RNAs often form complex secondary and tertiary structures, and in turn, interact with proteins and metabolites for diverse regulatory and catalytic functions (1–3). Double-stranded RNAs (dsRNAs) are essential components of RNA secondary structures (4–7). RNA triplex structures, formed between dsRNA regions and single-stranded RNA (ssRNA) regions, are emerging as an important tertiary structure motif (8–13). It is thus of great potential to develop ligands that can selectively and sequence-specifically recognize viral and cellular dsRNA regions as chemical probes and therapeutics.
Targeting dsRNAs through sequence-specific major-groove triplex formation is a promising strategy and numerous chemical modifications have been developed to enhance the binding affinity and specificity of Triplex-Forming Oligonucleotides (TFOs) (14–19). Peptide Nucleic Acids (PNAs) (20) typically have a neutral peptide-like backbone (Figure 1) and are advantageous with strong resistance against nucleases and proteases, and enhanced binding to natural nucleic acids. PNAs were originally designed to bind to double-stranded DNAs (dsDNAs) to form major-groove PNA·dsDNA (PNA·DNA–DNA, with the ‘·’ and ‘–’ representing Hoogsteen and Watson–Crick pairs, respectively, see Figure 1A–C for the base pairing interactions) triplexes. However, PNAs have been shown to form Watson–Crick duplexes with complementary DNA, RNA, or PNA in an antiparallel (with the C-terminus of a PNA aligned with 5′ end of DNA/RNA) or parallel (with the N-terminus of a PNA aligned with 5′ end of DNA/RNA) orientation (20–22). In addition, PNAs can be involved in the formation of parallel major-groove triplexes with various compositions including PNA·DNA–PNA, PNA·RNA–PNA, PNA·DNA–DNA, and PNA·RNA–RNA (13,23–36). It is interesting to note that short PNAs show selective binding to dsRNAs over dsDNAs, suggesting PNAs’ great potential as dsRNA-specific binders (13,28–37).
To recognize a Watson–Crick G–C base pair in an RNA duplex, one may design a major-groove C+·G–C triple (Figure 1A). However, to form a Hoogsteen C+·G pair often requires a relatively low pH (<6) for the protonation of the N3 atom of the C base (28,33,38). Modified PNA bases such as thio-pseudoisocytosine (L, Figure 1B) and 2-aminopyridine (M) have close-to-neutral pKa values for the corresponding nitrogen atom and thus allow enhanced recognition of a G–C pair in RNA duplexes through L·G–C and M·G–C base triple formation, respectively, at physiological pH (29,33,39). In addition, short modified PNAs incorporating L bases show selective binding toward dsRNAs over ssRNAs (Figure 1B, D, E) (33,36), suggesting that L-modified PNAs have a great potential for probing and targeting dsRNAs.
To recognize an inverted Watson–Crick C–G pair in an RNA duplex, a guanidine-modified PNA monomer (Q monomer) (34) or other monomers may be incorporated into PNAs through PNA·RNA–RNA triplex formation. Similar to L, Q is a cytosine derivative favoring Q·C–G triple formation, while destabilizing Watson–Crick-like Q–G pair formation (34–37). To recognize an inverted Watson–Crick U–A pair in an RNA duplex, a PNA strand incorporating an E (3-oxo-2,3-dihydropyridazine) base has been used through E·U–A base triple formation (39–41).
An unmodified T base has been incorporated into PNAs to recognize a Watson–Crick A–U pair through major-groove T·A–U base triple formation (Figure 1C) (13,20,28,38,39,42). Enhancing the recognition of Watson–Crick A–U base pairs is crucial due to the relatively high abundance of A–U pairs in dsRNAs. A substitution of the 5-methyl group in T by hydrogen or halogen atoms (F, Cl, Br, and I, designated as XU in Figure 1C) results in a decrease of the pKa of N3 nitrogen atom (43) and thus may strengthen the hydrogen bond formed involving the N3 atom. Thus, XU substitution of T in a PNA is expected to enhance Hoogsteen and Watson–Crick hydrogen bonding interactions (Figure 1C, F) (44,45). The relatively large halogen atoms may also contribute to base stacking interactions (44). Here, we report the binding properties of PNAs incorporating uracil and 5-halouracils (5-fluorouracil (FU), 5-chlorouracil (ClU), 5-bromouracil (BrU) and 5-iodouracil (IU) for the recognition of A–U base pairs in dsRNAs. We also tested if incorporating adjacent L and XU bases can cooperatively enhance the binding affinities of dsRNA-binding PNAs for targeting microRNA-198 (miRNA-198 or miR-198) hairpin precursor (46,47) and HIV-1 −1 ribosomal frameshift inducing mRNA hairpin (48–50).
