Properties of N 4 -Methylated Cytidines in miRNA Mimics

Introduction

MicroRNAs (miRNAs) are natural RNAs of approximately 20 nucleotides (nt) in length that regulate the expression of a large proportion of the genome (Bushati and Cohen, 2007). MiRNAs are expressed in a complex biogenesis (Krol et al., 2010). Primary miRNA transcripts (pri-miRNA) are cropped in the nucleus by a RNase Drosha to generate pre-miRNAs, hairpins of approximately 60 nt with an imperfectly base-paired stem, a loop of 10–30 nt and a short 3′ overhang of a few nt. The stem of a pre-miRNA constitutes two arms (denoted 5p and 3p), one or both of which may be functionally competent, mature miRNAs. In the cytoplasm pre-miRNAs are processed by a second RNase Dicer, which excises the loop and loads the mature ∼22-nt miRNAs into the miRNA induced silencing complex (miRISC) together with an argonaute protein. This ribonucleoprotein complex is the active species, which binds to often-conserved target sites in the 3′ untranslated regions (UTRs) of mRNA substrates via Watson-Crick (W-C) recognition of nt 2–8 at the 5′ end of the miRNA (the “seed”). Frequently, miRNAs are members of families sharing a common seed sequence and, depending on their expression, permit them to function redundantly (Bushati and Cohen, 2007). Consequently, experiments conducted with miRNA mimics often result in subtle phenotypic changes making them especially vulnerable to artifacts.

RNA chemistry has progressed significantly to meet the demands of biologists for chemical tools to help understand the function of miRNAs. This includes the synthesis of long RNAs (Shiba et al., 2007), RNAs labeled with functional groups (Overhoff et al., 2009), RNAs for screening (Neubacher and Arenz, 2009), and synthetic siRNAs and miRNAs as genetic tools. Much effort has been devoted by commercial suppliers to maximizing the performance of synthetic RNAs in vitro, particularly with respect to stability, potency, and sequence selectivity. Whereas the rules of W-C reliably predict the base-pairing properties of oligoribonucleotides in vitro, their delivery into cells can produce unexpected effects. An immense population of expressed RNA sequences provides a choice of binding partners, albeit with partial complementarity, which compete with the intended mRNA or miRNA target. Such interactions can be of sufficient stability and durability to cause unintended biological consequences—known collectively as sequence-selective “off-target effects”—which confound the interpretation of experiments (Jackson and Linsley, 2010). This is a particular issue when faced with the often-subtle responses in miRNA biology, and consequently, appropriate controls are necessary to guard against artifacts. A commonly used type of control reagent, which was developed for the antisense and RNAi fields, is that of the mismatched oligoribonucleotide, in which 2 to 3 non-terminal nucleotides of the parent sequence are permutated (e.g., A→C). The minor difference in sequence between the parent and control oligoribonucleotides ensures that their global properties (including off-target effects) remain similar, while a reduced binding affinity of the control for the target RNA leads to an attenuated biological effect. We have previously employed mismatch controls to great effect in projects using chemically modified antisense and siRNA agents in vivo (Dorn et al., 2004) and also in analytical applications (Beuvink et al., 2007). However, it is difficult to apply a mismatch-control strategy when working with miRNA mimics. A mismatch substitution in the seed region drastically reduces the affinity of the miRNA for its mRNA target. However, it also generates a new miRNA seed sequence with its own associated transcript target set (Jackson et al., 2006). In contrast, permutation of nucleotides outside of the seed usually has no functional consequences (Esau et al., 2006).

We began a search for a new class of negative RNA control for miRNA experiments with the following criteria: the control should (1) not constitute a new seed sequence, (2) be accepted by the miRISC, (3) cause a significant loss of binding affinity to target RNAs, and (4) be synthesized conveniently into oligonucleotides on solid phase. In this work we describe the synthesis and properties of miRNA mimics containing methylated cytidines in the seed region.

Materials and Methods

Results and Discussion

A number of modified nucleosides with methylated nucleobases have been described that disable W-C pairing (e.g., N¹-methylG, N³-methylU) (Micura et al., 2001; Chiu and Rana, 2003). Methylation of the 4-amino group of cytidine yields a nucleoside with modified base-pairing properties.

