Diesel Exhaust and Allergen Modulate Mirna and RNA in Intact Human Epithelium

Online Data Supplement

Diesel exhaust and allergen modulate miRNA and RNA in intact human epithelium

Christopher F. Rider, PhD1, Masatsugu Yamamoto, MD1, Oliver P. Günther, PhD2, Jeremy A Hirota PhD1,3, Amrit Singh, BS3,Scott J. Tebbutt, PhD3, Chris Carlsten, MD, MPH1,3,4

Supplementary Methods

Participant Characteristics

Healthy human volunteers aged between 19 and 49 were screened using spirometry and tested for sensitivity to house dust mite (HDM; D. pteronyssinus), birch andTimothy/Pacific grass. Eighteen non-smoking volunteers participated, who were not subject to any of the following exclusion criteria:

(1) Pregnancy/breastfeeding

(2)Use of inhaled corticosteroids

(3) Use of bronchodilator medication more than 3 times per week

(4) Unstable asthma symptoms

(5) Presence of co-existing medical conditions (as assessed by the primary investigator)

(6) Participation in another study that involves taking medications

(7) Regular use of antihistamines, non-steroidal anti-inflammatories anticoagulants, acetylsalicylic acid (ASA) or decongestants

(8) Unstable asthma, i.e. an exacerbation in previous 2 weeks

(9) Allergy to Lidocaine, Fentanyl or Versed

Supplemental table E1 includes details of the 15 participants who successfully completed this study, including age, provocative methacholine concentration resulting in a 20% drop in FEV1 (PC20) and the PC20 dose response slope (DRS).

Bronchial brushing collection and processing and NanoString®Assays

Forty-eight hours after each exposure,participants underwent a second bronchoscopy during which bronchial washes (BW; collected from the first 40ml of saline instilled) and bronchoalveolar lavages (BAL; from 100 ml subsequently instilled)were collected from the middle lung lobes where saline or allergen had previously been delivered. Bronchial brushes were collected from the right and left lower lung lobes and stored at -80oC. Bronchial brushes consistently contained mostly epithelial cells during in-vitro testing (FAS=97.3%, FAA=98.7%, DES=95.4%, DEA=98.7%). Qiagen miRNeasy Mini Kits were used to extract miRNA and mRNA, according to the manufacturer’s protocol. We assessed RNA quality using an Agilent 2100 bioanalyzer with Eukaryote total-RNA Nano chips. MicroRNA and mRNA expression was determined by NanoString® Technologies (Seattle, US) using nCounter™ Human miRNA v2 (includes 800 miRNAs) and Human Immunology v1 (includes 511 mRNAs) panels respectively.

Inflammatory Marker Analysis

BAL cells were Wright-Giemsa stained and counted under a microscope using a haemocytometer. We spun volumes of BAL containing ~30,000 cells onto slides using a cytospin (Cytospin 3; Shandon, Pittsburgh, PA) and stained using a Harleco Hemacolor (Wright-Giemsa) kit (Millipore; Etobicoke, Canada). At least 400 cells were counted on each slide and percentages of eosinophils and bronchial epithelial cells (BEC) recorded by two independent observers, after which mean values were calculated.

Data Analysis

Sixty samples (15 participants x 4 conditions) were collected, but degraded samples were excluded based on examination of the Agilent® chromatograms by multiple authors. Full analysis was therefore performed on 58 samples of miRNA (from 15 participants, 13 with all 4 conditions) and 52 samples of mRNA (from 13 participants, 10 with all 4 conditions).Inflammatory markers, miRNA and mRNA expression data wereanalyzed in R (version 3.2.0), as detailed in Fig. 1B.Normalization of mRNA and miRNA datasets combined a correction factor calculated from positive control probe counts to represent platform-associated sources of variation, with a normalization factor derived from control probe-adjusted data to account for sample input variations. NanoString®'s default top 100 probes procedure34 was modified in two ways to determine the second factor: 1) all identified outlier samples were first excluded and, 2) probes with a coefficient of variation that was 1.5 times the inter-quartile range above the 75% quantile across samples were removed, to avoid disproportionally large effects of individual probes on normalization factors. We filtered variables to remove noise by first defining a global background threshold for absence/presence as the maximum of all sample-specific background thresholds. Next, a ‘flooring’ procedure was applied that set all expression values below the global background threshold to that value. Finally, only variables where at least one exposure group had a minimum of 75% of samples above the background threshold passed filtering.

Initial analysis was performed using a robust linear model implemented using the limma35 R package to evaluate the effects of individual conditions: 1) diesel exhaust plus allergen (DEA), 2) diesel exhaust plus saline (DES; saline was the diluent control for allergen), and 3) filtered air plus allergen (FAA); each relative to 4) filtered air plus saline (FAS). Corrections for false discovery rate were performed using Storey’s q-value correction, using the qvalue R package36. Subsequently, univariate analysis using a mixed effects model implemented in ‘lme4’37 was performed, to determine which miRNAs and genes were significantly modulated by DE exposure and/or allergen instillation. Data was plotted using GraphPad Prism version 6 and various plotting functions in R. Gene targets of miRNAs showing significantly increased expression were examined within the miRTarBase database38. Integrative analysis of miRNA, miRNA and inflammatory markers was performed using mixOmics39 and sparse partial least squares (PLS) used to simultaneously reduce dimensions and select important variables, yielding the principle components. PLS components are derived from mathematical procedures that use an orthogonal transformation to model the covariance structures of two sets of variables into multiple dimensions.These were ranked according to which projections contained the greatest correlations and the first ten used in plotting heatmaps. As a second approach to showing this data, network diagrams were constructed for elements having a correlation of ≥0.7. Finally, gene ontology of the genes with q-values of ≤0.2 following allergen instillation was examined using gene annotation tool to help explain relationships (GATHER)40.