Reagents and solvents used were obtained from commercial sources and used without further purification. All organic reactions were monitored with the use of thin layer chromatography (TLC) using aluminum sheets silica gel 60 F254 (Merck). Compounds were purified by flash column chromatography using silica gel with ethyl acetate/petroleum ether mixture as the eluting solvent. All 1H and 13C NMR spectra were obtained at room temperature on 300, 400 (100 MHz, 13C) or 500 MHz Bruker spectrometer. The chemical shifts (δ) are shown in parts per million (ppm). The residual solvent peaks were used as references for the 1H (chloroform-d: 7.26; dimethyl sulfoxide-d6: 2.50) and 13C (chloroform-d: 77.0; dimethyl sulfoxide-d6: 39.5) NMR spectra. The mass spectra of the compounds were obtained via liquid chromatography-mass spectrometry with electrospray ionization source (LCMS-ESI) and high-resolution mass spectrometry (electron ionization) (HRMS-EI). Reverse-phase high performance liquid chromatography (RP-HPLC) purified RNA and DNA oligonucleotides were purchased from Sigma-Aldrich Singapore.
Synthesis of PNA monomers
The detailed synthesis procedures for the PNA monomers are shown in Supplementary Data (Supplementary Figures S1–S15). The PNA monomers were synthesized based on the previously reported methods (Scheme 1) (51–53). Commercially available uracil or 5-halouracil was reacted with chloro- or bromo-acetic acid in the presence of potassium hydroxide to yield respective compounds 1–5. Compounds 1–5 were coupled with PNA backbone ethyl N-(2-Boc-aminoethyl)glycinate in the presence of Hexafluorophosphate Benzotriazole Tetramethyl Uronium (HBTU) and N-methylmorpholine (NMM), in Dimethylformamide (DMF) to give respective compounds 6–10. Finally, the ethyl ester group was hydrolyzed by using aq. LiOH in water–THF solvent. The formed lithium salt of carboxylic acid was neutralized and the mixture was acidified using diluted HCl for the desired PNA monomers 11–15.
Synthesis of PNA oligomers
The PNA thymine (T) and cytosine (C) monomers containing the standard aminoethyl glycine backbone were purchased from ASM Research Chemicals. The PNA monomer L was synthesized following our previously reported method (33). PNA oligomers were synthesized manually using Boc chemistry via a Solid-Phase Peptide Synthesis (SPPS) protocol (37). 4-Methylbenzhydrylamine hydrochloride (MBHA·HCl) polystyrene resins were used. The loading value used for the synthesis of the oligomers was 0.3 mmol/g and acetic anhydride was used as the capping reagent. Benzotriazol-1-yl-oxytripyrrolidinophosphonium hexafluorophosphate (PyBOP) and N,N-diisopropylethylamine (DIPEA) were used as the coupling reagent. The oligomerization of PNA was monitored by Kaiser test. Cleavage of the PNA oligomers was done using trifluoroacetic acid (TFA) and trifluoromethanesulfonic acid (TFMSA) method, after which the oligomers were precipitated with diethyl ether, dissolved in deionized water and purified by RP-HPLC using water–CH3CN–0.1% TFA as the mobile phase. Matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) analysis was used to characterize the oligomers (Supplementary Table S1 and Figure S16), with the use of α-cyano-4-hydroxycinnamic acid (CHCA) as the sample crystallization matrix. The extinction coefficients of XU were assumed to be the same as that of T in calculating the extinction coefficients of the PNA oligomers (37).
Non-denaturing polyacrylamide gel electrophoresis
Non-denaturing (12%) polyacrylamide gel electrophoresis (PAGE) experiments (37) were conducted with an incubation buffer containing 200 mM NaCl, 0.5 mM EDTA, 20 mM HEPES at pH 7.5, 7.7 or 8.0. The loading volumes for the samples containing RNA and DNA hairpins were 20 or 25 μl and 10 μl, respectively. The samples were prepared by snap cooling of the hairpin, followed by annealing with PNA oligomers by slow cooling from 65°C to room temperature and incubation at 4°C overnight. Prior to loading the samples into the wells, 35% glycerol (20% of the total volume) was added to the sample mixtures. 1× Tris–borate–EDTA (TBE) buffer, pH 8.3 was used as the running buffer for all gel experiments. The gels were run at 4°C at 250 V for 5 h. The gels were then stained with ethidium bromide for 30 min and imaged by the Typhoon Trio Variable Mode Imager.
