were reduced in volumeto 1–2 ml, filtered, and transferred to a volumetric flask. Thevolume was adjusted to 5 ml with dichloromethane and tolueneto yield a final dichloromethane–toluene ratio of about 1 : 1.Extracts were transferred to leak-proof sealed glass vials andstored at 4 °C.

For bioassay analysis, concentration series of the filter extractswere prepared in dimethylsulfoxide (DMSO, extra pure, Merck,Darmstadt, Germany). Aliquots of each extract (15–1000 μl) weretransferred to glass vials, and the solvent was evaporated at50 °C under a constant stream of nitrogen. At the point of dry-ness, DMSO was added (5–20 μl). Samples in DMSO were storedin the dark at 4 °C. Assay analysis was performed within a fewdays after sample preparation.

For GC/HRMS analysis, 50 ng D9-1-hydroxypyrene (Chiron, Trond-heim, Norway) were added to aliquots (2–3 ml) of the filterextracts as an internal standard. Dichloromethane was evapo-rated and extracts were adjusted with toluene to approximately3 ml. The samples were extracted with 4 ml of aqueous 0.5 M potas-sium carbonate solution. For acetylation, the toluene phase wasremoved and 0.25 ml of acetic anhydride (Fluka, Buchs, Switzer-land) was added. After shaking, samples were left for 30 min atroom temperature. Subsequently, aqueous extracts were shakentwice with 3 ml of hexane. The hexane extracts were pooled andconcentrated to approximately 10 μl.

In Vitro Reporter Gene Assay

The Estrogen Responsive-Chemically Activated Luciferase geneeXpression (ER-CALUX) assay is an in vitro reporter gene assayused to detect estrogen receptor (ER)-mediated gene expression(Legler et al., 1999). The assay is based on T47D human breastadenocarcinoma cells that are stably transfected with the ER-controlled luciferase reporter gene construct pEREtata-Luc. Inthese T47D.Luc cells, ER agonists induce luciferase gene expres-sion in a dose-dependent manner and are indirectly detectedvia luciferase activity. The measured overall ER-mediated activityintegrates agonistic and antagonistic interactions exhibited byall ER ligands present in a sample.

T47D.Luc cells were obtained from BioDetection Systems (BDS,Amsterdam, The Netherlands). Cell culture, assay analysis andquality assurance/quality control were performed as describedin detail elsewhere (Wenger et al., 2008a). T47D.Luc cells wereseeded in 96-well microtiter plates (Nunc, Rochester, NY, USA) ata density of 10 000 cells per well, as determined with a cell count-ing chamber. Cells were maintained in 0.1 ml assay medium,which consisted of D-MEM/F-12 phenol red free medium (Gibco,Paisley, UK) supplemented with NaHCO3 (1.26 g l−1), 1% (v/v)non-essential amino acids, and 5% (v/v) dextran-coated charcoal-treated fetal bovine serum (FBS, Australian origin). After 24 hof incubation (37 °C, 5% CO2, 100% relative humidity), mediumwas renewed and cells were incubated for another 24 h. Then,medium was removed and cells were exposed in triplicate to0.1 ml exposure medium. This was prepared by individuallydiluting PM1 extracts, field blanks, filter blanks, solvent controls,and the reference compound 17β-estradiol (E2, 0.6–60 pM), alldissolved in DMSO, 1000-fold in assay medium. The final DMSOconcentration in the wells was 0.1% (v/v). Triplicate analysis ofselected PM1 extracts was repeated in a second, independentexposure experiment. Cytotoxicity of the tested samples wascontrolled by microscopic evaluation of cell morphology. After

an exposure time of 24 h, luciferase activity was measured asrelative light units (RLU) on a microplate luminometer (MLX,Dynex, Chantilly, USA). The limit of quantification (LOQ) wascalculated as 10 times the standard deviation of the solventcontrol. The average LOQ of all plates measured in this studywas 1.1 ± 0.6 pM E2 (0.3 pg E2 ml−1). To test whether the detectedluciferase activity was ER-mediated or not, an ER antagonist [ICI182 780 (Howell et al., 2000); Tocris Bioscience, Avonmouth, UK]was mixed with E2 (3 and 10 pM) and with PM1 extracts (BER 1/26, BER 1/28, PAY 1/26). This yielded exposure medium con-taining 10 nM ICI 182 780 (Legler et al., 1999; van Lipzig et al.,2005) and 0.2% DMSO (v/v).

Estrogenic activity of five hydroxy-PAHs (1-hydroxynaphthalene,2-hydroxynaphthalene, 2-hydroxyphenanthrene, 2-hydroxychrysene,D9-1-hydroxypyrene; Chiron, Trondheim, Norway) was tested usingthe ER-CALUX assay. It was assumed that T47D.Luc cells and theER do not discriminate between native 1-hydroxypyrene and itsdeuterated form, which was purchased for GC/HRMS analysis.Serial dilutions of 1-hydroxynaphthalene (1.4–208 μM in well), 2-hydroxynaphthalene (1.4–208 μM), 2-hydroxyphenanthrene (0.05–51 μM), 2-hydroxychrysene (0.04–41 μM) and D9-1-hydroxypyrene(0.44–44 μM) were added in triplicate to T47D.Luc cells at afinal DMSO concentration of 0.1% (v/v). For each hydroxylatedPAH, a dose–response curve was determined in two or threeindependent experiments, applying an exposure time of 24 h.The reference compound E2 and solvent controls wereincluded on all plates used for assay analysis. Relative potency(REP) factors that express the estrogenic activity of the singlehydroxy-PAHs relative to the reference compound E2 werecalculated.

