Adenosine-to-Inosine Editing of MicroRNA-487b Alters Target Gene Selection After Ischemia and Promotes NeovascularizationNovelty and Significance
Rationale: Adenosine-to-inosine editing of microRNAs has the potential to cause a shift in target site selection. 2′-O-ribose-methylation of adenosine residues, however, has been shown to inhibit adenosine-to-inosine editing.
Objective: To investigate whether angiomiR miR487b is subject to adenosine-to-inosine editing or 2′-O-ribose-methylation during neovascularization.
Methods and Results: Complementary DNA was prepared from C57BL/6-mice subjected to hindlimb ischemia. Using Sanger sequencing and endonuclease digestion, we identified and validated adenosine-to-inosine editing of the miR487b seed sequence. In the gastrocnemius muscle, pri-miR487b editing increased from 6.7±0.4% before to 11.7±1.6% (P=0.02) 1 day after ischemia. Edited pri-miR487b is processed into a novel microRNA, edited miR487b, which is also upregulated after ischemia. We confirmed editing of miR487b in multiple human primary vascular cell types. Short interfering RNA–mediated knockdown demonstrated that editing is adenosine deaminase acting on RNA 1 and 2 dependent. Using reverse-transcription at low dNTP concentrations followed by quantitative-PCR, we found that the same adenosine residue is methylated in mice and human primary cells. In the murine gastrocnemius, the estimated methylation fraction increased from 32.8±14% before to 53.6±12% 1 day after ischemia. Short interfering RNA knockdown confirmed that methylation is fibrillarin dependent. Although we could not confirm that methylation directly inhibits editing, we do show that adenosine deaminase acting on RNA 1 and 2 and fibrillarin negatively influence each other’s expression. Using multiple luciferase reporter gene assays, we could demonstrate that editing results in a complete switch of target site selection. In human primary cells, we confirmed the shift in miR487b targeting after editing, resulting in a edited miR487b targetome that is enriched for multiple proangiogenic pathways. Furthermore, overexpression of edited miR487b, but not wild-type miR487b, stimulates angiogenesis in both in vitro and ex vivo assays.
Conclusions: MiR487b is edited in the seed sequence in mice and humans, resulting in a novel, proangiogenic microRNA with a unique targetome. The rate of miR487b editing, as well as 2′-O-ribose-methylation, is increased in murine muscle tissue during postischemic neovascularization. Our findings suggest miR487b editing plays an intricate role in postischemic neovascularization.
Neovascularization is the body’s natural repair mechanism to restore blood flow to ischemic tissues. Neovascularization is comprised of angiogenesis and arteriogenesis. Both angiogenesis and arteriogenesis are highly multifactorial processes, and clinical trials aimed at stimulating neovascularization in patients with ischemic disease have been relatively unsuccessful.1,2 However, in recent years, we, and others, have shown that microRNAs simultaneously regulate multiple facets of neovascularization, acting as a master switch.3,4 MicroRNAs are endogenous single-stranded RNA molecules of ≈22 nucleotides long that downregulate expression of their target genes. A single microRNA can influence the expression of hundreds of genes and it has been shown that complementarity of the target gene to the microRNA’s seed sequence, nucleotide +2 to +8 from the 5′ end of the microRNA, is crucial for target recognition.5
The role of microRNAs in vascular disease is well established. They have been shown to influence complex vascular pathological processes, including atherosclerosis, restenosis, aneurysm formation, and neovascularization.4,6 The vasoactive microRNA miR487b is upregulated during chronic hypertension and aneurysm formation in rats.7 Furthermore, we have shown that miR487b as well as 3 other microRNAs, all of which are transcribed from a polycistronic microRNA gene cluster located on the long arm of human chromosome 14 (14q32), directly affect blood flow recovery after ischemia in mice.3 According to TargetScan (http://www.targetscan.com), however, miR487b has only 16 conserved putative target genes, whereas most other microRNAs have hundreds of predicted conserved targets.8 Despite this, miR487b is still highly conserved in mammals (Figure 1A) and broadly regulates neovascularization.
Like all RNAs, microRNAs are subject to modifications, some of which result in sequence editing, which may alter their function. About 90% of all RNA editing events are caused by adenosine-to-inosine (A-to-I) editing.9 A-to-I editing is catalyzed by the adenosine deaminase acting on RNA family (ADARs) which binds to double-stranded RNA and hydrolytically deaminates adenosine residues changing it to an inosine.10 Unlike adenosine, inosine preferentially binds to cytidine and is therefore interpreted as guanosine by the cellular machinery. This can have several consequences, ranging from destabilization of the RNA molecule by affecting the secondary structure of the RNA, to changing the sequence of the functional part of the RNA.11 Dysregulation of A-to-I editing has been shown to contribute to various diseases, including vascular disease.12
Another common RNA modification is 2′-O-ribose-methylation (2′OMe), which has been shown to be interdependent with A-to-I editing.13–15 2′OMe is conserved among all major classes of eukaryotic RNA and is guided by small nucleolar RNAs.16 2′OMe of an adenosine residue was shown to protect the adenosine from A-to-I editing in vitro.13–15
Primary microRNAs (pri-microRNAs), including several 14q32 pri-microRNAs,17,18 can be A-to-I edited by ADARs.11 Editing of a pri-microRNA can influence microRNA maturation, but more interestingly, editing of the seed sequence can lead to microRNA redirection and the subsequent recognition of an altered set of target genes, or targetome. This type of editing has been shown to have far-reaching consequences on biological processes. For example, unedited mature 14q32 miR376a* was shown to promote invasive growth of glioblastoma cells, whereas edited mature miR376a* suppressed this phenotype.19 Furthermore, retargeting of miR376a through editing of the seed sequence was shown to play a role in uric acid regulation in select brain tissues.20 To date, however, no occurrences of microRNA editing during vascular remodeling have been reported.
