STATISTICAL METHODS TO ASSESS THE SUSCEPTIBILITY TO PARTICULATE MATTER AND HEALTH EFFECTS MEDIATED BY MICRORNASCARRIED IN PLASMA EXTRACELLULAR VESICLES

ABSTRACT Air pollution exposure is a major problem worldwide and has been linked to many diseases. PM10 is one of the components of air pollution and it includes a mixture of compounds. Several studies suggest that PM produces significant effects on respiratory and cardiovascular system, in relation to acute as well as chronic exposure. This process has been extensively studied, but to date it has not yet been fully understood. Ambient particles have been shown to produce a strong inflammatory reaction, and beside pro-inflammatory mediators, cell-derived membrane Extracellular Vesicles (EVs) are also released. EVs (particularly microvesicles) might be the ideal candidate to mediate the effects of air pollution, since potentially they could transfer miRNAs, after internalization within target cells through surface-expressed ligands, enabling intercellular communication in the body. Another gap in our current knowledge regarding PM-related health effects is the identification of susceptible subjects. Recent research findings pointed out obesity as a susceptibility factor to the adverse effects of PM exposure partly due to an increase in particle absorption. According these findings, our hypothesis is that, EVs might be the ideal candidate mechanism to mediate the effects of air pollution, since potentially they could be produced by the respiratory system, reach the systemic circulation and lead to the development of endothelial dysfunction. Moreover, EVs after internalization within target cells through surface-expressed ligands, may transfer miRNAs enabling intercellular communication in the body. Finally, obese individuals might represent one of the best population to investigate the effects of environmental air particles on several molecular mechanisms and, as a final objective, on cardiovascular and respiratory parameters. The main proposal of this research project is to develop the appropriate statistical methodology to address the following specific aims: • Aim 1. Determine whether exposure to air particles and PM-associated metals can modify EVs in plasma in terms of miRNAs content. • Aim 2. Determine whether the changes found in ECVs (Aim 1) are associated with respiratory, cardiac and inflammatory outcomes such as: single breath carbon monoxide diffusing capacity DLcoRapp, Forced expiratory volume in the 1st second FEV1, Forced Vital Capacity FVC, Heart Rate, Sistolic Blood Pressure SBP, Diastolic Blood Pressure DBP, C-Reactive Protein CRP, and Fibrinogen. • Aim 3. Investigate the potential role of miRNAs as mediators of the effect of PM10 exposure on respiratory, cardiac and inflammatory outcomes listed in Aim2. We used a cross-sectional study investigating the effects of particulate air pollution on a population of susceptible overweight/obese subjects, recruited in Lombardy Region, Italy. The population study will include 2000 overweighed/obese (BMI between 25 and 29.9 is considered overweight and an adult who has a BMI of 30 or higher is considered obese subjects, recruited at the Center for Obesity and Weight Control (Department of Environmental and Occupational Health, University of Milan and IRCCS Fondazione Ca’Granda – Ospedale Maggiore Policlinico). We will follow a two-stage, split sample study design. The first (discovery) stage involves genome-wide miRNA expression profiling, by means of OpenArray technology, among 1000 of the aforementioned 2000 participants (the first 1000 subjects consecutively recruited at the Center for Obesity and Weight Control). The second (replication) stage involves a replication analysis of the top 10 miRNAs that resulted from the first stage. At December 31, 2013 (first stage) we recruited 1303 subjects, 87% of whom living in the province of Milan. At April 2015 we recruited a total of 1786 evaluable subjects. Due to technical problems the replication data were not available for statistical analysis at the time of the layout of the thesis. Different normalization strategies on miRNAs expression data were evaluated and compared in different set of miRNAs: Endogenous U6, Global Mean and Mean of 4 more stable miRNAs. The performance of the different normalization strategies was assessed by: (1) evaluating their ability to reduce the experimental induced (technical) variation, (2) determining their power to extract true biological variation. We showed for large scale miRNA expression profiling Global Mean normalization strategy outperforms the other normalization strategy in terms of:  better reduction of technical variation: - lower % of miRNAs differentially expressed before and after FDR adjustment - lower Fold change range;  more accurate appreciation of biological changes. - higher % of miRNAs differentially expressed before and after FDR adjustment; - higher Fold Change range; PM10 exposure assessment is based on daily PM10 concentration estimates by the FARM model (the flexible air quality regional model), a three-dimensional Eulerian grid model for dispersion, transformation and deposition of particulates, capable to simulate PM10 concentration. By means of ArchGis software the residential address of each subject was georeferenced and the resulting map was superimposed on the map of FARM Model. In this this way to each subject was attributed: (a) the estimated daily exposure of the cell containing their residential address; (b) the exposure of the cell containing the address of the Center for Obesity and Work; (c) the daily average exposure for Milan, calculated as the average of the 22 cells that falls into the city boundaries. Since in each run of OpenArray were simultaneous analysed up to 4 OpenArray plates, identified by a barcode, for a total of 12 samples (3 per plate) it was possible identify an hierarchical data structure with three levels: sample level (level-1), barcode level (level-2) and run level (level-3). In order to verify the association between miRNAs expression and PM10 we developed a three-levels hierarchical linear model (HLM) using the MIXED procedure in SAS. The following list of first 10 top miRNAs were identified: miR_106a_002169, miR_152_000475, miR_181a_2__002317, miR_218_000521, miR_27b_000409, miR_30d_000420, miR_652_002352, miR_92a_000431, miR_25_000403, miR_375_000564. Simple mediation models were applied in order to investigate the role of miRNAs expression as potential mediator on the effect of PM10 on respiratory, cardiac and inflammatory outcomes such as: single breath carbon monoxide diffusing capacity DLco, Forced expiratory volume in the 1st second FEV1, Forced Vital Capacity FVC, Heart Rate, Sistolic Blood Pressure SBP, Diastolic Blood Pressure DBP, C-Reactive Protein CRP, and Fibrinogen. 95% BC bootstrap Confidence intervals for Indirect effect were estimated. Finally, Multiple Parallel mediation models were applied in order to investigate the role of a set of miRNAs expression identified by means of simple mediation models as potential set of parallel mediator on the effect of PM10 on respiratory, cardiac and inflammatory outcomes. A significant indirect effect of PM10 on: - DLcoRapp, was found through the following mediators: mir_106a_002169, mir_152_000475, mir_218_000521 expression; - FEV1Rapp was found through the following mediators: mir_27b_000409 mir_30d_000420 mir_92a_000431 mir_181a_2_002317 mir_218_000521 expression; - FVCRapp was found through the following mediators: mir_27b_000409, mir_92a_000431 and mir_181a_2_002317 expression; - Heart Rate was found through the following mediator: mir_218_000521 expression; - Sistolic Blood Pressure was found through the following mediator: mir_92a_000431 expression; - CRP was found through the following mediator: mir_106a_002169 and mir_652_002352 expression. - Fibrinogeno was found through the following mediator: mir_375_000564 expression. Finally, the total indirect effect of PM10 exposure: - on DLcoRapp obtained summed the indirect effects across all mediators: mir_106a_002169, mir_152_000475, and mir_218_000521 expression is statistically different from zero; - on FEV1Rapp obtained summed the indirect effects across all mediators: mir_27b_000409 mir_30d_000420 mir_92a_000431 mir_181a_2_002317 mir_218_000521 expression is statistically different from zero; - on FVCRapp obtained summed the indirect effects across all mediators mir_27b_000409, mir_92a_000431 and mir_181a_2_002317 expression is statistically different from zero; - on CRP obtained summed the indirect effects across all mediators mir_106a_002169 and mir_652_002352 expression is statistically different from zero.