Comparison of the Chemical Characteristics and Toxicity of PM2.5 Collected Using Different Sizes of Cyclones

Cyclone sampling devices have been helpful in assessing the toxic effects of fine particulate matter (PM2.5). The particle collection efficiency of sampling devices is critical. This study investigated the effect of cyclone size on particle size, chemical composition, and particle toxicity. Three cyclones with different inner diameters (12–68 mm) were tested for penetration using an aerodynamic particle sizer, fluorescent polystyrene latex, and a differential mobility analyzer. The elemental and water-soluble ion compositions of the particles collected by different cyclones were compared. An evaluation of the particles’ toxicity was conducted by comparing the results of dithiothreitol (DTT), limulus amebocyte lysate (LAL), and cell exposure assays. The experimental evaluation showed a 50% cut-size of the cyclones between 0.17–0.28, 0.34–0.36, and 0.70 μm for the small, medium, and large cyclones, respectively. To collect PM2.5 and evaluate separation performance in the real environment, the small and large cyclones were selected according to the particle penetration and flow rate. A comparison of chemical composition and enrichment factor values found that the particles in the small cyclone samples contained smaller and more anthropogenic sources than those in the large cyclone samples. The oxidative potential (OP) measured by the DTT assay of the samples collected using the small and large cyclones differed across sampling periods and associated with the transition metals. The viability of human epithelial A549 cells after exposure to the collected particles using the cyclones was different across sampling periods and associated with OP. The endotoxin concentrations measured in the LAL assay were found only in the large cyclone samples; they affected the estimated level of cytokine based on IL(interleukin)-6 release from human leukemia monocytic (THP-1) cells derived macro-phage-like cells. Regardless of the size, the cyclone techniques used in this study to collect aerosol particles would be a powerful tool for a detailed evaluation of particle toxicity.


INTRODUCTION
Epidemiological studies have linked the inhalation of fine particulate matter (PM 2.5 ) with increased respiratory (Harkema et al., 2004) and cardiovascular (Lippmann, 2014;Pope et al., 2004) diseases that can cause premature death (And-erson et al., 2012).Major ongoing studies of atmospheric aerosol particles have investigated physicochemical characteristics and toxicity using acellular and cellular assays (in vitro and in vivo) to identify the origins of ambient particles and assess their effect on the environment and human health (Patel and Rastogi, 2021;Lin and Yu, 2019;Fang et al., 2016;Diaz et al., 2011;Godleski et al., 2011).
In significant studies of toxicological or chemical assays of ambient air, filter-based sampling is commonly used to collect PM 2.5 samples because of its high collection efficiency (Honda et al., 2017;Roper et al., 2015;Yang et al., 2014).However, the filter-based method requires crude particle extraction when the particle is attached to the filter, which may lead to a loss of PM components and alter particle toxicity (Chowdhury et al., 2018;Roper et al., 2018;Van Winkle et al., 2015;McWhinney et al., 2013;Pan et al., 2004).Moreover, the levels of the components in the filter samples can be changed by unexpected artifacts due to the adsorption of gases, chemical reactions, volatile compounds, and other factors in the ambient air (Hasegawa, 2016;Kameda et al., 2010).
To avoid the issues related to filter sampling, researchers have investigated cyclone devices to collect PM samples globally (Honda et al., 2021;Youn et al., 2021;Du et al., 2020;Okuda et al., 2015;Rule et al., 2010;Gussman et al., 2002;Vanderpool et al., 2001).Exposure to cyclone-collected particles may cause stronger biological responses than those observed after exposure to extracted particles from filter materials and induce lung inflammation (Honda et al., 2021).A powder sample can be directly subjected to in vitro and in vivo assays exposed to cells and experimental animals without losing of components in the extraction process or unexpected artifacts during sample collection (Ko et al., 2021;Sagawa et al., 2021;Chowdhury et al., 2019;Ogino et al., 2019Ogino et al., , 2017;;Onishi et al., 2018;Okuda and Isobe, 2017;Okuda et al., 2015;Rule et al., 2010).
In general, cyclone efficiency for ambient aerosols is enhanced by the gas inlet velocity and cyclone dimensions, with the 50% cut size of a cyclone (Dp 50 ) decreasing as the particle collection efficiency increases (Cooper and Forrest, 2010;Wang et al., 2004).This presents that as the cyclone efficiency increases, the abundance (mass per unit weight) of larger particles in PM decreases and vice versa.In recent years, the toxic effects of airborne particles have been strongly linked with distribution of size and their compositions because the peak of the mass concentration and toxicity of each particle in the ambient air are found at different particle diameters (Fang et al., 2017;Fujitani et al., 2017;Voutsa et al., 2015).This emphasizes the importance of particle size distribution in assessing the adverse effects of airborne aerosols on human health, but also the relevance of PM compositions that respond with highly toxic potential (e.g., transition metals, carbonaceous materials, and bioaerosols) (Faraji et al., 2018;Fujitani et al., 2017;Vuong et al., 2017;Voutsa et al., 2015;Charrier and Anastasio, 2012;Schins et al., 2004).Thus, it is critical to focus on the interdependence of the inlet velocity and cyclone dimensions when sampling for particle toxicity analysis.Previous studies have shown almost no data on impacts of cyclone size and inlet velocity on particle toxicity analyses (Youn et al., 2021;Du et al., 2020;Lin et al., 2018;Peters et al., 2001;Chen et al., 1999).In this work, we have investigated to continue our previous research and contribute to achieve more reliable data when using cyclones in particle toxicity assessment to evaluate of health effects (Kurihara et al., 2022;Honda et al., 2021;Ogino et al., 2019;Onishi et al., 2018;Okuda et al., 2007).
This study aimed to evaluate the separation performance of different cyclone sizes and impacts of them on the chemical compositions of the samples and toxicity analyses.Three cyclones (large, medium, and small) were tested for penetration using an aerodynamic particle sizer, fluorescent polystyrene latex, and a differential mobility analyzer to investigate the optimal cyclone size for a reliable evaluation of the health effects of PM 2.5 .Two cyclones (small and large) with different separation efficiency were selected and simultaneously collected PM 2.5 samples throughout the year.The ion species of the particles collected using the two cyclones were compared to evaluate the separation characteristics of the cyclones.For the detailed comparison of the samples, elemental analysis, oxidative potential (OP), endotoxin analyses, and cell exposure were conducted primarily to evaluate the relationship between the particle toxicity and separation performance of the cyclones.

