In vitro assessment of the pulmonary toxicity of particulate matter emitted during haze events in Chiang Mai, Thailand via investigation of macrophage responses

Chiang Mai (Thailand) experiences severe haze pollution in the dry season (December–April) each year mainly due to local and regional biomass burning (e.g. of agricultural land). A major component of the haze is airborne particulate matter (PM). During haze events, biomass burning is likely to be the dominant source of PM emissions, and at other times emissions from traffic dominate. The hazard of traffic derived PM has been extensively investigated previously but there are uncertainties regarding the toxicity of PM emitted from biomass burning. The toxicity of PM10 samples collected during and after haze events in Chiang Mai in 2020 was compared in vitro in J774.1 macrophages as they are responsible for the clearance of inhaled particles. Diesel exhaust particles and ultrafine carbon black were included as benchmark particles as they have been commonly used as a surrogate for PM. Cytotoxicity was evaluated 24 h post exposure at concentrations of 3.9–125 µg ml−1. Cytokine production (tumour necrosis factor alpha (TNF-α), interleukin (IL)-6, IL-1β, macrophage inflammatory protein (MIP-2)) was assessed and cell morphology visualised using light and scanning electron microscopy. The hydrodynamic diameter, zeta potential and endotoxin content of all particles was assessed as well as the metal content of PM samples. All particles induced a concentration dependent decrease in cell viability and increased TNF-α and MIP-2 production. Only PM samples stimulated IL-6 production and only non-haze PM caused IL-1β production. No change in IL-10 production was detected for any particle. PM samples and DEP caused vacuole formation in cells. The concentrations of endotoxin and metals were highest in non-haze PM, which may explain why it induced the greatest inflammatory response. As non-haze PM was more toxic than haze PM, our results indicate that the source of PM emissions can influence its toxic potency and more specifically, that PM emitted from biomass burning may be less toxic than PM emitted from traffic.