Gas Chromatography/High Resolution Mass Spectrometry

A mixture of unlabeled hydroxy-PAHs (1-hydroxynaphthalene,2-hydroxynaphthalene, 2-hydroxyphenanthrene, and 2-hydroxy-chrysene) and D9-1-hydroxypyrene was acetylated and used as aquantification reference. Acetylated hydroxy-PAHs were sepa-rated by gas chromatography (Fisons Instruments HRGC Mega 2,Rodano, Italy) on a capillary column (PS086, 20 m × 0.28 mm, filmthickness 0.15 μm) with hydrogen as carrier gas at 50 kPa. Detec-tion and identification were achieved by high resolution massspectrometry (Thermo Finnigan MAT 95, Bremen, Germany) inelectron ionization mode (GC/EI-HRMS). Samples (2 μl) were injectedin splitless mode at an injector temperature of 260 °C usingan A200S autosampler (CTC Analytics, Zwingen, Switzerland).The temperature program for the GC/EI-HRMS analysis was asfollows: 1 min isothermal at 110 °C, increased at 8 °C min−1 to300 °C, then isothermal for 5 min. The transfer line was held at aconstant temperature of 280 °C and the ion source was oper-ated at 180 °C. The electron energy was set at 70 eV and themass spectrometer was tuned to a resolution of around 10 000.Data was acquired in multiple-ion-detection (MID) mode. Thesignals [M]+, [M − C2H2O]+ and [M − C3H3O2]+ of acetylated hydroxy-analogs of naphthalene, phenanthrene–anthracene, pyrene–fluoranthene and chrysene, and of the acetylated internalstandard D9-1-hydroxypyrene, were recorded. Detected hydroxy-PAHs were quantified against the internal standard. To estimatethe contribution of the detected hydroxy-PAHs to the totalER-mediated activity, GC/HRMS data were converted into E2-CEQconcentrations by applying molar REPEC20 E2 values previouslydetermined in the ER-CALUX assay and corrected for each com-pound’s molecular weight.

In vitro estrogenicity of ambient particulate matter

J. Appl. Toxicol. 2009; 29: 223–232 Copyright © 2008 John Wiley & Sons, Ltd. www.interscience.wiley.com/journal/jat

227

Data and Statistical Analysis

A three-parametric sigmoidal regression model was fitted tothe dose–response data of E2 using a least-squares approach, asspecified elsewhere (Wenger et al., 2008a). Luciferase activityinduced by PM1 extracts was expressed as E2-CALUX equivalent(E2-CEQ) concentrations. This was done by applying the regres-sion equation of E2 to solve for the E2 concentration exhibitingthe activity of the sample concentration chosen for quantifica-tion. Dose–response data of the tested hydroxy-PAHs was pro-cessed in the same way as data of the E2 dilution series. The fittedcurves of E2 and hydroxy-PAHs were used to determine REPfactors. To this end, the concentration of E2 at a defined responselevel (i.e. EC20, EC50 and EC80) was divided by the concentra-tion of the test hydroxy-PAH exhibiting the same level of activityas E2 (i.e. activity corresponding to 20, 50 and 80% of the maxi-mum E2 response). In this way, REPEC20 E2, REPEC50 E2 and REPEC80 E2

were calculated. Furthermore, the classical REPEC50 was calculatedby dividing the EC50 of E2 by the EC50 of the test compound.

Assay data were compared with meteorological data andwith concentrations of hydroxy-PAHs and several air pollutants.Correlation coefficients (Pearson, r) and linear regressions werecalculated using Systat version 10 (SPSS Inc., 2000).

Results

Estrogenic Activity of PM1 from an Urban and a Rural Site

Two PM1 extracts form the urban site in Bern (BER 1/26, BER 1/28)and one extract from the rural site in Payerne (PAY 1/26) were usedto measure full dose–response curves in the ER-CALUX assay andto assess the range of quantifiable sample concentrations. Forthe three extracts, estrogenic activity was observed at concen-trations between 8 and 600 μg PM1 per ml of exposure medium(Fig. 3). These amounts of particles corresponded to 0.1–14 m3

of sampled air. Figure 3 shows that both PM1 extracts sampledon 26 January induced a maximum response of ~30% and theone sampled on 28 January reached ~60%, relative to 60 pM

17β-estradiol (E2), the highest tested concentration of the refer-ence compound. Higher induction levels were not achieved

due to cytotoxic compounds present in the PM1 extracts. Toxiceffects on T47D.Luc cells were observed starting at concentra-tions of 200–300 μg PM1 ml−1. To verify that measured luciferaseactivities were ER-mediated and not induced by other pathways,we mixed a known ER antagonist (ICI 182 780, 10 nM) with E2(3 and 10 pM) and PM1 extracts (BER 1/26, BER 1/28, PAY 1/26).Expression of the marker protein was induced by E2 and PM1extracts, but was completely inhibited after addition of theantagonist (Fig. 4). Changes in cell morphology pointing to cyto-toxic effects (i.e. vacuolization, abnormal cell shape, damagedmembranes, detached cells) exerted by the tested mixtures werenot observed. Hence, we concluded that E2 and PM1 extractsexhibited ER-mediated activity in the ER-CALUX assay.

Based on the dose–response curves determined, the quantifi-able range for all PM1 samples could be estimated. For quantifica-tion, sample concentrations were chosen that induced luciferaseactivity in the range between the limit of quantification (LOQ)and 30% of the maximum induction level of E2 (Wenger et al.,2008a). As shown by Fig. 5, estrogenic activity was detected andquantified in all 24 samples that were tested. E2 CALUX equiva-lent (E2-CEQ) concentrations were in the same range at theurban and the rural site. Concentrations in Bern varied between4.5 and 23 ng E2-CEQ per gram of PM1 with an average of 9.0 ±5.6 ng E2-CEQ g−1 (supporting information, Table S1). Concentra-tions in Payerne ranged from 2.4 to 22 ng E2-CEQ g−1 with anaverage of 9.4 ± 6.8 ng E2-CEQ g−1. At both sites, similar medianconcentrations of 7.3 and 7.4 ng E2-CEQ g−1 were found. Theobserved ranges of estrogenic activity correspond to 0.08–1.25and 0.07–0.77 pg E2-CEQ per m3 of sampled air in Bern andPayerne, respectively (Table S1).

With the exception of one day (27 January), the temporal trendsof estrogenic activity were well correlated between the urbanand rural site (r = 0.92), as were the PM1 concentrations (r = 0.95).Linear regressions with r 2 = 0.84 for estrogenic activities andr 2 = 0.90 for PM1 concentrations were calculated (Figs 2 and 5).This indicates that the sampling sites did not represent indepen-dent air quality situations during the smog episode studied. Thehighest E2-CEQ concentrations were observed at both sites onthe same two days. These were 21 and 23 ng E2-CEQ per gram

Figure 3. Dose–response curves of 17β-estradiol (E2), two samples ofparticulate matter (PM1) from Bern (BER 1/26, BER 1/28) and one PM1sample from Payerne (PAY 1/26) determined in the ER-CALUX assay withan exposure time of 24 h. Concentrations are given in grams E2 or PM1per millilitre of exposure medium.