Whereas mono-methylation (N⁴-methylcytidine; mMC) permits a regular G-C base pair if the methyl group adopts a trans conformation with N³ (Fig. 1A, B), dimethylation (N⁴,N⁴-dimethylcytidine; dMC) disrupts the W-C base pair and therefore would be expected to lower the strength of the pairing interaction considerably (Fig. 1C). We hypothesized that a dMC-nucleotide in the seed of a miRNA would likely stack into the A-form helix and base-pair weakly with G in the passenger strand, jutting its dimethylamino group into the major groove, thereby permitting loading into the RISC in analogous fashion to a G-U wobble (Gu et al., 2011). However, binding to the mRNA in the subsequent targeting step would be so weakened by the partial base pair in the seed that a significant loss of silencing activity would likely be experienced, similar to that of a mismatch control. As most miRNAs contain at least 1 C-nt in the seed, it is not necessary to expand the repertoire of bases beyond C.

OPEN IN VIEWER

FIG. 1.

Watson-Crick pairing of RNAs containing guanine with natural and methylated cytidines. (A) Canonical G-C base pair. (B) Guanine N⁴-methylcytosine (mMC) base pair. (C) One possible conformation of a guanine N⁴,N⁴-dimethylcytosine (dMC) base pair. (D) Melting profiles of duplexes comprising RNAs from Table 1: (: RNA duplex [1-2]; : RNA duplex [1-3]; : RNA duplex [1-4]); : RNA duplex [1-5]. (E) The thermodynamic T_M of duplexes is indicated by arrows, determined from the theoretical equilibrium curve f(T).

An additional attractive feature of the dMC modification is that it could be conveniently incorporated into RNAs using the convertible nucleoside approach (MacMillian and Verdine, 1990), employing a 4-triazolcytidine phosphoramidite during automated synthesis. Using dimethylamine instead of gaseous methylamine during oligonucleotide deprotection would then create the dMC at the desired position in the strand. This versatile synthesis strategy offers also the opportunity to incorporate other functionality into the major groove of miRNAs for a variety of chemical biology applications.

The convertible nucleoside, 4-triazolcytidine phosphoramidite, was synthesized as previously described (Shah et al., 1994). Solid phase RNA synthesis was performed under standard conditions. The 4-triazole group in selected cytidines of RNA sequences was then substituted by monomethyl- or dimethylamine with concomitant deprotection, while still on the CPG solid support. Removal of RNA protecting groups under standard conditions uses gaseous methylamine, thus the synthesis of mMC-containing RNAs was achieved without changes to the usual protocol. Substitution of triazole with dimethylamine was performed using aqueous dimethylamine, followed by gaseous methylamine to remove any traces of remaining base and phosphodiester protecting groups. RNAs were purified by reverse phase high performance liquid chromatography (RP-HPLC) and then detritylated. Modified sequences were characterized by electrospray mass spectrometry. Synthesis yields for unmodified RNAs from a 50-nmol synthesis were in the range of 6–8 nmol, whereas yields of dMC-modified and mMC-modified RNAs were 1–2 nmol and 1–10 nmol, respectively.

A series of RNAs (Table 1) were assembled to determine the consequences of incorporated N-methyl cytidines on binding affinity in short RNA duplexes. Control sequences bearing standard mismatch permutations were prepared for comparison. Melting curves from RNA- and modified RNA-duplexes (Fig. 1D) were transformed into an associated fraction plot calculated from the upper and lower baselines (Fig. 1E), and T_M were calculated for f=0.5 (Table 2). Values of ΔG° were obtained from a plot of ln K_a versus 1/T and the Gibbs equation. The complementary 14-nt duplex [1-2] yielded a T_M of 61.2°C. A single C→A mismatch base-change at position 12 (duplex [1-3]) decreased the T_M by 8.9°C, corresponding to a ΔΔG° of 2.2 kcal/mol, which is within the typical range for an A-G mismatch depending on the flanking base pairs (Davis and Znosko, 2007). The melting curve of the mMC-containing RNA duplex [1-5] is similar to that of the parent duplex [1-2] (T_M's of 60.9°C and 61.2°C, respectively). Previous reports of the thermodynamic properties of RNA duplexes containing N⁴-methylcytidine appear to be inconsistent. The results of studies on cytosine nucleobases using nuclear magnetic resonance spectrometry (Engel and Hippel, 1974) showed that the N⁴-methyl group favors a syn-conformation with N³, and therefore, forcing the anti-conformation necessary for a W-C base pair would be an energy destabilizing element during hybridization. Indeed, a considerable loss of binding affinity was found for a single mMC modification close to the center of an 11-nt RNA duplex (Allerson et al., 1997). However, N⁴-methylcytidine in a self-complementary 13-nt RNA duplex showed a considerable increase in duplex stability (Micura et al., 2001), whereas in an imperfect RNA duplex from the HIV rev-response element N⁴-methylcytidine caused no change in T_M (Grasby et al., 1995), consistent with our data. It is likely, therefore, that the base pairs flanking an mMC-G base pair determine whether the modification is stabilizing or destabilizing, as is also the case for mismatched base pairs (Kierzek et al., 1999; Davis and Znosko, 2007).