Circular dichroism (CD) spectroscopic characterization was carried out by using a JASCO model J-1500-150 spectropolarimeter and a quartz cell (optical path length: 10 mm). The concentrations of RNA and PNA are 2 and 4 or 20 μM, respectively. RNA was snap cooled from 95°C in an incubation buffer containing 100 mM NaCl, 10 mM sodium phosphate, pH 7.0. The RNA hairpin was annealed with PNA by incubating at 65°C for 10 min followed by slow cooling to room temperature. The scanning rate was 50 nm/min and the data were averaged from six scans. The measurements were done at room temperature.
UV-absorbance-detected thermal melting
UV-absorbance-detected thermal melting experiments were conducted using the Shimadzu UV-2550 UV-Vis spectrophotometer with the use of an 8-microcell cuvette. The absorbance at 260 nm was recorded with the temperature increasing from 15 to 95°C followed by the temperature decreasing from 95 to 15°C. The temperature ramp rate is 0.5°C/min. The optical path length of the 8-microcell cuvette is 1 cm. The incubation buffer is 200 mM NaCl, 0.5 mM EDTA, 20 mM NaH2PO4, at pH 7.5. All samples contain 5 μM RNA and 5 μM PNA in 130 μl buffer. The samples containing the ssRNA and PNA were annealed by slow cooling from 95°C to room temperature, followed by incubation at 4°C overnight. Data were normalized at high temperature and the melting temperatures were determined based on the Gaussian fits of the first derivatives of the curves.
Confocal microscopy studies
HeLa cells (1 × 105) were plated in a petri-dish in 500 μl Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS). The cells were grown at 37°C, 5% CO2 for 24 h. The carboxyfluorescein-labeled PNAs were added to the cell cultures, followed by incubation for 24 h. A final concentration of 0.5 mg/l Hochest 33258 were added and incubated for 1 h. The cells were washed twice with PBS. Fresh OPTIMEM medium was added and the cells were immediately visualized on 63× objective of Zeiss LSM 800 confocal microscope using 405 and 488 nm lasers for Hoechst and carboxyfluorescein, respectively.
Computational modelling of PNA·RNA–RNA triplex structures
The crystal structure (PDB ID: 1PNN) of a PNA·DNA–PNA triplex (54) was used as a starting structure for the modelling of the PNA·RNA–RNA triplex structures. The backbones of the Watson–Crick PNA–DNA duplex were replaced with the RNA backbone using Discovery Studio 2016 (Dassault Systèmes, San Diego). The geometries of the base triples (C·G–C and T·A–T) without backbone attached were also built based on the crystal structure, with the C·G–C and T·A–T triples replaced with L·G–C and T·A–U triples, respectively, using Discovery Studio 2016. The new base triple structures were optimized under B3LYP/6-31G* scheme using Gaussian 09 (Gaussian, Inc., Wallingford, CT, USA). The optimized base triples were then used to replace the base triples in the original PNA·DNA–PNA crystal structure to obtain the starting structure for a PNA·RNA–RNA triplex structure containing the PNA sequence of AcNH-TLTLTTTL-CONH2 with the N-terminal amine capped by an acetyl group. The RNA duplex contains 11 bp (corresponding to base pairs from A4–U29 to U14–A19, see Figure 2A) with the sequences of (5′AGAGAGAAAGU3′ and 5′ACUUUCUCUCU3′). We then applied energy minimization and molecular dynamics (MD) simulations to optimize the triplex structure using GROMACS 5.1 package (55). The force field used for RNA residues was Amber99bsc0 with χOL3 modification (56). The force field parameters for unmodified PNA can be retrieved from REDDB database (57). For the modified PNA residues, the atomic charges can be fit using RED server (58), and the topological parameters were adapted from Amber10 (University of California, San Francisco) Generalized Amber Force Field (GAFF) library (59). Unless otherwise noted, we applied the base–base hydrogen bond restraints throughout the simulations.