Figure 4. Estrogen receptor (ER)-mediated luciferase induction of17β-estradiol (E2) and PM1 extracts (BER 1/26, BER 1/28, PAY 1/26) andinhibition by an ER antagonist (ICI 182 780, 10 nM). Data are expressed aspercent of luciferase induction of 3 pM E2 (ER-CALUX, exposure time24 h). Shown is the mean of a triplicate determination in one exposureexperiment (relative standard deviation < 15%).

D. Wenger et al.

www.interscience.wiley.com/journal/jat Copyright © 2008 John Wiley & Sons, Ltd. J. Appl. Toxicol. 2009; 29: 223–232

228

of PM1 on 28 January and 16 and 17 ng E2-CEQ g−1 on 7 February.On 27 January, twofold increased ER-mediated activity wasobserved in Payerne vs Bern (i.e. 22 and 11 ng E2-CEQ g−1). Inter-estingly, PM1 levels were relatively low at the days with elevatedmass-based estrogenic activity compared with other samples(Figs 2 and 5). This may indicate that accumulation of inactivecompounds led to a dilution of particle-bound xenoestrogens. Noassociation was found between estrogenic activity and weekday.

Estrogenic Activity of Hydroxylated PAHs

Five hydroxylated PAHs were tested in the ER-CALUX assay for theirpotency to induce ER-mediated gene expression. Figure 1 displaystheir chemical structures and relative potency values. Figure 6shows the dose–response curves of E2 and the five tested hydroxy-PAHs. E2 induced a maximum response at 60 pM. This concentra-tion was used as a reference for 100% effect. The EC50 value forE2 was 3.7 ± 0.6 pM (mean ± standard deviation), as obtained from12 independent exposure experiments with triplicate determina-tions. Two of the tested compounds, namely, 2-hydroxychryseneand 2-hydroxyphenanthrene, achieved a maximum ER-mediatedactivity distinctly above the highest luciferase induction levelof E2. This large effect relative to the reference compoundwas observed at 277 ± 92 and 228 ± 5% of 60 pM E2 for 2-hydroxychrysene and 2-hydroxyphenanthrene, respectively(Table 1). As illustrated by Fig. 6, 2-hydroxynaphthalene reacheda similar maximum activity as E2, whereas 1-hydroxypyrene and1-hydroxynaphthalene exhibited only 50–60% of the responseof 60 pM E2. EC50 values of the hydroxylated PAHs ranged from0.5 to 35 μM (Table 1). Cytotoxicity was not observed by micro-scopic visualization of the cells within the concentration rangesused for 1-hydroxynaphthalene and 2-hydroxynaphthalene. Allother tested hydroxy-PAHs were cytotoxic at higher concentrations,

which possibly decreased the maximally achievable ER-mediatedactivity. Cytotoxic effects started to be visible at 41 μM for 2-hydroxychrysene, at 51 μM for 2-hydroxyphenanthrene and at44 μM for 1-hydroxypyrene (D9). EC50 values and maximuminduction levels (i.e. efficacies) were determined from two orthree separate experiments, and are summarized in Table 1.

Different levels of maximum induction were observed for thefive hydroxylated PAHs. Therefore, several REP factors were deter-mined for each hydroxy-PAH instead of a single point estimate.The reported series of REP factors (Table 1) indicate the potentialrange of uncertainty in REP estimation due to non-parallel dose–response relationships (Villeneuve et al., 2000), as observed withthe tested hydroxy-PAHs and E2 (Fig. 6). REPs based on EC50sand on concentrations inducing 20, 50 and 80% of the maxi-mum response of E2 were calculated and are listed in Table 1.These values show that the tested hydroxy-PAHs were 5 × 104

to 3 × 107 times less potent than E2 to induce expression of ER-controlled genes in T47D human breast adenocarcinoma cells.The following order of relative estrogenic potency was found:2-hydroxychrysene > 2-hydroxyphenanthrene > 1-hydroxypyrene> 2-hydroxynaphthalene > 1-hydroxynaphthalene. Compoundswith a hydroxy group at position 2 were in general more potentthan those with a hydroxy group at position 1, as illustrated byFig. 1. Furthermore, the size of the molecule and the spatialarrangement of the fused benzene rings (i.e. linear, angular, clus-ter) affected the estrogenic potency of the tested hydroxy-PAHs.2-Hydroxyphenanthrene and 2-hydroxychrysene display an angu-lar arrangement of three and four rings, respectively, and werenearest to the structure of E2 among the tested compounds.

Contribution of Hydroxylated PAHs to the Observed Activity of PM1

1-Hydroxynaphthalene, 2-hydroxynaphthalene and 2-hydroxy-phenanthrene were detected in all PM1 extracts by GC/HRMSanalysis, whereas 2-hydroxychrysene, the most potent of the testedhydroxy-PAHs, was not detected with the method applied. 1-Hydroxypyrene could not be identified unambiguously, becauseno reference material was available. However, in the mass

Figure 5. Estrogenic activity of particulate matter (PM1) collected at anurban (Bern) and a rural site (Payerne) in Switzerland. Activity is shown asnanograms 17β-estradiol equivalents (E2-CEQ) per gram of PM1 (ER-CALUX,24 h exposure). Bars represent the mean ± standard deviation of one ortwo independent exposure experiments each with triplicate determina-tions. The insert indicates that estrogenic activities on a day-by-day resolu-tion were well correlated between the sites (linear regression, r2 = 0.84).N.m., not measured; n.d., not detected; empty circle in inset, outlier.

