Virus strain and infection of animals
All animal experiments with 6–8 week old female BALB/c mice were approved by the Institutional Animal Care and Use Committee (IACUC), National University of Singapore (protocol 050/11), and fulfilled all listed criteria in the ARRIVE guidelines (Additional file 4). Mice were housed in ABSL-2 facilities in ventilated cages according to IACUC guidelines. Animals were divided into 4 groups each consisting of 4 mice, i.e. control uninfected and infected groups at 7 and 15 dpi. The two time-points were chosen based on the kinetics of influenza infection, i.e. when repair of primary viral damage commences and certain repair cell types initially appear at 7 dpi, whereas repair of inflammatory damage begins after viral clearance and when repair cell types peak at 15 dpi [10, 12]. Mice were anesthetized with a mixture of 7.5 mg/ml ketamine and 0.1 mg/ml medetomidine, and infected intra-tracheally with a sub-lethal dose of 23 plaque-forming units of influenza A/Puerto Rico/8/1934(H1N1) or PR8 virus (in 50 μl volume) to ensure that all mice recovered from the infection. Control mice were given phosphate-buffered saline (PBS) intra-tracheally. Mice were monitored daily for weight loss from infection and weight gain from recovery. Upon reaching the stipulated time-point, mice were euthanized in a CO2 chamber, and the lungs harvested for histology and RNA extraction.
Murine lung tissue processing and extraction of miRNAs and mRNAs
Approximately 10% of each lobe of lung was snap-frozen at -80°C to preserve RNAs. The lungs were then homogenized in Trizol solution (Qiagen, Hilden, Germany) using a gentleMACS tissue dissociator (Miltenyi Biotec, Bergisch Gladbach, Germany), and total RNA was immediately extracted using the miRNeasy miRNA extraction kit (Qiagen). The isolated total RNA samples containing miRNAs were then subjected to miRNA and gene microarray analyses.
The remaining 90% of each lung was fixed in 4% formalin, dehydrated in ascending ethanol concentrations, embedded in paraffin, sectioned into 4-μm slices, and stained with hematoxylin and eosin (H&E). Ideally, the lungs could be inflated prior to fixation. Notwithstanding this, highly satisfactory assessment of the infected areas was achieved by an experienced pulmonary pathologist to include: inflammatory cell infiltration (polymorphs and lymphocytes); septae, hyaline and fibrin; hemorrhage; edema (for tissue damage); and new progenitor cells and re-epithelization (for tissue repair and replacement). This evaluation was then correlated with the molecular analyses. Scoring of damage to ascertain the homogeneity of the infection was also conducted for mouse lungs at 7 dpi based on a modified scoring system, i.e.% infected × [alveolar hemorrhage + 2(alveolar inflitrates) + 3(fibrin) + alveolar septal congestion] . The scoring was carried out based on the worst-affected area of the lungs, which then yielded the percentage of the lung area showing similar scores. Since mice at 15 dpi were recovering and could not be scored with the same system, the percentage of lung area affected by the infection was obtained instead to ascertain the homogeneity of damage.
Paraffin-embedded blocks of lung tissues were also sectioned into 5-μm slices for IHC analyses. The slides were dewaxed with two washes of xylene, rehydrated with two washes of 100% ethanol, and one wash each of 90%, 70% and 50% ethanol, followed by two washes of deionized water. Slides were then subjected to antigen retrieval in boiling sodium citrate buffer with Tween 20. Slides were further permeated with 0.025% of Triton X-100 for 10 min, before being blocked with 3% BSA for 2 h. Primary antibodies for PCNA (SC-9857, Santa Cruz Biotechnology, Dallas, TX) and SP-C (SC-13979, Santa Cruz Biotechnology) at 100× dilution in 3% BSA were used for co-staining dividing and repairing cells and AT2 pneumocytes, respectively. Following overnight incubation at 4°C in the dark, the slides were washed with Tris-buffered saline thrice for 5 min each, and stained with secondary antibody containing Alexa Fluor 488 for PCNA and Alexa Fluor 568 for SP-C (Molecular Probes, Grand Island, NY) at room temperature for 1 h in the dark. Slides were then washed and mounted in colloidal gold Antifade mounting medium with DAPI (Molecular Probes). Slides were then concurrently scanned with the high-resolution MIRAX MIDI system equipped with fluorescence illumination (Carl Zeiss, Jena, Germany) for the complete lung image to quantify dividing AT2 pneumocytes. Stained sections were also visualized with a FX-1000 confocal microscope at higher magnification.
