The VEGF IRESes are differentially susceptible to translation inhibition by miR-16

Abstract

Experiments with EMCV (Encephalomyocarditis virus) internal ribosome entry sites (IRESes) have shown that microRNAs (miRs) are unable to inhibit IRES driven translation. However, it is accepted that miRs can inhibit translation through multiple mechanisms, only some of which require interaction with the 5′ cap structure. In this report, we first validate the targeting of miR-16 to a predicted binding site in the VEGF 3′UTR. We developed a series of experiments to ascertain whether or not miR-16 can inhibit translation of transcripts driven by either of the VEGF IRESes. Our results indicate that cellular IRESes can be classified as both sensitive and insensitive to miR control. While VEGF IRES-A activity was not altered by miR-16 targeting to the 3′UTR, IRES-B was susceptible to miR-16 inhibition. Taken together with previous results that show that IRES-B selectively translates the CUG initiated VEGF-121 isoform, we can conclude that the existence of two differentially susceptible IRESes in the VEGF 5′UTR leads to even more complex regulatory control of VEGF isoform production. This study demonstrates for the first time the inhibition of cellular IRES driven translation by a miR.

INTRODUCTION

microRNAs (miRs) are short single stranded regulatory RNAs (∼21–23 nucleotides [nt]) that regulate various cellular and molecular processes. They are transcribed from genomic loci by RNA polymerase II and are exported from the nucleus as short hairpin precursors that are cleaved to generate their mature form. Through associations with proteins from the RNA-induced silencing complex (RISC), such as Argonaute (Ago), miRs can both promote mRNA 5′ to 3′ degradation and deadenylation, as well as inhibit protein translation through complex and controversial mechanisms (Filipowicz et al. 2008). Translation inhibition by miRs can involve regulation of both translation initiation (Humphreys et al. 2005; Pillai et al. 2005; Mathonnet et al. 2007; Wang et al. 2008) and steps downstream of initiation (Lytle et al. 2007). miRs have also been proposed to promote proteolysis of the nascent peptide (Nottrott et al. 2006) and ribosome drop off during elongation (Petersen et al. 2006). Translation inhibition by miRs can be abrogated through replacement of the 5′m7GpppG cap structure with a 5′ApppG cap (which makes a transcript incapable of binding eIF4E) and increased concentrations of eIF4F (which promotes initiation through interaction with the 5′ cap). These two findings suggest that miRs should be ineffective against IRES driven translation. In fact, results obtained with viral IRESes are contradictory; the EMCV IRES being insensitive and the HCV and CrPV IRESes, sensitive (Petersen et al. 2006; Kiriakidou et al. 2007; Mathonnet et al. 2007). It remains to be seen whether or not cellular transcripts, translationally controlled by IRESes, can be inhibited by microRNAs. While some viral IRESes may have evolved to avoid any negative regulation by cellular factors, it is likely that cellular IRESes, which are normally found in genes encoding critical growth regulatory genes, are regulated by the miR system in order to further control gene expression. While several genes that are regulated through IRES-dependent mechanisms have been validated as miR targets (myc, Bcl-2, CAT-1, connexin-43, etc.), nothing is known concerning the effects of miRs on translation driven by these IRES elements.