Figure 6. Induction of ER-mediated luciferase activity by 17β-estradiol(E2) and five hydroxylated PAHs (2-HO-Ch, 2-hydroxychrysene; 2-HO-Ph,2-hydroxyphenanthrene; 1-HO-Py, 1-hydroxypyrene; 2-HO-Np, 2-hydroxy-naphthalene; 1-HO-Np, 1-hydroxynaphthalene) in T47D.Luc cells after anexposure time of 24 h. For each compound, a representative example oftwo or three independent exposure experiments is shown. Each pointrepresents the mean ± standard deviation of triplicate determinations.

In vitro estrogenicity of ambient particulate matter

J. Appl. Toxicol. 2009; 29: 223–232 Copyright © 2008 John Wiley & Sons, Ltd. www.interscience.wiley.com/journal/jat

229

chromatograms at m/z 260.0832, 218.0726 and 189.0699 of allPM1 extracts an abundant signal was observed at a retention timesimilar to acetylated D9-1-hydroxypyrene. Because of the corre-sponding mass pattern, this signal probably represents themolecular ion and the typical fragments of an acetylated hydro-xypyrene or an isomer thereof. Furthermore, besides the identified2-hydroxyphenanthrene, GC/HRMS data also point to the pres-ence of additional hydroxylated derivatives of phenanthrene oranthracene. At both sampling sites, 2-hydroxyphenanthrene wasthe most abundant of the identified hydroxy-PAHs (0.3–20 μg g−1

of PM1, 6–382 pg m−3 of air), followed by 2-hydroxynaphthalene(0.01–0.9 μg g−1, 0.2–43 pg m−3) and 1-hydroxynaphthalene (0.01–0.4 μg g−1, 0.2–8 pg m−3). GC/HRMS data are reported in detail inthe supporting information (Table S2). Interestingly, concentra-tions of 2-hydroxyphenanthrene, 2-hydroxynaphthalene and 1-hydroxynaphthalene were on average about twofold higher inPayerne than in Bern. An explanation might be that atmospherictransformation of PAHs leads to higher concentrations of hydro-xylated PAHs in aged air masses even in rural areas far from directemission sources. Hydroxy radicals (HO•) in the atmosphere aremainly formed by the photolysis of O3 (Vione et al., 2004). Con-centrations of O3 were higher in Payerne than in Bern (see thenext section). This possibly points to increased atmospheric for-mation of hydroxy-PAHs at the rural site.

REP factors based on concentrations exhibiting 20% of themaximum response of E2 were applied to estimate the activityof the three identified hydroxy-PAHs in the ER-CALUX assay(Fig. 7). REPEC20 E2 factors (Table 1) were chosen because the averageluciferase activity of the sample concentrations used for quanti-fication was 19 ± 8% of the activity of E2 60 pM. The contributionof the sum of 2-hydroxyphenanthrene, 2-hydroxynaphthaleneand 1-hydroxynaphthalene to the overall estrogenic activity was

Figure 7. Estimated estrogenic activity of the sum of three hydroxy-PAHs (2-hydroxyphenanthrene, 2-hydroxynaphthalene, and 1-hydroxy-naphthalene) expressed as picograms 17β-estradiol CALUX equivalents(E2-CEQ) per gram of particulate matter (PM1). The insert shows that thesummed E2-CEQ concentrations of the OH-PAHs only weakly correlated(r = 0.54) between the urban (Bern) and rural site (Payerne). A linearregression was not calculated. N.m., not measured; n.d., not detected;empty circles in inset, outliers.

Tabl

e 1.

EC50

s, ef

ficac

ies

and

rela

tive

estr

ogen

ic p

oten

cies

of f

ive

hydr

oxyl

ated

PA

Hs

Com

poun

dn

EC50

[M]

STD

EVEf

ficac

y (%

E2

max

)RE

P EC2

0 E2

(mol

ar)

STD

EVRE

P EC5

0 E2

(mol

ar)

STD

EVRE

P EC8

0 E2

(mol

ar)

STD

EVRE

P EC5

0 (m

olar

)ST

DEV

E212

3.7×

10−1

26.

3×

10−1

310

01.

01.

01.

01.

02-

HO

-Ch

24.

9×

10−7

3.4×

10−8

277±

921.

0×

10−5

2.8×

10−6

1.4×

10−5

8.1×

10−7

2.1×

10−5

1.8×

10−6

8.5×

10−6

2.6×

10−6

2-H

O-P

h2

3.7×

10−6

4.8×

10−7

228±

51.

1×

10−6

1.8×

10−7

1.4×

10−6

3.9×

10−7

2.3×

10−6

9.3×

10−7

8.7×

10−7

1.4×

10−7

1-H

O-P

y3

6.9×

10−6

1.6×

10−6

59±

163.

7×

10−7

7.7×

10−8

3.3×

10−7

4.5×

10−8

E max

< E

C80 E

25.

9×

10−7

1.7×

10−7

2-H

O-N

p3

1.7×

10−5

5.0×

10−7

103±

101.

5×

10−7

2.2×

10−8

2.0×

10−7

1.2×

10−8

2.8×

10−7

3.0×

10−8

2.0×

10−7

1.7×

10−8

1-H

O-N

p2

3.5×

10−5

1.1×

10−8

53±

76.

5×

10−8

5.0×

10−9

7.5×

10−8

E max

< E

C80 E

21.

1×

10−7

6.8×

10−9

EC50

s, ef

ficac

ies

and

rela

tive

pote

ncy

(REP

) fac

tors

repr

esen

t the

mea

n±

stan

dard

dev

iatio

n (S

TDEV

) of t

wo

or th

ree

inde

pend

ent 2

4h

expo

sure

exp

erim

ents

(n) p

erfo

rmed

with

the

ER-C

ALU

X as

say.

The

EC5

0 of

eac

h in

divi

dual

com

poun

d is

the

conc

entr

atio

n (M

) at w

hich

50%

of t

he o

bser

ved

max

imum

effe

ct w

as re

ache

d. E

ffica

cies

wer

e ca

lcul

ated

usi

ngth

e m

axim

um r

espo

nse

of 1

7β-e

stra

diol

(E2

max

) as

100%

effe

ct. M

olar

-bas

ed R

EP fa

ctor

s w

ere

calc

ulat

ed a

s fo

llow

s: R

EPEC

i E2

=EC

i E2/c

once

ntra

tion

of h

ydro

xy-P

AH

indu

cing

an

effe

ct o

f Ei, w

here

i=

20, 5

0 or

80%

of E

2 m

axim

um; R

EPEC

50=

EC50

E2/E

C50 h

ydro

xy-P

AH (2

-HO

-Ch,

2-h

ydro

xych

ryse

ne; 2

-HO

-Ph,

2-h

ydro

xyph

enan

thre

ne; 1

-HO

-Py,

1-h

ydro

xypy

rene

(D9);

2-H

O-N

p, 2

-hyd

roxy

naph

thal

ene,

1-H

O-N

p, 1

-hyd

roxy

naph

thal

ene)

.

