|Home | About | Journals | Submit | Contact Us | Français|
MicroRNAs (miRNAs) are an abundant class of 20–23-nt long regulators of gene expression. The study of miRNA function in mice and potential therapeutic approaches largely depend on modified oligonucleotides. We recently demonstrated silencing miRNA function in mice using chemically modified and cholesterol-conjugated RNAs termed ‘antagomirs’. Here, we further characterize the properties and function of antagomirs in mice. We demonstrate that antagomirs harbor optimized phosphorothioate modifications, require >19-nt length for highest efficiency and can discriminate between single nucleotide mismatches of the targeted miRNA. Degradation of different chemically protected miRNA/antagomir duplexes in mouse livers and localization of antagomirs in a cytosolic compartment that is distinct from processing (P)-bodies indicates a degradation mechanism independent of the RNA interference (RNAi) pathway. Finally, we show that antagomirs, although incapable of silencing miRNAs in the central nervous system (CNS) when injected systemically, efficiently target miRNAs when injected locally into the mouse cortex. Our data further validate the effectiveness of antagomirs in vivo and should facilitate future studies to silence miRNAs for functional analysis and in clinically relevant settings.
MiRNAs are an abundant class of non-coding RNA ranging from 20 to 23nt of length that are post-transcriptional regulators of gene expression. MiRNAs have been mainly associated with developmental processes in metazoa such as Caenorhabditis elegans or Drosophila melanogaster (1). However, evidence also suggests a role for miRNAs in a wide range of functions in mammals, including insulin secretion, heart, skeletal muscle and brain development (2,3). Furthermore, miRNAs have been implicated in diseases such as cancer (4) and hepatitis C (5), which make them attractive new drug targets. In contrast to the widely used RNAi technology using small interfering RNA (siRNA) duplexes, strategies to inhibit miRNAs have been less well investigated. Reverse-complement 2′-O-methyl sugar modified RNA is frequently being used to block miRNA function in cell-based systems (6). The use of miRNA inhibitors in mice, however, remains challenging. Recently, two independent approaches to silence miRNAs in vivo have been reported. Our group demonstrated silencing of miRNAs in mice based on RNA analogs termed ‘antagomirs’ (7). Antagomirs are RNA-like oligonucleotides that harbor various modifications for RNAse protection and pharmacologic properties such as enhanced tissue and cellular uptake. They differ from normal RNA by complete 2′-O-methylation of sugar, phosphorothioate backbone and a cholesterol-moiety at 3′-end. Antagomirs efficiently silence miRNAs in most tissues after three injections at 80mg/kg bodyweight (bw) on consecutive days. Esau et al. recently demonstrated a different strategy to silence miRNA(miR)-122 in mouse liver (8). They inhibited miR-122 following a four-week treatment protocol with two injections per week of 12.5–75mg/kg bw of an antisense oligonucleotide (ASO) harboring a complete 2′-O-methoxyethyl and phosphorothioate modification.
Ultimately the efficient use of oligonucleotides to target miRNAs for functional studies and therapeutic use will be dependent on characterizations based on miRNA targets in vivo. Moreover, more efficient strategies could be developed through a more detailed understanding of the mechanisms of miRNA silencing. In this study, we begin to address these questions using regulation of endogenous miR-122 targets and fluorophore-tagged antagomirs.
The single-stranded RNAs and modified RNA analogs (antagomirs) were synthesized as previously described (7). The oligonucleotides used in this study are listed in Table 1 and a schematic representation of the chemical modifications is shown in Figure 1. Quasar-570 (Q570) phosphoramidite from Biosearch Technologies was coupled to the 5′-terminal under standard solid-phase phosphoramidite synthesis conditions to obtain fluorophore-tagged antagomir 122 and mm-antagomir-122. Extended 15-min coupling of quasar-570 phosphoramidite was carried out at a concentration of 0.1M in CH3CN in the presence of 5-(ethylthio)-1H-tetrazole activator followed by standard capping, oxidation and deprotection afforded labeled oligonucleotides. The Q570 conjugated sequences were HPLC purified on an in-house packed RPC-Source15 reverse-phase column. The buffers were 20mM NaOAc in 10% CH3CN (buffer A) and 20mM NaOAc in 70% CH3CN (buffer B). Fractions containing full-length oligonucleotides were pooled and desalted. Analytical HPLC, CGE and ES LC-MS established the integrity of the compounds. For duplex generation, equimolar amounts of miR-122 and antagomir were heated in 1×PBS at 95°C for 5min and slowly cooled to room temperature.
