Development of Long Noncoding RNA-Based Strategies to Modulate Tissue Vascularization

Background Long noncoding ribonucleic acids (lncRNAs) are a subclass of regulatory noncoding ribonucleic acids for which expression and function in human endothelial cells and angiogenic processes is not well studied. Objectives The authors discovered hypoxia-sensitive human lncRNAs via next-generation ribonucleic acid sequencing and microarray approaches. To address their functional importance in angiogenic processes, several endothelial lncRNAs were characterized for their angiogenic characteristics in vitro and ex vivo. Methods Ribonucleic acid sequencing and microarray-derived data showed specific endothelial lncRNA expression changes after hypoxia. Validation experiments confirmed strong hypoxia-dependent activation of 2 intergenic lncRNAs: LINC00323 and MIR503HG. Results Silencing of these lncRNA transcripts led to angiogenic defects, including repression of growth factor signaling and/or the key endothelial transcription factor GATA2. Endothelial loss of these hypoxia-driven lncRNAs impaired cell-cycle control and inhibited capillary formation. The potential clinical importance of these endothelial lncRNAs to vascular structural integrity was demonstrated in an ex vivo model of human induced pluripotent stem cell–based engineered heart tissue. Conclusions The authors report an expression atlas of human hypoxia-sensitive lncRNAs and identified 2 lncRNAs with important functions to sustain endothelial cell biology. LncRNAs hold great promise to serve as important future therapeutic targets of cardiovascular disease.


Immunoprecipitation (IP) experiments
Ea.Hy926 cells underwent IP using mouse control IgG (Santa Cruz, sc-2025) and mouse eIF4A3 IgG (Merck-Millipore, #05-1527). For IP, antibodies were coupled to Dynabeads™ (Invitrogen). 10 million cells were used for IP and separated for control and eIF4A3 pulldown group. Pulldown of proteins was performed at 4°C for 1 h. After extensive washing steps, total RNA and protein fraction was taken from magnetic beads and used for further analysis (qPCR, Western Blot). Proliferation studies Cells were seeded in 96-well cell culture plates and transfected liposomally with siRNA/ LNA™ GapmeR one day after seeding. To measure proliferative capacity in siRNA or LNA™ GapmeR-modulated cells, a WST-1 proliferation assay (Roche) was applied. 48 h after transfection medium was changed and replaced by medium containing WST-1 reagent according to the manufacturer´s instructions. Finally, WST-1 absorbance was measured at 450 nm. Additionally, a colorimetric BrdU ELISA kit from Roche (#11647229001) was applied. Standard procedures were performed according to manufacturers´ instructions.

Capillary tube formation assay
The ability of HUVECs to form capillaries was analyzed on growth factor reduced Matrigel™ (BD Bioscienes). Therefore, siRNA/ LNA™ GapmeR-modulated cells were seeded at a density of 15.000 cells/8well chamber slide. Capillary formation was monitored 4-6 h after seeding and quantified as cumulative capillary (tube) length with NIS-Elements BR software (Nikon).

Scratch wounding migration assay
The migration capacity of siRNA-/LNA™ GapmeR-modulated cells was analysed by the use of culture-inserts (Ibidi). Therefore, 15.000 cells (HUVECs) or 30.000 cells (Ea.Hy926) were placed to each side of culture-inserts 24 h after transfection. 24 h later, plastic patches were removed and microscopic images were captured after 0, 4, 6 and 24 hours. Wound area was calculated by the use of NIS-elements BR software (Nikon). Migration index was calculated by the formula: area (0 h) -area (6 h)/area (0 h). Analysis of binucleic cells siRNA LINC00323-003 modulated cells and controls were 4% PFA-fixed after 48 h of transfection. DAPI-staining was performed afterwards to count the number of binucleic and total cells via the use of NIS-elements BR software (Nikon).

Apoptosis staining
Liposomally transfected cells were subjected to Annexin V/7-AAD staining applying FlowCellect Annexin Red Kit (Merck Millipore) and analyzed on a Guava easyCyte (Merck Millipore). Apoptotic cells were considered to be Annexin V positive and 7-AAD negative. In parallel to apoptosis detection, cell size was determined after lncRNA modulation.