1 Evaluation of the Separation Characteristics of
Cyclones Using Particle Penetration Measurement The particle size distribution measurement of the cyclone sampling devices, expressed using the equation for particle penetration (Okuda et al., 2015), is a significant factor in assessing the health hazards of PM (Grahame and Schlesinger 2010;McCreanor et al., 2007).This is because the particles of major health-hazardous particles, such as transition and heavy metals and watersoluble and insoluble fractions are contained in fine region of typical atmospheric aerosols (De Kok et al., 2006).The increase in flow rate decreases the particle penetration through cyclones (Youn et al., 2021).Therefore, penetration measurement of the cyclones is important to assess their performance when the flow rate increases.
The cyclones and their dimensions, and the pump models in the sampling system are shown in Table 1.The instruments included large (HVS3, CS3 Inc., Sandpoint, ID), medium (URG-2000-30EH, URG), and small (URG-2000-30EHB, URG) cyclones.Individual cyclones were tested by the following three experiments for measuring particle penetration: measurement of particle size distribution using aerodynamic particle sizer (APS), particle number counts of each monodisperse particle classified using differential mobility analyzer (DMA), and measurement of fluorescence intensity using fluorescent monodisperse polystyrene latex (PSL) particles with and without a cyclone.
The flow rates for the cyclones were designed experimentally according to the pumps to investigate the impacts of cyclone size on the collection efficiency and chemical compositions, without getting greater pressure loss.In order to determine the inlet velocity, the selection of a pump is critical.From engineering point of view, it is useful any strong pump to overcome high pres-sure drop of cyclone.From economical point of view, the cost of pump is important.From the environmental and analytical point of view, the use of an oil-free pump is required to prevent sample contamination (Cooper and Forrest, 2010;Wang et al., 2004).In this study, we chose the pump (Table 1) according to economical, environmental and analytical points of view, in particular, the environmental aspects.Therefore, according to economic and environmental points of view, an affordable oilfree pumps were chosen and operated them in the maximum power in order to obtain the particulate samples as much as possible.To measure particle penetration through the cyclones, the airflow rates were set at 100 L/ min (small), 150 L/min (medium), and 1,200 L/min (large).

1. 1 Particle Penetration of the Cyclones
Measured Using APS An APS spectrometer (Model 3321, TSI Inc., USA) measured the aerosol size distributions in the ambient air from 0.5 to 20 μm and analyzes the time-of-flight of individual particles in an accelerating flow field (Volckens and Peters, 2005;Peters and Leith, 2003).The experimental system for measuring the separation characteristics of a cyclone using the APS is shown in Fig. 1.The particle size distribution in the air flow before and after the cyclones were measured using the APS.This measurement was conducted six times for 6 min at each flow rate of the cyclones.the penetration of particles with a size of 0.5 μm or smaller through the cyclones.The system for measuring the cyclone's separation characteristics using fluorescent PSL particles (Polyscience Inc., Warrington, PA, USA) is shown in Fig. 2. The fluorescent PSL particles were supplied by an atomizer and collected with a filter for 20 min, each with and without a cyclone.The filter sample without a cyclone was designated as a reference filter, and the filter sample after the cyclone was designated as a backup filter (BF).The flow rate for the cyclones and reference filters was measured using a mass flow meter and set to the same value before each test.The flow rate of the introduced clean air was carefully adjusted that the flow of the air contained the PSL particles was 2 L/min, with and without each cyclone.Moreover, we confirmed that the number concentration of the generated PSL particles did not differ significantly with or without cyclones using a condensation particle counter (CPC, Model 3750).Fluorescent particles of 0.1, 0.2, 0.5, and 1.0 μm were supplied.The fluorescent PSL particle solution in the atomizer was diluted 20 times with ultra-pure water from the fluorescent solution of PSL to make it 0.05 mL/mL.The filters collected the PSL particles dissolved in ethyl acetate and shaken at 150 rpm for 120 min.After that, the solution was filtered through a quartz fiber filter to eliminate impurities, and the obtained solution was measured using a spectrofluorophotometer (RF-6000, Shimadzu Corporation, Japan).The fluorescent PSL particle concentration was measured at 441 nm of the excitation wavelength, 470-510 nm of fluorescence wavelength, 600 nm/min scan speed, 1.0 nm of the data interval, auto sensitivity, 5.0 nm of spectral bandwidth excitation side, and 5.0 nm of fluorescence side.