Given the narrow range of predicted miR487b target genes mentioned above, we hypothesized that miR487b undergoes A-to-I editing under ischemic conditions. We show here that the first nucleotide of the seed sequence of pri-miR487b is indeed subject to both A-to-I editing and to 2′OMe in mice and in human primary vascular cells. Using the hindlimb ischemia (HLI) model, we also show that the rate of miR487b A-to-I editing, as well as methylation, is increased under ischemic conditions. The resulting edited mature miR487b has a unique targetome and acts as a proangiogenic microRNA.
The data that support the findings of this study are available from the corresponding author on reasonable request.
Terminology for miR487b and its modified forms are summarized in the Table.
Unilateral HLI was induced in C57Bl/6 mice by electrocoagulation of the left femoral artery proximal to the superficial epigastric arteries. Mice were euthanized by cervical dislocation and the adductor and gastrocnemius muscles were excised en bloc and snap frozen on dry ice before (T0) and at 1, 3, and 7 days (T1, T3, and T7 respectively) after induction of HLI.
MiR487b A-to-I Editing
To identify and quantify pri-miR487b editing, we performed complementary DNA (cDNA) sequencing and total pri-miR487b PCR amplification from cDNA followed by selective digestion of only unedited miR487b using PfeI restriction enzyme (Thermo Scientific). Individual TaqMan miR assays (Applied Biosystems) were used to quantify wild-type miR487b (miR487b-WT) and edited miR487b (miR487b-ED), using a standard kit and a custom kit (targeting AGUCGUACAGGGUCAUCCACU), respectively.
Reverse-Transcription at Low dNTP Concentrations Followed by Quantitative-PCR
To identify and quantify 2′OMe of miR487b, we performed reverse-transcription at low dNTP concentrations followed by quantitative-PCR RTL-Q, a modification to the endpoint PCR based RTL-P method, described by Dong et al21 (Online Figure II).
Selective miRNA overexpression achieved in cultured primary human vascular cells and in ex vivo aortic segments by adding synthetic pre-miR487b-WT or pre-miR487b-ED to culture medium and comparing it to a control pre-miRNA. In vitro scratch wound healing assays and ex vivo mouse aortic ring assays were performed as described previously.3,22
A detailed description of methods and associated references can be found in the Online Data Supplement.
Discovery of In Vivo A-to-I Editing of the Seed Sequence of Pri-miR487b
To identify editing sites in pri-miR487b, we used Sanger sequencing. cDNA chromatograms of pri-miR487 consistently contained a secondary guanine peak which was absent in chromatograms obtained from genomic DNA, indicating the occurrence of A-to-I editing (Figure 1B). A-to-I editing occurs on the second nucleotide of the mature miR487b sequence, which is the first nucleotide of the microRNA’s seed sequence, changing the seed from 5′-AUCGUAC-3′ to 5′-IUCGUAC-3′ (Figure 1A).
Validation and Quantification of Pri-miR487b Editing in Muscle Tissues
pri-miR487b contains a digestion site for PfeI, which is abolished by A-to-I editing. PfeI’s specificity and efficiency was confirmed by digestion of amplified synthetic DNA of wild-type and edited pri-miR487b (Figure 2A).
Next, we investigated pri-miR487b editing during different stages of vascular remodeling. Amplified cDNA from gastrocnemius and adductor muscle tissues before (T0) and at 1, 3, and 7 days (T1, T3, and T7 respectively) after HLI were digested using PfeI. Even after more than 10× overdigestion, an uncut band remained visible in all samples, confirming the occurrence of A-to-I editing (Figure 2B).
In the ischemic gastrocnemius muscle, the percentage of edited pri-miR487b increased by almost 2-fold 1 day after surgery (T0=6.7±0.4% versus T1=11.7±1.6%; P=0.02), but was decreased at 3 and 7 days after surgery (5.0±0.4% and 3.8±0.1%; P=0.04 and P=0.0005 versus T0, respectively; Figure 2C).
In the adductor muscle, however, the percentage of edited pri-miR487b was unchanged at T1 (2.8±0.1%) compared with T0 (3.7±0.8%, P=0.3), but decreased at T3 (1.21±0.4%, P=0.05 versus T0). At T7, percentage of edited pri-miR487b was increased to 8.0±1.3% (P=0.03 versus T0; Figure 2C).
Mature miR487b Editing in Muscle Tissue and Human Primary Vascular Cells
Next, we determined if edited pri-miR487b is processed into mature miR487b-ED. We used a custom TaqMan rt/qPCR kit to quantify mature miR487b-ED. We found that the edited pri-microRNA is indeed processed, in part, to form a mature microRNA, with percentages of edited miR487b ranging from 1% to 4% of the total mature miR487b in both ischemic gastrocnemius and adductor muscles (Figure 2D). The resulting mature microRNA, miR487b-ED, is an entirely new microRNA, with a unique seed and mature microRNA sequence according to miRbase (http://www.miRbase.com).