All animals were maintained in a C57Bl/6J background on a 12-h light–dark cycle in a pathogen-free animal facility at Rockefeller University. Six-week-old mice received on three consecutive days tail vein injections of saline or different RNAs in 0.2ml per injection at normal pressure. Liver tissue was harvested 24h after the last injection or as otherwise indicated.
For cerebral injections, mice were anaesthetized and antagomir-16 was injected into the left frontal cortex (~800ng). Injections of equal volume PBS into the contralateral area of the right hemisphere served as controls. After 72h mice were sacrificed, blood was removed through systemic perfusion of the left ventricle with PBS and a ~0.4cm3 area surrounding the injection site excised from the cortex.
Total RNA was isolated using the Trizol protocol with ethanol precipitation and northern blot analysis was performed using formamide-containing PAGE as previously described (7).
Extraction of total RNA, synthesis of cDNA, and PCR were performed as previously described (7). Primer sequences and annealing temperatures are shown in supplementary materials.
Mice were perfused with ice-cold PBS through the left ventricle and ~100mg liver tissue excised. Cells were fractionated on continuous sucrose density gradients from 0.4 to 2M as described earlier (9). Fractions were separated on 14% PAGE containing 8M urea and 20% formamide. Concentration of 5′-Q570-labeled antagomir in liver fractions was measured using an fmax spectrophotometer from Molecular devices.
For immunofluorescence of P-bodies and antagomirs, 1mg of Q570-labeled antagomirs were injected in 0.2ml and normal pressure on Day 1, followed by injection of 50μg of a Gfp-GW182-expressing DNA-plasmid (10; kindly provided by Dr E.K. Chan, Gainesville, Florida, USA) in 2ml PBS and high pressure on Day 2. On Day 3 mice were anesthetized and perfused through the left ventricle with 2% paraformaldehyde. Livers were incubated overnight at 4°C in 4% paraformaldehyde, followed by 16-h incubation in 30% sucrose/PBS (v/v). Frozen sections (7μm) were mounted on glass slides and analyzed using a laser-scanning microscope.
Results are given as means. Statistical analysis was performed with Student's t-test, and the null hypothesis was rejected at the 0.05 level.
The antagomir-122 chemistry includes six phosphorothioate backbone modifications. Two phosphorothioates are located at the 5′-end and four at the 3′-end. To test whether the number of phosphorothioates is critical for the ability of antagomir-122 to silence miR-122, we compared four different antagomir-122 molecules that only differ in the number of phosphorothioate modification (P=S). Injection of antagomir-122at 3×20mg/kg bw with no P=S into mice did not influence miR-122 levels in the liver (Figure 2A). The addition of a single P=S did not alter miR-122 levels either as compared to the antagomir-122 with six P=S (Figure 2B). Intriguingly, complete P=S modification of the antagomir-122 did not further increase the effect on miR-122 levels (Figure 2B). These results demonstrate that phosphorothioates are important for antagomir-122 function, however, complete replacement of P=O by P=S decreases its efficiency. The result can be explained by the enhanced stability provided by the terminal P=S-linked antagomir and the reduced thermodynamic stability of the fully modified P=S antagomir duplex with targeted miRNA.