1. 3 Particle Penetration of Cyclones Measured
Using the DMA System Fig. 3 shows the DMA system (DMA-5180, SHIBA-TA) for measuring particle penetration through the cyclones.An Am-241 neutralizer between the diffusion dryer and the DMA was also used.Test particles were generated using an atomizer using 4% ammonium sulfate solution.The flow rate of the introduced clean air is the same as in the above method using the fluorescent PSL particles.The particles were classified in order by the DMA in a size range from 100 nm to 500 nm in diameter.The classified monodisperse particles with one diameter were measured the number concentration by CPC at front of and back of the cyclone to calculate the penetration rate.In addition, we used an  isokinetic sampling probe to avoid particle separation by the small flow rate of the CPC against that of pump.

2 Parallel Field Sampling Using the Large and
Small Cyclones Total ten sets of PM 2.5 were simultaneously collected from 2 April 2020 to 14 April 2021 using the small cyclone (URG-2000-30EHB) and the large cyclone (HVS3) on the rooftop at the Yagami campus of Keio University, Yokohama, Japan.Detailed information about the sampling site was provided in a previous study (Honda et al., 2021).Both sampling cyclones were selected as having the most productive flow rate and separation performance.A schematic diagram of the experimental cyclone sampling system of PM 2.5 samples is shown in Fig. 4. The flow rates of the small and large cyclones were 100 L/min and 1200 L/min, respectively.The samples were set between 21 and 60 days depending on the atmospheric concentration of PM 2.5 to collect adequate amount of samples for detailed analysis.

3 Sample Analysis
The samples were analyzed to determine total watersoluble ions using ion chromatography as described in the following section.The ion species were compared to evaluate the separation characteristics of the small and large cyclones for individual elements, because the total ions are the predominant chemical components of the ambient aerosol particles (Okuda et al., 2018;Okuda et al., 2015).Moreover, the ion components essentially comprise the secondary aerosols formed during warm and cold periods (Shen et al., 2008;Song et al., 2006).Two sets of PM 2.5 samples out of all the samples collected between 10 December 2020 and 6 January 2021 and 10 March and 14 April 2021 were chosen for detailed analysis (elemental and water-soluble ion analyses, dithiothreitol (DTT) and limulus amebocyte lysate (LAL) assays, and cytotoxicity assays using human epithelial A549 cells and macrophage-like cells).The ion contents of the two samples were different enough to evaluate the effects of composition on the detailed toxicity analysis.

4 Water-Soluble Ion Analysis
The powder sample weighed about 2 mg was extracted using ultra-pure water.The sample solution was analyzed using an ion chromatography (IC) system (Dionex AS-AP, ICS-2100 and ICS-1100, Thermo Fisher Scientific Inc., USA) for anion species (Cl -, NO 3 ) and cation species (Na + , NH 4  750, 6200, 4437, 323, 260, 154, 229, and 463 ppmw, respectively; the LOD value was calculated with standard deviation multiplied by five of the replication analyses of standard solution.The precision (coefficient of variation) for five times repetition analysis of standard solution was less than 7%.

5 Detailed Comparison of the Particles
Collected by the Large and Small Cyclones Using Chemical Composition and Toxicity Analysis

5. 1 Elemental Analysis
Inductively Coupled Plasma-Mass Spectrometry (ICP-MS, iCAP TQ, Thermo Fisher Scientific) was used to analyze the samples in the Atmospheric Environment Innovation Center at Seikan Kensa Center Inc. for ICP-MS analysis.The powder sample was weighed at about 1 mg and decomposed in 5 mL nitric acid, 1 mL hydrogen peroxide, and 2 mL hydrofluoric acid with a microwave decomposition device.The decomposed liquid was transferred to a 50 mL centrifuge tube, and the volume was adjusted to 50 mL.Finally, the sample solutions were analyzed.The quantification of metals (e.g., Al, Ti, V, Cr, Mn, Fe, Cu, Zn, Ni, Pb) was performed.The standard solution was diluted 10, 100, and 1000 times with a mixed solution of 5% nitric acid and 1% hydrofluoric acid to drive for an appropriate concentration for the target element.The LOD for Al, Ti, V, Cr, Mn, Fe, Cu, Zn, Ni, and Pb was 40, 1.9, 0.1, 0.4, 0.4, 5.0, 0.8, 1.4, 0.26, and 0.3 ppmw, respectively.Recovery rate of the target elements ranged from 87% to 97%, measured using standard reference materials (CRM).

5. 2 Identification of Metal Emission Sources
Enrichment factors (EFs) were used to assess the contribution of anthropogenic emissions to the atmosphere for each metal in PM 2.5 ( Jing et al., 2022;Ny and Lee, 2010).The reference element is often taken to be Al and Fe (for terrestrial sources) and Na (for oceanic sources) (Ny and Lee, 2010;Zoller et al., 1974;Chester and Stoner, 1973).Moreover, Ti and Mn have been used successfully (Jing et al., 2022;Obiajunwa et al., 2002;Loska et al., 1997).In this study, we avoid using of Al or Fe as reference elements because the small and large cyclones are made of aluminium and stainless steel, respectively.We focused on assessing the enrichment levels of V, Cr, Mn, Fe, Cu, Zn, Ni, and Pb in samples because the transition metals mainly come from anthropogenic activities (Chen et al., 2010;Okuda et al., 2007;Song et al., 2006).Therefore, Ti was used as the reference element in the following equation to avoid Al and Fe, which may be contaminated by cyclones, and to focus on heavy metals and transition metals: where C i and C Ti refer to the abundance of each element and Ti in PM 2.5 and the Earth's crust.The average crustal concentrations were obtained from a study by Taylor (1964).