Percent mature miR487b-ED in gastrocnemius was low before surgery but increased by more than 2-fold afterward (T0=1.3±0.5% and T3=3.1±0.2%; P=0.02). Percent mature miR487b-ED in the adductor muscle, however, followed an opposite trend, as percentage editing decreased from 3.3±0.4% at T0 to 1.0±0.2% at T3 (P=0.05 versus T0; Figure 2D). Observed changes in percentage miR487b-ED after HLI are predominantly caused by changes in amount of miR487b-ED, rather than by fluctuations in miR487b-WT expression (Online Figure III).
To exclude the possibility that the influx of circulating bone marrow derived cells expressing miR487b-ED could have caused these effects, we measured miR487b in the cDNA of murine bone marrow. Neither miR487b-WT nor miR487b-ED were expressed in the bone marrow (data not shown).
Editing of human pri-miR487 and mature miR487b at the same position was confirmed in human umbilical artery fibroblasts (HUAFs). In cultured HUAFs, both pri-miR487b and miR487b were found to be edited by ≈11% (Online Figure IVA and IVB; Figure 2F). RNA binding protein immunoprecipitation performed on HUAF lysate showed pri-miR487b enrichment in both the ADAR1 and ADAR2 pull-down fractions when compared with the negative control IgG pull down, confirming complex formation of ADAR enzymes with pri-miR487b (Figure 2G).
Mature miR487b-WT and miR487b-ED expression were also measured in human umbilical artery smooth muscle cells (HUASMCs) and human umbilical vein endothelial cells (HUVECs). We found that fibroblasts had the highest expression of total pri-miR487b and total mature miR487b, as was expected from previous studies.7 Expression was 5-fold lower in endothelial cells (Figure 2E; Online Figure IVF). With ≈7.5% mature miR487b-ED in both HUASMCs and HUVECs, percentage mature miR487b editing was also highest in HUAFs (11%; Figure 2F; P=0.02 and P=0.01 versus HUASMCs and HUVECs respectively). Consistent with this finding, expression of ADAR1 and especially ADAR2 appeared higher in HUAFs compared with HUASMCs and HUVECs (Online Figure IVC and IVD).
2′OMe of the 2+ Adenosine in Pri-miR487b
Besides A-to-I editing, RNA molecules are frequently subject to 2′OMe. As it has been shown that 2′OMe of an adenosine residue in mRNA protects it from A-to-I editing in vitro,13–15 we hypothesized that pri-miR487b can be 2′-O-methylated by the methyltransferase fibrillarin. Fibrillarin-IP showed that pri-miR487b is indeed enriched in the fibrillarin pull-down fraction, compared with negative control IgG pull down (Figure 3A). We then performed reverse-transcription at low dNTP concentrations followed by quantitative-PCR (RTL-Q) to quantify the estimated specific 2′-O-ribose-methylated fraction (EMF) in the gastrocnemius and adductor muscle before and after HLI induction. Interestingly, we found specific 2′OMe (2′OMeA) of pri-miR487b at the exact same adenosine residue that is subject to editing (+2A), in all sample types (Figure 3C and 3D), which corresponds with the adenosine position where A-to-I editing occurs. We observed a similar baseline of specific 2′OMe in both adductor and gastrocnemius muscle of ≈33% EMF before surgery (32.8±14% and 33.6±12%, respectively). After femoral artery ligation, the EMF appeared to increase initially in the ischemic gastrocnemius muscle (T1=53.6±12%; P=0.27), but decreased again by day 3 and 7 (23±11% and 17±9%; P>0.1 and P=0.03 versus T1, respectively; Figure 3C). In the adductor muscle, the EMF of pri-miR487b also appeared to increase 1 day after surgery to 43.8±11% (P=0.55), but on T3 2′OMe had almost been abolished (EMF of 3.2±1.6%; P=0.02 versus T1), to recover again by day 7 (EMF of 29.0±12%; P>0.1 versus T3; Figure 3D). Therefore, specific 2′OMeA of pri-miR487b followed the same, and not the hypothesized inverse, pattern as pri-miR487b A-to-I editing.
2′OMe of the same adenosine residue was confirmed for human pri-miR487b. We observed an EMF of 24±7% in HUAFs, an EMF 9.5±9% of in HUASMC, and an EMF of 31±19% in HUVECs (Figure 3B). Relative fibrillarin expression measured in these different cell types were consistent with these differences in average pri-miR487b methylation (Online Figure IVE).
ADAR1, ADAR2, and Fibrillarin in miR487b Expression, Editing, and Methylation
To gain further insights into miR487b editing and methylation and their potential interplay, ADAR1, ADAR2, and fibrillarin were individually repressed in HUAFs using short interfering RNAs. Knockdown of ADAR1 or ADAR2 both resulted in significantly decreased amount of pri-miR487b editing compared with control short interfering RNA (Figure 4A and 4B; P=0.02 and P=0.05, respectively), demonstrating that both ADAR1 and ADAR2 can edit pri-miR487b. This reduction was not observed in percentage mature miR487b editing, indicating that microRNA processing contributes to determine the final levels of mature edited microRNAs (Figure 4C).