Vascular endothelial growth factor A (VEGFA or VEGF) is an essential growth and survival factor for endothelial cells. It plays a major role in physiological and pathological angiogenesis through its ability to stimulate growth of new blood vessels from nearby capillaries (Ferrara 2005). Through alternative splicing, the highly conserved VEGF gene can produce various protein isoforms, with the three principal forms consisting of 121, 165, and 189 amino acids. VEGF is translated from two start codons, each of which is regulated by an independent IRES (Huez et al. 2001). The classical AUG codon is regulated by IRES-A and the CUG codon, which is 539 bases upstream, is regulated by IRES-B (Fig. 1A; Huez et al. 1998, 2001). VEGF has also been reported to posses an internal promoter (Pages and Pouyssegur 2005). The importance of VEGF expression regulation is further highlighted by the fact that either a 50% reduction, through a heterozygous null mutation, as well as a two- and three-fold increase, though deletion of the 3′UTR in a single allele, results in embryonic lethality (Carmeliet et al. 1996; Ferrara et al. 1996; Miquerol et al. 2000). In this report, we validate a predicted miR-16 target site in the full length VEGF 3′UTR and show that miR-16 negatively regulates VEGF translation through this site. In addition, we demonstrate that translation initiated from the VEGF IRES-B is inhibited by miR-16, indicating that at least some cellular IRESes are sensitive to miR regulation.

FIGURE 1.

miR-16 regulates expression of endogenous VEGF (A) VEGF mRNA is schematized with its alternative splicing and its two IRESs B and A. Predicted miR-16 responsive element (MRE) is represented within VEGF 3′UTR with its position. (B) Northern blot analysis to detect mature human miR-16 on untransfected HeLa cells (lane 1), transfected cells with 100nM of LNA16 (lane 2), and with 100 nM of exogenous miR-16. (C) qRT-PCR analysis to detect mature miR-16 realized in the same conditions as in (B). (D) Elisa assays were performed to determine miR-16 effect on VEGF secretion in cell media. Measured amounts were normalized to mock experiment. Bars represent standard deviation and asterisks depict statistically significant differences compared to mock (**: p < 0.0005; Student's t-test). (E) VEGF mRNA was quantified by qRT-PCR (normalization to 18S rRNA and to mock).

RESULTS AND DISCUSSION

We were intrigued by the potential regulation of VEGF by miR-16 (Accession No. MI0000070) through a microRNA responsive element (MRE) located ∼260 bases downstream of the translation stop in the VEGF 3′UTR (Fig. 1A), which has been predicted by multiple computational programs. To validate the regulation, we needed a cell line that expresses VEGF, together with miR-16, to address the effect of variation in miR-16 on VEGF translation. As shown in Figure 1B, we detected endogenous and transfected miR-16 in HeLa cell extracts by Northern blotting and found that transfection with a locked nucleic acid antisense to miR-16 (LNA16) drastically reduced the level of endogenous miR-16. These results, validated by quantitative real-time polymerase chain reaction (qRT-PCR), are shown in Figure 1C. ELISA was used to determine if the level of miR-16 affected the level of VEGF. While transfection of miR-16 leads to a 50% reduction in VEGF secretion (p < 0.0005), transfection of LNA16 led to a more than 60% increase in VEGF production (Fig. 1D). To confirm that miR-16 acts on VEGF translation and not transcription or mRNA stability, we checked VEGF mRNA levels by qRT-PCR (Fig. 1E). Neither miR-16, LNA16, nor a scrambled control miR had an effect on VEGF mRNA levels that correlates with its effect on protein levels. Thus, miR-16 levels inversely correlate with VEGF protein levels, implicating that VEGF is translationally controlled by miR-16. In addition, the 50% residual VEGF expression in miR-16 transfected cells (Fig. 1D) suggests that VEGF translation can still take place even in the presence of high concentrations of miR-16.

To confirm that miR-16 regulates VEGF translation through the predicted target site in the 3′UTR (Fig. 1A), we created Renilla luciferase (RL) reporter constructs with either the full length human VEGF 3′UTR or one of two mutant constructs where the target site was either deleted or mutated within the seed region (Fig. 2A). Reporter activity, normalized to the activity of the cotransfected Firefly luciferase (FL), was quantified in HeLa cell extracts. The ratio (RL/FL) obtained with the complete 3′UTR was 6.9-fold lower in cells transfected with miR-16, compared to a scrambled miR control (p < 0.05). Conversely, transfection of LNA16 led to a 4.2-fold increase in the normalized RL activity. Both deletion of the miR-16 target site, and mutation of the seed region, rendered the reporter completely insensitive to miR-16 and LNA16 transfections (Fig. 2B). While these experiments demonstrate that the miR-16 binding site within the 3′UTR of VEGF is required for translation inhibition by miR-16, they do not address whether or not this regulation is functional in the context of the IRES-containing 5′UTR of VEGF.