5. 3 DTT Assay
The reactive oxygen species (ROS) generation ability (or OP) of PM is one of the comprehensive and realistic toxicity indicators which reflects physicochemical properties (e.g., size, surface area, chemical composition) of PM.Redox-active particles deplete antioxidants, damag-ing biomolecules, and leading to acute and chronic diseases (Gao et al., 2020;Delfino et al., 2011).In particular, the dithiothreitol-measured OP of aerosol particles has been extensively used worldwide as a suitable indicator for evaluating the oxidation properties of ambient particles (Wang et al., 2019).First, to measure the OP of PM 2.5 , a phosphate buffer (0.1 M Na 2 HPO 4 + 0.1 M KH 2 PO 4 ) at pH 7.4 was treated with Chelex resin to remove trace metals used as a solvent (Charrier and Anastasio, 2012).Then, the powders of DTT and 5,5-dithiobis-(2-nitrobenzoic acid) (DTNB, FUJIFILM Wako Pure Chemical Corporation, Japan) were dissolved in the phosphate buffer to prepare 1.0 mM DTT and 6.25 mM DTNB solutions.The powder samples weighed about 2 mg were sonicated in 19 mL of the phosphate buffer for 15 min at a temperature <15°C.Next, 1.9 mL of each sample solution was transferred to a microtube, and 2 μL of Tween-20 solution (10% in H2O, BioVision, Inc., USA) was added.The reaction continued at 37°C for 30 min in a Thermo block (ND-M01, Nissin, Japan).A total of 100 μL of the 1.0 mM DTT solution was added to the microtube.After 6 min of the start of the reaction with the DTT solution, 20 μL of the DTNB solution was added quickly to develop color (the reaction stopped there) and put in a refrigerator.More detailed information about the preparation of the standard solution and measurement absorbance is given in the previous study (Kurihara et al., 2022).The precision of the DTT assay was checked using the reference material CRM#28.The sample's DTT consumption rate (μM/mg/min) was calculated as the regression slope between DTT concentration and time.LOD was 0.09 μM/mg/min; the LOD value was calculated on the standard solution curves set using a linear regression model (Piehler et al., 2020).

5. 4 LAL Assay
Endotoxin (a component of the exterior cell membrane of Gram-negative bacteria) is a group of toxic and proinflammatory compounds of lipopolysaccharide (LPS) responsible for releasing and activating cytokine production (Carty et al., 2003;Heinrich et al., 2003).A certain type of bacteria causes symptoms involving airway inflammation, fever, and lung function decline due to their strong immune-modulating properties (Heinrich et al., 2003;Reed and Milton, 2001).Endotoxins were assayed using the quantitative kinetic chromogenic LAL method.Before experiments, all glassware such as triangular flasks, test tubes with aluminum caps, and reservoirs were steril-ized at 250°C for 2 hours.First, a 0.05% Tween solution was prepared using Tween-20 (10% in H2O, BioVision, Inc., USA) and endotoxin-free distilled water/LAL reagent water (Otsuka Pharmaceutical Factory, Inc., Tokushima, Japan).Then, 2 mg of each powder sample was dissolved in 4 mL 0.05% Tween solution at room temperature using a horizontal shaker for 1 hour and sonicated at low temperature for 15 min.An endotoxin standard solution was prepared at 7 concentrations (1, 0.5, 0.25, 0.1, 0.05, 0.025, 0.0125 EU mL -1 ) using a control standard endotoxin (Associates of Cape Cod. Inc., MA, USA).After 7 min of centrifugation at 5000 rpm, 50 μL of supernatant per sample, blank (LAL reagent water), and standard endotoxin solution was transferred to a 96-well microplate.Then 50 μL of a lysate reagent (Pyro Color-MP: Chromogenic Diazo-Coupling Kit, Associates of Cape Cod. Inc., MA, USA) was added to the microplate using an 8-series pipette and the microplate was stirred manually to mix.The plate was immediately put on an ice pack after warming in 96-well shaking incubators at 37.0°C for 25 min.After the incubation was completed, the PyroColor Diaza Reagent DIA150-MP (Associates of Cape Cod.Inc.) was added to each well.A total of a 50 μL of HCl solution, ammonium sulfamate and N-(1-Naphthyl) polyethylenediamine (NEDA) were added to each well using an 8-unit pipettor and stirred.Absorbance was measured at 540 nm using a microplate reader (AS ONE: MPR-A100).Endotoxin concentrations were expressed as endotoxin units per milligram of sample (EU mg -1 ).The correlation coefficient of a calibration curve was taken to 0.980 or more to confirm the experiment's accuracy.The accuracy of the DTT assay was checked using the reference materials CRM#28 and CRM#30 (Urban Aerosols and Gobi Kosa Dust, NIES, National Institute for Environmental Studies Onogawa, Tsukuba, Japan).LOD was 0.1 EU mg -1 ; the LOD value was calculated using a linear regression model fitted to the standard solution curves (Piehler et al., 2020).

