Seasonal Trends, Profiles, and Exposure Risk of PM 2.5 -bound Bisphenol Analogs in Ambient Outdoor Air: A Study in Shanghai, China

Fine particulate matter (PM 2.5 ) possesses a larger surface area, which enables hazardous chemicals to adsorb. The particle can lodge deep in the lungs and bronchi of humans, causing diverse cardiovascular and respiratory diseases. PM 2.5 exposure has significant socioeconomic repercussions as well as an increased risk of mortality. Some features of PM 2.5 components have yet to be fully comprehended. PM 2.5 -bound bisphenols (BPs), which mostly originate from the


INTRODUCTION
Following a rapid urbanization and industrial expansion, China has suffered a dramatic increase in air pollution over the last three decades, and particulate matter (PM) has long been a public concern in large cities. PM comprises a diversity of components including polycyclic aromatic was measured alongside its five analogs within the outdoor PM2.5 samples, with the principal goals of determining BPs' profiles, evaluating seasonal patterns, assessing exposure risks, and examining the influence of weather conditions on BPs' concentration. In addition, a comparative analysis of our findings with former reports on air particulate BPs in China (Table 2) was also considered, and played an intrinsic role when interpreting the results. Furthermore, some pivotal insights for the progress of research on atmospheric BPs were also unveiled.

Sampling
PM2.5 samples were collected for a 24 h period at 100 L min -1 daily between July 5, 2019, and November 6, 2020 from two sites ( Fig. 1) in Shanghai (China) using a medium volume air sampler (TH-150C, China) with a flow rate of 100 L min -1 . One of the sites was a rooftop of a 4-story building in Xuhui campus of East China University of Science and Technology (ECUST) (31°08'43"N, 121°25'31"E). The sampler was placed 15 m above the ground, and the sampling inlet was placed about 1.5 m above the instrument's placement floor far from whatever interference. ECUST's

ECUST,Xuhui 2. Pudong 55 Lingshan road
Xuhui campus is located in the southwestern part of Shanghai. It is an urban residential district exempted from high-rise skyscrapers, the reason why it is not subjected to the industrial pollution originating from its close neighborhoods. Two main roads (Inner ring and outer ring) entangle the site at approximately 2000 km away from each. Humin road is located to the north and is 1000 m away from the site, while the central ring road is situated 1.5 km from the site. These are the major traffic networks. More minor-duty roads (Meilong and Lao Humin) are 600 m to the south and 300m to the east respectively. The predominant wind direction in Shanghai is North-West in winter, and South-East in summer. The second sampling site was at Pudong (55 th Lingshan Road) (23°51'21.384"N, 120°39'49.464"E). Prior to sampling, quartz membrane filters (Whatman, grade QM-A, pore size: 2.2 mm, width: 90 mm, UK) onto which PM2.5 was adsorbed, were baked (450°C for about 5 hours) using a muffle furnace (Thermo scientific™ Lindberg/Blue M™, USA) in order to get rid of background contamination. Thereafter, they were conditioned for 24 hrs at a temperature of 25 ± 1°C and relative humidity of 40 ± 5%, and then weighed using a highresolution scale (CPA-26P Sartorius, German). Subsequently, they were wrapped separately in aluminium foil, and sealed in polyethylene zip bags before their deployment to the site. To avoid the damaging which could possibly occur subsequent to the sampling, the samples were stored within a medium at a temperature below 20°C until the extraction.

Extraction and Cleaning
The previously reported method by Xue et al. (2016) was used as a reference with minor modifications (Fig. S1). Because analytes can sorb onto glassware, only pre-baked (450°C for 5 hours) glasses were utilized all along the process (Berkner et al., 2004). Field blanks of quartz fibrous membrane filters (QFFs: 90 mm, UK) were prepared alongside filters containing PM2.5 samples. Firstly, the outer edges of the QFFs which didn't contain PM2.5 were removed and discarded. The remnants were then chopped into tiny pieces (1 cm × 1 cm), and placed into 40 mL glass tubes which had been pre-rinsed 3 times in methanol. Afterward, a 100 µL volume of a 1 ppm solution of 13 C12-BPA in methanol was spiked in each of the glass tubes, before a 25 mL volume of ethyl acetate (HPLC grade) was added (Xue et al., 2016). Subsequently, the glasses were covered with tiny pieces of Al foil, and capped tightly. Both flasks (one with a blank and the other with the actual PM2.5 sample) were ultra-sonicated (water bath temperature: 55°C) for 30 minutes. The mixture in each glass was then centrifuged (Eppendorf 5804, Hamburg, Germany) at 4000 rpm for 5 min, and the extract was isolated from the residue using glass pipettes before being transferred into an eggplant bottom-shaped glass that had been pre-rinsed 3 times with methanol. To ensure a thorough extraction of analytes, the solvent addition, sonication, and centrifuging steps were replicated twice with the residue samples (Nakazawa et al., 2014). This time, 20 mL and 15 mL of ethyl acetate were sequentially utilized, and the resulting extracts were lumped with the initial extract. The mixture was concentrated using a rotary evaporator (Water bath: 55°) under vacuum condition until miniscule volumes (~1 mL) remained. The extracts were transferred to the vials using glass pipettes which had been pre-rinsed with methanol. To prevent analyte loss following pipetting, each of the glasses was rinsed three times with a small volume of methanol, and the solution was added to the extract in the vial. The extracts were concentrated to dryness by blowing over them a gentle stream of nitrogen air. Each of the extracts was then re-suspended in 1 mL of methanol and vortex mixed. The extracts were ultimately filtered using glass micro syringes (1000 µL), before being transferred into fresh vials for liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis.