Introduction
In Thailand the seasonal burning of biomass, such as forests, agricultural land and crop residues, causes severe haze (smoke) pollution (Phairuang et al 2017, Kliengchuay et al 2018, Punsompong and Chantara 2018, Ramakreshnan et al 2018, Moran et al 2019, Kawichai et al 2020, Kawichai et al 2021, Kongpran et al 2021. For example, each year Chiang Mai, a city in northern Thailand, experiences severe haze air pollution during the dry season (December-April) as a consequence of the burning of agricultural land and waste (Pengchai et al 2009, Tsai et al 2013, Punsompong and Chantara 2018, Ruchiraset and Tantrakarnapa 2018.
A major component of haze pollution is airborne particulate matter (PM) , Pavagadhi et al 2013, Pongpiachan et al 2015, George et al 2020. Air quality is most commonly assessed (globally) via measurement of daily or annual mean levels of PM 10 (particles with a diameter of <10 µm) (WHO 2021). There is substantial epidemiological evidence that PM 10 can exacerbate or cause various adverse health outcomes such as cardiovascular disease (CVD), respiratory disease (e.g. chronic obstructive pulmonary disease, asthma, respiratory infections, lung cancer), low birth weight, neurological disorders (e.g. dementia) as well as metabolic diseases (e.g. diabetes) (e.g. Pope and Dockery 2006, Heusinkveld et al 2016, Alderete et al 2018, Johnston et al 2019. In addition, PM from outdoor air pollution is classified as a Group 1 human carcinogen (IARC 2015). In Thailand, PM is associated with an increase in respiratory and CVD morbidity (e.g. pneumonia admissions, respiratory and cardiovascular outpatient visits), lower birth weight as well as an estimated 50 000 deaths annually (Ruchiraset and Tantrakarnapa 2018.
Whilst PM 2.5 is of greatest concern to human health, in Thailand monitoring of PM 10 levels is most common and thus we focused on the toxicity of PM 10 in our study. During haze events in Thailand the levels of PM 10 frequently exceed World Health Organisation (WHO) guideline limits (annual mean of 15 µg m −3 and 24 h mean of 45 µg m −3 ) and the (120 µg m −3 ) daily PM 10 standard set by the Thailand Pollution Control Department (PCD) (Pardthaisong et al 2018, Kongpran et al 2021, WHO 2021. For example Tsai et al (2013) demonstrated that average PM 10 levels during haze events in Chiang Mai in 2010 were ∼157 µg m −3 and ∼79 µg m −3 at other times. Similarly Pengchai et al (2009) demonstrated that daily PM 10 levels reached 182 µg m −3 during the dry season in Chiang Mai in [2005][2006]. Furthermore, in Chiang Rai (another Northern province in Thailand) daily PM 10 levels reached 479 µg m −3 during haze events in 2011-2018 associated with biomass burning during the dry season (Kliengchuay et al 2021). It is therefore concerning that over the last decade haze events have become more severe and frequent in Thailand and other countries in South East Asia (Pardthaisong et al 2018, Kliengchuay et al 2018, Kawichai et al 2020, Kawichai et al 2021. Climate change and the warming meteorological conditions in South East Asia are likely to lead to increased flammability of the fuel sources which is likely to increase the prevalence of biomass burning, and as a consequence the frequency and severity of haze events (Keywood et al 2013). As a result of transboundary winds, pollutant emissions from other countries can also contribute to haze events in Thailand, and the haze pollution emitted in Thailand is not restricted to the northern provinces but also extends to other parts of Thailand and neighbouring countries, causing transboundary pollution (Ramanathan and Crutzen 2003, Sirimongkonlertkun 2018, Chomanee et al 2020, Phairuang et al 2020.
Many research studies have focused on investigating the health impacts of PM that originated from fossil fuel combustion (e.g. emissions from traffic). By contrast, relatively fewer studies have assessed the toxicity of PM derived from different sources of biomass burning (e.g. agricultural crop burning, forest fires/wildfires, and use of biomass as a fuel for domestic cooking and heating) (reviewed by Sigsgaard et al 2015, Johnston et al 2019. WHO air quality guidelines are currently based on the mass concentration of PM 10 (and PM 2.5 ) and do not distinguish between the potential differential toxicity of PM emitted from different sources (WHO 2021). However, there is concern that regulations based only on total PM mass concentrations may not be sufficient to protect human health (Stanek et al 2011, Cassee et al 2013, Bell et al 2014, Li et al 2019. As biomass burning is expected to be an important source of outdoor PM emissions worldwide (Sigsgaard et al 2015) it is timely to assess the toxicity of PM emitted from different sources. This will allow for the effective development of environmental policy and introduction of intervention strategies in different sectors (e.g. agriculture) which aim to improve outdoor an indoor air quality. Importantly, there is already evidence from in vitro and in vivo studies that PM emitted from sources other than fossil fuel combustion, such as biomass burning (e.g. wildfires, wood combustion) can exhibit toxicity (e.g. Karlsson et al 2006, Kocbach et al 2008, Wegesser et al 2010, Franzi et al 2011, Park et al 2018. In the present study, we compared the toxicity of PM 10 samples collected in Chiang Mai (Thailand) during and after haze events in 2020 to identify if the emission source influenced PM toxicity. During haze events PM is primarily emitted from biomass burning, and at other times (non-haze) emissions from traffic dominate. We selected diesel exhaust particles (DEPs) and ultrafine carbon black (ufCB) as benchmark particles for our study as their toxicity has been extensively investigated in the existing literature and they are commonly used as surrogates to represent PM in experimental studies (e.g. Li et al 1999, Stoeger et al 2006, Turner et al 2015. It is of benefit that the physico-chemical properties of these benchmark materials have been extensively characterised in previous studies (e.g. Stone et al 1998, Li et al 1999, Singh et al 2004, Stoeger et al 2006. Alveolar macrophages are resident immune cells in the lung and constitute the first line of defence against inhaled particles (Wiemann et al 2016). In this study, the murine J774A.1 macrophage cell line was used as a surrogate of alveolar macrophages as they have been extensively used in in vitro investigations assessing the pulmonary toxicity of various particles (e.g. Brown et al 2004, Zhou and Kobzik 2007, Rothen-Rutishauser et al 2010, Clift et al 2011, Yue et al 2012, Ma et al 2015, Boyles et al 2018. All particles were evaluated for their ability to induce cytotoxicity, changes in cellular morphology and pro-inflammatory responses in macrophages in vitro as a measure of their toxicity. The physico-chemical properties of PM are critical to their toxic potency and can be influenced by many factors, including the source of emission, meteorological conditions and atmospheric composition (Reid et al 2005, Kelly andFussell 2012). Although particle size is key to the pathogenicity of PM, the absorption of toxic compounds to the surface of particles is also of high importance. In particular transition metals (e.g. Fe, Ni, Zn, Al), bacterial endotoxins (lipopolysaccharide (LPS)) and polycyclic aromatic hydrocarbons (PAHs) can become associated with PM to enhance its toxicity (e.g. O'Grady et al 2001, Soukup and Becker 2001, Takano et al 2002, Reid et al 2005, Kongpran et al 2021. Accordingly, in parallel to the in vitro assessment of PM toxicity, the physico-chemical properties of the haze and non-haze PM samples were characterised (size (hydrodynamic diameter), metal and endotoxin content) to identify which particle properties may confer toxicity.

