背景和目的

既往研究提示，大气细颗粒物（Fine particulate matter，PM_2.5）短期暴露影响神经酰胺（Ceramide，Cer）水平，但大气PM_2.5长期暴露对神经酰胺水平的影响尚不明确，并且这种影响是否会进一步导致动脉粥样硬化性心血管疾病（Atherosclerotic cardiovascular diseases，ASCVD）的发生尚不清楚。本研究旨在基于我国动脉粥样硬化性心血管疾病风险预测研究（Prediction for ASCVD Risk in China，China-PAR）的两项队列，采用病例队列设计，鉴定与大气PM_2.5长期暴露相关的神经酰胺分子，探讨这些神经酰胺分子对新发ASCVD风险的影响及其在PM_2.5长期暴露致ASCVD发生中的中介效应。

材料和方法

研究对象来自China-PAR项目的两项队列，包括中国心血管健康多中心合作研究（International Collaborative Study of Cardiovascular Disease in Asia，InterASIA）以及中国心血管流行病学多中心协作研究-1998（China Multi-Center Collaborative Study of Cardiovascular Epidemiology-1998，ChinaMUCA-1998）。InterASIA和ChinaMUCA-1998分别建立于2000-2001年和1998年，并于2007-2008年、2012-2015年、2018-2020年进行面对面随访调查。考虑到神经酰胺检测所使用的生物样本，本研究将2007-2008年调查作为研究基线。排除失访者、基线前已患不稳定型心绞痛、急性心肌梗死或脑卒中者、血样不足者，剩余14496名研究对象构成该病例队列研究的原始队列。以约10%的比例随机抽取1444名研究对象作为子队列人群，剔除PM_2.5、神经酰胺水平缺失者，将1427名子队列人群纳入分析。以上述14496名研究对象随访至2020年12月31日的541名新发ASCVD病例作为病例组，剔除PM_2.5、神经酰胺水平缺失者，共530名新发ASCVD病例（80例同时属于子队列）与子队列人群组成病例队列研究人群。新发ASCVD病例定义为首次发生非致死性急性心肌梗死、不稳定型心绞痛、冠心病死亡、致死性或非致死性缺血性脑卒中。利用基线调查时采集的空腹≥10小时的外周静脉血标本，分离出血浆样本，采用基于超高效液相色谱-串联质谱技术的靶向脂质组学方法，对27种神经酰胺水平进行定量分析。由于神经酰胺水平呈现右偏态分布，在统计分析中对神经酰胺浓度进行基于秩的逆正态变换（Rank-based inverse normal transformation，INT）。采用基于卫星遥感反演技术构建的2000-2016年高精度（1km×1km）网格化月均PM_2.5数据集，估算研究对象基线调查前三年的PM_2.5年均水平。首先，基于子队列人群，采用线性回归模型分析与PM_2.5暴露相关的候选神经酰胺，以双侧检验中错误发现率（False discovery rate，FDR）校正P值（即Q值）<0.05判定显著性。其次，采用Prentice加权Cox比例风险回归模型，分析候选神经酰胺与新发ASCVD风险的关联，计算变换后神经酰胺每增加1个标准差（Standard deviation，SD）对应的ASCVD发病风险比（Hazard ratio，HR）和95%置信区间（Confidence interval，CI），以双侧检验P<0.05判定显著性。进而，采用中介分析探索候选神经酰胺在PM_2.5长期暴露与ASCVD发病风险关联中的中介作用，识别关键的中介因子。将PM_2.5长期暴露与新发ASCVD风险关联的总效应分解为直接效应和神经酰胺中介的间接效应，中介比例为间接效应在总效应中的占比。采用Bootstrap法（N = 1000）计算总效应、间接效应和中介比例的95%CI，以双侧检验P<0.05判定显著性。主要分析模型均为多变量校正模型，校正变量为年龄、性别、地理区域、城乡居住地、队列来源、吸烟、饮酒、体力活动、膳食情况、体质指数（Body mass index，BMI）、高血压、糖尿病、血脂异常、社会经济地位（Socioeconomic status，SES）和ASCVD家族史。为检验结果的稳健性，通过调整PM_2.5暴露窗口期、额外校正气象因素、校正其他脂质协变量以及排除随访一年内新发ASCVD病例等方式进行敏感性分析。

结果

病例队列研究共纳入1877名研究对象，包括1427名子队列人群与530例新发ASCVD病例，中位随访时间为10.53年。多变量线性回归模型表明，在检测到的27种神经酰胺中，Cer(d18:0/15:0)、Cer(d18:0/20:0)、Cer(d18:0/22:0)、Cer(d18:0/24:0)、Cer(d18:1/16:0)、Cer(d18:1/18:1)、Cer(d18:1/20:0)、Cer(d18:1/22:0)、Cer(d18:1/23:0)、Cer(d18:1/24:0)、Cer(d18:2/20:0)、Cer(d18:2/22:0)、Cer(d18:2/23:0)、Cer(d18:2/23:1)、Cer(d18:2/24:2)共15种神经酰胺水平随PM_2.5长期暴露浓度升高而升高（Q<0.05）。采用从2000年1月至基线调查年份的PM_2.5年均浓度、额外校正平均气温和相对湿度等气象因素进行敏感性分析，结果保持稳健。