We also determined the optimal nucleotide length of antagomirs for silencing endogenous miR-122 levels in vivo. Addition of two nucleotides or shortening of antagomir-122 by two nucleotides did not significantly alter its efficiency (Figure 2C). However, silencing of miR-122 was abolished at 3×20mg/kg bw when the length of antagomir-122 was reduced to 19nt. Together, these results further elicit the requirement of optimum number of phosphorothioate modifications and minimum length of antagomirs for their biological function in vivo. The tendency for improved activity of 25-mer antagomir can be explained on the basis of improved thermodynamic binding affinity of the 25 mer, which should also have higher biostability from exonucleases for the core 23 mer.
To investigate the optimal dose- and time-dependency of antagomir-122, we analyzed miR-122 levels as well as mRNA levels of endogenous miR-122 targets. Northern blots show that optimal reduction of miR-122 levels is achieved at antagomir concentrations between 3×40 and 3×80mg/kg bw (Figure 3A). The effect of 3×40mg/kg on miR-122 levels is stable for at least 8 days (=5 days after the last injection) (Figure 3B). The expression of miR-122 targets correlated with the reduction of miR-122 levels in northern blots and showed highest upregulation at antagomir concentrations between 3×40 and 3×80mg/kg bw (Figure 3A). All targets analyzed showed stable upregulation for at least 5 days after the last injection (Figure 3B). Together, the results demonstrate that robust and lasting upregulation of miR-122 targets is achieved at antagomir concentrations between 3×40 and 3×80mg/kg bw as early as 24h after the last injection.
We tested the impact of different mismatch numbers in the antagomir sequence on miR-122 levels and miR-122 targets. Four mismatches, two mismatches or a single mismatch at position 19 was sufficient to prevent downregulation of miR-122 and upregulation of three different miR-122 targets (AldoA, Tmed3 and Hfe2) as measured by RT-PCR (Figure 4A). However, single nucleotide mismatches at two different positions (nt1 or nt11) did not prevent downregulation of miR-122 levels or target regulation (Figure 4B). These data demonstrate that antagomirs can exhibit high sequence specificity. However, discrimination at the single nucleotide level is position-dependent and has to be tested for each microRNA sequence that is being targeted.
We asked whether antagomir-mediated miRNA silencing involves degradation of the miRNA by analyzing the ability of antagomirs to induce degradation of preformed duplexes. We synthesized duplexes of antagomir-122 and a synthetic miR-122 that harbored modifications to protect against different RNAse activities. MiR-122 was either protected at the outside (‘out’) with a phosphorothioate modification to protect against exonucleases or at two consecutive internal positions (nt13 and nt14 of miR-122; ‘in’) using 2′-O-methyl sugar modification (Figure 5A) to protect against endonuclease activity. Injection of both types of duplexes led to the appearance of degradation products of the synthetic miR-122 (Figure 5B). These degradation products did not appear when the duplexes were directly analyzed on the polyacrylamide gels or after they had been subjected to the RNA isolation protocol (Figure 5B, ‘spiked control’). Furthermore, the spiked control data demonstrate that the synthetic miRNA was not lost during the isolation procedure. These data show that both types of stabilized miR-122 that were bound to the antagomir-122 had been degraded. Accordingly, both types of protected duplexes upregulated three different miR-122 targets (Figure 5B). We conclude that antagomir-mediated-silencing of miRNAs involves target degradation. However, this process does not depend on exonuclease activity and seems to differ from the RNAi pathway.
To localize antagomirs and miRNA in subcellular compartments, we used 5′-Q570-labeled antagomir-122 or Q570-labeled mm-antagomir-122. Q570-labeling did not impair antagomir-122 function, although silencing efficiency was slightly decreased (data not shown). First, we fractionated liver homogenates from mice that had been treated with Q570-mm-antagomir-122 by ultracentrifugation on sucrose gradients. Northern blot analysis of various fractions showed a single peak of tRNA in fraction 2 (Figure 6A). In contrast, miR-122 and mm-antagomir-122 localized both to two peaks, fraction 2/3 and fraction 7/8 (Figure 6B). We next investigated whether co-localization of antagomirs and miRNAs involves the P-body compartment. To visualize P-bodies in mouse liver in vivo we used a GFP-expressing construct (GFP-GW182) that has previously been demonstrated to act as a marker for the P-body compartment (10). We overexpressed GFP-GW182 in liver using high-pressure high-volume tail vein injections and analyzed GFP- and Q570-fluorescence with laser-scanning microscopy. Q570-labeled antagomirs were exclusively localized in the cytosol and distinct from P-bodies (Figure 6C). There was no overlap between these two compartments. Together, we conclude that antagomirs and miRNA interact in a cytoplasmic compartment upstream of P-bodies.