PM sampling
Ambient PM 10 samples were collected from an urban area of Chiang Mai (Chiang Mai University (18 • 44 ′ N, 98 • 57 ′ E)) using a Gent Air sampler (IAEA standard) with 20 l min −1 flow rate, with a 47 mm polytetrafluoroethylene (PTFE) filter during March to April 2020 (haze) and September to October 2020 (non-haze).
Average 24 h PM 10 levels from the Thailand PCD monitoring station during the sampling time ranged from 97.3 to 149.3 µg m −3 and 11.1-28.5 µg m −3 for haze and non-haze periods respectively. The PCD monitoring station is located by a busy road in Chiang Mai and is approximately 3 km from the sampling site.

PM extraction from filters
PM was extracted from PTFE filters with methanol and the total mass of extracted PM was assessed by gravimetric analysis, using a Sartorius micro balance. The PTFE filter was placed particle side down in a glass Petri dish containing 1.5 ml of methanol, sealed with parafilm and sonicated for 5 min, using a bath sonicator. After sonication, the filter was turned upside down and sonicated again for 5 min. The methanol (containing PM) was removed and transferred into a pre-weighed glass vial. The same filter was again placed particle side down in a glass Petri dish and the whole process was repeated. The PTFE filter and a glass vial-containing PM were left to dry inside a drying cabinet at 60 • C, for 24 h, to remove the methanol. The extracted PM mass from the filter was determined by the glass vial weight difference before and after extraction. Each time the extraction protocol was performed, a blank PTFE filter was treated in the same way as the sample filters, to assess the toxicity of material released from the filter solely due to the extraction method (termed blank filter extracts) and a blank filter control included in all experiments. The glass vial-containing PM and blank filter extracts were stored at 4 • C until needed.

Particle dispersion protocol
DEP was purchased from National Institute of Standards and Technology and ufCB was obtained from Degussa (Printex P90, Degussa, Germany). PM and ufCB particles were prepared according to the protocol described by Jacobsen et al (2010). Briefly, PM and ufCB stock suspensions of 1 mg ml −1 were prepared in water containing 2% FBS and vigorously vortexed. Due to the hydrophobic nature of the DEP, DEP was first pre-wetted with 100% ethanol, at a concentration of 6 mg ml −1 and vigorously vortexed according to the method of Jensen et al (2011). Next, water containing 2% FBS was added to give final DEP concentration of 1 mg ml −1 . PM, ufCB and DEP stock suspensions (1 mg ml −1 ) were sonicated for 16 min using a bath sonicator (32-38 kHz at 20 • C). After sonication particles were diluted to the required concentrations in complete cell culture medium.

Particle characterisation 2.2.1. Dynamic light scattering (DLS)
The hydrodynamic diameter and the polydispersity index (PDI) of all particle suspensions were determined by DLS, using a Zetasizer Nano-ZS (Malvern Instruments Ltd., UK). Particles were dispersed as described above and diluted in complete cell culture medium, at a concentration of 125 µg ml −1 and measurements made immediately after preparation.

Determination of endotoxin content
The endotoxin content of particle samples (haze and non-haze PM, ufCB and DEP) was assessed using the chromogenic limulus amebocyte lysate test (QCL-1000TM Assay, Lonza, Walkersville, MD). Particles were suspended at a concentration of 62.5 µg ml −1 in complete cell culture medium and endotoxin levels assessed according to the manufacturer's instructions.

Cytotoxicity assays
Alamar Blue, 5-carboxyfluorescein diacetate, acetoxymethyl ester (5-CFDA, AM) and Neutral red assays were used to measure the loss in metabolic, cellular esterase activity and lysosome integrity, respectively, in J774A.1 cells exposed to particles for 24 h.
J774A.1 cells were seeded in a flat bottom 96 well cell culture plate at a density 0.84 × 10 5 cells cm −2 and incubated for 24 h at 37 • C, 5% CO 2 . Cells were then exposed to particles (haze and non-haze PM, ufCB and DEP) at final concentrations ranging between 3.9 and 125 µg ml −1 . The negative control consisted of complete cell culture medium only, while 0.01% Triton X-100 was used as a positive control. In addition, cells were exposed to blank filter extracts to identify whether components released from the PTFE filter during the extraction process had an effect on cell viability. After a 24 h incubation at 37 • C with 5% CO 2 , the supernatants were collected and stored at −80 • C, to be used for cytokine analysis, and cytotoxicity was assessed according to the protocol described in Connolly et al (2015).

