A total of 47,232 small RNA clone sequences distributed over 65 tissues, including the kidney glomerulus, were obtained from a recent survey 16. Two compendia with microarray data from mouse tissues, including lung 1718, were downloaded from the NCBI Gene Expression Omnibus repository. To ensure consistent mapping between datasets, clone/probe sequences were re-annotated against miRBase release 10.1 19 using a proprietary Matlab (Mathworks Inc. Natick, MA, USA) script. For each mature miRNA, a P-value for over-representation in the glomerulus library compared to the other tissues was calculated using Fisher's exact test. Likewise, P-values for differential expression in the lung compared to remaining adult tissues were determined using the Student's t-test. The t-test provides a useful metric of differential tissue expression, although the formal requirements for the underlying distribution of the data may not be completely met 20. Genomic localization of miRNAs was evaluated using data derived from the UCSC browser (July 2007 assembly) 21.

Microvascular fragments were isolated from mouse tissues and embryonic stem cell cultures using mechanical and enzymatic digestion followed by incubation with magnetic Dynabeads coated with anti-CD31 (anti-platelet endothelial cell adhesion molecule (PECAM)). The procedure was performed essentially as described previously 22. All mice were adult (8 to 12 weeks old) males, either wild-type C57BL/6 or Pdgfb^ret/ret backcrossed for seven generations onto a C57BL/6 background 23. RNA from vascular fragments and remaining tissue was prepared using miRNeasy Mini spin columns (Qiagen, Hilden, Germany). Samples were quantified with a NanoDrop spectrophotometer (Thermo Scientific Corporation, Waltham, MA, USA) and cDNA was synthesized using equal amounts of RNA in each reaction (High-Capacity Reverse Transcription Kit or MicroRNA Reverse Transcription Kit, Applied Biosystems, Foster City, CA, USA). Expression levels were determined using pre-designed TaqMan assays (Applied Biosystems) on a 7900 HT real-time PCR system, according to the manufacturer's instructions. Relative levels were calculated using the 2^-Ct method. Fli1 mRNA levels were determined using SYBR Green quantitative qPCR (95°C, 55°C, 72°C, 40 cycles) using the following primers: 5'-TATCAGATCCTGGGGCCAAC-3' and 5'-CTCATCAGGGTCCGTCATTT-3'.

The murine embryonic stem cell line R1 24 was routinely cultured on growth arrested mouse embryonic fibroblasts in stem cell medium composed of DMEM-Glutamax (Invitrogen, Carlsbad, CA, USA) supplemented with 25 mM HEPES pH 7.4, 1.2 mM sodium pyruvate, 19 mM monothioglycerol (Sigma-Aldrich, St. Louis, MO, USA), 15% fetal bovine serum (Gibco/Invitrogen, Carlsbad, CA, USA), and 1,000 U/ml leukemia inhibitory factor (Chemicon International/Millipore, Billerica, MA, USA). EBs were generated by aggregation of stem cells in hanging drops in the absence of leukemia inhibitory factor, as described previously 25. Briefly, EBs were collected after 4 days and seeded into 12-well dishes onto a layer of 0.9 ml solidified collagen type I solution composed of Ham's F12 medium (Promocell, Heidelberg, Germany), 6.26 mM NaOH, 20 mM HEPES, 0.117% NaHCO₃, 1% Glutamax-I (Gibco) and 1.5 mg/ml collagen I (PureCol, Advanced BioMatrix, San Diego, CA, USA). Immediately thereafter, a second layer of 0.9 ml collagen solution was added on top and allowed to polymerize. After 3 hours, 0.9 ml of stem cell medium supplemented with vascular endothelial growth factor A (VEGFA; PeproTech, Rocky Hill, NJ, USA), at a final concentration of 30 ng/ml, was added to induce angiogenic sprouting. The medium was replaced every second day. EBs were excised from the gels at day 14 and immediately processed for isolation of CD31+ vascular fragments, as described above. NG2+ cells were isolated with the same protocol using a rabbit anti-rat NG2 antibody (Chemicon; AB5320), after depletion of CD31+ cells from the cultures.

