Toxic organic contaminants in airborne particles responsible for negative health effects

: Several studies indicate a strong correlation between airborne particulate matter (PM) and adverse health outcomes, as a function of particle size. Traditionally, the study of PM has focused on particles less than 10 or 2.5 μm in aerodynamic diameter (Da); however, recent evidence links particles less than 0.1 μm in Da and cardiovascular disease. In this study, we monitored 66 organic contaminants in size-fractionated PM samples collected in outdoor environments. The data obtained were analyzed with Principal Component Analysis (PCA), which showed a correlation be-tween sites and classes of pollutants, allowing the identification of characteristic emission sources of each monitored site.


Introduction
Particulate matter regional deposition in human respiratory tract is primarily based on the particle aerodynamic size [1].In particular, ultrafine particles (UFs, less than 0.1 μm in aerodynamic diameter) may penetrate blood circulation, reach the inner organs [2][3], and distribute the toxic substances adsorbed on their surface.UFs have a greater surface area to mass ratio compared to larger particles, providing a larger area to adsorb potentially toxic chemicals [4].Both the size and the composition of the airborne particles are closely related to the sources.In this optic, the study of ultrafine, fine, and coarse trace organic compositions can enhance the identification of different emission sources, characteristic of specific environments, and provide important insight into their contribution to adverse human health effects.Receptor models, applied to organic chemicals extracted from particles of different size collected in specific areas, can contribute to the identification of possible characteristic sources [5].Among the receptor models, chemical mass balance, [6] positive matrix factorization [7] and principal component analysis (PCA) [5][6][7][8] are widely used to obtain source apportionment results.The aim of the present study was to identify characteristic or common emission sources in three workplaces and in an urban outdoor area.We applied the PCA to a data set, including 66 organic contaminant concentrations, analyzed in coarse (C), fine (F) and ultrafine (UF) particles, collected in the three workplaces and in the urban outdoor area.The concentrations were reported in ng per mg of PM sampled.The compounds analyzed belonged to six different groups of organic contaminants: polycyclic aromatic hydrocarbons (PAHs) and their derivatives (nitro-PAHs and oxy-PAHs), polychlorinated biphenyls (PCBs), polybrominated diphenyl ethers (PBDEs) and Novel Brominated flame-retardants (N-BFRs), selected for their toxicity and persistence in the environment.

Sampling
PM at different sizes was collected in four areas located in central Italy, in the same period of autumn 2017, to prevent weather conditions from affecting distribution of PM in outdoor atmospheres [9].Particle collection was performed: in an outdoor urban area (RM); in an outdoor wastewater treatment plant (WWTP), during different phases of the process; in two different sites within an intensive livestock farming activity, nearby a shed used as stable (COW) and close to a shed for the grinding, mixing, and storing of feeds (FEED).

Analytical Methods
The analytical procedure was extensively described in a previous paper [10].Briefly, filters were extracted by an accelerated solvent extractor ASE 200 (Dionex Sunnyvale, CA, USA) with DCM/n-Hexane (1:1) (two cycles) followed by ethyl acetate (two cycles) at 100°C and 1500 psi.The extracts were evaporated and re-dissolved with 50 μL of toluene until analytical determinations.
The analysis of PCBs, BDEs, NBFRs, PAHs, and PAH derivatives (oxy-PAHs and nitro-PAHs) were performed by gaschromatography-mass spectrometry (GC-MS).An HP 7890-B gas chromatograph fitted with an HP 7693 autosampler and coupled with an HP 5977B single quadrupole mass-selective detector (Agilent Technologies, Palo Alto, CA) was used for GC-MS analysis both in electronic ionization and in negative chemical ionization.Selective ion monitoring was used in both MS configurations.The injector temperature was set at 280 °C and the samples (1 μL) were injected in splitless mode.The helium carrier gas was set at a constant flow rate of 1.0 mL/min.Analytical methods were described in detail in previous studies [10][11][12][13][14].

Organic compound concentrations
The method of standard addition, to minimize matrix effects, was used for the quantitative analysis of the 66 organic compounds.Each of the three extracts from ultrafine, fine and coarse particles were split in five aliquots and added with multistandard solutions of all analytes at growing concentrations.The absolute value of the x-intercept corresponded to the unknown concentration of the analyte.The concentrations of compounds belonging to the same group were added up to have one result for each class of organic species in each of the three fractions.Figure 1 shows the distribution of all groups of organic substances in the three fractions of PM in each of the four sampling site.All classes of compounds have a higher concentration per unit of PM at the FEED site in all the three particulate fractions; as regards UF, F and C fractions, the distribution of organic substances is proportional, but the finer is the fraction, the more concentrated are the organic substances, as can be seen by the different y-axis scales [15].As expected, PAHs and derivatives contribute more significantly to the organic composition of particles, than other classes of compounds, due to a higher number of emission sources [16].In particular, photochemical production is the source of the majority of nitro-and oxy-PAHs, as evidenced by the higher concentrations of those compounds produced in the atmosphere as secondary pollutants: 9-2-and 3-nitro-phenanthrenes and 9-fluorenone in all the PM fractions (data not shown).