Cytokine secretion
Levels of cytokines in the supernatant of J774A.1 cells exposed to sub-lethal concentrations (31.3, 62.5 and 125 µg ml −1 ) of the tested (haze and non-haze PM, ufCB and DEP) particles or blank filter extracts were analysed using a Luminex® mouse cytokine 5-plex assay (Magnetic Luminex assay, R&D Systems Inc., USA), according to the manufacturer's guidelines. The pro-inflammatory cytokines, tumour necrosis factor alpha (TNF-α), interleukin (IL)-1β, IL-6 and macrophage inflammatory protein (MIP)-2, were chosen as they play a crucial role in the initiation and amplification of immune responses within the lung following particle exposure (Driscoll 2000, Sahin andWasmuth 2013), whereas the anti-inflammatory cytokine, IL-10, promotes the resolution of inflammation and healing (Kocbach et al 2008). Supernatants from untreated cells (exposed to complete cell culture medium) served as a negative control, and cells were exposed to 10 ng ml −1 LPS (endotoxin) were included as a positive control.

Assessment of cellular morphology 2.3.1. Light microscopy images
J774A.1 cells were seeded in a flat bottom 96 well cell culture plate at a density 0.84 × 10 5 cells cm −2 and incubated for 24 h at 37 • C, 5% CO 2 . Cells were then exposed to particles (haze and non-haze PM, ufCB and DEP) at a concentration of 125 µg ml −1 or blank filter extracts. Control cells were exposed to complete cell culture medium only. After 24 h incubation at 37 • C images of the cells were obtained using light microscopy.

Statistical analysis
All data are expressed as mean ± standard error of the mean (SE). Statistical analysis was performed using a one-way analysis of variance ANOVA, followed by Tukey multiple comparison post-test. The data was tested for normal distribution and if necessary, data sets were transformed. All statistical analyses were performed using Minitab 18 software. A p value of <0.05 was considered to be significant. Experiments were repeated at least three times. All figures were generated using Graph Pad Prism.

Cytotoxicity
The ability of the particle panel to cause cytotoxicity was assessed using three assays; the Alamar Blue, 5-CFDA, AM and Neutral Red Assays. All assays showed that all particles reduced the viability of J774A.1 cells in a concentration dependent manner, 24 h post exposure (figure 1). Since interference of the DEP and ufCB particles with the Neutral red assay was observed, this assay was not performed for these particles. More specifically, DEP and ufCB readly adsorbed the Neutral Red reagent giving false increases in viability (data not shown).
Exposure of J774A.1 cells to non-haze and haze PM induced a significant increase in cytotoxicity in all three assays only at the highest concentration tested of 125 µg ml −1 (figures 1(A)-(C)). ufCB particles caused cytotoxicity at concentrations ⩾31.3 µg ml −1 in the Alamar Blue, and 5-CFDA, AM assays (figures 1(A)-(C)). DEP caused significant cytotoxicity at concentrations ⩾62.5 and ⩾31.3 µg ml −1 , when measured using the Alamar Blue (figure 1(A)) and 5-CFDA, AM (figure 1(B)) assays respectively. Overall levels of cytotoxicity were low following exposure of cells to all particles, however the toxicity of the particles can be ranked: ufCB > DEP > haze PM > non-haze PM. No cytotoxicity was observed in cells exposed to extracts from blank filters (data not shown).

Cytokine production
To evaluate the inflammatory response induced by the different particle samples, levels of TNF-α, IL-6, IL-1β, IL-10 and MIP-2 production were assessed (figure 2). A concentration dependent increase in TNF-α and MIP-2 production was observed for all particles (figures 2(A) and (D)). Non-haze PM and DEP samples caused a significant increase in TNF-α and MIP-2 levels at all concentrations tested, whereas haze PM and ufCB stimulated significant increases at concentrations of 62.5 and 125 µg ml −1 , respectively (figures 2(A) and (D)). Only non-haze PM caused a significant increase in IL-1β production (figure 2(C)). In addition, the levels of IL-6 were below the detection limit for both the control, DEP and ufCB (data not shown), but IL-6 production was detected for non-haze and haze PM. The concentrations of IL-10 were below the detection limit of detection for all particles and the control (data not shown). The production of TNF-α, IL-1β and MIP-2 following exposure to non-haze PM was significantly greater than that observed for the other particles (figure 2). LPS was included as a positive control and significantly increased TNF-α, IL-6, IL-1β and MIP-2 production, compared to control cells (figure 2). Extracts from blank filters had no effect on cytokine production by J774A.1 cells (data not shown). The capacity of the particles tested to stimulate a pro-inflammatory response could be ranked: non-haze PM > haze PM > DEP > ufCB.