LC-MS/MS Analysis
The analysis parameters have been reported elsewhere (Xue et al., 2016) with minor modifications. Shimadzu Prominence Modular HPLC system (Shimadzu, Kyoto, Japan) having an autosampler, system controller, and a binary pump was utilized for chromatographic separation. The Shimadzu LCMS-8050 electrospray triple quadrupole-mass spectrometer (ESI-MS/MS) (Shimadzu, Kyoto, Japan) helped with analyzing the analytes. The used column was Acquity UPLC ® CSH™ C18 1.7 µm. The mobile phase consisted of Milli-Q water that contained 0.03‰ of ammonium hydroxide (B) and methanol (A). The injection volume was 10 µL. To separate the analytes, a gradient elution mobile phase was set at a flow rate of 0.4 mL min -1 starting at 5% (v/v) A, held for 1 min, taken to 99% A within 4 min, held for 1 min, then returned to 5% A at 6.1 th min, and maintained for 2 min for an overall run time of 8 min. The MS/MS system was operated in multiple reaction monitoring (MRM) negative ion mode. Nitrogen was utilized as a collision gas and nebulizer. The flow rates for the nebulizing, heating and drying gases were 3 L min -1 , 10 L min -1 and 10 L min -1 consecutively. The interface and DL temperatures were 300°C and 250°C respectively. The heat block temperature was 400°C, and the even time equaled to 0.026 seconds. The identification of BPs was brought by comparing mass spectra with default library data as well as the data for the used standards.

Quality Assurance/Quality Control (QA/QC)
The isotope dilution method concerning the internal standard ( 13 C12 BPA), helped to conduct quality control (QC) and quality assurance (QA) for bisphenols. To evaluate the performance of the method, factors such as repeatability, linearity, recovery, matrix effect and sensitivity were cautiously monitored before and after sample measurement. There wasn't a malfunction of signal suppression or enhancement in electrospray ionization (ESI) as there wasn't a coelution of matrix compounds observed in the procedure. The target compound was identified by referring to the retention times proportionate to the internal standards. The compound also helped correct the effects associated with the matrix compounds. For the computation of each target analyte, the multi-level (7 points) linear calibration curves were utilized and possessed R 2 > 0.9. To construct the curves, each target analyte's concentration-response factor (a ratio between peak area of analyte and peak area of IS) was plotted against the response-dependent concentration factor (a ratio between concentration area of analyte and concentration area of IS). In order to check if the instrument drifted in response factors, a progressive injection of calibration standards was necessitated after every batch of 10 samples. There was also a check for carryover of analytes by injection of pure solvent (methanol) after the same number of samples. The calculated recoveries (average ± SD) for the matrix-spike (n = 10) varied from 71 ± 6% to 83 ± 10%, when the recoveries for blank spike (n = 13) ranged from 89 ± 3% to 93 ± 5% (Table S3). Three randomly picked samples were repetitively analyzed (7times) to ensure repeatability of the method. The standard deviations for the calculated concentrations were: BPA (0.04), BPB (0.05), BPP (0.02), BPF (0.04), BPS (0.06), and BPAF (0.04) ( Table S3). The instrumental quantification limits (LOQs) which varied from 0.0005 ng m -3 to 0.0007 ng m -3 were determined referring to the lowest concentration point of the calibration curves with a signal/noise (S/N) > 10. The same procedure was utilized to determine method quantification limits (MLOQs) which varied from 0.02 ng m -3 (BPAF) to 5.00 ng m -3 (BPF). The instrumental detection limits (LODs) which varied from 0.006 ng m -3 (BPAF) to 1.515 ng m -3 (BPF) were obtained using the least concentrated standard solutions with S/N = 3 (Table S3).