The starting PNA·RNA–RNA triplex structure was first solvated in a periodic boundary box, followed by an energy minimization, an NVT (constant temperature and volume) equilibrium, and NPT (constant temperature and pressure) equilibrium step, with the final temperature set to be 298K and pressure set to be 1 atm. To further optimize the backbone conformation, we applied a 40-ns simulation protocol (298 K→398 K→298 K→398 K→298 K), with all the base-base hydrogen bonds restrained. Thereafter, a 100-ns product MD was performed, with the base-base hydrogen bonds of the terminal A14–U19 pair (corresponding to the base pair near the hairpin loop) restrained. The clustered triplex structure was obtained from the last 10 ns of the MD simulation. Discovery Studio 2016 was used to build the triplex structures containing the other PNA sequences by replacing the corresponding PNA bases without further optimization.
Substitution of T with U or halouracils in an unmodified PNA enhances binding to a dsRNA
We carried out non-denaturing PAGE to characterize the triplex formation (Supplementary Figures S17–S28). An unmodified PNA (P1, Figures 1G and 2, Table 1) has been previously shown by non-denaturing PAGE to bind to an RNA hairpin rHP1 (Kd = 5.1 μM at 200 mM NaCl, pH 7.5) (33). In this study, we used a decreased concentration of rHP1 (0.25 μM instead of 1 μM as used previously) and obtained a slightly decreased Kd (2.1 μM, Figure 3A, Table 2). The replacement of the single thymine at position 3 in P1 by uracil (U-3) results in an enhanced binding toward rHP1 (with the Kd value decreased from 2.1 to 1.2 μM at pH 7.5, Table 2, Supplementary Figures S17 and S18). Replacing two T residues at the third and fifth positions in P1 with U bases (U-3,5) further enhances the binding affinity toward rHP1 with a Kd value of 0.6 μM at pH 7.5 (Figure 3B, Table 2). The observed stabilizing effect is consistent with the fact that a hydrogen bond (formed between N3 of T and N7 of A) in the PNA·RNA Hoogsteen T·A can be enhanced by shifting the monomer N3 pKa value from ∼9.8 for T toward neutrality for U (Figure 1C) (43,44).
It is of note that a destabilizing effect was previously observed in forming a DNA·dsDNA triplex upon DNA T to DNA U substitution in the Hoogsteen strand (44,60). The T-to-U destabilizing effect in a DNA Hoogsteen strand was attributed to the loss of the favorable hydrophobic effect and stacking involving the methyl groups of T residues in the major groove (44,60). It is likely that the favorable stacking and hydrophobic effect of the methyl groups of T residues within a PNA strand are not as significant in PNA·dsRNA triplexes. It is also possible that the stacking interactions in a PNA·RNA2 triplex may be different from a DNA·DNA2 triplex. Previous studies suggest that the binding of the third strand to an RNA or DNA duplex to form a major-groove triplex induces a subtle conformational change. The resulting structure of the parental RNA or DNA duplex is different from either the A-form or B-form structure, although the base-base Watson–Crick pairing geometries remain largely the same (8,10,61–64). It is thus conceivable that the stacking interactions in a PNA·RNA2 triplex may be different from RNA·RNA2 or DNA·DNA2 triplex.
The 5-halouracil substitution in PNAs may result in enhanced base stacking, in addition to strengthened hydrogen bonding interaction. We tested the effects of single halouracil substitutions at varied positions (positions 3, 5, and 6, Figure 2F, Table 1) of the unmodified 8-mer PNA P1 on PNA·RNA2 triplex formation. The T residues of PNA P1 at positions 3, 5 and 6 are flanked by two cytosine residues, cytosine and thymine, and two thymine residues, respectively. Among all the singly-modified PNAs at position 3, BU-3 (with BrU modification) shows the largest enhancement in binding to rHP1 (Kd = 0.4 μM at pH 7.5, Table 2, Supplementary Figures S17 and S18). It is probable that the bromine atom in a BrU base may have optimized electronegativity and atom size, for enhanced Hoogsteen hydrogen bonding (through pKa reduction) and base stacking interactions, respectively (44). The stabilization effect of a single BrU base substitution is essentially position independent, as the Kd values for PNAs with singly modified BrU at positions 3, 5, and 6 are within a narrow range of 0.4-0.5 μM. Thus, BrU may be incorporated in between T and/or C residues in a PNA for enhanced binding to dsRNAs.
However, the binding affinity toward rHP1 of singly-modified PNAs containing FU residue was found to be position dependent (Table 2, Supplementary Figures S17 and S18). The PNA containing a FU modification at position 3 (FU-3, with FU in between two cytosine, CFUC) has a Kd value of 1.1 μM, whereas, the PNA with the FU residue at position 5 (FU-5, flanked by cytosine and thymine, CFUT) or 6 (FU-6, flanked by two thymine residues, TFUT) has a tighter binding (Kd = 0.3 μM). It is likely that a fluorine atom has favorable interactions with an adjacent T residue, e.g., a C–H···F–C hydrogen bond or favorable electrostatic interaction (65–67) may form between the F atom and the methyl group of a T residue on the C-terminal side (see the computationally modelled triplex structure below).