Morphological changes in macrophages following particle exposure
Images of J774A.1 cells exposed to the particles were taken using light and SEM to assess the impact of the particles on cell morphology and to visualise the interaction of the particles with cells. Both light microscopy and SEM revealed that control cells (exposed to cell culture medium) had a normal, spherical morphology and a uniform size (figures 3(A) and 4(A)). Similarly, cells exposed to the blank filter extract exhibited the same morphology as control cells (data not shown).
numerous, small vacuoles whereas other cells had one large vacuole. Using SEM a collapsed cell membrane is evident in cells exposed to haze and non-haze PM, and DEP which is likely an artefact due to dehydration of large cytoplasmic vacuoles during SEM sample preparation (figure 4). No vacuoles were observed in cells exposed to ufCB. Using light microscopy particle interactions with cells could be observed. Particles were associated with all cells exposed to ufCB and DEP (figures 3(B) and (C)), however it is not possible to confirm whether Figure 2. Pro-inflammatory cytokine production. J774A.1 cells were exposed to haze and non-haze derived PM, DEP and ufCB for 24 h, at concentrations of 31.3, 62.5 and 125 µg ml −1 . Control cells were exposed to cell culture medium only and LPS (10 ng ml −1 ) was included as a positive control. Cell supernatants were analyzed for the presence of TNF-α (A), IL-6 (B), IL-1β (C), and MIP-2 (D) using a Luminex assay. Data are expressed as mean cytokine concentration (pg ml −1 ) ± SE (n = 6-7). Significance indicated by * = p < 0.05, * * = p < 0.01, * * * = p < 0.001, * * * * = p < 0.0001, when compared to untreated cells and $ = p < 0.05, $$$ = p < 0.001, when compared to other particle treatments. particles are associated with the exterior or interior of the cells using light microscopy. Similarly, particles were observed to be associated with cells exposed to haze and non-haze PM (figures 3(D) and (E)), although particle interactions were less than those observed for ufCB and DEP.

DLS
The Z-average hydrodynamic diameter and PDI of particles that were suspended in complete cell culture medium were measured using DLS (table 1). DEP and ufCB formed relatively monodispersed suspensions,  with an average hydrodynamic diameter of 296 and 220 nm, respectively (table 1). By contrast, suspensions of haze and non-haze derived PM had a wider size distribution of particles as indicated by the higher PDI values and higher levels of variability displayed for their hydrodynamic diameter (table 1).

Endotoxin content
Haze and non-haze PM displayed high levels of endotoxin (figure 5). ufCB and DEP showed lower levels of endotoxin, compared to PM samples. The endotoxin content of the blank filter extract was below the limit of detection for the assay (<0.1 EU ml −1 ).

Figure 5.
Endotoxin content of particle samples. Particles were suspended at a concentration of 62.5 µg ml −1 in endotoxin-free water and assessed for endotoxin content, using the endpoint chromogenic LAL assay. Data are expressed as mean endotoxin content ± SE (EU ml −1 ) (blank filter n = 1; non-haze PM n = 6; haze PM n = 5; ufCB n = 3 and DEP n = 3). Significant differences indicated by * = p < 0.05.

Elemental analysis of PM samples
The metal content of non-haze and haze PM samples was assessed, and a summary of the results are presented in figure 6. Al, Fe, P and Zn were present at the highest concentration in both haze and non-haze PM. In addition, levels of P, Cu and Zn were significantly higher in non-haze PM samples. Analysed metals were below detection limit in blank filter extracts (data not shown).

Discussion
To the best of our knowledge we compared the toxicity of haze and non-haze PM 10 samples from Chiang Mai (Thailand) for the first time using an in vitro macrophage model to better understand if the source of PM 10 emissions influences its toxic potency. During haze events in Chiang Mai PM is primarily emitted from biomass burning (e.g. burning of agricultural land) whereas at other times emissions from traffic dominate. We observed that both haze and non-haze PM samples as well as the benchmark particles (ufCB and DEP) could cause a small reduction in cell viability, an increase in pro-inflammatory cytokine production and detrimental changes in cell morphology (e.g. vacuole formation). All particle samples tested in our study stimulated an increase in cytokine production, however the inflammatory response activated in macrophages was greatest for non-haze (traffic derived) PM, suggesting that it was more toxic than haze (biomass burning derived) PM and the reference particles (ufCB and DEP). The higher endotoxin and metal content of non-haze PM is likely to explain why it induced a greater inflammatory response than haze PM.