Data Analysis
IBM SPSS software (version 22.0) and Microsoft excel 2013 helped for statistical analysis. The concentrations inferior to the method's limit of quantification (MLOQ) were replaced with values equal to the half of MLOQ when determining the median and arithmetical mean (Xue et al., 2016). The fractions (%) of data with concentrations close to the detection limit were: BPA (7%); BPB (14%); BPP (65%); BPF (5%); BPS (6%) and BPAF (8%). For the conformation to normality, all the data underwent the logarithmic transformation (Wilson et al., 2007;Xue et al., 2016), and in this regard, the Kolmogorov Smirnov test (Table S4) was used together with the Normal probability plots (NPP) (Fig. S2). The assessment of the relationships between BPs' clusters was conducted using Pearson correlation analysis (Table S6), as the datasets were concluded to be normally distributed following the normality testing. The samples with BP concentration exceeding the limit of detection (LOD) were solely used during the correlation analysis (Xue et al., 2016). A oneway analysis of variance (One way-ANOVA) alongside a Tukey's Honestly Significant difference (Tukey HSD) test (Table S10), helped to examine the difference between groups of data. The correlation between two sets of data was considered as statistically significant in case p ˂ 0.05.

Estimating the Daily Intake and Health Risk
To evaluate non-cancer risks as a consequence of the inhaled PM2.5-bound BPs, we referred to a recently published approach by (Liu et al., 2021b) which is briefly outlined hereafter. Two indicators (hazard quotient, HQ and estimated daily intake, EDI) (U.S. EPA, 2000) were estimated as following: where, [BP] stands for PM2.5 load of BP analogue expressed in ng m -3 . EDI, (1) is the daily inhaled quantity for each BP analogue expressed in (ng kg BWt -1 Day -1 ). BWt, stands for a body weight of the exposed human, and IR depicts the inhalation rate in m 3 Day -1 . The summary for all the involved variables is available in Table S5. In order to determine HQ (2), we referred to the following equation: were, TDI (ng kg BWt -1 Day -1 ) (3), stands for the maximal tolerated daily intake for the chemical. In case the reference dosage (RefD) is available, it can substitute this variable. In case both RefD and TDI are not provided for a chemical, the estimation of TDI is undertaken following the Eq. (3) presented below: In which, UFs stand for uncertainty factors, and NOAEL depicts the non-observable-adverseeffect-level. The additional explanations concerning the above variables in Eq. (3) are detailed in the captions of Table S7.

BPs' Detection Level
The reporting on atmospheric BPs other than BPA remains scarce across the globe, and in particular, there isn't any pertaining to Shanghai city. Thus, for an explicit and comprehensive understanding of our results, the comparison of data was directed to other regions in China as well as other locations where the atmospheric BPs were previously measured. In general, BPs were unevenly detected in samples from both sites and various sampling periods, and there wasn't any sample free of all target BP analogues. The detection proportions for the general campaign, different locations as well as different sampling months are respectively illustrated in Fig. 2 and Table 1. Bisphenol A was the predominant analogue in all samples (88%) followed in decreasing order by BPF (82%); BPS (75%); BPAF (63.1%); BPP (15.5%) and BPB (9.5%) (Fig. 2). Similarly, the predominant detection frequencies for the three BPs (BPA: 100%, BPF: 98% and BPS: 86%) ( Table 2) were also observed in PM2.5 samples from China's Pearl River Delta region (Liu et al., 2021b). This also aligns with the reported PA and BPF detection frequencies (75.9% and 63.9% respectively) within particulate phase samples in indoor air from Albany (New York) (Xue et al., 2016).
Spatially, BPA demonstrated a higher detection frequency in Samples from Xuhui campus (91%), and it was the second most abundant (60%) after BPF (70%) in samples from Pudong (55 th Lingshan Road) (Fig. 2). Two BP analogues (BPB & BPP), were not identified within samples from this site. In fact, these analogues have been rarely identified in air samples. In addition, the sampling from this site occurred in summer, when there is a low possibility to detect BPs in suspended atmospheric samples. As it is illustrated in Table 1, the samples gathered in colder months (January and December), possessed the relatively higher detection frequencies (e.g., January 2020: DR(BPA) = 94.1%, DR(BPF) = 88.2%; December 2019: DR(BPA) = 100%, DR(BPF) = 83.3%). This is ascribed to the strong bonding linking bisphenols to the polymeric matrix which averts their direct evaporation to air during hotter months (Fu and Kawamura, 2010). In addition, we can mention the alteration of atmospheric equilibrium caused by the increase of local sources (e.g., solid waste burning) as well as the decline of photochemical reactions by hydroxyl radicals which occurs during colder months Salapasidou et al., 2011).