It is interesting to note that singly- (at position 3, 5, or 6) and doubly-modified PNAs (at positions 3 and 5) generally show similar Kd values at pH 7.5 (Table 2, Figure 4A, and Supplementary Figures S17 and S18), suggesting that the stabilizing effect of an internal XU modification may propagate beyond the nearest neighbor and the next nearest neighbor stacking partners in a triplex. Such non-nearest neighbor effect has been previously revealed for the formation of duplexes with other modifications and mutations by bulk thermal melting and single-molecule mechanical unfolding experiments (68–73).
We tested if the modified PNAs form pre-organized helix structure by CD spectroscopy as previously observed for a backbone-modified PNA (74). Our CD spectroscopic data suggest the formation of PNA·dsRNA triplexes (Supplementary Figure S29A), which is in agreement with previously reported results (23,28,38). However, a BrU-modified PNA alone does not show helicity, indicating that the PNAs are not preorganized as a helical structure for triplex formation. It is possible that a XU modification reduces the flexibility of PNAs resulting in a reduced entropic penalty for binding to an RNA duplex. Taken together, our results indicate that a single BrU or multiple BrU (separated by more than two residues) residues may be incorporated into a PNA for enhanced binding to dsRNAs.
Substitution of T with U or halouracils in an unmodified PNA reduces pH dependence of PNA·dsRNA triplex formation
Due to a relatively low pKa of 4.5 for N3 atom of a monomer cytosine, for a PNA containing one or more C bases, the formation of protonated C+·G–C base triple (Figure 1A) and PNA·dsRNA triplex is expected to be pH dependent. It is known that the apparent pKa of internal C residues in a TFO within an RNA·dsRNA triplex or a DNA·dsDNA triplex is shifted up near or above neutrality primarily driven by the triplex formation (8,75). Since substituting a RNA TFO with a PNA significantly enhances triplex formation (33), the apparent pKa of internal C residues in a PNA·dsRNA triplex may also be up shifted toward or above neutrality.
Our PAGE studies show that the binding affinity of PNA P1 to rHP1 decreases significantly upon increasing pH with the Kd values of 2.1, 15.6, and 27.0 μM, respectively, at pH 7.5, 7.7, and 8.0 (Figure 4B, Table 2, Supplementary Figures S17–S22), which is consistent with the fact that PNA P1 has two internal and one terminal C residues. It is of note that our PAGE data reveal relatively less significant pH dependent binding for the doubly-modified PNAs (Table 2, Figures 3 and 4B). For example, with the pH varied from 7.5 to 8.0, PNA BU-3,5 has Kd values ranging from 0.3 to 1.2 μM.
We have previously discovered (33) that a PNA incorporating a modified base thio-pseudoisocytosine (L) has an enhanced recognition of a G–C pair in an RNA duplex through L·G–C base triple formation at physiological pH, and selective binding toward dsRNAs over ssRNAs (Figure 1B, E). The pKa of a monomer L is around 9 (33,76,77). It is remarkable to note that, a PNA containing three L and no C residues (NH2-Lys-TLTLTTTL-CONH2) has a binding affinity (0.2 μM at pH 7.5) (34), similar to that of PNA BU-3,5 (NH2-Lys-TCBrUCBrUTTC-CONH2) containing two BrU and three C residues (0.3 μM at pH 7.5, Figure 3I, Table 2). The results indicate that upon substitution of T with BrU, a PNA·dsRNA triplex containing C+·G–C base triples may become more resistant against high-pH destabilization. It is likely that enhancing hydrogen bonding and stacking involving XU·A–U base triples (compared to PNAs without XU substitution) may in turn favor C+·G–C base triple formation and reduce the pH dependence by further shifting up the pKa of C residues in a PNA·dsRNA triplex through the coupling between stacking and hydrogen bonding interactions (78). However, the Kd value of PNA FU-3,5 binding to rHP1 is still relatively pH dependent 0.3 and 4.7 μM at pH 7.5 and 8.0, respectively), probably because FU stabilizes triplex formation mainly through enhancing base triple formation but not by shifting the pKa of adjacent C residues. Taken together, our results suggest that one may simply incorporate halouracils (e.g. BrU and ClU) next to the C bases in PNAs to enhance PNA·dsRNA triplex formation and reduces pH dependence.