In situ hybridization was performed using a 3' DIG-labeled miRCURY LNA probe to mouse miR-145 and miR-126 (Exiqon, Vedbaek, Denmark) as previously described 26. For dual detection of miR-145 and the pericyte marker NG2, the immunostaining was performed after development of the in situ signal. Slides were washed in phosphate-buffered saline, blocked with 3% donkey serum and 1% bovine serum albumin in phosphate-buffered saline, then incubated with rabbit anti-rat NG2 antibody (Chemicon; diluted 1/50) overnight at 4°C, washed in phosphate-buffered saline, then detected with Alexa488-conjugated donkey anti-rabbit IgG (Invitrogen; diluted 1/200).

Mouse vascular aortic endothelial cells (VAECs; Dominion Pharmakine, Derio-Bizkaia, Spain) were cultured in RPMI 1640 media (Sigma) supplemented with 10% fetal calf serum (Gibco), 1 μg/ml dexamethasone, 10 U/ml heparin, 50 U/ml penicillin/streptomycin and 75 μg/ml EC growth factor supplement (Sigma). For scratch wound migration assays, cells were transfected by electroporation (Nucleofector system, Basic Endothelial Cell Kit, Amaxa Inc/Lonza group ltd, Basel, Switzerland) using 0.5 μg of synthetic mature miR-145 double-stranded RNA (dsRNA; Pre-miR-145; Applied Biosystems) or negative control dsRNA (Stealth siRNA negative control; Applied Biosystems), seeded onto 6-well plates and cultured for 48 hours. Scratch wounds were generated in the cell monolayer using a pipet tip and each wound was photographed at 0 and 24 hours. Wound widths were evaluated blindly at both time-points and the average amount of closure was determined for each replicate transfection. VAEC proliferation was measured by quantification of 5'-bromo-2'-deoxyuridine (BrdU) incorporation. Cells were pulsed for 4 hours with 20 μM BrdU and DNA synthesis was determined using a colorimetric ELISA (Calbiochem/Merck, Darmstadt, Germany) according to the manufacturer's instructions. Absorbance was measured at dual wavelengths of 450 to 540 nm.

Migration of HUVECs in response to a stable gradient of VEGFA-165 (PeproTech; 0 to 50 ng/ml over a distance of 400 μm) or BJ-hTERT (human foreskin fibroblast) cells in response to platelet-derived growth factor (PDGF)-BB (0-20 ng/ml) was examined using a microfluidic chemotaxis chamber, essentially as previously described 27. HUVECs were transferred to 3-cm culture dishes coated with type A gelatin from porcine skin (Sigma) and were allowed to attach to the dish in EGM-2MV medium (Lonza) with serum and supplement growth factors. After 2 hours the medium was aspirated and the cells were transfected with 0.5 μg of Pre-miR negative control, Pre-miR-145, Anti-miR negative control or Anti-miR-145 (Applied Biosystems) using siPORT NeoFX (Ambion, Austin, TX, USA) in serum and growth factor free EBM-2 medium (Lonza) containing 0.2% bovine serum albumin. After 24 hours the gradient experiment was initiated. BJ-hTERT cells were cultured in minimal essential medium (MEM, Invitrogen) containing 10% fetal calf serum (Gibco), 1 mM sodium pyruvate (Gibco) and non-essential amino acids (Gibco). Cells were transfected using electroporation (0.5 μg, Nucleofector system, Amaxa) and were allowed to rest between 24 and 48 h before being seeded onto gelatin A-coated culture dishes and serum starved overnight, before onset of gradient. VEGFA-165 or the PDGF-BB gradients were generated in serum-free cell medium. Cell migration was tracked during 3 hours (HUVECs) or 4 hours (BJ-hTERT cells) using a Cell Observer System (Carl Zeiss AB, Stockholm, Sweden) fitted with a Zeiss Axiovert 200 microscope, an AxioCam MRm camera, a motorized X/Y stage, and an XL incubator with equipment for temperature and CO₂ control (Zeiss). Cells were kept in a humidified atmosphere of 5% CO₂ in air at 37°C during all experiments. AxioVision software (Zeiss) was used for time-lapse imaging and cell tracking.