ADAR1 and ADAR2 are also known to play independent roles in microRNA biogenesis and maturation.23,24 We found that repression of either ADAR1 or ADAR2 caused a 2-fold reduction in total pri-miR487b expression (Figure 4D; P<0.01 for both), but at mature miR487b level, only miR487b-WT was reduced after ADAR1 repression (Figure 4E; P=0.02). Repression of fibrillarin resulted in a 17% decrease in pri-miR487b 2′OMeA relative to control (P=0.07), indicating that the methylation of pri-miR487b is indeed fibrillarin-dependent 2′-O-methylation (Figure 4F). Reduced pri-miR487b 2′-O-methylation after fibrillarin knockdown had little effect on miR487b A-to-I editing, however, suggesting these pri-miR487b modifications are not directly interdependent (Figure 4B and 4C).
Interestingly, specific knockdown of ADAR1 or ADAR2 did not only reduce A-to-I editing in both cases, but also significantly increased pri-miR487b 2′OMeA by 1.4-fold compared with control (P<0.001 for both). Analysis of fibrillarin expression revealed that knockdown of ADAR1 or ADAR2 causes a 3-fold increase in fibrillarin expression (Figure 4G; P<0.001 and P=0.03 versus control, respectively). Correspondingly, short interfering RNA-mediated fibrillarin knockdown also caused significant increases in ADAR1 and ADAR2 expression (Figure 4G; P=0.01 versus control for both). Taken together, these findings suggest that pri-miR487b 2′-O-methylation and A-to-I editing are not directly interdependent, but instead that there is a direct inverse correlation between ADAR1 and ADAR2 expression on the one hand and fibrillarin expression on the other.
Comparison of Putative Targets of Edited and Unedited miR487b
Changes in the seed sequence of a microRNA may shift its targetome. Using Diana-MR-microT software,25,26 we found that there was only an ≈15% overlap between the miR487b-WT and miR487b-ED targetomes (Figure 5A). The overlapping target genes all contained separate binding sites for both miR487b-WT and miR487b-ED in their 3′UTR.
Pathway enrichment analysis using the PANTHER software (Protein Analysis Through Evolutionary Relationships; http://pantherdb.org)26a revealed enrichment for CCKR (cholecystokinin receptor) signaling, the Ras pathway, the PDGF (platelet-derived growth factor) signaling pathway, and insulin/IGF (insulin-like growth factor) associated pathways (P=0.004, P=0.023, P=0.036, and P=0.049, respectively) within the miR487b-WT’s murine targetome, but none for its human targetome (Figure 5A; Online Table IIIA). The targetome of miR487b-ED on the other hand contained robust enrichment for several pathways in both man and mouse. Specifically, miR487b-ED’s murine targetome contained enrichment for the Wnt signaling pathway and the Cadherin signaling pathway (both P>0.00001). Furthermore, within miR487b-ED’s human targetome Wnt signaling was also enriched (P=0.004) in addition to the angiogenesis pathway, the PDGF signaling pathway, and insulin/IGF associated pathways (P=0.004, P=0.015, and P=0.041, respectively). Numerous publications could be found linking each enriched pathway to neovascularization, ischemia, and vascular disease for all enriched pathways with the exception of the CCKR signaling (Online Table IIIB).
Validation of Target Shift Induced by miR487b Seed Sequence Editing
Next, we set out to test whether the single nucleotide change in the seed sequence of miR487b indeed causes a shift in target site selection. We performed dual luciferase reporter gene assays using endogenous and mutated putative miR487b binding sequences from IRS1 (insulin receptor substrate 1) and BMP1 (bone morphogenetic protein 1).
The 3′UTR of IRS1 contains 1 predicted binding site for miR487b-WT and 1 predicted binding site for miR487b-ED (Online Figure V). Dual luciferase reporter gene assays showed that miR487b-WT caused an 11±6% inhibition of luciferase activity of the endogenous IRS1 sequence (P=0.3; Figure 5B), which is similar to previously published results.7 MiR487b-ED, on the other hand, significantly inhibited luciferase activity by 54±2% (P<0.0001) through binding to endogenous IRS1 3′UTR. When the first putative binding site of endogenous IRS1 3′UTR was mutated so that both sites became putative target sites for miR487b-ED, miR487b-WT had no effect when compared with the control pre-miR (P=0.95), whereas knockdown by miR487b-ED increased to 67±2% compared with the control (P<0.0001; Figure 5B). Conversely, when the second putative binding site was mutated instead, so that both sites became putative target sites for the miR487b-WT, miR487b-WT inhibited luciferase activity by 49±2% (P<0.0001), whereas miR487b-ED had no effect (P=0.9).