Cytotoxicity
All particles tested (haze PM, non-haze PM, ufCB, DEP) induced a concentration dependent reduction in J774A.1 cell viability, although in all cases the levels of cytotoxicity were low. The cytotoxicity of the particles to macrophages in vitro was ranked: ufCB > DEP > haze PM > non-haze PM. It is established that as particle size decreases particle toxicity typically increases due to their larger surface area, and that the ultrafine component of air pollution (particles with a diameter of <100 nm) is likely to drive its toxicity (Seaton et al 1995, Brown et al 2001, Duffin et al 2002. Therefore, as ufCB contains the highest proportion of ultrafine particles within the samples that were tested it is not surprising that it is the most cytotoxic particle as the DEP and PM samples had a wider size distribution. Cytotoxicity is commonly selected an as endpoint in in vitro toxicity studies as it allows the toxicity of different particles to be compared and enables the identification of sub-lethal concentrations of particles to test when assessing cell responses such as cytokine production. We assessed cytotoxicity using three different approaches and all assays confirmed that the particles under investigation caused a concentration dependent reduction in cell viability. It is common to calculate the lethal concentration (LC) causing 50% cell death (LC 50 ) when assessing the cytotoxicity of paticles in vitro. However the LC 50 was not reached within the concentration range investigated for any of the particles tested and the use of higher concentrations to calculate an LC 50 was not deemed to be physiologically relevant.
We observed that the cytotoxicity assays varied in their sensitivity. More specifically, the greatest toxic effects for all particles were observed when the Alamar Blue assay was used, and the Neutral Red assay was the least sensitive assay. These findings align with the existing literature (e.g. Brown et al 2018) and may inform which cytotoxicity assay is prioritised in future studies to ensure that the most sensitive assay is selected. It is established that particles can interfere with biochemical assays (e.g. Stone et al 2009) as, for example, the colour of the particles can contribute to the absorbance or fluorescence readings used to quantify cytotoxicity, or the large surface area of particles means that assay components can adsorb onto their surface to compromise the function of the assay. In our study, ufCB and DEP interfered with the NR assay due to the adsorption of the neutral red reagent to the particle surface and thus this assay was omitted from the cytotoxicity screen for these materials.

Cytokine production
It is established that the ability of PM 10 to activate an inflammatory response is critical to its pathogenicity (reviewed by Stone et al 2017, Johnston et al 2019. Accordingly, investigation of cytokine production allowed the pro-inflammatory response activated by haze and non-haze PM to be evaluated in vitro. Pro-inflammatory cytokines (TNF-α, IL-1β, IL-6 and MIP-2) were selected due to their role in the initiation and amplification of pulmonary inflammatory responses following particle exposure (Driscoll 2000, Sahin andWasmuth 2013), whereas IL-10 promotes the resolution of inflammation and healing (Kocbach et al 2008). We observed that both haze and non-haze PM, as well as ufCB and DEP stimulated TNF-α and MIP-2 production by macrophages. Haze and non-haze PM increased IL-6 production. Non-haze PM also increased the secretion of IL-1β. Our results therefore suggest that all particles activate a pro-inflammatory response, which aligns with the findings of existing studies where PM samples derived from traffic or biomass burning sources, ufCB and DEP stimulate the increased expression or production of pro-inflammatory cytokines in vitro from pulmonary cells including; macrophages (e.g. Leonard et al 2000, Karlsson et al 2006, Jalava et al 2007, Kocbach et al 2008, and bronchiolar and alveolar epithelial cells (e.g. In our study non-haze PM induced a greater level of cytokine production than haze PM. Thus, our findings suggest that urban (traffic derived) PM may elicit a greater inflammatory response than biomass burning derived PM when exposed to cells at an equivalent mass concentration. Whilst some studies agree with the observation that traffic derived PM elicits greater pulmonary toxicity than biomass burning derived PM (e.g. George et al 2020, Pang et al 2020). Interestingly, Mazzoli-Rocha et al (2008) demonstrated that biomass burning derived PM exhibited similar toxicity to traffic derived PM when histology (e.g. influx of neutrophils) and lung mechanical parameters were used to investigate their pulmonary toxicity in vivo (mice) following intranasal administration. The experimental design (e.g. (cell/animal) model used, particle concentrations tested, time point, cellular response investigated) and emission source of PM is different across existing studies which may explain why their findings are not in agreement. For example, biomass burning covers a range of different activities and the biomass type can influence its toxicity (e.g. Kim et al 2019, Martin et al 2021).
Existing studies have demonstrated that PM emitted from different sources varies with respect to its physico-chemical properties (e.g. particle size, charge, composition (e.g. metal, PAH and endotoxin content), oxidative potential) which can influence its toxic potency (e.g. Steenhof et al 2011, Park et al 2018, Marchetti et al 2019, reviewed by Kelly and Fussell 2012. Thus, we will now explore what characteristics of the PM samples we investigated in our study may drive the responses observed. We observed that cytokine production by cells was greater for the PM samples, when compared to ufCB and DEP. This is likely due, in part, to the higher endotoxin content of the PM samples. Endotoxin present on the particle surface is likely to activate toll like receptors on the macrophage cell surface which triggers the expression of pro-inflammatory cytokines (reviewed by Miyata and van Eden 2011). For example, existing studies have demonstrated that cytokine production by cells following exposure to PM 10 is partly mediated by endotoxin (Soukup andBecker 2001, den Hartigh et al 2010).
Epidemiological and experimental studies also provide evidence of the key role of metals (especially vanadium, zinc, iron, copper and nickel) in the toxicity of PM 10 (Schaumann et al 2004, Chen andLippmann 2009). The higher metal content of non-haze PM could therefore explain why it stimulated the greatest inflammatory response. More specifically, the redox activity of metals associated with PM is likely to cause the activation of redox sensitive transcription factors such as nuclear factor kappa B (NfκB) to stimulate pro-inflammatory cytokine expression (e.g. Jimenez et al 2000, reviewed by Miyata and van Eden 2011). For example, den Hartigh et al (2010) observed that a metal chelator (deferoxamine mesylate), and antioxidant (N-acetyl-L-cysteine) lowered PM mediated cytokine expression by primary human monocytes which indicates that metals associated with PM may stimulate ROS production in cells to stimulate cytokine production.
Whilst not measured in this study, it is likely that PAHs bound to the particle surface contribute to the observed toxicity of PM. Biomass burning can lead to increased PAH emissions (e.g. Chiu et al 2011), with the fuel type and combustion conditions able to influence the type and level of PAH formation (Zosima et al 2016). Indeed, Kongpran et al (2021) demonstrated that PAH levels (particle-bound and gas phase) were ∼26 times higher during haze events than during non-haze periods in 2018 in Northern provinces in Thailand. Similarly, Kawichai et al (2020) observed that PAH levels were significantly higher in Chiang Mai (Thailand) when there were high levels of biomass burning in 2017 when compared to levels measured during periods of low biomass burning. In addition, Wiriya et al (2013) demonstrated that particle bound PAH levels increased in Chiang Mai (Thailand) during the dry season (i.e. haze period) likely as a consequence of open (biomass) burning (wildfires). Previous in vitro investigations showed that PAHs enhance particle-induced inflammatory responses including the release of pro-inflammatory cytokines as well as the stimulation of NFκB (e.g. Bonvallot et al 2001, Kocbach et al 2008. Thus, future studies should compare the PAH content of PM samples emitted from different sources in parallel to investigating their toxicity.