FIGURE 2.

miR-16 directly interacts with a site in the VEGF 3′UTR. (A) Sequences of the miR-16 responsive element (MRE) in the wild-type and mutant (deleted/mutated) reporter constructs. (B) All constructs were co-transfected with synthetic miR-16 or its antisense LNA. Renilla luciferase activities are normalized to the cotransfected firefly luciferase (FL) activities. Data are presented relative to controls transfected with scrambled miR (*: p < 0.05; Student's t-test).

VEGF translation has been shown to be driven by two independent IRESes in cell lines (Akiri et al. 1998; Huez et al. 1998; Stein et al. 1998) or in transgenic animal models (Bornes et al. 2007). We wanted to evaluate the negative effect of miR-16 on the VEGF 3′UTR in the context of the VEGF 5′UTR. To this end we generated reporter constructs driven by either VEGF IRES-A, IRES-B, or the EMCV IRES. (Fig. 3A). Consistent with results obtained by Mathonnet and Pillai (Pillai et al. 2005; Mathonnet et al. 2007), no miR-16 dependent changes were observed in constructs containing the EMCV IRES (Fig. 3B). While the VEGF IRES-A provided results similar to the EMCV IRES (Fig. 3C), reporters containing IRES-B showed a 6.2-fold decrease in activity when co-transfected with miR-16. Northern blots were performed to confirm the integrity of all reporter transcripts (data not shown). This decrease was dependent on the presence of the miR-16 site and co-transfection with LNA16 showed the expected opposite effect and increased RL activity 3.5-fold (Fig. 3D). These findings were surprising since IRES-A and IRES-B have been previously shown to have similar efficiencies in several systems, both in vitro and in vivo (Bornes et al. 2007; Bastide et al. 2008). It is likely that miR sensitivity could be a useful test in the classification of cellular IRESes but further studies are required before any general conclusions can be drawn. Nevertheless, the fact that IRES-B, and not IRES-A, is susceptible to miR-16 control suggests that there are distinct differences in the mechanisms employed by these two IRESes, with IRES-A sharing properties with the EMCV IRES in terms of miR susceptibility.

FIGURE 3.

VEGF IRESes are differentially responsive to miR-16. (A) Schemes of RL-VEGF reporter bearing wild type VEGF 3′UTR fused to the RL coding region and even IRES-A, IRES-B, or EMCV-IRES upon the Renilla coding region. A-RL-DEL, B-RL-DEL, and EMCV-RL-DEL correspond to the same constructs deleted from their miR-16 site. (B) Reporter activity of EMCV-RL-VEGF and EMCV-RL-DEL, (C) Reporter activity of A-RL-VEGF and A-RL-DEL, (D) Reporter activity of B-RL-VEGF and B-RL-DEL. (B–D) Renilla luciferase was normalized to activity of the coexpressed firefly luciferase (FL), *: p < 0.001; Student's t-test.

In order to test if miR-16 is able to control IRES-B driven translation independent of the proximity of the IRES to the 5′ cap structure, we generated bicistronic reporter constructs using the full length VEGF 3′UTR (Fig. 4). Again, our results demonstrate that miR-16 efficiently inhibited IRES-B and had no effect on either IRES-A or the EMCV IRES (Fig. 4A). The similarity in the data obtained with both monocistronic and bicistronic experiments suggest that negative regulation by miR-16 is not sensitive to the vicinity of the IRES to the 5′ cap site. These results confirm that IRES-B driven VEGF translation, which leads to initiation at the CUG codon and production of the diffusible VEGF-121 isoform, can be negatively regulated by miR-16. They also suggest that our ELISA assays, which monitor endogenous VEGF production, are indicative of the ability of miR-16 to inhibit IRES-B driven translation in actual VEGF transcripts.