BMP1 is predicted to have 6 separate binding sites for miR487b-ED (Online Figure VI). Binding of miR487b to BMP1’s first 3 putative binding sites (part A) was assessed separately from binding to its last 3 target sites (part B), which confer an estimated 70% and 30% of the combined binding strength, respectively. Luciferase activity clearly reflected the difference in predicted binding strength, with miR487b-ED repressing luciferase activity of BMP1 part A by 42±2% and only 19±2% for part B (Figure 5C; P<0.0001 for both). miR487b-WT on the other hand had no effect on luciferase activity of either BMP1 constructs containing endogenous miR487b-ED binding sites. Next, a single nucleotide was mutated in each binding site, so that all sites within the BMP1 3′UTR sequences became putative target sites for miR487b-WT instead. miR487b-WT repressed luciferase activity of the mutated BMP1 part A by 17±1% (P<0.0001), but had no effect on the weaker sites of mutated BMP1 part B (P=1.0). miR487b-ED had no effect on the mutated BMP1 part B either, but was still able to cause a mild repression of mutated BMP1 part A luciferase activity (7±1%; P<0.01).
Specific Regulation of Endogenous Targets by miR487b-WT and miR487b-ED
To confirm endogenous target regulation by miR487b-WT and miR487b-ED is also selective, we overexpressed miR487b-WT or miR487b-ED in HUAFs and HUVECs and measured the expression of a subset of putative target genes. Expression of miR487b-WT targets DNAJC9, B3GALNT2, and MAP2K4 were repressed after miR487b-WT overexpression compared with control in both HUAFs and HUVECs, but not after miR487b-ED overexpression (Figure 6A through 6C). RPS6KB1 and BMP1 on the other hand only contain miR487b-ED binding sites and were repressed after miR487b-ED overexpression in both HUAFs and HUVECs, but not after miR487b-WT overexpression (Figure 6D and 6E). Furthermore, endogenous IRS1, with 1 binding site for miR487b and miR487b-ED each, was repressed after overexpression of either miR487b-WT and miR487b-ED, again in both HUAFs and HUVECs (Figure 6F).
Taken together, these finding show that editing of miR487b does change binding site selection and that miR487b-WT and miR487b-ED regulate different targetomes.
Expression of miR487b-WT and miR487b-ED Targetomes in the Adductor Muscle
A microRNA influences complex biological processes by fine-tuning the expression of its broad targetome rather than causing strong downregulation of single genes.27 We examined miR487b-WT and miR487b-ED targetome gene expression by using an existing data set of whole-genome expression microArray performed on total adductor muscle mRNA of mice after induction of HLI.28 Average expression of the 1353 detected genes within targetome for miR487b-WT was repressed at T1 by 0.8±0.3% (P=0.004) and at T7 by 0.6±0.3% (P=0.03) compared with total detectable genome expression (Online Figure VII). For the miR487b-ED targetome, however, average expression of its 1521 detectable genes was repressed at all time points after HLI, with 0.6±0.2% (P=0.04) repression at T1 and 0.7±0.2% at T3 and 0.9±0.1% at T7 (P=0.003 and P=0.0005, respectively). Averaged expression of the 232 overlapping target genes of miR487b-WT and miR487b-ED was also repressed at all time points after surgery, but approximately twice as strongly as that of the total edited targetome (T1=1.4±0.4%, T3=1.8±0.5%, and T7=1.6±0.6%, P=0.04, P=0.0005, and P=0.008 versus total, respectively).
Next, we filtered putative target genes based on a Diana binding score similar or better than that of the endogenous binding sites that were confirmed using luciferase assays (Diana binding score>0.55). A higher binding score indicates an increased likelihood or efficiency of target gene suppression by the microRNA. Expression patterns of the restricted miR487b-WT targetome followed a similar pattern to its unrestricted one, but did not differ significantly any more from total genome expression (Figure 5D). Binding score restriction of the miR487b-ED targetome on the other hand, only resulted in loss of average expression knockdown at T7, whereas increasing repression by more than 4-fold at T1 to 3.1±0.8% (P=0.02) and by 3-fold at T3 to 2.8±0.7% (P=0.009).
Causal Involvement of miR487b-ED and Not miR487b-WT in Neovascularization In Vitro
To examine the effects of miR487b-ED, we overexpressed miR487b-WT or miR487b-ED in mouse aortic segment ex vivo (Online Figure VIIIA and VIIIB). We observed an almost 3-fold increase in the number of neovessels formed in the miR487b-ED treated aortic segments compared with control, whereas miR487b-WT overexpression did not increase sprout formation (Figure 7A; P=0.002 and P=0.25, respectively). Cell type specific triple stainings of aortic segments confirmed that the sprouting neovessels are lined with endothelial cells, as described previously.22 The endothelial cells were surrounded by a highly organized layer of fibroblasts, underlining the importance of fibroblasts in neovascularization. Smooth muscle cells were less abundant and more disorganized in all 3 groups, miR487b-WT, miR487b-ED, and control (Figure 7B).
Next, we investigated the effects of miR487b-WT or miR487b-ED overexpression on cell migration, using a scratch wound healing assay in both HUAFs and HUVECs. Overexpression of the microRNAs in both HUAFs and HUVECs is shown in Online Figure VIIIC through VIIIE. Treatment with miR487b-ED significantly increased scratch wound closure in HUAFs by 50±9% and in HUVECs by 13±4% compared with control (Figure 7C and 7D; P=0.006 and P=0.04, respectively), whereas miR487b-WT treatment did not (P=0.2 and P=0.11 for HUAF and HUVECs, respectively).