Cell morphology
Light and electron microscopy revealed that PM could stimulate cytoplasmic vacuole formation in macrophages. Whilst not commonly reported in the published literature investigating the toxicity of PM, similar observations have been made in studies which have investigated responses of other cell types to PM (e.g. Piao et al 2018, Lawson et al 2020 or nanoparticles (Wu et al 2015). Cellular vacuolisation is known to be associated with various forms of cell death including methuosis, paraptosis and autophagy (Maltese and Overmeyer 2014). The origin of the vacuoles observed following exposure to PM is not known, but could be investigated in the future.
Light microscopy images suggested that all particle types may be internalised by cells, with the greatest level of cellular interaction observed for ufCB and DEP. However, it is not possible to confirm the uptake of particles by cells using light microscopy as the resolution is often not sufficient to visualise individual particles. It is also challenging to assess the uptake of carbon particles by cells using other forms of microscopy that have greater resolution (e.g. transmission electron microscopy) as their appearance can be similar to sub-cellular structures and it is not possible to confirm their presence based on elemental analysis due to the high carbon content of cells. Flow cytometry has been used quantify particle uptake by cells (e.g. Stringer and Kobzik 1998), and this approach could be explored in future studies. However, it was not possible to quantify particle uptake in this study due to the limited quantities of PM that were available to test.

