Identification of new biomarkers for Acute Respiratory Distress Syndrome by expression-based genome-wide association study

The ARDS datasets were collected from Gene Expression Omnibus (GEO, NCBI, ( http://​www.​ncbi.​nlm.​nih.​gov/​geo) public functional genomics data repository. The GEO database was searched for “acute lung injury,” which returned 60 hits, and “lung injury,” which returned 487 hits (as of September 2014). Most of these entries were individual samples that duplicated samples, which were already combined into existing data sets. The elimination of these duplicates reduced the obtained data to 48 sets (Additional file 1: Table S1). These sets were further filtered down to 31 data sets using the following criteria: 1) the array samples must represent genome wide studies by the established microarray platforms with the number of interrogated sequences >5000, thus excluding custom platforms designed for a specific organ or pathway, which introduces tissue or biological process bias; 2) the experimental settings must contain untreated (either placebo or sham) and unmodified (wild type) subjects; 3) biological replicate n ≥ 3 per each sham or experimental group; 4) proper signal distribution as identified by the significance of microarray (SAM) plot.

Authors believe that all data uploaded from GEO involving human subjects have been obtained with the approval of appropriate ethics committees by Silva et al. and Eltzschig et al. (Additional file 1: Table S1).

The eGWAS was conducted as described previously [ 4]. Briefly, to estimate differences between groups of samples from ARDS subjects and sham controls, raw postquantitation microarray data were reanalyzed by SAM 2.0 software. Then the P values from the number of positive/negative experiments for each gene and sum of the number of positive/negative experiments for all other genes were calculated using a 2 × 2 chi-square or a Fisher’s exact test. The probe IDs across different microarray platforms for mouse, rat, and human were linked using the array information library universal navigator (AILUN) tool ( http://​ailun.​stanford.​edu). The probes that remained unmatched by the AILUN tool were linked to the AILUN created mouse-rat-human dataset via their gene symbol entries.

Given that canine and human samples were represented by only 1 and 2 experiments, respectively the species-related weighting algorithm [ 7] was not applicable.

The pathway analysis, which identifies the most relevant biological processes to a specific list of candidate genes, was conducted using the Ingenuity Pathways Knowledge Base tool (IPA, Ingenuity Systems, Inc., Redwood City, CA.) as described previously [ 3].

PubMatrix (multiplex literature mining tool) analysis was conducted as described previously [ 7]. We restricted our search to human symbols approved by HUGO Gene Nomenclature Committee (HGNC), which were enriched by all aliases and former (discontinued) symbols for selected candidate genes ( http://​www.​genenames.​org).

All mouse experiments were approved by the University of Missouri Kansas City Institutional Animal Care and User Committee. To confirm the unversal nature of the response to the ARDS stimuli by our candidates, we employed the mouse septic model - cecal ligation and puncture (CLP), which was not present in the GEO data collection (Additional file 1: Table S1). We selected the most commonly used time point (24 h after injury) for the CLP septic model [ 8]. Animals ( n = 6) were anesthetized and laparotomy performed. The cecum was exposed and ligated. In three mice, the cecum was punctured twice using 16G needle and then squeezed to extrude contents in a 2 mm of fecal amount. Mice ( n = 3) without the puncture were used as controls (shams). Wounds were closed. After 24 h mice were sacrificed and the lungs were collected for histological, real-time RT-PCR, and Western blot studies. All mouse experiments were approved by the University of Missouri Kansas City Institutional Animal Care and User Committee.

For histological studies, lungs were perfused first with PBS followed by 4 % paraformaldehyde, and then embedded in paraffin. 5 μm thick slices were prepared and stained with H&E according to standard protocol.

The real time PCR of mRNAs extracted from lung tissue was conducted as previously described [ 7]. Slight modifications were made according to the manufacturer’s new protocol. Briefly, the 384-well microtiter plate setting of a ViiA™ 7 Real-Time PCR System (Applied Biosystems) was employed. TaqMan® Predeveloped Assay Reagent mouse β-actin (REF 435234E, probe dye VIC-MGB) was used as an internal control for normalization. TaqMan® Gene Expression assay for mouse Clec4e and Cd300lf were purchased from Applied Biosystems Inc. (Mm 01183703_m1 and Mm 00467508_m1, respectively). All experimental protocols were based on manufactures’ recommendation using the TaqMan® Universal Master Mix II (P/N 4440039). A relative quantitative method was used to calculate corresponding transcript levels relative to actin expression. The significance of the obtained difference in signal between sham and CLP mice was assessed using t-tests where p < 0.05 was considered significant.

Immediately after euthanizing the mice, the lungs were perfused with PBS prior to tissue proteins extraction with lysis buffer. Protein concentration was determined by BCA assay (Thermo Scientific, prod # 23227). Equal total proteins (20 μg) for each lung cell lysate sample were analyzed by SDS-PAGE followed by Western blotting on PVDF membrane. Anti-CD300LF (Proteintech Group, Inc, cat# 13334-1-AP) antibody was used. Monoclonal anti-actin antibody (Santa Cruz, cat. # SC-47778) was used as a loading control. Protein signal was revealed by FluorChem M chemiluminescence (Proteinsimple™).