In this study, we show that the seed sequence of miR487b is subject to A-to-I editing by ADARs both in mice and in human primary cells, which results in a novel microRNA, miR487b-ED. The rate of miR487b editing is increased during active neovascularization after ischemia. Additionally, we found fibrillarin-dependent 2′OMe of miR487b at the same adenosine residue that is subject to A-to-I editing in murine muscle tissue and human primary cells. Although postischemic miR487b editing and 2′OMe patterns are similar, in vitro mechanistic studies suggest an inverse correlation between expression of the ADAR proteins on the one hand and fibrillarin on the other. In vitro luciferase reporter gene assays and target validation experiments demonstrated that miR487b editing results in a complete shift in target site selection. In contrast to miR487b-WT, the putative targetome of miR487b-ED is actively repressed during effective neovascularization after ischemia and is enriched for neovascularization-associated pathways. MicroRNA overexpression experiments demonstrated that miR487b-ED causes a 2-fold increase in neovascularization compared with miR487b-WT in an ex vivo aortic ring sprouting assay and also increases wound healing in both human vascular endothelial cells and fibroblasts.
To our knowledge, we are the first to demonstrate A-to-I editing of the angiomiR miR487b. Furthermore, our study provides the first evidence of altered microRNA editing during vascular remodeling and ischemia in general. In a recent study by Nigita et al,29 the authors detected similar percentages of mature microRNA-seed editing of 7 microRNAs in small-RNAseq data sets of breast adenocarcinoma cell line MCF-7 cultured under normoxic and a time course of hypoxic conditions. For most of these microRNAs, a trend of increasing microRNA-seed editing with increasing hypoxia was observed, however, the study’s setup lacked power to establish a significant correlation. Our study used a hypothesis-driven methodology, opposed to a discovery-based method of RNAseq, which allowed us to focus on a specific region. Because of this, we were able to show statistically significant changes, not only in mature microRNA, but also in pri-microRNA editing.
The positive trend of microRNA-seed editing with increasing hypoxia found by Nigita et al29 corresponds to our findings, where we demonstrate increased mature miR487b editing in the strongly ischemic gastrocnemius muscle, which displays angiogenesis after femoral artery ligation. In contrast, we observed decreased miR487b editing in the adductor muscle. This can most likely be attributed to the fact that the adductor experiences predominantly increased shear stress instead of ischemia, causing mainly arteriogenesis instead of angiogenesis after femoral artery ligation.30,31 We observed that baseline percentages of pri-miR487b and mature miR487b editing also differed between both muscles. This is consistent with previous findings showing that levels of both total and site specific A-to-I editing, as well as pri-miRNA maturation, can vary strongly per tissue.17,32–36
A-to-I editing is directed by the enzymes ADAR1 and ADAR2. We demonstrated that pri-miR487b associates with both ADAR1 and ADAR2 and is also edited by both enzymes. ADARs seem to modify specific adenosines to inosines in short and imperfect dsRNA substrates like pri-microRNAs, however, they do not require strict sequence specificity.11,37,38 Therefore, it is not directly possible to contribute the increase in miR487b editing after ischemia to either ADAR enzyme. Neither could we observe a correlation between Adar1 or Adar2 expression and percentage of editing. In the adductor muscle, Adar1 mRNA levels were increased at day 1, and Adar2 levels were increased at day 3 after femoral artery ligation (Online Figure IX), whereas observed pri-miR487b editing dropped by 3-fold. It has been shown before that changes in ADAR expression or activity do not always correlate with frequencies of specific editing.39,40 As a result, additional regulatory mechanisms, potentially including 2′OMe, are thought to modulate editing of specific adenosine residues.
Based on previous in vitro findings that 2′OMe of an adenosine in specific mRNAs can protect it from deamination by ADARs,13–15 we hypothesized that the adenosine we showed to undergo editing in pri-miR487b, can also be 2′-O-ribose-methylated. In this study, we demonstrate that pri-miR487b associates with 2′-O-ribose-methyltransferase fibrillarin. Using reverse-transcription at low dNTP concentrations followed by quantitative-PCR, a modified 2′OMe detection method that allows for quantitative estimation of specific nucleotide 2′OMe, we showed that pri-miR487b is indeed 2′-O-ribose-methylated in murine muscle tissues and in human primary cells at the same adenosine residue that is subject to editing. However, like A-to-I editing, EMF appeared to increase at day 1 after induction of HLI in the gastrocnemius and decreased again by day 3 and 7. Furthermore, in the adductor muscle, EMF of pri-miR487b also followed the same pattern as A-to-I editing, because both decreased from day 1 to day 3 after surgery, only to increase again by day 7. These data suggest a positive correlation between A-to-I editing and 2′OMe rather than the previously reported negative correlation. However, further in vitro mechanistic studies showed that a reduced pri-miR487b 2′OMe after fibrillarin knockdown had little effect on pri-miR487b A-to-I editing, suggesting that for the specific case of pri-miR487b, these modifications are not directly interdependent. We did uncover an inverse correlation between expression of ADARs and fibrillarin in vitro; knockdown of ADAR1 or ADAR2 causes a 3-fold increase in fibrillarin expression, and knockdown of fibrillarin results in significant increases in both ADAR1 and ADAR2 expression. To our knowledge, this is a novel finding, which may help explain the previously reported inverse correlation between A-to-I editing en 2′O-ribose methylation of RNAs.13–15
