摘要
化学组成是大气细颗粒物(atmospheric fine particulate matter,PM_(2.5))健康危害的重要因素,但目前缺乏对于PM_(2.5)效应组分尤其是复杂有机污染物的系统认识.效应导向是系统识别效应组分的关键策略,这亟需稳健可靠的筛选方法.研究针对2016~2018年北京PM_(2.5)样品,采用人肺上皮细胞模型,以暴露后的活性氧和5种细胞因子为效应指标开展研究.针对有机组分间的多重共线性,组分与效应指标之间复杂的线性、非线性或交互关联,研究建立了以单污染物模型为基础、结合偏最小二乘回归和贝叶斯核机器回归的复合筛选策略,以兼容高维有机组分与细胞效应间的广义回归关联,从61种多环芳香组分识别出10种稳健的PM_(2.5)效应组分,包括优先控制的高环多环芳烃(PAHs)和具有湾区结构的烷基PAHs,并从模型和效应层面交互验证,验证了复合筛选策略的模型稳定性、外延性及结果可靠性,降低假阳性的概率.本研究针对真实的PM_(2.5)样品,在高维有机组分与效应的广义关联特征下,解决稳健筛选的难题,为科学识别PM_(2.5)中关键效应组分提供了关键的模型方案.
Air pollution is a major public health risk,with fine particulate matter(PM_(2.5))being the dominant contaminant.The chemical composition is an important factor influencing the health effects of PM_(2.5),but it is extremely complex with high spatial and temporal variability,especially the organic component.Organic components are commonly lipid-soluble and bioavailable,and can cause complex effect endpoints through a variety of mechanisms.Therefore,the identification of key organic effector components is crucial for the systematic understanding of PM_(2.5) health effects.Effect-orientated offers a potential solution for the identification of toxic components from a large number of signals(either known or unknown pollutants)through statistical methods based on specific effect indicators.Establishing robust and reliable screening methods for high-dimensional organic components database,and their generalized associations with health effects,is critical for effect-orientated studies.In this study,an in vitro exposure assay was conducted with human pulmonary epithelial cells(BEAS-2B)and PM_(2.5) samples.A total of 122 PM_(2.5) samples were collected during 2016-2018 in Beijing,China and the organic components were soxhlet-extracted.Gas chromatography tandem triple quadrupole mass spectrometry(GC-MS/MS)was utilized for the targeted detection of polycyclic aromatic compounds.Reactive oxygen species and five cytokines,including IL(interleukins)-1β,IL-6,IL-8,TNF(tumor necrosis factor)-αand VEGF(vascular endothelial growth factor),were used as effect indicators.The exposure duration(12 h)and exposure concentration(100μg/mL)of PM_(2.5) extracts exposed to BEAS-2B cells were determined by cell activity and cell membrane permeability.Significant seasonal differences were observed for all cellular effectors,except IL-6,after exposure to the same mass concentration of PM_(2.5) extracts,suggesting that PM_(2.5) chemical composition is a pivotal factor influencing the cellular effects.Based on the multiple collinearities among organic compounds and the complex linear,nonlinear,or interactive associations among PM_(2.5) components and effect indicators,we established a screening strategy based on the single-pollutant model.The model combined partial least-squares(PLS)regression and Bayesian kernel machine regression(BKMR),and identified generalized regression associations among high-dimensional organic compounds and cellular effects.PLS regression enables to solve the multicollinearity between organic components,investigate the associations between multiple components and cellular effects,and identify the key components.BKMR can probe the non-linear associations and interactions between multiple components and cellular effects,explore the overall associations between components and effects through dose-effect profiles,and identify key effector components.This multiple screening strategy utilizes the cross-validation between multi-models and multi-effecters to effectively reduce the probability of false positives and increase the robustness and reliability of the results.Ten robust effect components were identified,screened from 61 target pollutants,including prioritized controlled hypercyclic polycyclic aromatic hydrocarbons(PAHs)and alkylated PAHs containing a bay structure.Furthermore,the stability,extensibility,and reliability of this strategy were validated by(1)elaborating the stability of the model through self-validation of the analyzed samples,(2)elaborating the extensibility of the model through other years of data,and(3)discussing the robust effector components through the available toxicological or epidemiological literatures on their mechanisms and health effects.In summary,this study established a cross-validated strategy to address the challenge of robust screening under the generalized associations among high-dimensional organic components and effects of PM_(2.5) samples.Effect-oriented studies based on stable and reliable multi-models provide a modeling solution for the identification of key toxic components in real environmental pollutants,such as PM_(2.5).
作者
柳金明
史咲頔
钟沛文
李艾霖
邱兴华
Jinming Liu;Xiaodi Shi;Peiwen Zhong;Ailin Li;Xinghua Qiu(College of Environmental Sciences and Engineering,State Key Joint Laboratory of Environmental Simulation and Pollution Control,and State Environmental Protection Key Laboratory of Atmospheric Exposure and Health Risk Management of Environmental Protection,Peking University,Beijing 100871,China)
出处
《科学通报》
北大核心
2025年第22期3650-3658,共9页
Chinese Science Bulletin
基金
国家重点研发计划(2022YFC3702704)
国家自然科学基金(42293324,41961134034)资助。
关键词
细颗粒物
有机组分
筛选策略
细胞效应
效应导向分析
fine particulate matter
organic components
screening strategy
cellular effects
effect-directed analysis