Oligonucleotides (65 bp) harboring wild-type or mutated miR-145 binding sites from the mouse Fli1 3' UTR (Additional data file 1) were annealed and ligated into the HindIII and SpeI sites of the pMIR-REPORT CMV-firefly luciferase reporter vector (Applied Biosystems). All constructs were verified by sequencing. HEK293 cells were seeded onto 24-well plates at a density of 50,000 cells/well and cultured overnight in DMEM (10% fetal calf serum) without antibiotics. Cells were transfected with 60 ng of pMIR-REPORT, 8 ng of pRL-SV40 renilla luciferase control vector and 10 pmol of Pre-miR negative control or Pre-miR-145 using Lipofectamine 2000 (Invitrogen) and luciferase activity was assayed after 48 hours using the Dual-Luciferase Reporter System (Promega, Madison, WI, USA).

Long mouse and human Fli1 3' UTR fragments were amplified by PCR using the following primers (numbers indicate the position starting from the stop codon): 5'-AACTAACACCAGTTGGCCTTC-3' and 5'-CGTCAGGAGTGTCTGAGTTTG-3' (1-704); 5'-GCTTCTTCTAGCTGAAGCCCATC-3' and 5'-GTCAAATTATTTTACAACATGG-3' (3-1,391). Amplimers were cloned in psiCHECK-2 (Promega) to generate Renilla luciferase-3' UTR reporter constructs. Basal expression of firefly luciferase from the same plasmid served as an internal control. HEK293T cells seeded in 96-well plates were cotransfected with plasmid (50 ng per well) and synthetic miRNA (0.25 to 2.5 pmol per well; Biomers, Ulm, Germany) using Lipofectamine 2000. Luciferase activity was assayed 24 hours after transfection as described 28. Nucleotides 2 and 4 in the seed region of three predicted miR-145 sites within the mouse 3' UTR fragment were mutated using Multisite-Quickchange (Stratagene/Agilent, Santa Clara, CA, USA). Results represent Renilla/firefly luciferase ratios from four independent experiments performed in triplicates. Statistical significance was evaluated using Student's t-test.

VAECs were electroporated with either Pre-miR-145 or Pre-miR negative control as described above. Nuclear extracts were prepared using the CelLytic NuCLEAR kit (Sigma) at 72 hours post-transfection. Western blotting was performed using a Fli1 antibody (Sc-356; Santa Cruz Biotechnology, Santa Cruz, CA, USA) at 2 mg/ml and ECL reagents (Amersham Biosciences/GE Healthcare Bio-Sciences, Uppsala, Sweden). As a loading control, the membrane was stripped and reprobed using a lamin A/C antibody (Sc-7293, Santa Cruz Biotechnology) at a dilution of 1/1,500. Densitometric analysis was performed using ImageJ software.

Protein-coding genes with selective expression in the microvasculature were identified in a recent study based on their enrichment in the lung and in the kidney glomerulus 29. Differential expression in both of these endothelium-rich tissues minimized contamination by epithelial transcripts and permitted identification of numerous known and novel microvascular markers. Here we applied a similar strategy to identify candidate microvascular-enriched miRNAs. Data were gathered from three different sources: a set of small RNA sequence libraries of varying sizes covering 65 mouse tissues, including the glomerulus 16, and two compendia with microarray data from adult mouse tissues, including lung 1718 (Figure 1a). miRNAs were scored for enrichment in glomerulus and lung and this formed the basis of our selection (Additional data file 2).

Among those with favorable scores in this analysis, miR-126-3p and miR-126-5p (the two mature forms of miR-126) stood out as strongly enriched in both glomerulus and lung (Figure 1b). Several other miRNAs also appeared as promising candidates for selective vascular expression, including miR-145, miR-30d, miR-23b and miR-24 (within the dashed lines in Figure 1b). miRNAs connected by thick grey lines in the figure are co-localized in the genome (<10 kb) and likely derive from the same polycistronic transcript 30.