The seed sequence of miR487b-ED is different from any known microRNA in humans and mice, meaning that because of A-to-I editing, an entirely new microRNA with a novel targetome is created. Through luciferase reporter gene assays and validation of a subset of target genes, we showed that editing of the seed sequence of miR487b does indeed completely shift its target site selection. In an identical context, miR487b-ED induced stronger luciferase silencing than miR487b-WT binding. Furthermore, we found more active repression of miR487b-ED’s targetome in the adductor muscle after HLI, than miR487b-WT, especially when we focus on putative targets with a strong Diana binding score. Unlike for miR487b-WT’s targetomes, both human and murine miR487b-ED targetomes were significantly enriched for multiple pathways that are associated with angiogenesis, arteriogenesis, and vascular disease. The murine targetome of miR487b-WT did contain enrichment for several neovascularization-related pathways, as was to be expected based on our previous findings,3,7 but also displayed robust enrichment for CCKR signaling, which is associated with gastrointestinal peptide hormones cholecystokinin and gastrin instead.41 Surprisingly, no enriched pathways were found within the miR487b-WT’s human targetome. The enriched pathways within human and murine miR487b-ED targetomes all contain obvious ties to neovascularization and vascular disease, including angiogenesis-related signaling and the insuline/IGF pathway in humans and the cadherin signaling pathway in mice, in addition to Wnt signaling, for which enrichment seems to be conserved across both species. Interestingly, Wnt signaling has been implicated to promote vascular remodeling and even cardiovascular regeneration.42
We performed microRNA overexpression experiments in an ex vivo aortic ring sprouting assay and an in vitro scratch wound healing assay to examine the functional effects of miR487b-WT and miR487b-ED on angiogenesis. We found that miR487b-ED treatment resulted in an increase in both neovessel formation and scratch wound healing, whereas miR487b-WT did not. These results reflect the stronger enrichment of angiogenesis-associated pathways within the miR487b-ED putative targetome. Combined with miR487b-ED’s higher target gene silencing efficacy, our findings suggest that editing of miR487b plays a central role in neovascularization.
As the expression of edited mature microRNAs is low compared with their unedited counterparts, it is difficult to estimate the clinical significance of A-to-I editing of microRNAs. However, the potential for a clinical relevance is undeniable, as illustrated by the seed sequence editing of, for example, miR455. Highly metastatic melanoma cells were shown to have reduced ADAR1 levels resulting in reduced miR455 editing, yielding increased inhibition of the CPEB1 gene, a known tumor suppressor gene.43 Furthermore, overexpression of unedited miR455 was shown to promote tumor growth and lung metastasis, whereas overexpression of edited miR455 suppressed this phenotype. With regards to cardiovascular disease and vascular remodeling, Stellos et al12 were the first to demonstrate that A-to-I editing can control gene expression in human atherosclerotic diseases. They showed that editing of the cathepsin S mRNA, which encodes a cysteine protease associated with angiogenesis and atherosclerosis, is increased in plaques from atherosclerotic patients with increased expression as a result. Interestingly, they also showed that ADAR1 controls HUVEC function and that ADAR1 expression is strongly increased in and around plaques from atherosclerotic patients. Although their focus was on the editing of long RNAs with double-stranded regions because of presence of inverted Alu repeats, this study underlines the importance of A-to-I editing in general. We provide new evidence that A-to-I editing also controls the expression of numerous other genes in angiogenesis and ischemia by regulation of microRNA targeting.
In conclusion, we demonstrate that the angiomiR miR487b undergoes A-to-I editing and 2′OMe in the seed sequence of the microRNA, creating a completely new microRNA unlike that of any known human or murine microRNA. The amount of editing is significantly increased after ischemia in mice. Editing of miR487b causes a complete shift in the targetome of the microRNA. The miR487b-ED-targetome is enriched for angiogenesis-associated pathways and functional assays demonstrated that miR487b-ED is proangiogenic, whereas miR487b-WT is not. Moreover, we found that miR487b-ED’s targetome is actively repressed during neovascularization after ischemia. Our findings suggest that editing of miR487b is plays an intricate role in the regulation of postischemic neovascularization.
We thank M. Eggenkamp and D. van den Homberg for their technical support.
Sources of Funding
This study was supported by a grant from the Dutch Heart Foundation (Dr E. Dekker Senior Postdoc, 2014T102).
In November 2017, the average time from submission to first decision for all original research papers submitted to Circulation Research was 11.99 days.
The online-only Data Supplement is available with this article at http://circres.ahajournals.org/lookup/suppl/doi:10.1161/CIRCRESAHA.117.312345/-/DC1.
- Nonstandard Abbreviations and Acronyms
- adenosine deaminase acting on RNA
- A-to-I editing
- adenosine-to-inosine editing
- complementary DNA
- estimated 2′-O-ribose-methylated fraction
- hindlimb ischemia
- human umbilical arterial fibroblast
- human umbilical smooth muscle cell
- human umbilical venous endothelial cell
- edited miR487b
- wild-type miR487b
- primary microRNAs
- Received November 3, 2017.
- Revision received December 19, 2017.
- Accepted December 27, 2017.
- © 2017 American Heart Association, Inc.