Concentration of BPA
The determined concentrations for six target BPs for the overall campaign are illustrated in Fig. 3, and Table 1 summarizes data for the respective sampling months. The concentration snapshots for the respective sampling seasons are presented in Fig. 4 and Table S1, while the clustered profiles for identical seasons are depicted in Fig. 5) The previously reported BPs concentrations in airborne particles in China are summarized in Table 2. According to results from the analysis, the concentrations of BPA in all PM2.5 samples varied from less than MLOQ to 7.52 ng m -3 (Median: 2.40 ng m -3 , Mean: 2.75 ng m -3 ). The mean BPA concentration is slightly below the previously reported concentration (Mean: 4.77 ng m -3 ) in Shanghai (Table 2) which underpins the formerly reported efficient control over the atmospheric pollution in China . The mean BPA concentration is also one order of magnitude lower than the previously reported mean BPA concentration in total suspended particles (TSP) from the Chinese city of Xian (47 ± 41 ng m -3 )  as well as mean BPA concentrations (Summer: 5.9 ± 1.4 ng m -3 to Winter: 20 ± 2.7 ng m -3 ) measured earlier in atmospheric particulates (PM2.5) from the southern Chinese city of Guangzhou . The maximal BPA concentration is higher but in the same order of magnitude as the lately reported maximum BPA concentration (1.65 ng m -3 ) in PM2.5 sampled from Pearl River Delta (China) ( Table 2). In fact, Pearl River Delta region is a hub of casual e-waste recycling and dumping on a global scale and reasonably in China (Gu et al., 2010;Zeng et al., 2016). Provided that that e-waste incineration and dumping is the principal source of BPA in this region (Fu and Kawamura, 2010), and the strict regulations in place since January 2018 (Zeng et al., 2016) which preclude the informal e-waste handling (recycling) and importation in China, the environmental concentration of BPA may have been gradually declining in Pearl river Delta region. As for Shanghai, there is a bigger industrial complex and commercial activities (Huang et al., 2012), which apparently generate a higher quantity of plastic wastes. The maximum BPA concentration is also one order of magnitude below the reported BPA concentration in PM10 (0.03-47.3 ng m -3 ) sampled from urban atmosphere of Thessaloniki (Greece) (Salapasidou et al., 2011) as well as mean BPA concentration (18 ± 14 ng m -3 ) found in total suspended particles (TSP) sampled from the Indian city of New Delhi . The survey conducted on ubiquity of BPA in atmosphere by Fu and Kawamura (2010) confirmed the occurrence of BPA within atmospheric aerosols gathered from different regions in China and elsewhere, with the concentrations varying between 0.001 and 17.4 ng m -3 which is consistent with BPA concentration range in this study.  In view of PM2.5 samples collected in different months, the highest BPA concentrations were measured in January of 2020 and December of 2019 with concentration ranges (Mean) from below MLOQ to 7.5 ng m -3 (3.4 ng m -3 ), and from 1.43 to 7.14 ng m -3 (3.16 ng m -3 ) ( Table 1). Given the relatively lower average temperature (7.8°C and 9.7°C respectively), the concentration of hydroxyl radicals (OH) available to oxidize BP molecules was lower within these months. Additionally, the dropping temperature leading to the higher condensation of BPs on particlephase, underpins the increased BPs' concentration in PM2.5 samples collected within colder months. Previously, Matsumoto et al. (2005)