PNAs incorporating modified T residues retain sequence specificity in binding to dsRNA and show no appreciable binding to dsDNA
We next tested the sequence specificity of the XU modified PNAs. No triplex formation was observed at pH 7.7 between modified PNAs and mutated hairpins with an A–U pair in rHP1 (opposite to PNA position 5) substituted with a C–G pair (rHP5), a G–C pair (rHP6), or a U–A pair (rHP7) (Figure 2, Supplementary Figures S23 and S24). Our CD data confirm that PNA BU-3,5 can sequence specifically bind to rHP1 with no binding to rHP2 (Supplementary Figure S29A). rHP2 has the G9–C24 pair in rHP1 replaced with a C–G pair. The results suggest that the XU modified PNAs have a high sequence specificity through XU·A–U and C+·G–C triple formation. Furthermore, P1 and the XU modified PNAs show no triplex formation at pH 7.7 with a DNA hairpin (dHP1, Figure 2E, Supplementary Figure S25), which is consistent with previously reported results (28,33,34). The preferential binding of PNAs to an RNA duplex over a DNA duplex may suggest that the relatively deep and narrow RNA duplex major groove is more compatible for accommodating a PNA. Taken together, it is promising to incorporate XU residues into PNAs for the enhanced and sequence-specific recognition of A–U pairs in RNA duplexes.
Substitution of T with U or halouracils in an unmodified PNA enhances binding to ssRNAs
It is expected that, similar to stabilizing a triplex by incorporating Hoogsteen XU·A pairs, incorporating Watson–Crick XU–A pairs would also stabilize Watson–Crick PNA–RNA duplex (45). We tested by thermal melting the binding of the XU-modified PNAs to ssRNAs through Watson–Crick duplex formation (Supplementary Figures S30 and S31). Compared to P1, the singly- and doubly-modified PNA sequences have higher melting temperatures (based on the heating curves), with ClU and BrU modifications being the most stabilizing (Figure 4C, D).
Compared to FU base, ClU and BrU may be more ideal for optimizing the coupled stacking and hydrogen bonding interactions (78,79) for the sequences tested here, with the PNAs containing pyrimidine residues only. The stabilization effect of XU substitution on PNA–RNA duplex was also observed previously for a different PNA sequence (with the XU residue flanked by two G residues), although relatively less variation among FU, BrU, and ClU was observed (45). It is interesting to note that previous NMR studies suggest that a Watson–Crick XU–A pair is locally more dynamic with enhanced base pair opening kinetics and proton transfer within a base pair (80–83). Thus, enhanced local dynamics (as revealed by NMR) may contribute favorable entropy to the global stabilizing effect (observed in the UV-absorbance detected thermal melting experiment). We observed a relatively large hysteresis between the heating and cooling curves in the thermal melting experiment (Supplementary Figures S30 and S31). A comparison of the Tm values of the cooling curves suggests a relatively less significant stabilizing effect of XU substitution. The results may indicate that XU substitutions in a PNA strand stabilize a Watson–Crick PNA–RNA duplex mainly through enhancing molecular interactions in the duplex and thus decreasing unfolding kinetics.
Enhancing PNA–RNA Watson–Crick duplex formation may result in the invasion of a RNA–RNA duplex. Our PAGE data suggest that lowering NaCl concentration from 200 to 10 mM results in enhanced PNA·RNA–RNA triplex formation, but no strand invasion complex formation (Supplementary Figure S28A, B). The PAGE data suggest that addition of 2 mM MgCl2 results in the destabilization of PNA·dsRNA triplexes (Supplementary Figure S28C, D). However, PNA BU-3,5 still shows a relatively tight binding even in the presence of 2 mM MgCl2 (Kd values of 0.9 μM with 2 mM MgCl2 versus 0.3 μM without 2 mM MgCl2 (see Figure 3I and Supplementary Figure S28D)). DNA hairpin dHP1 shows no binding to the PNAs at 200 mM NaCl, but the formation of strand invasion complexes at 10 mM NaCl (Supplementary Figure S28E). Thus, substitution of T with XU is a promising strategy for enhancing the recognition of dsRNAs over dsDNAs at physiological conditions.
