Functional Screening Identifies MicroRNAs as Multi-Cellular Regulators of Heart Failure

Heart failure (HF) is the leading cause of death in the Western world. Pathophysiological processes underlying HF development, including cardiac hypertrophy, fibrosis and inflammation, are controlled by specific microRNAs (miRNAs). Whereas most studies investigate miRNA function in one particular cardiac cell type, their multicellular function is poorly investigated. The present study probed 194 miRNAs –differentially expressed in cardiac inflammatory disease – for regulating cardiomyocyte size, cardiac fibroblasts collagen content, and macrophage polarization. Of the tested miRNAs, 13%, 26%, and 41% modulated cardiomyocyte size, fibroblast collagen production, and macrophage polarization, respectively. Seventeen miRNAs affected all three cellular processes, including miRNAs with established (miR-210) and unknown roles in cardiac pathophysiology (miR-145-3p). These miRNAs with a multi-cellular function commonly target various genes. In-depth analysis in vitro of previously unstudied miRNAs revealed that the observed phenotypical alterations concurred with changes in transcript and protein levels of hypertrophy-, fibrosis- and inflammation-related genes. MiR-145-3p and miR-891a-3p were identified to regulate the fibrotic response, whereas miR-223-3p, miR-486-3p, and miR-488-5p modulated macrophage activation and polarisation. In conclusion, miRNAs are multi-cellular regulators of different cellular processes underlying cardiac disease. We identified previously undescribed roles of miRNAs in hypertrophy, fibrosis, and inflammation, and attribute new cellular effects to various well-known miRNAs.


MiRNA selection
MiRNA expression profiling in patients and small animal models of inflammatory cardiac disease (patients diagnosed with viral myocarditis (VM), patients which underwent cardiac transplantation (HTX), and three cardiac disease animal models: Coxsackievirus B3 induced-VM, mouse HTX, and the ZSF1 rat model of diastolic heart failure) was the basis for the selection of 194 differentially expressed inflammatory miRNAs. Dataset 1 shows all miRNAs selected for in vitro screening based on a significantly altered expression in one or more cardiac disease models, indicating a possible regulatory role in cardiac disease progression and development. Both miRNA strands were included in the screen when the miRNA profiling data did not specify which miRNA strand (-3p/-5p) was measured.