Concentrations of BPs other than BPA
As depicted in Fig. 3 and Fig. 4, a total number of five BP analogues were analyzed alongside BPA. Three BPs (BPF, BPS and BPAF) demonstrated the highest concentration in PM2.5 samples with the ranges (Mean; Median) of < MLOQ to 6.32 ng m -3 (2.4 ng m -3 ; 2.0 ng m -3 ), < MLOQ to 4.61 ng m -3 (0.3 ng m -3 ; 0.03 ng m -3 ) and < MLOQ to 3.98 ng m -3 (0.8 ng m -3 ; 0.3 ng m -3 ) respectively (Fig. 3). The maximal concentrations are respectively 1-3 orders of magnitude higher than the reported concentration ranges (BPF: < MLOQ to 0.359 ng m -3 , BPS: < MLOQ to 0.01 ng m -3 and BPAF: < MLOQ to 0.0087 ng m -3 ) of BPs in PM2.5 from Pearl River Delta, even though the decreasing order for the maximum concentrations is the same (BPF > BPS > BPAF) (Liu et al., 2021b). This may indicate the degree to which these analogues are prioritized when it comes to substituting BPA at industrial scale. Prior to this study back in 2014, Xue et al. (2016) attempted to determine the concentrations of three BP analogues (BPF, BPAF and BPS) within indoor air particulates from Albany, New York (USA), and the reported overall mean concentrations for BPF and BPS (3.4 ng m -3 and 0.13 ng m -3 respectively) were in the same order of magnitude as the mean concentrations for the same BP analogues in our study. However, the mean concentration for BPAF (0.03 ng m -3 ) was one order of magnitude below the mean concentration in our study. The ranges and averages concentrations for two BP analogues (BPB and BPP) (Fig. 3) were the lowest for either the overall campaign (From 0.001 to 3.83 ng m -3 , 0.129 ng m -3 ; from 0.007 to 3.66 ng m -3 , 0.141 ng m -3 ) (Fig. 3) or both sampling sites (Table 1). In combination with their least detection frequency for all the campaign and both sampling sites (Fig. 2), we may conclude that these two analogues are less opted as the suitable replacements of BPA in China. This may be evidenced by the previous study on 240 total dietary samples from several Chinese regions including Shanghai (DRBPB: 0%) which concluded that" Chinese people are not exposed to BPB through environmental or ingestion pathways" (Yao et al., 2020). Earlier before, similar observations had been reported during the studies on breastmilk and urine samples (Niu et al., 2017;Yang et al., 2014). Regarding the different sampling periods, we noticed that the samples collected within colder months exhibited the highest concentrations for all BP analogues, and the peak concentrations were found in January of 2020, which matches the finding with BPA (Table 1). Within this month, three BP analogues (BPF, BPS and BPAF) possessed the utmost concentrations with the ranges (Mean; Median) from below MLOQ to 6.3 ng m -3 (Mean: 2.5 ng m -3 ; Median: 2.1 ng m -3 ), < MLOQ to 4.6 ng m -3 (Mean: 0.7 ng m -3 ; Median: 0.51 ng m -3 ) and < MLOQ to 3.9 (Mean: 1.15 ng m -3 ; Median: 0.62 ng m -3 ) respectively.

Seasonal Trends
Considering BPA levels in PM2.5 samples collected in different seasons, there was a clear difference in concentration between cold and warm seasons (Table S1; Fig. 4). In fact, the colder seasons were noted to exhibit relatively higher mean BPA concentrations (winter, 2019: 3.55 ng m -3 & winter, 2020: 3.42 ng m -3 ) compared to the hotter season (Summer, 2019: 1.61 ng m -3 ). This is in line with the earlier report on BPA concentration in urban areas of China for winter (2.0-20 ng m -3 ) and summer (1.0-10 ng m -3 ) periods . It is also consistent with the measured concentration of BPA in PM2.5 from the urban atmosphere of Cordóba, Argentina (Graziani et al., 2019), and within airborne particulates from Netherland and Italy (Cecinato et al., 2017) where the higher BPA levels were measured within colder months.
To shed light on the causes of fluctuating seasonal BPA concentrations other than their aforementioned tight bonding to the polymeric matrix , we can also consider climatic events taking place in warm periods such as a higher sunlight intensity, elevated mixing layer's height, raised temperature, and occasionally the rain washing the atmospheric particles, as the probable sources for a decline in BPs concentrations frequently observed in hotter periods (Lee et al., 2001;Xia et al., 2013). The decrease in mean BPA concentrations noticed between the samples collected in winter and those gathered in the summer of 2019, is roughly equal to 54.4%. This proves that the control measures concerning curbing the emission of air pollutants have been effective in China. Specifically with bisphenols, the imposed termination of informal recycling and importation of e-waste from developed countries which has been in effect since January 1 st 2018 in China (Zeng et al., 2016) (Table 3), there was a higher mean BPA concentration within air samples collected in the Fall of 2020 compared to samples gathered in the Fall period of the previous year. Although lockdown-induced restrictions had paralyzed the economy, the aftermath of the outbreak marked by a quick resumption of social economic activities might have been linked with a somewhat higher mean concentration of BPA in air samples gathered in the Fall of 2020.
Taking into account BP analogs other than BPA, previous studies (Liu et al., 2021b;Xue et al., 2016) in which the measurements were undertaken in ambient outdoor air samples, did not discuss the influence a seasonal shift may have on BP concentrations. In this study, the same tendency observed for BPA was similarly reiterated with other BP analogs, meaning that the concentrations were seemingly higher in PM2.5 samples collected in colder seasons (Fall and winter) than samples gathered in the hotter season ( Fig. 4; Table S1). As shown in Fig. 5, BPF concentrations in samples collected in three investigated seasons (Winter, Fall, and summer), exceeded the concentrations for the remaining 4 analogs with the highest concentration found in PM2.5 samples collected in the winter of 2020, and varying from below MLOQ to 6.32 ng m -3 (Mean: 2.47 ng m -3 ; Median: 2.14 ng m -3 ). The differences in mean concentrations between winter and summer samples for three frequently identified BP analogs excluding BPA (BPF, BPAF and BPS), were roughly 39%, 71% and 79.3% respectively, when the decreasing orders of the mean concentrations for BP analogs other than BPA in winter, fall and summer samples, were:  (Fig. 4). The similarity of decreasing magnitude of mean concentrations for three BPs (BPF, BPS and BPAF) in all seasons, may be a typical motive corroborating the choice of these BPs over the other two analogs (BPP and BPB) for the industrial replacement of BPA in China.