其次，与子队列人群相比，新发ASCVD病例人群年龄更大、BMI水平更高、男性、当前吸烟、体力活动不理想、高血压患者、糖尿病患者、血脂异常人群、较低社会经济地位人群占比较高。Prentice加权Cox比例风险回归模型表明，上述神经酰胺中，Cer(d18:1/20:0)、Cer(d18:1/22:0)和Cer(d18:1/24:0)水平与ASCVD发生风险呈显著正相关。经INT变换的Cer(d18:1/20:0)、Cer(d18:1/22:0)、Cer(d18:1/24:0)水平每增加1个SD，新发ASCVD风险分别升高15.3%（HR = 1.153，95%CI：1.015-1.308）、15.5%（HR = 1.155，95%CI：1.011-1.320）和16.8%（HR = 1.168，95%CI：1.006-1.355）。排除随访一年内新发ASCVD病例、校正其他脂质协变量的敏感性分析中，上述神经酰胺仍与ASCVD发生风险呈显著正相关。

中介分析表明，PM_2.5暴露水平升高可通过升高Cer(d18:1/20:0)、Cer(d18:1/22:0)、Cer(d18:1/24:0)三种神经酰胺水平，增加ASCVD发生风险。基于多变量校正模型分析发现，经INT变换后的Cer(d18:1/20:0)、Cer(d18:1/22:0)、Cer(d18:1/24:0)水平每增加1个SD对应的中介比例分别为6.58%、13.97%和11.41%（P<0.05）。采用2000年1月至基线调查年份的PM_2.5年均浓度进行敏感性分析发现，中介分析结果保持稳健，中介比例略微减小。

结论

本研究发现，大气PM_2.5长期暴露可干扰神经酰胺代谢，导致多种神经酰胺水平升高。其中，Cer(d18:1/20:0)、Cer(d18:1/22:0)和Cer(d18:1/24:0)三种神经酰胺水平升高与ASCVD发病风险增加相关，并且介导了大气PM_2.5长期暴露与新发ASCVD风险的关联。研究结果揭示了神经酰胺在PM_2.5长期暴露致ASCVD发生发展中的作用，未来仍需进一步研究明确神经酰胺在这一过程中的作用机制。

Objective

Previous studies have suggested that short-term exposure to ambient fine particulate matter (PM_2.5) could alter ceramide (Cer) levels. However, the effects of long-term PM_2.5 exposure on ceramide levels remain unclear, as does whether this effect further contributes to the development of atherosclerotic cardiovascular diseases (ASCVD). Hence, this study aims to identify the ceramides associated with long-term PM_2.5 exposure concentration, then evaluate whether these candidate ceramides are associated with incident ASCVD risk, and further assess the potential of ceramides to mediate the increases of incident ASCVD risk following PM_2.5 exposure, relying on a case-cohort study from two cohorts of the Prediction for ASCVD Risk in China (China-PAR) project.

Methods

The case-cohort study was derived from two cohorts in the China-PAR project, including the International Collaborative Study of Cardiovascular Disease in Asia (InterASIA) and the China Multi-Center Collaborative Study of Cardiovascular Epidemiology-1998 (ChinaMUCA-1998). The InterASIA and ChinaMUCA-1998 cohorts were established in 2000-2001 and 1998, respectively, and face-to-face follow-up surveys were conducted in 2007-2008, 2012-2015, and 2018-2020. Given the availability of biological samples for ceramide testing, the survey in 2007-2008 was designated as the baseline in this study. After excluding participants who were lost to follow-up, those with prevalent unstable angina, acute myocardial infarction, or stroke before the baseline survey, and those without sufficient blood samples, a total of 14 496 participants constituted the original cohort for this case-cohort study. Then 1444 subjects were randomly selected as the subcohort population with a 10% sample proportion. After excluding those without PM_2.5 and ceramide levels, 1427 subcohort populations were included in the analysis. Meanwhile, the case-cohort study consisted of 540 incident ASCVD cases from the above 14496 subjects followed up to December 31^st, 2020. After excluding those with missing PM_2.5 and ceramide levels, 530 incident ASCVD cases (80 cases included in the subcohort) were included in the analysis. Incident ASCVD cases were defined as the first occurrence of unstable angina pectoris, non-fatal acute myocardial infarction, fatal coronary heart diseases, and non-fatal or fatal ischemic stroke. Using the targeted lipidomic techniques based on ultra-performance liquid chromatography coupled to tandem mass spectrometry (UPLC-MS/MS) techniques, a total of 27 ceramides were measured quantitatively based on the plasma samples isolated from fasting (≥10 hours) peripheral venous blood samples collected at baseline. Since the concentrations of ceramides showed a right-skewed distribution, the rank-based inverse normal transformation (INT) was applied to ceramide concentrations in statistical analyses. A high-precision (1km×1km) gridded monthly average PM_2.5 dataset from 2000 to 2016 was constructed by satellite-based remote sensing inversion technology. Based on the dataset, the long-term PM_2.5 exposure concentration of each subject was evaluated by the average annual PM_2.5 levels of the three years before baseline survey.