Isolation and culture of primary cells
Rat neonatal ventricular cardiomyocytes (nRCMs) and rat neonatal cardiac fibroblasts (nRCFs) were isolated by enzymatic dissociation from the hearts of 1-3 day old Wistar rats and cultured as described previously 17 . Briefly: After decapitation, neonatal rat pups were dissected using a cut through the sternum and hearts were collected under semi-sterile conditions. Subsequently, atria were removed and the ventricles were divided into 8-12 equal parts. Ventricles were consecutively digested with a mixture of 0.3 mg/ml Collagenase (Sigma #C2674) and 0.3 mg/ml Pancreatin (Sigma # P3292) in ADS buffer (120 mM NaCl, 5 mM KCl, 0.8 mM MgSO 4 , 0.5 mM KH 2 PO 4 , 0.3 mM Na 2 HPO 4 , 20 mM HEPES, 5.6 mM Glucose, pH 7.35) at 37°C. Every 20 min the supernatant was collected, suspended in 10% NBCS, and 10 ml fresh enzyme solution was added to the residual tissue. After 5 rounds of incubation, cell solutions were pooled, centrifuged, and re-suspended in nRCM plating medium (DMEM #11966, 17% M199 medium, 10% HS, 5% NBCS) and pre-plated onto 162cm Corning Costar cell culture flasks (Sigma #CLS3151). After incubation in a humidified 37 °C /5% CO2 incubator for 1 h, the supernatant, containing mainly nRCMs, was collected and cells therein were counted manually. Adherent nRCFs remaining in the flask were cultured for 2 more days in nRCF medium (DMEM #22320, 10% FBS).
Bone marrow-derived macrophages (BMDMs) were generated as previously described 18 , in brief: Bone marrow cells were isolated from 12 week old C57BL/6-N mice and cultured in RPMI supplemented with 15% L929 conditioned medium (containing M-CSF) in petri dishes for 7 days to generate BMDMs. Cells were lifted for plating at day 8 using cell scrapers.
Culture, transfection, and stimulation of primary cell culture. nRCM: nRCMs were seeded into 1% gelatin coated 96-well black, clear bottom, culture plates (Corning #3603) at a density of 40,000 cells/well in 100µl seeding medium. Prior to cell seeding, 0.171µl Lipofectamine 2000 (Invitrogen #11668-019) transfection reagent (diluted 700x in OPTIMEM) was added to the wells in combination with mirVana mimics (Life-Technologies) according to manufacturer's protocol to end up with a final concentration of 10nM. The day after isolation, cells were washed and medium was replaced for 100µl serum-free experimental medium (DMEM 11966, 20% M199 medium) to starve cells for 24h hours. After starvation, cells were stimulated for 72 hours with 5µM phenylephrine (PE) (Sigma #P6126) or treated with PBS as control.
nRCF: nRCFs were lifted using trypsin/EDTA 48 hours after isolation, washed, centrifuged and seeded into uncoated 384 wells µclear plates (Greiner #781092) in 30 µl RCF medium containing 10% FBS. A seeding density of 2000 cells/well was used. After 24h and prior to transfection, cells were washed extensively using an automated washer dispenser (EL406, BioTek) and 30 µl low serum (0.1% FBS) nRCF medium was added. Lipofectamine RNAiMax (Invitrogen) was mixed with OPTIMEM and mirVana mimics (Life-Technologies) according to manufacturer's protocol and 5 µl of the mix was added to the wells, to give a final concentration of 20 nM for mimics and a dilution factor of RNAiMAX of 700x. After 24h of starvation, cells were stimulated with 10 ng/ml TGFβ1 (Peprotech hum. Recombinant TGFβ1 # 100-21) or vehicle (PBS).
BMDM: BMDMs were lifted using a scraper after 8 days of differentiation. Subsequently, cells were counted and seeded at a density of 2000 cells/well into uncoated 384 wells µclear plates (Greiner #781092) in 30 µl BDMD medium. After 24h, medium was refreshed prior to transfection.
Viromer Green (Lypocalyx, Halle, Germany) was mixed with Buffer F and mirVana mimics (Life-Technologies) or miRCURY LNA Power inhibitors (Exiqon) according to manufacturer's protocol and 5 µl of the mix was added to the wells, to give a final concentration of 20 nM mimics and a dilution factor of 1800x for the transfection reagent. 24h after transfection, cells were stimulated with 20 ng/ml IL-4, 20 ng/ml IFNγ (Peprotech), 50 ng/ml LPS (Sigma) or vehicle (PBS).
In the present study we used human miRNA mimics for transfection into rat and murine primary cells. Effect of inter-species differences is limited as sequence analysis showed that 118 out of the 194 miRNAs display a 100% sequence match between human and rodent and only 9 miRNA mimics showed a mismatch in the seed sequence (Supplementary Table S7), potentially leading to differences in their actual mRNA-targets.
For the screening experiment, image acquisition was performed using an ImageXpress Micro automated high-content screening fluorescence microscope (Molecular Devices) at ×10 magnification with FITC, CY3 and DAPI filter blocks. For the hypertrophy, fibrosis, and inflammation screen, a total of respectively 25, 16, and 9 images were acquired per wavelength per well.
Image analysis was performed using the eCognition software (Definiens, platform (Munich, Germany). Number of nuclei was determined based on Hoechst staining for all three cell types. nRCM and nRCF cell size was quantified using an algorithm that recognized α-tubulin stained area. For nRCFs, collagen1α1 positive area was quantified as a measure of collagen production. For BMDMs, cell roundness was determined as a percentage of cells detected with a length-width ratio larger then threshold. NFκB nuclear translocation was based on p65 positive area in nucleus (Hoechst) and cytoplasm (α-tubulin stained area) and expressed as ratio of nuclear/cytoplasmic p65 staining.  Table S8). The details of the sequences and thermal cycling conditions were according to the standard protocol. Data were acquired and analysed with IQ5 software (Bio-Rad, U.S.). The ΔΔCt method was used to analyse obtained Ct Values and make the mRNA levels relative to the appropriate control groups and corrected for housekeeping gene.

Identification of miRNA target genes
To study the mechanism underlying the regulatory function of the 17 miRNAs proven to have a multicellular function, we identified shared common target genes. Rather than using in silico algorithms presenting predicted miRNA target genes (many of which appear to be not functional in validation studies), we used the publically available data deriving from the study of Spengler et al. 65 . This study used AGO2 crosslinking immunoprecipitation coupled with high throughput sequencing (HITS-CLIP) of bound RNA interaction sites, resulting in the detection of 4000 cardiac AGO2 binding sites across more than 2200 target transcripts. Each single AGO2-interacting transcript site is matched with a complementary seed sequence of cardiac-expressed miRNAs. We identified all targeted transcripts matching the seed sequence of the 17 miRNAs with a multi-cellular function, resulting in the identification of 1290 target genes of which 15 genes can be targeted by at least 7 out of the 17 selected miRNAs (Supplementary Table S6).

Data analysis and statistics
For all analysed miRNA samples, mean and standard deviation of individual wells over 3 replicate plates was determined. Inter-plate variability was corrected by scaling each data point to the corresponding negative control mean (negative control mimic transfected cells). The effect of miRNA transfection on different read outs was determined via calculation of the log 2 fold change of the sample mean over the mean of the negative control mimic-transfected cells within the same condition and statistically tested using an unpaired T-test.
In all three screens, identification of miRNA mimic-induced phenotypical changes was based on a statistically significant (p<0.05) sample mean log 2 fold change of the main read out over negative control and deviating more than 2x the STDEV from the negative control mean. The main readouts per screen were cell size for the hypertrophy screen, number of nuclei and collagen area for the fibrosis screen, and cell roundness and NFκB nuclear translocation for the inflammation screen. Only hit selection for the NFκB nuclear translocation read out in the inflammation screen was based on statistically significant mean log 2 fold change over negative control, deviating more than 1x the STDEV from the negative control mean.

Supplementary Figures
Supplementary Figure S1. Hypertrophy screening using primary neonatal rat cardiac myocytes  Supplementary Table S2