Study novelty, limitations and future directions
Existing in vitro and in vivo studies have assessed the toxicity of ambient PM collected from different locations, however the majority of existing studies have tested PM samples from the EU or US (e.g. Wegesser et al 2009, Danielsen et al 2011, Franzi et al 2011. Whilst epidemiology studies have been conducted in Thailand to investigate the adverse health effects of different sources of PM (Ruchiraset and Tantrakarnapa 2018, at this time no published studies could be identified that investigated the toxicity of traffic or biomass burning derived PM samples collected in Thailand using in vitro or in vivo models. Our study revealed that the toxicity of non-haze (biomass burning derived) PM samples collected in Thailand (Chiang Mai) was greater than that of haze (traffic derived) PM, which suggests that PM emitted from traffic may be more toxic. Only a limited number of studies have directly compared the toxic potency of different PM sources in the same study and interestingly often have conflicting results with regards to what emission source is more toxic (e.g. Karlsson et al 2008, Corsini et al 2013, Miousse et al 2015. Jin et al 2016, Longhin et al 2016. Thus, more research studies are needed which directly compare the toxicity of PM emitted from different sources. Within our study, whilst one emission source of PM (e.g. traffic or biomass burning) will dominate when the haze and non-haze PM samples were collected there is unlikely to be only one source of PM emissions during each sampling period. For example, during haze events the main source of PM emissions in Chiang Mai will be biomass burning, but there may also be a more limited contribution from other sources (e.g. traffic). Thus, future studies could quantify the levels of a tracer of wood burning (e.g. levoglucosan) in PM samples to identify whether biomass burning contributes to PM emissions (Schkolnik andRudich 2006, Fuller et al 2014). Alternatively, PM emissions from biomass burning or diesel engines can be generated in a laboratory setting (e.g. Steerenberg et al 1998, Corsini et  . Whist we measured the hydrodynamic diameter and zeta potential of PM suspensions in relevant biological media to the hazard assessment, and quantified the metal and endotoxin content of PM samples it was not possible to assess the PAH content and oxidative potential due to limited quantities of samples available and due to the fact that the research was performed during the COVID-19 pandemic which restricted what laboratory work could be performed. It is therefore recommended that future studies should perform a more comprehensive assessment of the physico-chemical properties of the PM samples under investigation to allow better identification of what PM properties confer toxicity. In our study, comparisons of PM toxicity were made on a mass basis and did not take into account the exposure level of PM during each sampling period. The levels of PM were higher during haze events in Chiang Mai than those measured in the non-haze period. Thus, a risk assessment which encompasses both hazard and exposure data would be required to identify the potential detrimental impact of PM on health. Mitschik et al (2008) reviewed in vitro studies which had investigated the pulmonary toxicity of PM and recommended that PM concentrations of 50-100 µg ml −1 and exposure times of 9-24 h were used. The concentrations of PM we selected therefore aligns well with their recommendations and the concentration ranges used in the published literature.
Our study focused on investigating the toxicity of PM 10 collected during haze and non-haze periods, however future studies should assess the toxicity of PM 2.5 samples collected from Thailand, as PM 2.5 is of greater concern to health.
We used a (mouse) J774A.1 macrophage cell line in our study due to the prominent role that macrophages play in the clearance of inhaled particles (Wiemann et al 2016). Previous studies have shown that macrophage cell lines can be used as part of tiered testing strategies to minimise the requirement to perform in vivo testing as they can provide a good prediction of in vivo short term inhalation hazard potential (Wiemann et al 2016). Whilst there are morphological and functional similarities between macrophage cell lines (e.g. J774.A1 cells) and primary human or rodent macrophages, there are also differences (Chamberlain et al 2009, Boyles et al 2015. For example, it has been observed that whilst macrophage cell lines (such as J774.A1 cells) can be activated by chemical stimuli (e.g. LPS) they often exhibit lower levels of cytokine expression and production than primary cells and they can differ with respect to their cell surface markers and morphology (Chamberlain et al 2009). Accordingly, the use of primary human or rodent (alveolar) macrophages could be considered in future studies to further probe the suitability of using cell lines to predict PM toxicity. However, the use of primary cells has more ethical implications, and is often more costly and time consuming and the use of cell lines allows a rapid, cost effective screen of particle toxicity to be performed. Importantly, our research is aligned with the principles of the 3Rs (to reduce, refine, and replace animal testing) and the use of an in vitro model to evaluate the toxicity of PM samples has the added benefit that it makes testing more ethical, cheaper and faster than in vivo studies.

Conclusion
Using a macrophage in vitro model we have demonstrated that haze and non-haze PM collected from Chiang Mai, Thailand can elicit toxicity. There is evidence from our research that non-haze (traffic derived) PM causes a more potent inflammatory response when compared to haze (biomass burning derived) PM, which may reflect the higher concentration of endotoxin and metals found in non-haze PM. In this study the toxicity of PM 10 samples was compared at an equivalent mass concentration and the same duration of exposure. However haze events can result in large increases in PM concentration which greatly exceed WHO guideline levels and the duration of human exposure to different sources of PM can vary. For example, in our study in the levels of PM 10 in Chiang Mai reached 149 µg m −3 during haze events, whereas the maximal level observed at other times was 28.5 µg m −3 .
Currently, PM regulations in Thailand and internationally that aim to protect human health are based on mass concentrations of PM without consideration of the source of PM emissions. A greater understanding of whether the toxicity of PM emitted from different sources varies will provide an evidence base that informs the introduction of appropriate policy intervention strategies to improve the health of the population of Thailand, as well as other countries. This will lead to positive outcomes on the economic development and welfare of the global population as air pollution is associated with a spectrum of adverse health effects and economic losses (e.g. due to; lost productivity as a consequence of absences from work, increased medical costs associated with the adverse health impacts of PM, increased premature deaths and less tourism).

Data availability statement
The data generated and/or analysed during the current study are not publicly available for legal/ethical reasons but are available from the corresponding author on reasonable request.