5. 5 Cell Viability and Cytokine Measurements
The cytotoxicity of the collected particles was evaluated by the viability of human epithelial A549 cells and the IL (interleukin)-6 production of macrophage-like cells differentiated from human leukemia monocytic (THP-1) cells.A549 cells derived from human alveolar cell carcinoma of lung cancer cells are frequently used in toxicity research to evaluate the biological effects of particles on human respiratory systems (Jia et al., 2017).Macro-phage cells, which are antigen-presenting cells, are one of the most potent producers of proinflammatory cytokines in the airways and lungs and play roles as the driver of local and systemic inflammatory responses.Moreover, exposure of atmospheric particles to macrophage cells has also confirmed a significant release of IL-6, one of those important cytokines (Jia et al., 2017;Hiraiwa and van Eeden, 2013).
The A549 cells used to measure this viability were purchased from Japanese Cancer Research Resources Bank ( JCRB), National Institutes of Biomedical Innovation, Health and Nutrition, Japan.And they were maintained with Minimum Essential Medium (MEM, Thermo Fisher Scientific, Waltham, MA, USA) including 10% deactivated fetal bovine serum (FBS, Origin South America, CCP-FBS-BR-500, Cosmo Bio Co., Ltd., Tokyo, Japan), 1% penicillin-streptomycin solution (PS, FUJIFILM Wako Chemicals Corporation, Osaka, Japan), and 1% MEM Non-Essential Amino Acids (Thermo Fisher Scientific, Waltham, MA, USA) at 37°C in a humidified atmosphere containing 95% air and 5% CO 2 .The cells were seeded with a concentration of 1,000 cells/well on 96 well plates 24 hours before the particles were exposed.It should be noted that the small seeding concentration and culture time in this study differ from the typical cell density exposed in the subconfluently cell state.However, although the actual effect may be overestimated, the toxicity of the particles can be accurately characterized due to be grasped with higher sensitivity.Incubated cells were exposed for 24 hours to a solution suspended in the above-prepared culture medium at particle concentrations of 7.5, 75, and 150 μg mL -1 .The viability of exposed cells was measured by the Premix Water-soluble tetrazolium salt-1 (WST-1) Cell Proliferation Assay System (MK-400, TAKARA BIO INC., Shiga, Japan).Following a standard protocol in this analysis, we added 10 μL of WST-1 solution to each well, incubated for 4 hours for the reaction, and measured their absorbance at 450 nm and 630 nm.Cell viability was calculated as a ratio and compared with the control cells.Each concentration was repeated in three wells (n = 3).
To provide additional data on PM 2.5 cytotoxicity, inflammatory mediator release, cytokines (IL-6) were measured after treating of macrophage-like cells with PM 2.5 .IL-6 is promptly and transiently produced during infections and tissue injuries and stimulates acute-phase responses and immune reactions.Moreover, high-level IL-6 exerts pathologically on chronic inflammation and autoimmunity (Tanaka et al., 2014).Before exposure, THP-1 was treated with phorbol 12-myristate 13-acetate (PMA, FUJIFILM Wako Chemicals Corporation, Osaka, Japan) for differentiation into functional macrophages (Starr et al., 2018).Human monocytic leukemia THP-1 cells purchased from the JCRB were maintained in RPMI-1640 (FUJIFILM Wako Chemicals Corporation, Osaka, Japan) containing 1% deactivated FBS, 1% PS and stored in a humidified atmosphere containing 95% air and 5% CO 2 at 37°C.In this measurement, THP-1 cells were seeded at 4.8 × 10 4 cells/well on 12 well plates with PMA (200 nM).After incubating for 72 hours, the medium was replaced with a PMA-free medium and additional incubated for 72 hours until the sample is exposed.The exposure time of the sample was set 24 hours, similar to the cell viability described above.The levels of IL-6 in the centrifuged supernatants were measured by Human IL-6 enzyme-linked immunosorbent assay (ELISA) Kit (LBIS Human IL-6 ELISA Kit, AKH-IL6, FUJIFILM Wako Shibayagi, Gunma, Japan) 24 hours after particle exposure, according to the manufacturer's instructions.The absorbance was observed under the microplate reader at 450 nm and 630 nm.LOD was 0.98 pg/mL; the LOD value was calculated on the standard solution curves set using a linear regression model (Piehler et al., 2020).The amount of cytokine release is determined by the concentration from the absorbance of the measured sample by their calibration curve that was obtained the absorbance measured by seven concentrations of IL-6 standard regent (i.e., positive control).For each experiment, the mean values were obtained from the replicate analysis (e.g., n = 3).

5. 6 Statistical Analysis
T-test was used to compare the differences between cell viability and cytokine levels for a similar dose-manners.*p<0.05implies statistically significant and **p <0.01 is highly significant.