Targeting miR-198 hairpin precursor structure
We next incorporated BrU into PNAs to target miR-198 hairpin precursor structure (46,47). We designed a series of 6-mer PNAs for the recognition of a duplex region of miR-198 hairpin precursor (Figure 5A–G). Our PAGE results suggest that substitution of the two T residues in miR-C,T with BrU (miR-C,BU) results in an enhanced binding at pH 6.0 (from >20 μM to 12.4 μM), although no binding was observed at pH 7.5 (Figure 6A–D, Supplementary Figure S26). It is probable that simultaneous protonation of three consecutive C residues may be unfavorable with relatively low apparent pKa values of the three C residues.
Substituting all four C residues in miR-C,T with L (miR-L,T, see Figure 1B, E for the structure of L) results in an enhanced binding at both pH 6.0 and 7.5 (Figure 6A, B, E, F, Supplementary Figure S26). Consistent with our previous results (33,34), minimal pH dependence was observed for the L-modified miR-L,T. Thus, a neutral L residue is advantageous in substituting consecutive C residues to enhance the recognition of consecutive G–C pairs. However, further replacing the T residues in miR-L,T with BrU (miR-L,BU) destabilizes the PNA·RNA–RNA triplex (from 2.1 μM to 5.6 μM at pH 7.5, Figure 6E–H). The results indicate that adjacent L and BrU residues may not be cooperative in stabilizing a PNA·RNA–RNA triplex, possibly due to unfavorable stacking interactions in an LBrU step (see below the computationally modeled triplex structures and stacking patterns).
Our CD spectroscopic data confirm the formation of PNA·RNA–RNA triplex formation involving the relatively short 6-mer PNAs (Supplementary Figure S29B). Interestingly, both our UV-absorbance-detected thermal melting and CD data suggest that the short PNAs show no strong binding to miR-198 ssRNA (5′-GGGAGAUAGG-3′), a 5′ fragment of miR-198 hairpin precursor (Figure 5B, Supplementary Figures S29C and S32). Thus, short PNAs are promising for selectively targeting dsRNAs over ssRNAs.
Targeting HIV-1 ribosomal frameshift inducing mRNA hairpin
We incorporated BrU into PNAs to target HIV-1 viral mRNA hairpin structure (Figure 5H) critical for stimulating −1 ribosomal frameshift (48–50). Our previous PAGE data suggest that a short 6-mer PNA (NH2-Lys-LLTTLL-CONH2, HIV-T,T, Figure 5I) can bind to the HIV-1 frameshift inducing hairpin (Kd = 1.1 μM at 200 mM NaCl, pH 7.5) (33). Here, we used a decreased RNA hairpin concentration of 0.25 μM (instead of 1 μM) and obtained the binding affinities more accurately (Kd = 0.3 μM for HIV-T,T at 200 mM NaCl, pH 7.5, Figure 7A). To compare the binding affinities of the PNAs with various substitutions, we also measured the binding at pH 8.0 (Figure 7B, D, F, H, J). We made two PNAs with single BrU substitutions (HIV-BU,T (NH2-Lys-LLBrUTLL-CONH2) and HIV-T,BU (NH2-Lys-LLTBrULL-CONH2)) to probe the sequence dependent stabilization effect of BrU substitution (Figures 5J, K and 7C–F). PNA HIV-BU,T shows similar binding affinities compared to PNA HIV-T,T (with Kd values of 1.1 versus 1.4 μM at 200 mM NaCl, pH 8.0, Figure 7A–D). The results are consistent with our binding data for PNAs targeting miR-198 hairpin precursor (Figure 6E–H) and the potential unfavorable stacking interactions for an LBrU step present in PNA HIV-BU,T (see below the computationally modeled triplex structures and stacking patterns). Significantly, we observed enhanced PNA·RNA–RNA triplex formation for PNA HIV-T,BU compared to PNA HIV-T,T and HIV-BU,T (with Kd values of 0.4 versus 1.4 and 1.1 μM at 200 mM NaCl, pH 8.0, Figure 7B, D, F). The stabilization effect of PNA HIV-T,BU compared to PNA HIV-BU,T results mainly from the BrUL step in HIV-T,BU, compared to the LBrU step in HIV-BU,T. The BrUT and TBrU steps seem to have essentially the same stabilizing effect as observed for the rHP1-targeting PNAs containing no L residues (Table 2). Clearly, the thermodynamic effect of BrU substitution for T in L-containing PNAs is sequence dependent. One may stabilize a PNA·RNA–RNA triplex by incorporating a BrUL step instead of an LBrU step.