- van Oostrom MC,
- van Oostrom O,
- Quax PH,
- Verhaar MC,
- Hoefer IE
- Raval Z,
- Losordo DW
- Welten SM,
- Bastiaansen AJ,
- de Jong RC,
- de Vries MR,
- Peters EA,
- Boonstra MC,
- Sheikh SP,
- La Monica N,
- Kandimalla ER,
- Quax PH,
- Nossent AY
- Romaine SP,
- Tomaszewski M,
- Condorelli G,
- Samani NJ
- Nossent AY,
- Eskildsen TV,
- Andersen LB,
- Bie P,
- Brønnum H,
- Schneider M,
- Andersen DC,
- Welten SM,
- Jeppesen PL,
- Hamming JF,
- Hansen JL,
- Quax PH,
- Sheikh SP
- Agarwal V,
- Bell GW,
- Nam JW,
- Bartel DP
- Shafik A,
- Schumann U,
- Evers M,
- Sibbritt T,
- Preiss T
- Stellos K,
- Gatsiou A,
- Stamatelopoulos K,
- et al
- Vitali P,
- Basyuk E,
- Le Meur E,
- Bertrand E,
- Muscatelli F,
- Cavaillé J,
- Huttenhofer A
- Ekdahl Y,
- Farahani HS,
- Behm M,
- Lagergren J,
- Öhman M
- Kawahara Y,
- Zinshteyn B,
- Sethupathy P,
- Iizasa H,
- Hatzigeorgiou AG,
- Nishikura K
- Qi L,
- Song Y,
- Chan THM,
- Yang H,
- Lin CH,
- Tay DJT,
- Hong H,
- Tang SJ,
- Tan KT,
- Huang XX,
- Lin JS,
- Ng VHE,
- Maury JJP,
- Tenen DG,
- Chen L
- Kanellos I,
- Vergoulis T,
- Sacharidis D,
- Dalamagas T,
- Hatzigeorgiou AG,
- Sartzetakis S,
- Sellis TK
- Paraskevopoulou MD,
- Georgakilas G,
- Kostoulas N,
- Vlachos IS,
- Vergoulis T,
- Reczko M,
- Filippidis C,
- Dalamagas T,
- Hatzigeorgiou AG
- Nossent AY,
- Bastiaansen AJ,
- Peters EA,
- de Vries MR,
- Aref Z,
- Welten SM,
- de Jager SC,
- van der Pouw Kraan TC,
- Quax PH
- Hellingman AA,
- Bastiaansen AJ,
- de Vries MR,
- Seghers L,
- Lijkwan MA,
- Löwik CW,
- Hamming JF,
- Quax PH
- Vitsios DM,
- Davis MP,
- van Dongen S,
- Enright AJ
- Liu Y,
- Emeson RB,
- Samuel CE
- Tripathi S,
- Flobak Å,
- Chawla K,
- Baudot A,
- Bruland T,
- Thommesen L,
- Kuiper M,
- Lægreid A
- Hermans KC,
- Blankesteijn WM
Novelty and Significance
What Is Known?
Neovascularization is crucial for blood flow recovery after cardiovascular events and can be regulated by endogenous microRNAs, which each regulate the expression of multiple neovascularization-related genes.
MicroRNA-487b (miR487b) has very few putative target genes compared with other vasoactive microRNAs, but is still able to regulate blood flow recovery after hindlimb ischemia, as well as arterial wall integrity during chronic hypertension in vivo.
MicroRNAs can be subject to adenosine-to-inosine (A-to-I) editing by adenosine deaminase acting on RNA enzymes. Should A-to-I editing occur in the seed sequence of the microRNA, this could alter the microRNA’s target site recognition and hence its target gene selection.
What New Information Does This Article Contribute?
During ischemia a fraction of miR487b is edited by adenosine deaminase acting on RNA enzymes in its seed sequence, and the rate of editing is increased upon blood flow recovery after ischemia in vivo.
The same adenosine residue that is subject to editing, can also be 2′-O-ribose-methylated by fibrillarin. The methylation rate also increased during ischemia in vivo even though adenosine deaminase acting on RNA enzymes and fibrillarin were found to negatively influence each other’s expression in vitro.
The edited miR487b (miR487b-ED) binds to and represses a completely different set of genes than the unedited miR487b (miR487b-WT). Because of this switch in targetome, miR487b-ED now promotes angiogenesis and neovascularization, whereas miR487b-WT does not.
The 14q32 microRNA gene cluster has been shown to regulate the complex, multifactorial process of neovascularization by modulating expression of multiple, rather than a single, target genes. MiR487b is also able to regulate blood flow recovery after ischemia, even though, compared with other vasoactive 14q32 microRNAs, miR487b has a strikingly small putative targetome. We hypothesized that the miR487b targetome is expanded or altered during neovascularization through A-to-I editing of the seed sequence. We demonstrated that a fraction of the miR487b transcript, pri-miR487b, is subject to A-to-I editing by adenosine deaminase acting on RNA enzymes in both mice and humans. The edited pri-miR487b is processed into a mature proangiogenic microRNA with a novel, unique seed sequence, miR487b-ED. The rate of editing is increased during blood flow recovery after ischemia in vivo in mice. Our findings show that A-to-I editing can actively change microRNA-functionality during vascular remodeling. As many other microRNAs may be subject to editing as well, editing can have major implications for future microRNA-research and even for future microRNA-based therapeutics.