FIGURE 4.

(A,B) miR-16 regulates production of endogenous VEGF through IRES-B Western blot analysis of endogenous VEGF after transfection of miR-16 and its antagonist LNA16 under hypoxia (1%O₂). Relative intensity resulting from quantification of each band normalized to intensity of mock transfected.

We have previously demonstrated that VEGF-121 is exclusively expressed through initiation events utilizing IRES-B while VEGF-189 and 165 are translated via both IRESes (Fig. 1A; Bornes et al. 2004; Bastide et al. 2008). In light of our current findings, VEGF-121 should be the isoform that is the most sensitive to miR-16. We sought to confirm this hypothesis through Western blotting with hypoxia stimulated HeLa cell extracts, transfected with either miR-16 or LNA16. Indeed, in these transfections, miR-16 preferentially alters the level of the VEGF-121 isoform (Fig. 4B, blot and quantification panels). Measurement of the relative intensities of each band showed that while miR-16 decreased VEGF-165 and VEGF-189 by about 30%, VEGF-121 levels were 60% lower in miR-16 transfected cells and 60% higher when miR-16 is absent due to transfection of LNA16. Taken together with our previous ELISA data, we can now speculate that levels of VEGF that persist even in high concentrations of miR-16 are likely generated through IRES-A driven, miR-insensitive translation from the AUG codon. In addition, it is likely that the miR-16 dependent decrease in VEGF levels is mediated through IRES-B driven translation at the CUG codon.

In conclusion, our current study demonstrates for the first time that a microRNA can inhibit translation driven by a cellular IRES. While the miR-16 site in the VEGF 3′UTR can act in both cap-dependent and IRES-B-driven contexts, it is not able to control translation in the IRES-A context (Figs. 3, 4). The similarity between IRES-A and the EMCV IRES suggest that miR-sensitivity will prove to be an important means to classify and validate cellular IRESes. Such classification should lead to our ability to distinguish various different mechanisms of translational control that presently fall under the general term “cellular IRES.” Furthermore, our results are consistent with several reports demonstrating that miRs can repress translation at late stages of the initiation process, which are independent of 5′ cap recognition (Nottrott et al. 2006; Petersen et al. 2006; Nissan and Parker 2008). Once again, VEGF presents itself as a useful model to study complex systems of gene expression regulatory control. It will be important to see if the miR-16/VEGF regulation is the exception or the rule in terms of miR control of cellular IRES driven translation.

MATERIAL AND METHODS

Plasmid constructions

To generate the pRL-VEGF reporter, the VEGF 3′UTR was PCR amplified from HeLa genomic DNA using the RefSeq sequence NM_001025366, positions 1748–3625. This product was cloned into the XbaI site downstream of Renilla luciferase in the vector pRLCMV_AF025843 (Promega).

Reporter constructs are referred to as pA-RL-VEGF and pB-RL-VEGF. The EMCV IRES was PCR amplified and cloned into pA-RL-VEGF, replacing IRES-A, through digestion with BamHI and NcoI.

Deletion/mutation of the miR-16 site in the 3′UTR of VEGF was performed using site directed mutagenesis (QuickChange, Stratagene).

Bicistronic constructs were obtained by replacing the Nhe1-Apa1 (Renilla luciferase) fragment of the pRL-VEGF by a Xba1-Apa1 fragment of pCR31L, pCR8L, and pCREL that carries the two luciferases genes separated by an IRES, respectively IRES-A, IRES-B, and EMCV-IRES (Huez et al. 1998; Bornes et al. 2004).

PCR primer sequences are available upon request.