First, based on the subcohort population, the ceramides associated with the 3-year average PM_2.5 level were identified by linear regression models with the false discovery rate (FDR)-corrected P-value (Q-value)<0.05 in the two-sided test. Second, the associations between candidate ceramides and incident ASCVD risk were assessed by Prentice-weighted Cox proportional risk regression models. Results were presented as hazard ratios (HRs) and 95% confidence intervals (CIs) corresponding to per standard deviation (SD) increase in transformed ceramide concentrations. Furthermore, mediation analysis was performed to explore whether the above candidate ceramides could be a potential mediator of the association between PM_2.5 and incident ASCVD. The total effect of PM_2.5 on incident ASCVD risk was decomposed into direct effect and indirect effect mediated by the ceramides, with the mediation proportion defined as the proportion of the indirect effect in the total effect. The 95%CIs of the total effect, indirect effect, and mediation proportion were estimated using the bootstrapping approach with 1000 samples. A two-sided P-value<0.05 was considered statistically significant. The main analyses were based on the multi-adjusted models. Adjusted variables were age, sex, geographic region, residence, source cohort, smoking status, alcohol consumption, physical activity, dietary status, body mass index (BMI), prevalent hypertension, prevalent diabetes, prevalent dyslipidemia, socioeconomic status (SES), and ASCVD family history. To test the robustness of results, sensitivity analysis was performed by adjusting for the window period of PM_2.5 exposure, additionally adjusting for meteorological factors, adjusting for other lipid covariates, or excluding the ASCVD cases within 1 year of follow-up.

Results

The case-cohort study consisted of 1877 participants, including a subcohort population of 1427 and 530 new ASCVD cases, with a median follow-up of 10.53 years. Multivariate linear regression models revealed that levels of 15 ceramides increased in response to elevated PM_2.5 concentrations, including Cer(d18:0/15:0), Cer(d18:0/20:0), Cer(d18:0/22:0), Cer(d18:0/24:0), Cer(d18:1/16:0), Cer(d18:1/18:1), Cer(d18:1/20:0), Cer(d18:1/22:0), Cer(d18:1/23:0), Cer(d18:1/24:0), Cer(d18:2/20:0), Cer(d18:2/22:0), Cer(d18:2/23:0), Cer(d18:2/23:1), and Cer(d18:2/24:2) among all 27 ceramides (Q-value<0.05). The results using the annual average PM_2.5 exposure from January 2000 to the baseline survey year or additionally adjusting for 3-year average temperature and relative humidity remained robust. Besides, the incident ASCVD cases tended to be older, with a higher BMI level, and a higher proportion of male, current smokers, non-ideal physical activity, prevalent hypertension, diabetes mellitus, dyslipidemia, and low SES compared with the subcohort population. The Prentice-weighted Cox proportional risk regression models showed that, the levels of 3 candidate ceramides, including Cer(d18:1/20:0), Cer(d18:1/22:0), and Cer(d18:1/24:0), were positively associated with incident ASCVD risk. For per SD increment in transformed Cer(d18:1/20:0), Cer(d18:1/22:0), and Cer(d18:1/24:0) levels, the risk of ASCVD increased by 15.3% (HR = 1.153, 95%CI: 1.015-1.308), 15.5% (HR = 1.155, 95%CI: 1.011-1.320) and 16.8% (HR = 1.168, 95%CI: 1.006-1.355), respectively. Sensitivity analyses that excluded incident ASCVD cases within the first year of follow-up or further adjusting for other lipid covariates still illustrated the significant positive association between these ceramide levels and the risk of ASCVD. Mediation analysis further illustrated that increased PM_2.5 levels increased the risk of ASCVD by up-regulating the levels of related candidate ceramides. After adjusting for the above covariates, the mediated proportions of per SD increment of transformed Cer(d18:1/20:0), Cer(d18:1/22:0), and Cer(d18:1/24:0) levels were 6.58%, 13.97%, and 11.41%, respectively (P-value<0.05). The sensitivity analysis that changed the PM_2.5 concentration to annual average PM_2.5 concentration from January 2000 to baseline survey year also found the mediated effects of these ceramides with the mediated proportion decreasing slightly.

Conclusions

In this study, we found that long-term exposure to ambient PM_2.5 interfered with ceramide metabolism, resulting in up-regulation of several ceramide levels. Three ceramides related to PM_2.5 exposure, including Cer(d18:1/20:0), Cer(d18:1/22:0), and Cer(d18:1/24:0), were positively associated with incident ASCVD risk and further mediated the association between long-term ambient PM_2.5 exposure and incident ASCVD risk. The findings reveal the role of ceramides in the occurrence and development of ASCVD related to long-term PM_2.5 exposure, and further studies are still needed to clarify the mechanism of ceramide metabolism in the development of ASCVD.