D. Wenger et al.

www.interscience.wiley.com/journal/jat Copyright © 2008 John Wiley & Sons, Ltd. J. Appl. Toxicol. 2009; 29: 223–232

230

small, that is, 0.01–0.1% in Bern and 0.02–0.2% in Payerne. Mainly2-hydroxyphenanthrene was responsible for this contribution.The calculated E2-CEQ concentrations of the three hydroxy-PAHsdetected in the PM1 extracts are listed in the supporting informa-tion (Table S3). Median values of 2.8 and 4.7 pg E2-CEQ g−1 werefound in Bern and Payerne, respectively, again indicating thatlevels of the identified hydroxy-PAHs were higher at the ruralthan at the urban site. The individual congeners showed a verysimilar temporal trend at each site but not between sites. Hence,the summed E2-CEQ concentrations of the three hydroxy-PAHsonly weakly correlated (r = 0.54) between sites (Fig. 7). This wasin contrast to the strong correlation observed for overall estro-genic activities as well as for PM1 concentrations.

Comparison of Assay Data with Meteorological Data and Air Pollutant Levels

At the NABEL measuring sites, meteorological data and concen-trations of several air pollutants are recorded (Bruggisser et al.,2007). The meteorological data include temperature, humidity,air pressure, solar irradiation and wind direction and velocity.Measured air pollutants are NO2, NO, O3, CO, PM1 and PM10. InPayerne, SO2 was also measured during the respective samplingperiod. Numbers on traffic frequency were available for Bern.Meteorological data were similar at both sites. Temperaturesranged from −4 to 3 °C, relative humidity from 67 to 97%, airpressure from 949 to 968 hPa, solar irradiation from 17 to 86 W m−2

and wind velocity from 0.5 to 4 m s−1. An exception was the winddirection. The wind came from north/north-east or from south/south-west in Payerne and mainly from north, less frequentfrom south-east or south-west, in Bern. Concentrations ofNO2, NO, CO, PM1 and PM10 were higher at the urban thanthe rural site (supporting information, Table S4). For example,PM1 concentrations were 20–60% higher in Bern than in Pay-erne; with the exception of one day with similar concentrations.As indicated before, PM1 levels at the urban and the rural sitewere strongly correlated during the high air pollution episodestudied (Fig. 2). PM1 levels were up to five times higher than theyearly average at each site. Concentrations of O3 were generallylow but higher in Payerne (11 ppb; median value, n = 12) than inBern (4 ppb).

The overall estrogenic activity of PM1 collected in Bern andPayerne was compared with meteorological data, traffic frequency,concentrations of air pollutants and concentrations of the threedetected hydroxy-PAHs (sum). At both sites, only weak correla-tions were found. A positive correlation was calculated betweenestrogenic activity and hydroxy-PAHs (r = 0.4) in Bern. A negativecorrelation was found for wind velocity (r = − 0.5). In Payerne,positive correlations were found between estrogenic activityand SO2 (r = 0.6) and between estrogenic activity and hydroxy-PAHs (r = 0.4). Again, a negative correlation was obtained forwind velocity (r = −0.5).

DiscussionIn vitro estrogenicity was detected in dichloromethane extractsof PM1 collected at an urban (Bern) and a rural site (Payerne) inSwitzerland during a high air pollution episode in winter 2006.The tested PM1 extracts contained compounds that were ableto induce ER-mediated gene expression in T47D human breastadenocarcinoma cells. The PM1 extracts exhibited 30–60% ofthe maximum response of E2. This corresponds well with the study

by Clemons et al. (1998). They reported that 20–50% of the max-imum activity of E2 was induced in MCF-7 human breast cancercells by a dichloromethane extract of urban PM10. The authorssuggested that a 100% effect was not achieved due to the pres-ence of partial ER agonists. In our study, however, higher induc-tion levels were not observed due to cytotoxic effects exertedon the exposed cells by PM1 extracts from both urban and ruralsites. The detected estrogenic activities corresponded to E2 equi-valent concentrations of 2.4–23 ng E2-CEQ g−1. These amounts ofestrogenic compounds per gram of PM1 were distinctly lowerthan those reported by previous studies, which found 0.2–1.0 μgE2 equivalents (E2-EQ) per gram of PM10 collected in Toronto,Canada, (Clemons et al., 1998) and 0.2–0.8 μg E2-EQ g−1 of PM100collected in Wuhan, a Chinese city (Wang et al., 2004). However,the chemical compositions of particles from different locationsare not identical (Putaud et al., 2004). Furthermore, particlesfrom the same location but of different size distributions varyin chemical composition due to their different sources (Sioutaset al., 2005). PM1 mainly consists of aggregated combustion-derived particles and condensed vapors, whereas PM10 and largersize fractions also contain mechanically generated particles. Thevolume-based concentrations of estrogenic compounds associ-ated with PM1 from Bern and Payerne ranged from 0.1 to 1.3 pgE2-CEQ m−3. The previous studies reported 5–23 pg E2-EQ m−3

(Clemons et al., 1998) and 71–329 pg E2-EQ m−3 (Wang et al., 2004).In all three studies, the volume-based concentrations only reflectedthe estrogenic activity of particle-bound compounds and, there-fore, were dependent on particulate matter levels at the sam-pling sites. It has been shown that not only the particulatephases of ambient air exhibit estrogenic activity but also thegas phase contains considerable amounts of xenoestrogens(Klein et al., 2006).