1 Evaluation of the Separation Characteristics of
Cyclones Using Penetration Measurement Systems Fig. 5 shows the penetration curves for the three cyclones measured using the aerodynamic particle sizer (APS), fluorescent monodisperse polystyrene latex (PSL), and differential mobility analyzer (DMA) systems.Fig. 5a shows the penetration curves of the APS for the cyclones.The penetration of particles for cyclones with an aerodynamic diameter of 0.54 μm was 25% for the small cyclone, and 34% for the medium cyclone; with an aerodynamic diameter of 0.84 μm, it was 30% for the large cyclone.However, we must explain the measurement problems that lead to the uncertainty of this result by APS.The pressure dropped at the downstream of the cyclone.In particular, the small diameter of the cyclone at large flow rates to reduce the cutoff diameter can be expected to result in greater pressure loss.Therefore, it is possible that the instrument conditions were not sufficient under such reduced pressure for the suctioned air volume, for the flow ratios between sample and sheath flow, or for the correction of measurement results.In fact, the pressure when measured with a small cyclone was below 600 mbar, and traditional measurements at that pressure led to strange results.Fig. 5b shows the penetration curves of the PSL particles, whereby the 50% cut size of a cyclone (Dp 50 ) of the small, medium and large cyclones were 0.28 μm, 0.36 μm, and 0.7 μm, respectively.Dp 50 for the PSL system for the small and medium cyclones were almost similar.Therefore, we conducted additional measurements using the DMA system for the small and medium cyclones.The penetration measurement for the large cyclone was not conducted using the DMA system because the high flow rate reduces the number of particles and required considerable time.In Fig. 5c, the penetration curves of the DMA system for the small and medium cyclones are shown.The penetration curves of the DMA system crossed the line of Dp 50 at 0.17 μm and 0.34 μm for the small and medium cyclones, respectively.These three approaches for the evaluation of these cyclones showed similar tendency in their cutoff diameters, but with strictly different the diameters.As described above, the cut-off diameter, especially small cyclone, measured by APS could be needed to consider carefully about their accuracy.The other two methods also have potential differences such as the density of the used particles and the particle diameter defined based on the measurement method, and uncertainties such as the low resolution for the particle size of the fluorescent particles and the measurement error.Considering them, at least the results of these two approaches are rather consistent, and the results by APS also supported this.
In this study, the values of Dp 50 were from 0.17 μm to  0.28 μm for the small cyclone, from 0.34 μm to 0.36 μm for the medium cyclone and 0.70 μm for the large cyclone.The Dp 50 of the small cyclone measured using the APS, PSL, and DMA systems was smaller than that of the medium cyclone.
The aerodynamic diameters of the most hazardous and high DTT active species begin at 0.1 μm in a typical atmosphere (Fang et al., 2017).Moreover, previous studies corresponding to the mass size distribution of aerosol particles showed that particles in PM 2.5 were mainly distributed with the main peak in 0.3-1.8μm (Fang et al., 2017;Voutsa et al., 2015;Xue et al., 2014;Feng et al., 2009;Yao et al., 2002).In addition, the high flow rate allows the collection of large amounts of PM 2.5 samples in a short period.Thus, we chose small and large cyclones based on Dp 50 and the flow rate to collect the powder samples of PM 2.5 to compare the effects of their separation performance on the particle toxicity.

2 Comparison of the Chemical Composition of
the Particles Collected by Small and Large Cyclones In Fig. 6, water-soluble ionic components in atmospheric particulates are compared to differentiate the separation characteristics between the large and small cyclones.Fig. 6a shows that the abundance of total ions in the small cyclone samples was higher (1.2-3.2 times) than in the large cyclone samples.The mean quantity of total ions in PM 2.5 for all periods was 257,670 part per million weight (ppmw) for the small cyclone and 123,800 ppmw for the large cyclone.Comparing the composition of total water-soluble ions, SO 4 2-and NH 4 + exhibited significantly higher mass portions in the small cyclone samples than in the large ones, whereas Cl -and Ca 2+ were found to have higher mass portions in the large cyclone samples than in the small ones during all sampling periods.The decrease of NO 3 -in the cold period and the increase of SO 4 2-in the warm period occurred.This tendency was observed mainly in the small cyclone samples (Fig. 6a and 6b).In the samples of both cyclones, sulfate, nitrate, and ammonium were the most abundant ions studied; they accounted for 73.9-93.1% of the total ions for the small cyclone and 59.6-72.7% of the total for the large cyclone.
Ion chromatography analysis of atmospheric particulates collected using the small and large cyclones for oneyear sampling periods showed that the major components of total water-soluble ions in the small cyclone samples were primarily comprised of SO 4 2-during the warm period (2 April 2020 to 14 October 2020), and SO 4 2-and NO 3 -during the cold period (12 November 2020 to 14 April 2021) (Fig. 6).In the small cyclone samples, NO 3 -showed winter maxima, but SO 4 2-was highest in summer, consistent with other studies conducted with ordinary filter samplings (Okuda et al., 2018;Song et al., 2006).Seasonal variations of SO 4 2and NO 3 -were observed, especially in the small cyclone samples.It has been reported that fine particles in the air are enriched with secondary aerosols formed from the oxidation of SO 2 and NO x and neutralization of NH 3 under certain meteorological conditions, and in particular, ammonium sulfate tends to be generated in summer (warm season), whereas, nitrates tend to enrich in particular phase due to gas/particle partitioning depends on ambient temperature (Segalin et al., 2020;Okuda et al., 2018;Song et al., 2006;Yao et al., 2003).The major ions in the large cyclone samples comprised almost similar SO 4 2-and NO 3 -levels in both periods.The abundance of Cl -and Ca 2+ was more elevated than that in the small cyclone samples.This study showed that the small cyclone for smaller particles (e.g., secondary SO 4 2-, NO 3 -, NH 4 + ) had higher collection efficiency than the large cyclone.