Based on the above described in silico analyses, we chose to further characterize the expression of miR-126-3p (the predominant mature form of this miRNA, hereafter referred to as miR-126), miR-145, miR-30d, miR-23b, miR-24 and miR-23a; the latter being co-transcribed with miR-24 1. Microvascular fragments were isolated from adult mouse tissues using mechanical and enzymatic digestion followed by separation using anti-CD31 (PECAM)-coated magnetic beads. RNA was prepared from the fragments and the remaining tissue fractions. qRT-PCR analysis showed that miR-126 was highly differentially expressed in CD31+ fragments in all adult organs assayed, with fragment-to-surrounding tissue ratios ranging from 90 to 250. These ratios are in parity with the endothelial markers Cd31 and Kdr (Vegfr2/Flk1; Figure 2). The remaining miRNAs were also enriched in vascular fragments to varying degrees. In particular, miR-145 showed consistent and high differential expression in microvessels (24-, 7-, 75- and 18-fold for brain, muscle, skin and kidney, respectively). In addition, miR-23a and miR-24 were consistently differentially expressed, with enrichments ranging from 5- to 16-fold. Gapdh, included as a control, showed weak or no enrichment across the panel.

To evaluate miRNA expression in immature blood vessels, CD31+ microvascular fragments were isolated from mouse kidneys at embryonic day 14, as well as from VEGFA-induced angiogenic sprouts formed in EB cultures. miR-126 showed strong enrichment in CD31+ fractions from both tissues (Figure 3). miR-23a and miR-24 were enriched in sprouts from EBs but not in fragments from embryonic day 14 kidneys. miR-145, in contrast, was predominantly expressed in the leftover fractions. The pericyte marker Pdgfrb showed a similar pattern with strong enrichment in CD31+ fragments from adult tissues but not in embryonic vascular fragments (Figures 2 and 3), which suggests that miR-145 could be expressed by pericytes.

To test the hypothesis that miR-145 is expressed by pericytes, CD31+ fragments were purified from the brains of Pdgfb retention-motif mutant mice (Pdgfb^ret/ret) that lack a stretch of basic amino acids in the carboxyl terminus of PDGF-B. These mice display defective pericyte investment of microvessels 23. As expected, Pdgfrb mRNA levels were reduced in Pdgfb^ret/ret vascular fragments compared to wild-type mice (P = 0.001; Figure 4a). Expression of miR-145 was also reduced in mutant microvessels (P = 0.008), whereas no notable differences were observed for the other miRNAs. These results gave further support to the idea that microvascular miR-145 expression is derived primarily from pericytes.

As a complementary approach, CD31+ ECs and NG2+ pericytes were isolated from EB cultures. Expression levels were determined using qRT-PCR and the ratio of the signals from the two fractions was determined. In accordance with the pericyte markers, miR-145 expression was higher in NG2+ cells compared to CD31+ cells (Figure 4b). In contrast, the endothelial marker Cd31 and miR-126 were highly enriched in the CD31+ fraction.

Next, we performed in situ hybridization on tissue sections using probes specific to miR-145 and miR-126. As expected, miR-145 stained smooth muscle cells in larger vessels whereas miR-126 stained ECs (staining patterns in kidney arteries are shown in Figure 4c, d). In brain parenchyma, miR-145 showed staining in solitary scattered cells, consistent with expression in pericytes (Figure 4f). Double staining using an NG2 antibody confirmed co-expression of the two molecules in brain capillaries (Figure 4g) and in small caliber blood vessels in the kidney (Figure 4e). There was, however, no detectable expression of miR-145 in pericytes in the heart, where the expression was confined to arterioles and larger vessels (Figure 4h). Compared to miR-145, NG2 staining indicated larger areas in kidney and brain microvessels (Figure 4e, g). This suggests that miR-145 is expressed by a subset of pericytes. However, this could also be explained by NG2's subcellular distribution in pericyte processes that extend from the main cell body and cover the capillary cell surface.