Generation of stable lncRNA-overexpressing Ea.Hy926 cell lines
The endothelial cell line Ea.Hy926 was used to generate stable cell lines overexpressing LINCRNAs LINC00323-003 and MIR503HG. For lentiviral cloning, gene synthesis for LINCRNAs with 5´-AgeI and 3´-XbaI restriction sites was done by GeneArt® (Life Technologies). Subcloning of genes was then performed in pLV+ plasmid containing a bidirectional promotor to express GFP, puromycin selection cassette and LINCRNA simultaneously. Lentiviral particles were produced in HEK293T cells with the help of additional plasmids pMDL-g/pIIE, pRSV-rev, pCMV-VSVG, harvested from supernatant after 48 and 72 h and concentrated applying Amicon Ultra Centrifugal Filter Units with a molecular weight cut-off 100.000 (Amicon). Concentrated lentivirus was transduced at 1:10.000 to Ea.Hy926 cells and reached a transduction efficacy of ~30%. Puromycin selection was done at a concentration of 1 µg/ml after 72 h of transduction and kept for 7d in minimum.

Rescue assays applying lentivirus
Cells were transduced with GATA2 lentivirus at day of seeding and transfected with siRNA/GapmeR against lncRNA 24 h afterwards. After 48 h, cells underwent functional phenotyping (e.g. BrdU-ELISA). Microarray analysis 2 µg of total RNA from HUVECs cultured under normoxic or hypoxic conditions for 24 h was subjected to NCode™ Human Non-coding RNA microarray (Invitrogen) analysis. RNA of each group was pooled from triplicates to be hybridized on the microarray. Mircoarray analysis was performed using GeneSpring GX software.

RNA-Sequencing analysis
The resulting 318 million paired-end short reads (90bp + 90bp length) were evenly distributed across the six samples, with 51.4 million to 55.2 million per sample. The reads were mapped with a success rate of 97.5% on the hg19 genome via the STAR 2-pass method with default parameters (1). The information about potential splice junctions gathered by the first mapping pass from all sequenced libraries, were extracted and used to improve the second mapping pass. Exploiting the information of the mapped reads, RNA fragments were counted strandspecifically on the exon-level and subsequently summarized on the gene-level via featureCounts (2) using the protein-coding genes of the EnsEMBL GRCh37 (3) gene catalog for mRNA and the lncipedia 2.1 (4) catalog for lncRNAs. Genes were tested for differential expression using DESeq2 (5), comparing fragment counts of the three hypoxia libraries to the three normoxia libraries. Genes which resulted in Benjamini-Hochberg (6) adjusted pvalues < 0.05 were considered significant.

Mapping of MicroArray and RNA-Seq data
To compare the NCode microarray to the RNA-Seq results, the microarray probes had to be mapped on the EnsEMBL and lncipedia gene catalogs used for fragment counting in RNA-Seq analysis. First, the genomic coordinates of the 39,150 probe were transferred from the human genome assembly hg18 to hg19 (7) using liftover (8) with a success rate of more than 99.9%. Strand-specific overlaps between transferred probes and exons of protein-coding genes of the EnsEMBL gene catalog or the lncipedia 2.1 lncRNA catalog were identified via bedtools (9). This resulted in 17,444 probes mapping on 10,582 protein-coding EnsEMBL genes and 4,425 probes mapping on 3,594 lncipedia 2.1 lncRNAs.Since more than 4,500 protein-coding genes and more than 500 lncRNAs were covered by more than one probe, the median of log2 fold changes was calculated to get one value of the microarray for each gene.

GEO (Gene expression omnibus) accession numbers
Microarray and RNA-Seq datasets were uploaded to GEO at NCBI and grouped under a SuperSeries GSE70335. lncRNA homology search ECR browser was used to check for homology of identified lncRNAs. For this we used the default settings given (e.g. evolutionary conserved regions ECRs, minimum length 100 bp, minimum identity 70%) (see http://ecrbrowser.dcode.org/).

Statistical Analysis
For statistical analysis, GraphPad Prism (Version 4) was applied. In case of two groups, unpaired t-test was performed. Error bars in graphs indicate standard error of the mean (SEM). One experiment contained 3 replicates per group. Pearson correlation and Fisher exact test were performed in R version 3.1.2 (11). Asterisk mean: * = p < 0.05; ** = p< 0.01; *** = p < 0.001.