Certain hydroxylated derivatives of PAHs have been shown toexhibit estrogenic activity in different in vitro test systems (Charleset al., 2000; Fertuck et al., 2001; Kamiya et al., 2005; van Lipziget al., 2005; Hayakawa et al., 2007). These studies focused onhydroxylated metabolites of benzo[a]pyrene and chrysene. Anexception were the studies by Kamiya et al. (2005) and Hayakawaet al. (2007), which evaluated hydroxy-derivatives of several dif-ferent PAHs for their activity in an yeast-based assay. We testedfive two- to four-ring hydroxy-PAHs, namely 1-hydroxynaphthalene,2-hydroxynaphthalene, 2-hydroxyphenanthrene, 1-hydroxypyreneand 2-hydroxychrysene, for their potency to induce ER-mediatedgene expression in the ER-CALUX assay. The observed order ofrelative estrogenic potencies indicates that the position of thehydroxy group as well as the molecular size and structure seemto be crucial for the potency of hydroxy-PAHs to bind and acti-vate the ER. This is also suggested by data of other in vitro stud-ies (van Lipzig et al., 2005; Hayakawa et al., 2007). Hayakawa et al.(2007) observed the same order of potency for 2-hydroxychrysene,2-hydroxyphenanthrene, and 1-hydroxypyrene, whereas no sig-nificant estrogenic activity was observed for 1- and 2-hydroxy-naphthalene. However, the tested concentrations ranging from1 nM to 1 μM were probably too low. Our findings confirm anearlier observation regarding the large effect of 2-hydroxychrysenein the ER-CALUX assay (van Lipzig et al., 2005). An entirely ER-mediated effect of 400% and 200–300% relative to the maximumresponse of E2 was found after exposing T47D.Luc cells for 6 and24 h, respectively, and an EC50 of 3.2 × 10−7 M was reported. Weobserved a maximum effect of ~280% and a similar EC50, usingan exposure time of 24 h. Such superagonistic responses havealso been observed with some phytoestrogens, for example,

In vitro estrogenicity of ambient particulate matter

J. Appl. Toxicol. 2009; 29: 223–232 Copyright © 2008 John Wiley & Sons, Ltd. www.interscience.wiley.com/journal/jat

231

resveratrol (Gehm et al., 2004). Hence, the reported REP factorscan be used in subsequent studies that apply the ER-CALUX assayor related test systems to assess the estrogenic potential ofhydroxylated PAHs.

Hydroxy-PAHs are emitted during combustion processes (Kamiyaet al., 2005; Sidhu et al., 2005) or are formed in the atmosphereby secondary oxidation of PAHs (Vione et al., 2004). Therefore,hydroxy-PAHs might be distributed ubiquitously in the environ-ment. In all PM1 samples collected in Bern and Payerne, wedetected three of the five hydroxy-PAHs that we had tested inthe ER-CALUX assay. In a previous study, 1-hydroxypyrene and 2-hydroxyfluorene were detected in ambient particulate matterand concentrations of 6–63 and 4–49 pg m−3, respectively, werefound (Kishikawa et al., 2004). These concentrations are con-sistent with the concentration ranges found for the three hydroxy-PAHs detected in PM1 from Bern and Payerne. The summedE2-CEQ concentrations of 2-hydroxyphenanthrene, 2-hydroxynaph-thalene, and 1-hydroxynaphthalene contributed little, that is, 0.01–0.2%, to the overall estrogenic activity of PM1. However, thesethree hydroxy-PAHs represent only a very small selection of thelarge variety of hydroxylated PAHs possibly present in ambientair. Some additional hydroxy-PAHs with yet unknown structureand estrogenic potency have been detected in atmosphericparticles from both sites. Other compounds must also contri-bute to the observed in vitro estrogenicity of PM1. An abundantclass of compounds found in combustion emissions and ambi-ent air are PAHs (Sidhu et al., 2005; Klein et al., 2006; Heeb et al.,2008). It has been shown that some PAHs and/or their metabo-lites exhibit estrogenic activity (Clemons et al., 1998; Charleset al., 2000; Fertuck et al., 2001; van Lipzig et al., 2005). Hence, PAHsand/or their metabolites may contribute substantially to theobserved in vitro estrogenicity of ambient particulate matter. Inaddition, nitro-PAHs and other substituted PAHs may be involvedin PM1 estrogenicity. This is supported by a preliminary experiment(data not shown) demonstrating that 2-nitrophenanthrene inducedER-mediated gene expression in the ER-CALUX assay. Finally, sub-stances belonging to other classes of compounds may also con-tribute to the overall estrogenic activity, for example, phenols,phthalates or polychlorinated and polybrominated biphenyls(PCBs, PBBs; Sidhu et al., 2005; Klein et al., 2006). Further work isneeded to identify more of the ER ligands present in the atmo-sphere, to determine their environmental concentrations andestrogenic potencies, and to deduce their potential origin.

The range and the temporal trends of estrogenic activity weresimilar at both locations. This is consistent with an earlier study,which reported that PM estrogenicity did not differ betweenurban and rural sites in Canada (Klein et al., 2006). Our resultsindicate that PM1-associated estrogenic compounds were dis-tributed more or less homogenously within the planetary bound-ary layer over the Swiss plateau during the considered samplingperiod. The strong correlations of estrogenic activities (r = 0.92)and PM1 concentrations (r = 0.95) between Bern and Payerne sug-gest that these sites were connected during the hibernal high airpollution episode studied and did not represent independent airquality situations. However, not necessarily the same chemicalsgenerated the overall estrogenic responses observed with theurban and rural PM1 samples.