We conclude that miR-145 is selectively expressed in microvessel pericytes whereas the remaining miRNAs are expressed in ECs.

miRNA target prediction software was used to identify possible targets for miR-145. The highest-scoring predicted target using the miRanda algorithm 31 was the gene encoding the Ets transcription factor Friend leukemia integration 1 (Fli1). Fli1 also scored favorably using picTar 32 and TargetScan 33, the latter identifying four evolutionarily conserved miR-145 binding sites in the Fli1 3' UTR (Figure 5a).

To evaluate if the predicated sites can bind miR-145 and induce silencing, we generated a series of eight constructs, each consisting of a CMV-luciferase reporter followed by a portion of the Fli1 3' UTR containing a wild-type or mutated site. Transfection of a synthetic miR-145 mimic dsRNA (Pre-miR-145) into HEK293 cells significantly reduced reporter activity for all predicted sites (Figure 5b). Single base-pair mutations reduced or abolished the effect of miR-145 on reporter activity in all cases. An empty reporter vector, lacking a cloned target site in the 3' UTR, was not affected by miR-145 overexpression.

Constructs with either a full length human or a long (700 bp) fragment of the mouse Fli1 3' UTR were generated to evaluate the predicted sites in their natural sequence context. Cotransfection with a miR-145 mimic significantly reduced reporter activity compared to transfection without a miRNA or with an unrelated miRNA (Figure 5c). The effect was abolished when mutations were introduced in three out of four predicted sites in the mouse 3' UTR (Figure 5d).

An effect of miR-145 on endogenous Fli1 protein levels was demonstrated in VAECs. Western blot analysis 72 hours post-transfection of Pre-miR-145 showed that Fli1 protein levels were decreased compared to cells treated with Pre-miR negative control (Figure 5e). Since miRNAs can induce both translational repression and target mRNA degradation, we performed qRT-PCR to assess the expression of Fli1 mRNA after introduction of Pre-miR-145 or Pre-miR negative control. No significant reduction was observed (Figure 5f). Translational repression without mRNA degradation has been described for numerous miRNAs, and our findings are consistent with a previous report suggesting that miR-145 is primarily a repressor of translation 34. VAECs were also transfected with Anti-miR-145 in a loss-of-function experiment. This did not affect Fli1 levels (data not shown), which is consistent with low endogenous expression of miR-145 in this cell type (Additional data file 3).

In order to assess the role of miR-145, functional assays were performed in human foreskin fibroblasts, and in ECs that express Fli1. Cell migration is often guided by growth factor gradients in vivo. PDGF-BB is known to stimulate migration of several different cell types, including smooth muscle and fibroblasts 3536. It is also a key regulator of pericytes in vivo 37. We therefore investigated cell migration in response to a stable gradient of PDGF-BB using a microfluidic chemotaxis chamber. Human foreskin fibroblasts (BJ-hTERT) were transfected with Pre-miR-145 or Pre-miR negative control in gain-of-function experiments and with Anti-miR-145 or Anti-miR-control in loss-of-function experiments (expression levels of miR-145 in BJ-hTERT cells are presented in Additional data file 3). Individual cells were tracked using time-lapse microscopy during 3 hours 27. The average migrated distance per cell toward the high-end of the PDGF-BB gradient was reduced by more than 50% in Pre-miR-145 transfected cells, whereas migration perpendicular to the gradient was only slightly, and not significantly, reduced (Figure 6a). Similarly, migration towards the high-end of the gradient was reduced by Anti-miR-145 (Figure 6b). Migration perpendicular to the gradient was also significantly reduced by this treatment.

To investigate the effect of miR-145 on VEGFA-165-induced migration, HUVECs were cultured in a stable gradient of VEGFA-165. Control cells migrated consistently toward the high-end of the gradient, whereas Pre-miR-145-transfected cells exhibited a clear (>50%) reduction in migration in this direction (Figure 6c). Migration perpendicular to the gradient was not significantly reduced.

Migration was also evaluated on VAECs cultured in EC growth factor supplemented medium using a wound healing assay. Migration was reduced in Pre-miR-145 transfected cells compared to cells transfected with a dsRNA control (Stealth siRNA control, Invitrogen; Figure 6d). However, proliferation rate, as determined using a BrdU ELISA assay, was not affected. These findings point to a role for miR-145 in regulation of cell migration.