The T_M of the RNA duplex [1-4] containing a single dMC modification at C12 is 51.7°C, giving a ΔT_M of –9.6°C and a ΔΔG° of 3.4 kcal/mol in comparison to the parent duplex [1-2]. Thus, in this particular 14-nt duplex the dMC modification is more destabilizing than a standard mismatch by approximately 1 kcal/mol. Modification of base pairs at the terminal position of the RNA duplex (mismatch-[1-6], mMC-[1-8], dMC-[1-7]) barely affected the free energy of hybridization, as expected. Furthermore, melting curves and energy calculations of the doubly modified duplexes [1-9], [1-10] and [1-11] were similar to those of the single non-terminally modified sequences.

In summary, the dMC modification fulfills 3 of the important requirements for a negative control for miRNA experiments: (1) it does not constitute a new RNA sequence and therefore does not trigger a new mRNA target signature; (2) it is easily incorporated into RNAs; and (3) it causes a large destabilization of base-pairing affinity with complementary sequences, including, presumably, the mRNA substrates of miRNAs in RISC.

We next investigated the properties of two miRNAs containing N-methylated cytidines in their seed regions; specifically, whether they would be accepted into RISC. MiR-106a and miR-34a play important roles in regulating genes involved in cancer (O'Donnel et al, 2005; He et al., 2007). Natural mimics of miR-106a and miR-34a were prepared by annealing RNA sequences for their 5p- and 3p-strands as defined in the miRNA database miRBase (Fig. 2A; www.mirbase.org). A series of modified guide strands of miR-106a and miR-34a were synthesized, bearing a single mMC, a single dMC, and a C→A mismatch in their seed regions at positions 8 and 4, respectively (Table 1). These were also annealed to the appropriate passenger strand sequences. Finally, commercial miRNA mimics for miR-106a and miR-34a were acquired as well as 2 siRNAs: one targeting Renilla luciferase (siRen) to control for transfection efficiency and a correct functioning of the reporter gene, and an unrelated siRNA (siCon), which we have previously used to control for transfection toxicity (Dogar et al., 2011).

OPEN IN VIEWER

FIG. 2.

Methylated cytidines in microRNA (miRNA) duplexes are accepted into the RNA induced silencing complexes (RISC) and show varying levels of biological activity. (A) Structure of mimics miR-106a and miR-34a. (B) Schematic representation of miR-106a and miR-34a targeting complementary sites embedded in parts of the SIRT1 and P21 3′ untranslated regions (UTRs), respectively. (C) and (D) HeLa cells transfected with luciferase reporter plasmids shown in (B) were treated after 24 hours with increasing doses (0, 2, 9 and 36 nM) of analogs of miR-106a and miR-34a. Relative luciferase activity was measured 48 hours after plasmid transfections, and residual luciferase activity is plotted after normalization to that of the 0 nM treatment [mean of triplicate transfections±standard deviation (SD)]. (E) Schematic representation of miR-34a and miR-106a analogs targeted to luciferase reporter genes bearing a single miRNA target site from P21 and SIRT1, respectively in their 3′ UTRs. (F) and (G) HeLa cells transfected with luciferase reporter plasmids shown in (E) were treated after 24 hours with analogs of miR-106a and miR-34a. Relative luciferase activity was measured 48 hours after plasmid transfections, and residual luciferase activity is plotted after normalization to that of the 0 nM treatment (mean of triplicate transfections±SD). (H) Caspase 3/7 activity was measured from lysates of HeLa cells 72 hours after transfection with miR-34a analogs. Caspase 3/7 activity is plotted after normalization to that of the 0 nM treatment (mean of triplicate transfections±SD).

The double-stranded RNAs (dsRNAs) were tested in series in 3 types of cellular assays. Two dual-luciferase reporters, each carrying in their 3′ UTRs a single complementary target sequence to the miRNAs miR-34 and miR-106a assayed for the siRNA-like activity of the miRNAs (Fig. 2B). A second pair of reporters was constructed to assay their miRNA activity, each containing a unique site for miR-106a and miR-34a using a bona fide target sequence from 2 validated target sites in the 3′ UTRs of CDKN1A (P21) (Ivanovska et al., 2008) and SIRT1 (Yamakuchi et al., 2008), respectively. Finally, for the miR-34a series of reagents, an apoptosis assay was employed to show the functional consequences of methylated cytidines on the miR-34a-mediated induction of apoptosis in HeLa cells.