3 Detailed Comparison of the Particles
Collected by the Large and Small Cyclones Fig. 7 shows the elemental analysis results of the PM 2.5 samples collected using the small and large cyclones.The transition metals, such as Ti, V, Cr, Mn, Fe, Cu, Zn, Ni, and Pb, were compared.We have excluded Al data from further discussion due to possible contamination because the small cyclone was made of aluminum.The total metal abundance in the PM 2.5 collected from 10 December 2020 to 6 January 2021 and 10 March to 14 April 2021 was 6,504 and 18,516 ppmw for the small cyclone and 46,002 and 21,704 ppmw for the large cyclone, respectively.Fig. 7 is presented without Fe data because the difference between the metal components was somewhat invisible due to the high content of Fe compared with other metals in the samples of both cyclones.The abundance of Fe (out of nine metal components in total) collected from 10 December 2020 to 6 January 2021 and 10 March to 14 April 2021 was 59.1% and 72.9% for the small cyclone and 84.8% and 83.4% for the large cyclone, respectively.V, Cr, Ni, and Pb were determined to be less than 1% of the sum of the total metal components for both cyclones.Cu and V exhibited a higher mass portion in the small cyclone samples than in the large cyclone samples, whereas Fe and Ti were higher in the large cyclone samples than in the small cyclone samples for both periods.The abundance of Cr, Zn, Mn, Ni, and Pb in the small and large cyclone samples varied for the sampling periods.
Table 2 shows the crustal enrichment factors (EFs) (vs.Ti) of the samples collected using the small and large cyclones.The EF values of all elements (except Fe in samples collected from 10 December 2020 to 6 January 2021) were higher in the small cyclone samples than in the large cyclone ones; in particular, a significant difference was observed for V, Cr, Cu, Zn, and Pb.This suggests that the small cyclone samples had crustal elements that are much smaller than those in the large cyclone samples.The higher EF values exist in much smaller particles, indicating that they originated from more anthropogenic sources (Obiajunwa et al., 2002).
A comparison of the chemical compositions of PM 2.5 showed that the metals (e.g., V, Cr, Mn, Ni, Cu, Zn, Pb) exhibited a higher mass portion in the small cyclone samples than in the large ones, which were formed primarily from anthropogenic sources (Chen et al., 2010;Okuda et al., 2007;Song et al., 2006).EF values also showed that these elements contained more anthropogenic sources than soil.Moreover, SO 4 2-, NO 3 -, and NH 4 + ions with higher abundance were in the small cyclone samples, which originated from oxidation and neutralization under certain meteorological conditions in the form of secondary aerosols (Segalin et al., 2020;Song et al., 2006;Yao et al., 2003).The metals that exhibited higher content in the large cyclone samples than that of the small cyclones were light (e.g., Mg 2+ and Ca 2+ ) and crustal elements (e.g., Ti and Fe) (Hieu and Lee, 2010).This can be seen in the calculation of EFs (Table 2).The particles in the large cyclone samples contained coarse and more crustal sources than anthropogenic sources.In contrast, the particles in the small cyclone samples comprised smaller and more anthropogenic sources than crustal sources.

1 DTT Assay
The results of the redox activity of the PM 2.5 samples collected using the small and large cyclones are summarized in Fig. 8.The DTT loss rates were 0.21 and 0.96 μM/mg/min for the small cyclone samples, and 0.82 and 0.62 μM/mg/min for the large cyclone samples collected from 10 December 2020 to 6 January 2021 and 10 March 2021 to 14 April 2021, respectively.
In this study, the DTT loss rate of the samples collected using small and large cyclones differed in the sampling periods, consistent with other studies (Fang et al.,Fig. 7. Percentage variations of elements in total elemental components (except Fe) of the PM 2.5 collected using the small and large cyclones.Fujitani et al., 2017).Fujitani et al. (2017) found that transition metals in the sample were a significant contributor to DTT consumption (more than 80%).In our study, the difference in the DTT oxidization ability is likely associated with the difference between transition metal concentrations (e.g., Cu, Mn, Ni, Fe, Zn, Pb) in the samples.

2 Endotoxin Analysis
Fig. 9 shows the endotoxin concentrations in the samples collected using the small and large cyclones.The endotoxin content of the particles was 129.58 EU/mg and 60.49EU/mg in the large cyclone samples collected from 10 December 2020 to 6 January 2021 and 10 March to 14 April 2021, respectively.Endotoxins were not detected in the small cyclone samples.
The endotoxin level in the large cyclone samples was detected significantly high compared to that in the small cyclone samples.This is consistent with other studies (Khan et al., 2018;Schins et al., 2004).Khan et al. (2018) found that concentrations of endotoxin in coarse particles were significantly higher than those in fine particles in Japan.In this study, the abundance of smaller particles in the samples of the small cyclone was higher than in the samples of the large cyclone.This indicates that the distribution of small particles significantly decreased the endotoxin levels in the small cyclone samples.