Image analyses for quantification of dividing AT2 pneumocytes in the lungs
An automated computer algorithm was developed to count the number of total cells, PCNA-positive cells, and PCNA-positive AT2 pneumocytes. Nuclei were segmented by converting the DAPI fluorescence channel into black-and-white using manual thresholding of DAPI intensities, and removing areas less than 3 pixels in size. The number of nuclei was counted as the number of total cells. The PCNA fluorescence channel was next converted to black-and-white using manual thresholding of PCNA intensities. PCNA-positive nuclei were identified by selecting co-localized areas of nuclei and PCNA channel, and performing dilating-hole filling-eroding using “disk” operator with 3 pixels in size. The number of PCNA-positive nuclei was counted as the number of PCNA-positive cells. Finally, the SP-C fluorescence channel was converted into black-and-white using manual thresholding of SP-C intensities, and areas less than 2 pixels in size were ignored. The surrounding areas of PCNA-positive nuclei (1 pixel in radius) were then checked for SP-C signal in the SP-C channel. The number of PCNA-positive nuclei with surrounding SP-C signal was counted as the number of PCNA-positive AT2 cells. All image processing and computation algorithms were implemented using MATLAB with image processing toolbox (MathWorks, Natick, MA). The MATLAB codes are available upon request.
miRNA and gene microarray analyses
Extracted total RNAs containing miRNAs were subjected to quality control using a bio-analyzer to ensure proper RNA quality before the microarray experiments. Total RNA samples from 4 mice per group were then hybridized individually onto the respective microarray platforms (Agilent, Santa Clara, CA). The miRNA microarray was a Mouse miRNA, 8 × 60 K format platform (AMADID 38112), while the gene microarray was a SurePrint G3 Mouse GE, 8 × 60 K, 1 color format platform (AMADID 028005). A total of 32 arrays (i.e. 16 miRNA arrays and 16 mRNA arrays) were hybridized to total RNA samples extracted from the lungs of 4 infected mice and 4 control mice for each 7 and 15 dpi. The expression values of the array probes were calculated following the manufacturer’s standard array protocol. Only probes above the expression threshold of 20 (raw signal intensity) were used to derive expression values.
The expression data for the filtered probes of both miRNA (1,179 probes) and gene (55,681 probes) microarrays were normalized using quantile normalization, and were subjected to DE analyses using QPROT software (an extension of the QSPEC software for generic expression data with missing values). DE genes were identified using the Z-statistics threshold that controls the FDR at 5% . The DE miRNAs and genes were further analyzed below to shortlist candidates associated with active roles in pulmonary repair and regeneration.
Real-time RT-PCR validation of miRNA and gene expression
A total of 5 miRNAs and 20 genes that were significantly regulated at both 7 and 15 dpi were selected for real-time RT-PCR to validate their differential expression based on the microarray data. A stem-loop primer miRNA Universal TaqMan RT-PCR system (Applied Biosystems, Foster City, CA) was utilized to validate miRNAs, and normalized using murine SNORD68 small nucleolar RNA [46, 47]. The miRNA stem-loop primers, specific forward primers and universal reverse primer are listed in Additional file 5: Table S2. Real-time RT-PCR assays to validate 20 genes were performed using reverse transcription with random hexamers (Promega, Madison, WI), the SYBR Green Master mix and LightCycler real-time PCR system (Roche, Penzberg, Germany), normalized against the RPL13a gene (Additional file 6: Table S3).
Developmental pathway and TargetScan analyses
The DE genes identified above were filtered for repair-associated genes by the following criteria. Functional selection for repair association was performed manually using the function tree of the IPA software (Qiagen), where genes under developmental and repair functional groups were identified. These genes were then filtered again with the hypergeometric test and FDR control (the Benjamini-Hochberg procedure), and GO analysis . A concurrent miRNA TargetScan analysis at a PCT score of 0.5 was performed using the GeneSpring GX software (Agilent), and only miRNAs that target the filtered genes from the previous steps were selected. The selected genes (grouped as GO terms at P < 0.05) and miRNAs were then visualized in a network with Cytoscape software (v3.0.1).
Further refinement of miRNA candidate selection
To further narrow down the selected miRNA candidates, an additional filter system of miRNA raw abundance in control samples was taken into account. Therefore, a final set of miRNAs was selected based on their statistical significance, targeting related GO term members of gene analysis, highest or lowest raw abundance in control samples where the fold changes were deemed more significant. The selected miRNAs were then classified based on their expression changes and their implicated functions, and also whether they target genes directly associated with lung development and regeneration.
FDR calculation and GO enrichment analyses of the array dataset were performed using R statistics software for selection of significant DE miRNAs and genes. Student’s one sample t-test was performed on real-time RT-PCR verification and dividing AT2 pneumocyte quantification data with GraphPad and SPSS software. Results were represented as mean ± standard deviation (SD).