Correlations between BP's Concentrations
As demonstrated by one-way ANOVA (Table S9), a statistically significant difference at 0.05 was found between the mean concentrations of BPs. Considering the Tukey HSD test results (Table S10), the significant differences in mean concentrations were identified for the majority of coupled BP concentrations with the couples CBPB and CBPP; CBPB and CBPS, as well as CBPP and CBPS, being the only exceptions (Table S10). Due to the relevance of correlation coefficients when  finding the source of pollutants, the correlation analysis between the concentrations of BP analogs was carried out for all the campaign and various sampling seasons using Pearson's correlations (Table 3; Table S5). Prior to analysis, all the analytes had to be detected at least within 4 samples (Dueñas-Mas et al., 2019), and following a combination of data for the overall campaign, a significant positive correlation (r = 0.7; p = 0.03) was found between BPA and BPB which indicates a similar emission source and co-presence for these analogs. The same relationship was also found between BPB and BPP (r = 0.349; p = 0.001) as well as BPP and BPAF (r = 0.245; p = 0.024). Furthermore, there was a significant negative correlation between BPA and BPF (r = -0.264; p = 0.015). This denotes a decreasing industrial consumption of BPA in Shanghai and BPF may be an optional replacement. As it was reported earlier by Xue et al. (2016), there is a progressive substitution of BPA with BPF in epoxy resins especially utilized in the automation industry. China being atop the global market in vehicle purchasing and manufacture (MEE, 2010), and Shanghai a hub of the country's automotive industry with a steeply increasing vehicle ownership (3.2 millions in 2016) (NBS, 2017), there must be a huge consumption and the environmental emission of BPF. Giving more attention to samples in each season, the strong and positive correlations were observed between the concentrations of BPA and BPP within samples collected in Fall of 2019 (p < 0.05, r = 0.837), Summer of 2019 (p < 0.05; r = 0.876) as well as between BPP and BPF concentrations within samples collected in winter of 2020 (p < 0.05; r = 0.769). The earlier reports on atmospheric BP analogs have not discussed the correlation between these compounds, which somewhat, led to superficial conclusions regarding these analytes. Thus, we recommend further studies on airborne BPs, to look deeper into their relationships as they may be of paramount importance when tracing the origin of these contaminants. Nevertheless, a dearth of studies have endeavored to determine the relationship between BP analogs in other sorts of samples from China, and apparently, there are some deflections observed between correlation results from our study and the reported correlations in these reports. For instance, very low significant positive correlations were found between BPA and BPF (p < 0.05, r = 0.17), and between BPA and BPS (p < 0.05, r = 0.21) concentrations when BPs were examined within indoor dust samples collected from China and 11 other countries . Similarly, an insignificantly positive correlation (p < 0.05; r = 0.142) was found between the concentrations of BPS and BPA during the measurement of BPS in urine samples from China, 6 other Asian countries, and the US. This led to the conclusion that exposure sources and pathways of these BP analogues may be the same (Liao et al., 2012). Both analogs were once again found positively correlated (p < 0.01, r = 0.17) alongside other couples of BPs like BPF and BPA (p < 0.01, r = 0.23) along with BPS and BPAF (p < 0.05; r = 0.14), when investigating BPs contamination in 13 varieties of foodstuffs from 9 Chinese cities. Thus, these pairs of BPs were once again ascribed as possessing similar sources and exposure routes (Liao and Kannan, 2014). Additional to Pearson's correlation ranks, principal component analysis (PCA) was also utilized to determine correlations between the concentrations of BPs, and the overall degree of PC1 and PC2 was 45.13%. As depicted by (Fig. 6(A)), the concentrations of BPA and BPB are much more correlated which is similar to data in Table 3. Nevertheless, some other correlations are not replicated by the model which is contrary to results obtained via Pearson's correlation. The cause may probably be the use of non-normalized data unlike in Pearson correlation analysis. In addition, the size of dataset may also compromise the accuracy of data. Therefore, we suggest further studies on BPs to conduct rigorous sensitivity evaluation of both methods while studying the relationships between concentrations.