We previously described that systemic injections of antagomir-16 into tail veins of mice do not influence the steady-state levels of miR-16 in the brain. MiR-16 is ubiquitously expressed including neurons. We tested whether antagomir-16 can decrease miRNA levels in the brain when injected directly into the cortex of anesthetized mice. PBS injections into the contra lateral side of the same animal served as controls. A single injection of ~0.8μg of antagomir led to a robust decrease in miR-16 expression at 3 days after the injection (Figure 7). These results demonstrate that direct application of antagomirs can efficiently target miRNAs in tissues that cannot be reached through tail vein injections.
In this study, we characterized the inhibition of miRNAs with antagomirs in vivo. A detailed understanding of this process is needed to investigate miRNA function in mice and exploit their therapeutic potential. Our study provides a unique platform since its major read-out is based on the dose-and time-dependent regulation of several endogenous and validated targets of miR-122.
Specificity of drug-like oligonucleotides is important to minimize off-target effects and to discriminate between related miRNAs that sometimes differ by only a single nucleotide. Previous studies have demonstrated that locked nucleic acid (LNA) are sensitive to single mismatch in in-situ staining protocols of zebrafish embryos (11). Tissue-culture-based luciferase assays indicated that 2′-O-methoxyethyl oligonucleotides might have a similar specificity (12). In line with these data we show that the antagomir chemistry enables discrimination of a single nucleotide. Intriguingly, this effect depends on the position of the mismatch within the antagomir sequence. Nucleotide exchanges at the very 5′-end of the antagomir or in the center did not prevent downregulation of miR-122 levels in northern blots and upregulation of miR-122 targets. This observation is in agreement with the work published by Davis et al., who reported a similar result in luciferase-based assays (12). Asymmetry of a single nucleotide mismatch may therefore be more detrimental for targeting miRNAs than symmetric changes. These data are important for the design of antagomirs that target specific members of miRNA families or when off-target effects have to be considered.
The mechanism of oligonucleotide-mediated miRNA silencing is still unknown. Previous data from our group and others suggested that this process involves degradation of the miRNA in vitro (12) and in vivo (7,8). Northern blots of tissue samples treated with antagomirs fail to detect fragments of the targeted miRNA. This could be explained by cellular RNase activity that readily degrades them. We have previously demonstrated that increasing the cellular amount of endogenous miRNA by introducing duplexes of miRNA/antagomir leads to detectable degradation products (7). In this study, we used this approach to ask whether antagomir-mediated silencing of miRNA involves a RNA-induced silencing complex (RISC)-dependent cleavage mechanism. In the RNAi pathway, the siRNA duplex of passenger strand and guide strand is integrated into the RISC complex and the argonaute-2 (Ago2) protein subsequently cleaves the passenger strand across from the guide strand's phosphate bond between position 10 and 11 (13,14). This cleavage was inhibited by a single 2′-O-methylation of the passenger strand corresponding to nucleotide 11 of the guide strand (13). Hypothetically, antagomirs could cleave miRNAs within RISC with the antagomir acting as the guide strand. To test this, we injected miRNA/antagomir-duplexes into mice that harbored a 2′-O-methyl endonuclease protection of the microRNA corresponding to nucleotide 10 and 11 of the antagomir. However, endonuclease protection between nucleotides 10 and 11 did not prevent the degradation of the miRNA as demonstrated by abundant miRNA fragments in northern analysis, nor did it prevent the upregulation of miR-122 targets. Thus Ago2-mediated cleavage is unlikely to mediate this process. Similar results were obtained when the miRNA was protected at the outside positions using phosphorothioates, indicating that the miRNA targeting does not dependent on exonuclease activity either. However, the fact that miRNA/antagomir-duplexes regulate miRNA targets suggests antagomir recycling. The appearance of miRNA fragments of decreased length suggests that degradation is involved in this recycling process. Further elucidation of the mechanism of degradation will be important for optimizing miRNA silencing.