Our findings do not point to specific local sources of PM1 estro-genicity at the urban or the rural site. Comparisons of estrogenicactivity with concentrations of particulate matter and gaseousair pollutants and with meteorological data did not permit theelucidation of specific emission sources and formation processes

of atmospheric xenoestrogens. Substantial effects of air pressure,temperature, humidity and solar irradiation on PM1 estrogenicitycould not be observed. The examined sampling period and thevariation of these meteorological parameters were possibly toosmall to study the effects of these parameters. Some correlationof estrogenicity and wind velocity was noticed. Estrogenic activitydid not correlate with particulate matter concentrations (PM1,PM10). This might be due to the accumulation of inactive com-pounds (i.e. without activity in the ER-CALUX assay) during thehigh air pollution episode, leading to a dilution of estrogeniccompounds. As expected, concentrations of NO2, NO and COwere higher at the urban site than at the rural site. Thesecombustion-related gaseous pollutants as well as traffic fre-quency did not correlate with PM1 estrogenicity. On the otherhand, a positive correlation was found between SO2 and E2-CEQconcentrations, indicating that combustion processes influenceatmospheric concentrations of particle-bound xenoestrogens.However, several studies reported that diesel exhaust exhibitedestrogenic activity in vitro (Wang et al., 2003; Wenger et al.,2008a) and in vivo (Watanabe and Kurita, 2001). Using the ER-CALUX assay, we measured 550–740 and 1050–1825 pg E2-CEQper m3 of filtered and unfiltered diesel exhaust, respectively, whichincluded semivolatile estrogenic compounds and xenoestrogensassociated with the total particulate fraction of diesel exhaust(Wenger et al., 2008a). These concentrations were 440–26 100times greater than those found in the PM1 extracts of 0.07–1.25 pgE2-CEQ m−3. Hence, diesel combustion contributes in some extentto the overall estrogenic activity observed with ambient particu-late matter. However, other anthropogenic sources must alsobe present. In addition, a secondary formation of estrogeniccompounds in the atmosphere, based on parent compoundsemitted from multiple sources, probably contributes to the totalestrogenic response and may lead to a more homogenous dis-tribution of estrogenic activity in ambient air. In conclusion,further data is needed to study parameters and sources affect-ing the estrogenic potential of ambient particulate matter.

The average daily intake of air ranges from about 2 m3 per dayin very young infants to about 24 m3 per day in adults (Kleinet al., 2006). Based on our in vitro data, a person living in Bern orPayerne would inhale 1–10 pg E2 equivalents per day. In thisestimation only PM1-associated compounds are considered. Con-sequently, the total intake of estrogenic compounds by breathingand swallowing must be higher. To date, the potential biologiceffects of human exposure to atmospheric estrogens remain unclear.However, atmospheric particulate matter contains compounds thatare able to interact with ERs in vitro and potentially also interferewith estrogen-regulated pathways of humans and animals.

Supporting InformationSupporting information can be found in the online version ofthis article.

Acknowledgements

We thank all our collaborators at Empa for providing know-howand PM1 samples, in particular, R. Haag, C. Seiler, C. Zellweger-Fäsi,E. Wettstein, S. Bugmann, T. Bruggisser and V. Lanz. This study wassupported by the board of Empa and the Swiss National ScienceFoundation (SNSF, as a project within the NRP50 program“Endocrine Disruptors — Relevance to Humans, Animals andEcosystems”, grant no. 40504-104378).

D. Wenger et al.

www.interscience.wiley.com/journal/jat Copyright © 2008 John Wiley & Sons, Ltd. J. Appl. Toxicol. 2009; 29: 223–232

232

References Bruggisser N, Bruggisser T, Buchmann B, Bugmann S, Fischer A, Gehrig R,

Graf P, Hill M, Hueglin C, Nyffeler U, Reimann S, Schwarzenbach B,Seitz T, Steinbacher M, Weber R, Wettstein E, Zellweger C. 2007.Technischer Bericht zum Nationalen Beobachtungsnetz für Luftfremdstoffe(NABEL). Dübendorf, Switzerland: Empa, Swiss Federal Laboratories forMaterials Testing and Research. Available at: http://www.empa.ch/nabel.

Charles GD, Bartels MJ, Zacharewski TR, Gollapudi BB, Freshour NL, CarneyEW. 2000. Activity of benzo[a]pyrene and its hydroxylated metabolitesin an estrogen receptor-α reporter gene assay. Toxicol. Sci. 55: 320–326.

Clemons JH, Allan LM, Marvin CH, Wu Z, McCarry BE, Bryant DW,Zacharewski TR. 1998. Evidence of estrogen- and TCDD-like activitiesin crude and fractionated extracts of PM10 air particulate materialusing in vitro gene expression assays. Environ. Sci. Technol. 32: 1853–1860.

Damstra T, Barlow S, Bergman A, Kavlock R, Van Der Kraak G. 2002.Global Assessment of the State-of-the-science of Endocrine Disruptors.World Health Organization (WHO), International Labour Organisation(ILO), United Nations Environment Programme (UNEP). Available at: http://www.who.int/ipcs/publications/new_issues/endocrine_disruptors/en/.

De Rosa M, Zarrilli S, Paesano L, Carbone U, Boggia B, Petretta M, MaistoA, Cimmino F, Puca G, Colao A, Lombardi G. 2003. Traffic pollutantsaffect fertility in men. Hum. Reprod. 18: 1055–1061.

Delfino RJ, Sioutas C, Malik S. 2005. Potential role of ultrafine particlesin associations between airborne particle mass and cardiovascularhealth. Environ. Health Perspect. 113: 934–946.

Fertuck KC, Matthews JB, Zacharewski TR. 2001. Hydroxylated benzo[a]pyrenemetabolites are responsible for in vitro estrogen receptor-mediatedgene expression induced by benzo[a]pyrene, but do not elicit uterotrophiceffects in vivo. Toxicol. Sci. 59: 231–240.

Gehm BD, Levenson AS, Liu H, Lee E-J, Amundsen BM, Cushman M,Jordan VC, Jameson JL. 2004. Estrogenic effects of resveratrol in breastcancer cells expressing mutant and wild-type estrogen receptors: roleof AF-1 and AF-2. Steroid Biochem. Mol. Biol. 88: 223–234.

Hayakawa K, Onoda Y, Tachikawa C, Hosoi S, Yoshita M, Chung SW, Kizu R,Toriba A, Kameda T, Tang N. 2007. Estrogenic/antiestrogenic activitiesof polycyclic aromatic hydrocarbons and their monohydroxylatedderivatives by yeast two-hybrid assay. J. Health Sci. 53: 562–570.

Heeb NV, Schmid P, Kohler M, Gujer E, Zennegg M, Wenger D, Wichser A,Ulrich A, Gfeller U, Honegger P, Zeyer K, Emmenegger L, PetermannJ-L, Czerwinski J, Mosimann T, Kasper M, Mayer A. 2008. Secondaryeffects of catalytic diesel particulate filters: Conversion of PAHs versusformation of nitro-PAHs. Environ. Sci. Technol. 42: 3773–3779.