Influence of Meteorological Parameters on BPs' Profile
Meteorological parameters (temperature, relative humidity, atmospheric pressure, and wind speed) contribute to a vigorous change in atmospheric organic pollutants' accumulation, generation, diffusion, phase partitioning, and removal (Amarillo and Carreras, 2016). Therefore, it is worthwhile to examine their nexus with concentrations of atmospheric BPs. The evaluation was preceded by logarithmic transformations and normality testing of BPs concentrations. The temporal change of temperature and relative humidity (Figs. 7(C-D)), enormously defined the daily fluctuated concentrations of BPs. As depicted in Fig. S3 and Table S4, the normal distribution of concentrations were found, and so, Pearson correlation rank was selected for assessing correlation analysis. A significant negative correlation (p < 0.05; r = -0.362) was observed between BPA concentration and mean temperature. This is in agreement with a previous report (p < 0.001, R 2 = -0.55) between BPA and temperature in Córdoba, Argentina (Graziani et al., 2019). The similar correlations were found between BPF and BPS concentrations with the mean temperature (p < 0.05; r = -0.244 and p < 0.05; r = -0.337 respectively). Contrary to the previous report by Graziani et al. (2019), there weren't correlations observed between BP concentrations with other meteorological parameters. Thus, we suggest future works to undertake more studies for a clarification. Taking a look at data from PCA analysis (Figs. 6(B-C)), temperature has distinctively defined the concentration of BPs compared to other investigated meteorological factors. This reiterates the results obtained from Pearson correlation analysis.
Pollutants co-emitted alongside BPs during e-waste handling, are equally subjected to the effects of meteorological conditions which can determine their atmospheric concentration trends. Toxic compounds found in fume and fly ash from incomplete combustion of e-waste include POPs such as polybrominated diphenyl ethers (PBDEs), polycyclic aromatic hydrocarbons (PAHs) and dioxin-like chemicals (Lin et al., 2022). Higher concentrations of heavy metals (Cr, Pb, Zn, Ni and Cu) were also found in air particles samples from sites affected by open-burning e-waste and informal recycling in Moradabad (India). It was affirmed that their concentration (From 3.64 µg m -3 to 243.310 ± 22.729 µg m -3 ) surpassed by far the concentrations in adjacent areas (Gangwar et al., 2019). In Guiyu, a Chinese town renowned for its intense e-waste dismantling activities, higher concentrations of Zn (1038 ng m -3 ) and Cr (1161 ng m -3 ) were measured in TSP samples and exceeded levels in some other Asian Cities (Shanghai, Tokyo and Seoul) (Deng et al., 2006). In addition, higher concentrations of BPs' environmental congeners (poly chlorinated/brominated dioxins and furans, PCDD/Fs and PBDD/Fs) (Brzuzy and Hites, 1996) were measured in air samples (Lin et al., 2022). Similarly to this study, a higher concentration of polychlorinated dibenzofurans (PCDD/Fs) was found in air samples (Gas + particles) collected in cold and rainy season from a sub-alpine Italian region. The elevated level was caused by intense household burning and stagnant atmospheric conditions (low winds) (Castro-Jiménez et al., 2012). The concentrations of halogenated flame retardants (HFRs) utilized in electronic and electrical equipment as well as thermoplastics, were much more significant in air samples gathered in winter (February and December) than the samples collected in Summer at Bohai Sea (China) . In PM2.5 samples from an island situated at 66 km off shore East Shanghai, the reported concentrations of 11 PBDEs was significant in colder seasons (winter & Spring) than in hotter seasons (Autumn and summer) .

HYSPLIT-based Source Tracing
Backward trajectories modeled using NOAA's HYSPLIT model which tracks the passage, helped o elucidate the most likely regions from where BPs in air samples originated. Back trajectories (BTs) of air mass determined for 120 h at a rough interval of 6 hours, and various altitudes (100 m, 500 m and 1000 m) were considered. The trajectories are depicted in Figs. 7(A-B). As represented by Fig. 7(A), in January of 2020, the air masses arriving in Shanghai, were influenced by northerly and easterly winds which majorly came from Hebei and Shandong provinces, while in July of 2019, air masses arriving in Shanghai Pudong 55 Jingshan road (Fig. 7(B)) were mainly dominated by westerly wind passing by Zhejiang province which is among Chinese regions with a bulk of industries.