To address the subcellular compartment where interaction of miRNA and antagomir occurs we engineered flurophore-labeled antagomirs. Flurophore labeling of siRNA has previously been used to evaluate cellular uptake of siRNA (15,16). Q570-labeled antagomirs were cleared from the plasma at a t1/2 of ~30min (data not shown), which is considerably faster than the plasma-clearance of cholesterol-conjugated siRNA of ~90min (17). A striking overlap of the subcellular localization profiles of antagomirs and miRNAs by sucrose gradient ultracentrifugation analysis of liver homogenates indicates that they might share subcellular compartments. Antagomir localization within hepatocytes was strictly limited to the cytosol. This could explain why antagomirs did not influence steady-state levels of the nuclear precursors of miRNAs (7). We asked whether antagomirs could localize to P-bodies, since this compartment has been linked to the miRNA pathway. P-bodies are enriched in Ago 2 as well as mRNA that is targeted by miRNAs. It is, however, still unknown to which degree endogenous miRNAs localize to P-bodies. In any case, we did not observe any co-localization of antagomirs and P-bodies and conclude that the interaction of antagomirs and miRNAs occurs upstream of this compartment.
In this study, we also extended our previous analysis of the influence of different chemical modifications on antagomirs. Phosphorothioate modifications provide protection against RNase activity and their lipophilicity contributes to enhanced tissue uptake. On the other hand, phosphorothioates decrease the melting temperature of RNA duplexes (12) and have been shown to be general inhibitors of cellular RNAse activity (18). Indeed our results indicate a critical balance of the number of phosphorothioates within the antagomir chemistry. While a significant number of phosphorothioates increases efficiency, complete phosphorothioate modification decreased efficiency. Our data also demonstrate that antagomirs require >19-nt length for optimal function.
Lastly, we demonstrate that antagomirs can efficiently decrease miR-16 levels in mouse brain when injected locally. Systemic infusions of antagomir-16 do not change brain levels of miR-16, most likely because of an inability to cross the blood-brain barrier (7). Local injections of small amounts of antagomir-16 efficiently reduced expression of this miR-16 in the cortex. This inhibition was specific since the expression of other miRNAs was not affected and no alteration in miR-16 levels were measured in the contralateral hemisphere that was injected with PBS. These results suggest that miRNA-inhibitors could facilitate the elucidation of miRNA function in the CNS.
Different chemical modifications have been described to block miRNA function in vitro (6,12) or in vivo (7,8). However, detailed characterizations of these inhibitors are either lacking or based on heterologous miRNA target reporter assays (12). In this study, we used the expression levels of endogenous miR-122 targets as a read-out. We show that antagomirs can be used in a time- and dose-dependent fashion to study miRNA targets. Furthermore, we characterized antagomirs with regard to specificity, functional minimal-length requirements and effectiveness in the CNS following direct application. Our data support the conclusion that antagomirs will be powerful tools for further studies on miRNA function in mice and possibly non-rodent models and enhance the validation of miRNA inhibitors as potential therapeutics.
This work was supported by the National Institutes of Health (grant 1 P01 GM073047) and the Howard Hughes Medical Institute. Funding to pay the Open Access publication charge was provided by the Swiss Federal Institute of Technology.
Conflict of interest statement. M.S. and T.T. are SAB members of Aluylam Pharmaceuticals, T.T. is a co-founder.