3 PM 2.5 Cytotoxicity in A549 and Macrophage-
Like THP-1 Cells Fig. 10 shows the cytotoxic effects of PM 2.5 collected using large and small cyclones evaluated with A549 cells in MEM.The results show the ratio of the metabolic activity of exposed cells to the metabolic activity of unseeded cells.The highest toxicity of the large cyclone samples was greater than in the small cyclone samples, except for the samples collected from 10 March to 14 April 2021 at 75 μg/mL and 150 μg/mL of PM 2.5 concentrations.The difference between the small and large cyclone samples was insignificant (p<0.1) at 7.5 μg/ mL, significant (p<0.05) at 75 μg/mL and highly significant (p<0.01) at 150 μg/mL.Fig. 11 shows the estimated levels of cytokines using the IL-6 release measurement for macrophage-like cells in the collected PM 2.5 .The value of the control shows  IL-6 levels in the blank medium used in the experiments.The level of IL-6 released after exposure to the samples collected using the large cyclone was markedly higher than that observed after exposure to the samples collected using the small cyclone.An in vitro toxicity assay of PM 2.5 collected using the small and large cyclones was conducted on lung epithelial A549 cells.The A549 cell viability after exposure to the PM 2.5 collected from both cyclones decreased corresponding with the DTT loss rate, consistent with previous studies (Zou et al., 2016;Cao et al., 2015), which found that smaller particles can enter cells or induce ROS production and damage cells.
To evaluate whether the PM 2.5 collected using small and large cyclones had proinflammatory effects on macrophage-like THP-1 cells, the production of IL-6 was examined.The cytokine level estimated using IL-6 release measurement was higher in the large cyclone samples than in the small ones, which is supported by other studies (Gorbet and Sefton, 2005;Huang et al., 2004;Schins et al., 2004).Schins et al. (2004) found that the cytokine levels were higher in coarse particles than in fine ones due more to their biological components (e.g., endotoxin) than metal content.Gorbet and Sefton (2005) reported that Gram-negative bacteria cause an inflammatory response to local infection by stimulating the synthesis of inflammatory mediators (e.g., IL-6).Zou et al. (2016) showed that the proapoptotic effects of particles are unrelated to their proinflammatory properties, and water-soluble ions and individual soluble metals (V, Ni, Fe) were relatively inactive for apoptosis and inflammation in macrophages.
The chemical composition analyses of PM 2.5 collected using the small and large cyclones showed that the small cyclone samples had higher concentrations of water-soluble ions and transition metals, except for Fe and Ti, than the larger cyclone samples.The comparison of the samples collected using large and small cyclones significantly reflected the significant toxicity of the particles.ROS generation was associated with transition metals and they decreased the viability of the human lung epithelial A549 cells.The endotoxin concentration was detected only in the large cyclone samples and significantly affected the level of cytokines estimated using the IL-6 release measurement.

CONCLUSIONS
In this study, we investigated the effect of cyclone size on particle size, chemical composition analysis and toxicity assays (DTT, LAL, cell-level) of PM 2.5 .Three cyclones with different inner diameters (12 mm for small, 20 mm for medium and 68 mm for large cyclones) were tested using the APS, PSL, and DMA systems.The toxicity of the particles was evaluated using by comparing the results of chemical composition analysis, DTT, and LAL, and cell exposure assays.
The experimental values of Dp 50 measured using the APS, PSL, and DMA systems were from 0.17 μm to 0.28 μm for the small cyclone, 0.34 μm to 0.36 μm for the medium cyclone, and 0.70 μm for the large cyclone.To compare their separation performances, small and large cyclones were chosen based on Dp 50 and the flow rate to collect PM 2.5 .A comparison of the chemical compositions of PM 2.5 showed that the metals (e.g., V, Cr, Mn, Ni, Cu, Zn, Pb) exhibited a higher mass portion in the small cyclone samples than in the large ones.Moreover, SO 4 2-, NO 3 -, and NH 4 + ions with higher abundance were in the small cyclone samples.The metals that determined a higher content in the large cyclone samples than in the small ones were light (e.g., Mg 2+ and Ca 2+ ) and crustal elements (e.g., Ti and Fe).A comparison of the chemical composition and EF values showed that the particles in the small cyclone samples contained smaller and anthropogenic sources than those in the large cyclone samples.
The DTT loss rate of the samples collected using the small and large cyclones became different across the sampling periods, which were associated with the transition metals.Endotoxin contents were detected only in the large cyclone samples.The viability of A549 cells after exposure to the small and large cyclone samples was different across the cyclone samples and sampling periods associated with ROS generation determined in the DTT assay.IL-6 release was higher in the large cyclone samples than in the small cyclone samples, consistent with the endotoxin concentrations measured in the LAL assay.
The cyclone technique used in this study to collect aerosol particles would be a powerful tool for a detailed evaluation of particle toxicity.
2. 1. 2 Particle Penetration of Cyclones Measuredusing Fluorescent PSL Monodisperse Particles Additional experiments were required to determine

Fig. 1 .
Fig. 1.Schematic diagram of the APS system to evaluate the particle penetration of the cyclones.

Fig. 2 .
Fig. 2. Schematic diagrams of the experimental system to evaluate the performance of the cyclones using the fluorescent PSL monodisperse particles.

Fig. 3 .
Fig. 3. Schematic diagrams of the DMA system to evaluate the particle penetration of the cyclones.

Fig. 4 .
Fig. 4. Schematic diagram of the PM 2.5 sampling system for small and large cyclones.

Fig. 5 .
Fig. 5. Particle penetrations of the three cyclones measured using the (a) APS, (b) PSL monodisperse particles, and (c) DMA systems.The error bars were calculated according to the changing of particle number during measurement.Small, medium, and large are the small, medium and large cyclones, respectively.

Fig. 6 .
Fig. 6.Ion chromatography analysis of the ambient aerosol samples collected using the small and large cyclones from 2 April 2020 to 14 April 2021.The abundance of total ions in the PM 2.5 (a) and the percentage variations of ions in total water-soluble ion components of the PM 2.5 (b) are shown.

Fig. 9 .
Fig. 9. Concentration of endotoxin in PM 2.5 collected using the small and large cyclones (n = 3).

Fig. 11 .
Fig. 11.PM 2.5 concentration-dependent manners in cytokines estimated using IL-6 release measurement for macrophage-like cells (*p<0.05imply that the difference between the small and large cyclone samples was significant, n = 3).The samples were collected (a) from 10 December 2020 to 6 January 2021 and (b) from 10 March to 14 April 2021.

Table 1 .
Dimensions and specifications of the cyclones used in this study.

Table 2 .
Enrichment factors for the analyzed trace elements in PM 2.5 collected using the small and large cyclones.