Exposure Risks
BP analogues with DRs > 50% were solely considered when assessing the exposure risks. Using two indicators (Hazard quotient, HQ and Estimated daily intake, EDI), the risk estimation was conducted for children and adults (Liu et al., 2021b). As depicted in Fig. 8, the EDI potential was linearly associated with the concentration of BPs with BPA and BPF showing the highest median EDI for the adults (0.635 and 0.481 ng kg BWt -1 Day -1 respectively) and children (1.12 and 0.85 ng kg BWt -1 Day -1 respectively). The median EDI ranged from 1.32 × 10 -3 to 1.12 for children, and from 7.47E × 10 -4 to 6.35 × 10 -1 ng kg BWt -1 Day -1 for adults (Table S8). Some of the reported daily intakes in the previous studies on atmospheric BPs, are presented below (Table 4).
The range of HQs varied from 1.521 × 10 -8 to 7.02 × 10 -5 for children, and from 8.61 × 10 -9 to 3.98 × 10 -5 for adults. The cumulative medians of HQ for both adults and children (1.65 × 10 -5 and 2.91 × 10 -5 respectively) were found to be below one. Hence, the inhalation of PM2.5-bound BPs of our interest is likely to pose no or negligible threats to inhabitants of Shanghai. This is consistent with the reported risk indices for an array of plastic additives previously analyzed in PM2.5 samples from Pearl River Delta (China) (Liu et al., 2021b). Considering the contributions of individual BP analogues to the overall median HQ, BPA exhibited the highest proportion (97.08%) followed by BPF (0.88%) for both adults and children.
Following an isolation of samples gathered in winter hazy days, we have noticed the increase of BPs' concentration (from 0.03 to 2.1 ng m -3 ) with the rising of PM2.5 emission during hazy days. Similarly, the exposure risk slightly increased (HQ increased from 1.58 × 10 -5 for adults to 3.27 × 10 -5 for adults) compared to risks determined for non-hazy days. As two analogues (BPB and BPP) were sidelined when calculating both EDIs and HQs due to absence of reference doses, we therefore, recommend the forthcoming works to determine the reference doses for all BP analogues, in order to avoid uncertainty when estimating the holistic threat to humans as a result of the simultaneous exposure to numerous contaminants.   González et al., 2013) 8h-TWA (mg m -3 ) Finland (5 factories

FUTURE PROSPECTS
Due to the small number of published reports on atmospheric BPs, a couple of voids are yet to be filled in order to thoroughly comprehend the dynamics and characteristics of these pollutants. Some of works to be undertaken going forward are concisely presented below.
• BPs can stay longer on air particulates' surface and thus, they are prone to be taken to regions far away from their point sources by way of air currents, and subsequently, get deposited. However, the identification of their sources, and the intrinsic patterns for separating locallyemitted from long-range transported BPs, haven't been reported. In addition, the contribution of trajectory's geographic and topographic features to BPs' deposition and displacement should be well explained. • Like other air pollutants, meteorology may hugely impact the profiles and features of atmospheric BPs. Thus, more efforts are needed in order to deeply comprehend the existing interplays. • PM2.5-BPs are susceptible to be inhaled with other pollutants (e.g., Heavy metals, POPs) which undoubtedly results in combined synergistic health effects. Nonetheless, the developed exposure risk assessment approaches have been sidelining the other co-inhaled pollutants. We, therefore, suggest that upcoming research works should aim at developing integrated approaches capable to determine the potential health risks as a result of the simultaneous inhalation of BPs with other pollutants. • There is a lack of reports on airborne BPs from several regions of the globe which may cripple the efforts with respect to mitigation of their effects. Thus, we call upon atmospheric scientists to raise their awareness vis à vis airborne BPs and other EDCs with more attention taken to developing countries.

CONCLUDING REMARKS
Airborne BP analogs in PM2.5 samples temporally and spatially collected from two well isolated sites in Shanghai, were examined. As per results, BP levels were consistent with the previously reported data across the globe with BPA generally possessing the highest concentrations among the target BPs. Relatively higher concentrations of BPs, were found in samples collected within colder months (December, 2019 (n = 12): ∑BPs = 98.017 ng m -3 ); January, 2020 (n = 17): ∑BPs = 140.406 ng m -3 ), and the trend aligns with the formerly reported seasonal variation. Pertaining this scenario to China's environment, and taking BPA as a reference, the open incineration of municipal solid wastes is the major source, and its accumulation is significant in regard to atmospheric conditions during wintertime. The measured bisphenol concentrations were found to exceed the previously reported quantities in gas phase samples (Morin et al., 2015;Xue et al., 2016). This is compliant with their low volatility which obstructs their evaporation, and hence, increases their adsorption on PM's surface. A statistically significant difference was found between the mean concentrations of BPs, and there was a negative correlation between the daily average temperature and the concentrations of three BP analogs. The estimation of health risks revealed that the inhalation of PM2.5-bound BPs of our interest is not likely to threaten the lives of Shanghainese. However, the available risk estimation approaches may not be profoundly reliable due to the reference toxic values which are not harmonized in published reports. We, therefore, suggest future research works to be oriented to developing consensually synchronized reference values, which will ultimately play a vital role when determining the health risks associated with the inhalation of BPs in different regions of the globe.