Howell A, Osborne CK, Morris C, Wakeling AE. 2000. ICI 182,780(Faslodex™): Development of a novel, ‘pure’ antiestrogen. Cancer 89:817–825.

Kamiya M, Toriba A, Onoda Y, Kizu R, Hayakawa K. 2005. Evaluation ofestrogenic activities of hydroxylated polycyclic aromatic hydrocarbonsin cigarette smoke condensate. Food Chem. Toxicol. 43: 1017–1027.

Katzenellenbogen BS, Choi I, Delage-Mourroux R, Ediger TR, Martini PGV,Montano M, Sun J, Weis K, Katzenellenbogen JA. 2000. Molecularmechanism of estrogen action: selective ligands and receptorpharmacology. J. Steroid Biochem. Mol. Biol. 74: 279–285.

Kishikawa N, Morita S, Wada M, Ohba Y, Nakashima K, Kuroda N. 2004.Determination of hydroxylated polycyclic aromatic hydrocarbonsin airborne particulates by high-performance liquid chromatographywith fluorescence detection. Anal. Sci. 20: 129–132.

Klein GP, Hodge EM, Diamond ML, Yip A, Dann T, Stern G, Denison MS,Harper PA. 2006. Gas-phase ambient air contaminants exhibit significantdioxin-like and estrogen-like activity in vitro. Environ. Health Perspect.114: 697–703.

Lanz VA, Alfarra MR, Baltensberger U, Buchmann B, Hueglin C, Szidat S,Wehrli MN, Wacker L, Weimer S, Caseiro A, Puxbaum H, Prévôt ASH.2008. Source attribution of submicron organic aerosols during wintertimeinversions by advanced factor analysis of aerosol mass spectra. Environ.Sci. Technol. 42: 214–220.

Legler J, van den Brink CE, Brouwer A, Murk AJ, Van der Saag PT, Vethaak DA,Van der Burg B. 1999. Development of a stably transfected estrogenreceptor-mediated luciferase reporter gene assay in the human T47Dbreast cancer cell line. Toxicol. Sci. 48: 55–66.

Nilsson S, Mäkelä S, Treuter E, Tujague M, Thomsen J, Andersson G,Enmark E, Pettersson K, Warner M, Gustafsson J-Å. 2001. Mechanismsof estrogen action. Physiol. Rev. 81: 1535–1565.

Pope CA, Burnett RT, Thurston GD, Thun MJ, Calle EE, Krewski D,Godleski JJ. 2004. Cardiovascular mortality and long-term exposureto particulate air pollution: epidemiological evidence of generalpathophysiological pathways of disease. Circulation 109: 71–77.

Putaud J-P, Raes F, Van Dingenen R, Brüggemann E, Facchini M-C,Decesari S, Fuzzi S, Gehrig R, Hüglin C, Laj P, Lorbeer G, Maenhaut W,Mihalopoulos N, Müller K, Querol X, Rodriguez S, Schneider J, SpindlerG, ten Brink H, Tørseth K, Wiedensohler A. 2004. A European aerosolphenomenology — 2: chemical characteristics of particulate matter atkerbside, urban, rural and background sites in Europe. Atmos. Environ.38: 2579–2595.

Sidhu S, Gullett B, Striebich R, Klosterman J, Contreras J, DeVito M. 2005.Endocrine disrupting chemical emissions from combustion sources:diesel particulate emissions and domestic waste open burn emissions.Atmos. Environ. 39: 801–811.

Sioutas C, Delfino RJ, Singh M. 2005. Exposure assessment for atmosphericultrafine particles (UFPs) and implications in epidemiologic research.Environ. Health Perspect. 113: 947–955.

Šrám RJ, Binková B, Dejmek J, Bobak M. 2005. Ambient air pollution andpregnancy outcomes: A review of the literature. Environ. HealthPerspect. 113: 375–382.

van Lipzig MMH, Vermeulen NPE, Gusinu R, Legler J, Frank H, Seidel A,Meerman JHN. 2005. Formation of estrogenic metabolites ofbenzo[a]pyrene and chrysene by cytochrome P450 activity and theircombined and supra-maximal estrogenic activity. Environ. Toxicol.Pharmacol. 19: 41–55.

Villeneuve DL, Blankenship AL, Giesy JP. 2000. Derivation and applicationof relative potency estimates based on in vitro bioassay results.Environ. Toxicol. Chem. 19: 2835–2843.

Vione D, Barra S, De Gennaro G, De Rienzo M, Gilardoni S, Perrone MG,Pozzoli L. 2004. Polycyclic aromatic hydrocarbons in the atmosphere:monitoring, sources, sinks and fate. II: Sinks and fate. Ann. Chim.(Rome) 94: 17–32.

Wang J, Wu W, Henkelmann B, You L, Kettrup A, Schramm KW. 2003.Presence of estrogenic activity from emission of fossil fuel combustionas detected by a recombinant yeast bioassay. Atmos. Environ. 37:3225–3235.

Wang J, Xie P, Xu Y, Kettrup A, Schramm K-W. 2004. Differing estrogenactivities in the organic phase of air particulate matter collectedduring sunny and foggy weather in a Chinese city detected by arecombinant yeast bioassay. Atmos. Environ. 38: 6157–6166.

Watanabe N, Kurita M. 2001. The masculinization of the fetus duringpregnancy due to inhalation of diesel exhaust. Environ. HealthPerspect. 109: 111–119.

Wenger D, Gerecke AC, Heeb NV, Naegeli H, Zenobi R. 2008a. Catalyticdiesel particulate filters reduce the in vitro estrogenic activity of dieselexhaust. Anal. Bioanal. Chem. 390: 2021–2029.

Wenger D, Gerecke AC, Heeb NV, Zennegg M, Kohler M, Naegeli H,Zenobi R. 2008b. Secondary effects of catalytic diesel particulatefilters: reduced aryl hydrocarbon receptor-mediated activity of theexhaust. Environ. Sci. Technol. 42: 2992–2998.