Human VEGF ELISA

Ninety-six-well plates coated with anti-human VEGF monoclonal antibody were purchased from R&D Systems following the manufacturer's guidelines. HeLa culture supernatants were collected, and VEGF was bound by the immobilized antibody. After extensive washing, a peroxidase-linked polyclonal antibody recognizing VEGF was added to the wells; plates were again washed before addition of a substrate solution (TMB Substrate Reagent Set BD Biosciences). Plates were incubated for 20 min at room temperature and absorbance was measured at 450–570 nm. Each experiment, performed in triplicate, was repeated three times.

Real-time PCR

PCR amplifications were performed in a mixture containing 50 ng of cDNA and the Taqman universal PCR master mix (Applied Biosystems). The reaction assay was performed on Gene Amp 7900HT sequence detection system and results were treated with the SDS 2.2.2 software (Applied Biosystems). In each run of the assay, the levels of VEGF mRNA and Ribosomal 18S RNA were analyzed. Levels of VEGF mRNA are expressed relative to the concentration of 18S rRNA. 18S was selected as an endogenous RNA control to normalize for differences in the amount of total RNA. GenBank Accession number of the TaqMan probe used for VEGF cDNA amplification is: NM_001025366. The TaqMan Ribosomal RNA Control Reagents were used to detect the 18S ribosomal RNA (rRNA) gene. The integrity of the messenger was checked by performing three independent real-time PCRs using three sets of primers covering the VEGF mRNA from the exon 1 to the 3′UTR (data not shown).

miR Northern blotting

Northern blots were performed by using 10 μg of total RNA on denaturing PAGE. After blotting on Amersham Hybond-N+, LNA modified oligonucleotides complementary to miR-16, 5′-end-labeled using T4 polynucleotide kinase and γ-³²P ATP, were used as a probe.

miR qRT-PCR

Total RNA was isolated by Trizol. Purified RNA was retrotransribed using The TaqMan MicroRNA RT Kit in association with the probe contained in the TaqMan MicroRNA Assay (HSA-MIR-16-1). This kit provides a probe that permits miR-16-1 cDNAs amplification with the Taqman universal PCR master mix (Applied Biosystems). Assays were performed on Gene Amp 7900HT sequence detection system and results were treated with the SDS 2.2.2 software (Applied Biosystems).

Transfections

HeLa cells were transfected using Lipofectamine 2000 (Invitrogen) as suggested by the manufacturer. RL-VEGF reporter plasmid was cotransfected with a LucF expressing plasmid serving as transfection control. When indicated, these two constructs were transfected with 100nM of Pre-miR-16-1 miR Precursor Molecule (Ambion, Applied Biosystems), with 100 nM of LNA anti-miR-16 (CGCCAATATTTACGTGCTGCTA) (Sigma-Aldrich Chimie) or with Pre-miR miR Precursor Molecules—Negative Control #1 (Ambion, Applied Biosystems). All transfection experiments were performed in duplicates and repeated a minimum of three times.

Western blotting

Western blots were performed after 24 h of transfection of HeLa cells under hypoxia (hypoxic conditions: 37°C with 5% CO2, 94% N2, and 1% O2 in a hypoxic incubator [Binder GmbH]). 30 μg of protein sample were heated and separated by 12.5% polyacrylamide gel (PAGE) and transferred onto a nitrocellulose membrane. VEGF proteins were detected using the VEGF (A-20): sc-152 antibody (Santa Cruz Biotechnology) (dilution 1:200). The protein signal was normalized using an anti-β-actin monoclonal antibody (AC-15, Sigma-Aldrich Chimie) (1: 10,000). Signals were detected using a chemiluminescence ECL kit (Amersham Biosciences).

Luciferase assay

Quantification of RL and FL activities was achieved with a luminometer (Centro LB960, Berthold) using the Dual-Luciferase Reporter Assay (Promega), according to the manufacturer's instructions. All transfections and measurements were performed in triplicate and repeated at least three times.