Principal component analysis (PCA)
Multivariate statistical computations were performed using the statistical software R (R-project for statistical computing, Ver.3.0, 32-bit).A principal component analysis (PCA) was calculated on the matrix of data of figure 1, including the chemical concentrations (ng per mg of PM) of 14 PAHs, 9 Ox-PAHs, 10 Nitro-PAHs, 15 PCB, 7 PBDE and 11 NBFR, in the three PM dimensional fractions (UF<1, F>1 and C<10 μm), in the four different sampling sites.The data matrix was transformed by performing row and column autoscaling to overcome the scale variations between the examined variables [17].Two principal component analyses were run on all compounds, and on the individual classes of contaminants (PAHs and derivatives, PCBs, PBDEs and NBFRs).
In Appendix A, Table A1 lists compounds and site codes used in PCA development.

All organic compounds
Figure 2 shows the loading plot and the score plot for all target compounds in all sampling sites and in the three PM dimensional fractions.The two main components PC1 and PC2 explain 85.4% of total variance.The score plot highlights a well-defined cluster, including WWTP F<1 and UF<0.1 and RM UF<0.1, that shows a significant correlation with the group of compounds positioned at the bottom of the loading plot (blue circles).This area corresponds to data relating to some PAHs and derivatives characteristic of anthropogenic sources.Furthermore, FEED F<1 and UF<0.1 also seem to be influenced by specific contamination sources, even if different from those identified in WWTP and RM sites.However, it is not possible to find any correlation with specific compounds as, in the upper part of the loading plot, the compounds are unevenly distributed and no cluster is shown.
More information can be obtained by considering the PCA carried out on each of the three macro-classes of compounds (PAHs and derivatives, PCBs, PBDEs and NBFRs).

Individual classes of organic compounds
Figure 3 shows loading and score plots for each single class of compounds.Plots relating to PAHs and derivatives (figure 3A) highlight a significant correlation among WWTP F<1, UF<0.1, RM UF<0.1 and the high molecular weight PAHs (HMW-PAHs), some NO-and Oxy-PAHs (blue circles).As HMWPAHs have a pyrolytic origin, this result suggests combustion as main source of contamination in WWTP and RM.The NO-PAHs that most affect the sites are 7NBaA and 9NA, both of primary origin and due to direct emissions from combustion processes, and 2NF of secondary origin by photochemical mechanisms.The good correlation with 9,10-Phenanthrenequinone, 1,9-Benz-10-anthrone and BaA-7,12-dione, generated by both combustion processes and reactions in atmosphere, further supports the hypothesis of combustion as predominant source of pollution for WWTP and RM in the fine and ultrafine fractions of PM [12][13].Lastly, the PCA, carried out with PAH and derivative data, also highlights that WWTP and RM sites are influenced by similar emission sources in agreement with their location.
PCBs show a good correlation with feed site for fine and ultrafine PM fraction (Figure 3B).FEED F<1 correlates with 5 dioxin-like PCBs (126, 156, 157, 167 and 169) (purple square), while FEED UF<0.1 is influenced by others PCBs including also PCB28 and 138 (red circles).These two compounds, with PCB52, 101, 153 and 180 are considered indicators of contamination in food and feed and their sum includes about half of the total amount of no dioxin-like PCBs detected in these matrices [18].WWTP and RM sites do not seem to be influenced by PCB sources.PBDE and NBFR classes show a significant correlation with feed site for fine PM fraction (figure 3C).A well-defined cluster is highlighted in the upper part of the loading plot (green circles) related to the FEED<1.The compounds which characterize this group are BDE28 and 49, PBEBs and TBPs in agreement with Fernandes et al. [19] that have detected the presence of PBDEs and some NBFRs in animal feed at non-negligible concentrations.As well as for PCBs, WWTP and RM sites do not show any correlation with PBDE and NBFR.

Conclusions
Because epidemiological studies mainly link the carcinogenic effects of exposure to UF, F and coarse PM emissions to the organic fraction, samplings and organic speciation for the three fractions of aerosols in three working environments and in an urban one were investigated and submitted to principal component analysis.In this study, the results obtained showed that the finer is the PM fraction, the more concentrated are the organic substances.From PCA analysis, specific correlations between the two sites WWTP and RM show WWTP mainly affected by urban sources.On the contrary, FEED site seems to be influenced by specific contamination sources, different from those identified in all the other sites, causing extremely high concentrations of organic contaminants.Conflicts of Interest: "The authors declare no conflict of interest.""The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results".
Appendix A Publisher's Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.Copyright: © 2020 by the authors.Submitted for possible open access publication under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses /by/4.0/).

Figure 1 .
Figure 1.Distribution (ng per mg of PM) of all groups of organic substances in ultrafine, fine and coarse fractions of PM.

Figure 2 .
Figure 2. Loading plot and the score plot for all target compounds in all sampling sites and in the three PM dimensional fractions.

Figure 3 .
Figure 3. Loading and score plots of PCA carried out on the three classes of compounds: PAHs and derivatives (A), PCBs (B), PBDE and NBFR (C).

Table A1 .
Organic compound list and codes used in PCA development.