HeLa cells were co-transfected with reporter constructs and dsRNAs at 3 concentrations, and normalized residual luciferase activity was measured 48 hours later. In each experiment, the control siCon control showed little inhibitory activity, whereas siRen strongly reduced luciferase activity, indicating good transfection efficiency and no untoward toxicity. Unmodified miR-106a (wt-106a) and miR-34a (wt-34a) inhibited their respective targets efficiently, similar to commercially available mimics mimic-106a and mimic-34a (Fig. 2C, D).

MiR-106a and miR-34a containing C→A mismatches in the seeds (mm-106a and mm-34a, respectively) showed a heavily reduced activity in comparison to the wild-type counterparts. A siRNA guide strand employs its entire length in the recognition of an mRNA target sequence, and therefore, a single mismatched base is not always expected to abolish all silencing activity unless it is located at a critical position. The doubly methylated dMC-106a also showed a significant reduction in silencing activity compared to its wild-type analog, whereas dMC-34a only showed activity at the highest dose. MiRNAs with a single mMC in the seed (mMC-106a, mMC-34a) showed similar levels of activity to their unmodified counterparts, indicating that the mMC-G base pair is intact and that positioning of the N⁴-methyl into the major groove of the seed duplex region does not adversely affect the silencing activity of the RISC. This contrasts with miRNAs, which are modified at the minor groove binding sites in the seed, with for example 2′-O-methyl substituents on the ribose. Here the modification abolishes RISC activity, probably due to poor recognition by proteins of the RISC (Jackson et al., 2006; Manoharan et al., 2011).

The silencing activity of our seed-modified dsRNAs was most affected in assays reporting miRISC activity, as would be expected for a mechanism that depends on W-C recognition involving a seed region of only approximately 8 base pairs. MiR-106a and miR-34a have been shown to regulate conserved target sites in the 3′ UTRs of P21 and SIRT1, respectively. We constructed a luciferase reporter for miR-106a and miR-34a containing approximately 150 nt of these UTRs, including the validated target sites (Fig. 2E). Co-transfection of the reporter with siRen showed efficient silencing of luciferase, whereas the wild-type miRNAs wt-106a and wt-34a and the commercial mimics showed the typical low level of inhibition expected for miRNA-regulation of a single target site in a UTR (Fig. 2F, G). Both mismatched and dMC-modified miRNAs were inactive in this assay. In contrast, mMC-106a was slightly less potent and mMC-34a was slightly more potent than their wild-type counterparts, once again confirming that the placement of groups in the major groove is not detrimental to miRNA activity.

We have previously shown that miR-34a induces caspase 3/7 in HeLa cells, causing apoptosis (Dogar et al., 2011). We tested the miR-34a dsRNAs in the caspase assay and observed that the trend in activity against the SIRT1 reporter construct was mirrored at the functional level (Fig. 2H). Thus, mimic-34a, wt-34a, and mMC-34a were all able to induce caspase 3/7 in HeLa cells at 72 hours, whereas mm-34a and dMC-34a were not.

In summary, we have elaborated a method by which N-alkylated cytidines may be introduced into synthetic miRNA reagents using the convertible nucleoside approach. We have demonstrated that an N⁴-monomethylated cytidine in 2 short related RNA sequences does not adversely affect hybridization affinity, though this may depend on its flanking base pairs. We have shown with 2 examples that N⁴-monomethylated cytidine in the seed of a miRNA is accepted by the RISC and that the protrusion of the methyl group into the major groove does not hinder silencing of complementary targets or bona fide miRNA target sites. This bodes well for the introduction of other functionality into miRNAs as tools for biochemistry or chemical biology.

The introduction of the N⁴,N⁴-dimethylcytidine into an RNA perturbs the formation of a W-C base pair and lowers its binding affinity for a complementary sequence. We demonstrated that miRNAs with a single dMC in the seed region, both at position 4 (in miR-106a) and at position 8 (miR-34a), show reduced silencing activity at fully complementary target sites as well as at partially complementary miRNA target sites, probably from reduced binding affinity to the tested mRNAs. We cannot exclude effects on other predicted targets of these miRNAs without conducting a full transcriptome-wide analysis. In summary, the modified C-base satisfies all of the aforementioned criteria for the use in negative miRNA controls, most importantly that a new seed sequence with target signature has not been introduced.