Interestingly, PNA HIV-BU,BU (NH2-Lys-LLBrUBrULL-CONH2) has a weakened binding compared to HIV-T,T (with Kd values of 4.4 versus 1.4 μM at pH 8.0, Figure 7B, H). Clearly, the thermodynamic contributions of consecutive BrU residues in a dsRNA-binding PNA are not additive, likely due to a local distortion of the helical structure with altered stacking geometry (84), although Br atom has a radius similar to that of a methyl group in T (85). In addition, stacking of two consecutive BrU residues in a dsRNA-binding PNA might result in the keto-enol tautomeric shift of the BrU base (43,86,87), and thus weaken PNA·RNA–RNA triplex formation. It is interesting that incorporating two consecutive FU residues in PNA HIV-FU,FU (NH2-Lys-LLFUFULL-CONH2) shows a stabilizing effect in triplex formation as compared to PNA HIV-T,T (with Kd values of 0.8 versus 1.4 μM at pH 8.0, see Figure 7B, J). It is probable that the stabilization effect of FU substitution is mainly through enhancing base triple formation without significantly affecting stacking interactions with adjacent triples.
Extensive studies have been done for facilitating the cellular uptake bioactive PNAs and other peptides and nucleic acids (32,34,88–100). We focused on testing the effect of bromination of PNA on cellular uptake as previous studies suggest that fluorination of PNA T base does not significantly enhance cellular uptake of PNAs (101,102). Our confocal microscopic imaging studies show that the carboxyfluorescein-labelled PNAs HIV-T,T-cf and HIV-BU,BU-cf are both essentially cell impermeable (Supplementary Figure S33).
Computational modeling of PNA·dsRNA triplex structures
There is no high-resolution three-dimensional structure available for a PNA·dsRNA triplex. To gain further insights into the potential interactions responsible for the triplex formation stability, we computationally modeled the PNA·dsRNA triplexes using the existing crystal structure of PNA·DNA–PNA triplex as a starting structure (Figures 8 and 9A). We used the PNA sequence of AcNH-TLTLTTTL-CONH2 to model the initial PNA·dsRNA triplex structure, because an L·G–C triple is relatively more stable than a C·G-C triple. We consider the modelled triplex structure to be reasonably accurate for providing the structural insights into the potential molecular interactions responsible for a PNA·dsRNA triplex formation. Our modeled PNA·dsRNA triplex structure suggests that the methyl group and halogen atom in T and XU bases may be potentially stacked with an adjacent base on the N-terminal side (Figures 8 and 9). Such a stacking interaction may explain the (de)stabilizing effects experimentally observed in our study.
For example, a stabilizing effect may be expected if a base stack involves a CXU step (from N-terminus to C-terminus, with X = halogen atoms), because, compared to a CT step, a CXU step has an enhanced stacking in addition to improved Hoogsteen hydrogen bonding (Figure 9B, C). The modelled structures also imply that, for a XUT step (Figure 9C, D), favorable interactions may form between the T methyl group and XU halogen atom, e.g. a C–H···F–C hydrogen bond or favorable electrostatic interaction (65–67) as discussed above.
The modelled structures may also explain why an LBrU step is less favorable than a BrUL step (Figures 5–7). For example, in an LBrU step, charge-charge repulsion may be expected due to the overlap in a stacking geometry between the partially negatively charged N1 nitrogen atom of L and Br atom of BrU in the major groove (Figure 9D). Taken together, our experimental and computational modelling results suggest that stacking interactions are critical in stabilizing PNA·dsRNA triplex structures. One may incorporate BrU upstream of an L residue, but avoid having a BrU residue downstream of an L residue in designing dsRNA-binding PNAs.
In summary, substitution of T with U in PNAs can enhance the binding of short PNAs to dsRNAs through major-groove triplex formation, which is opposite to the destabilizing effect observed for DNA·DNA–DNA triplex formation (44,60). The 5-position halogenation of U can further enhance the triplex formation (with BrU substitution being the most stabilizing), presumably due to the optimized effects of improved hydrogen bonding and stacking interactions. Our results also suggest that one may incorporate a single BrU or multiple nonconsecutive BrU residues into C base-containing PNAs to enhance PNA·dsRNA triplex formation and reduce pH dependence. The reduced pH dependence may be due to an upshift of the apparent pKa of C bases through favorable stacking interactions (78,79) with halouracil bases. It is significant that incorporating a BrUL step in a PNA (from N-terminus to C-terminus) can cooperatively enhance PNA·dsRNA triplex formation. Incorporating an LBrU step in a PNA, however, may not stabilize PNA·dsRNA triplex formation (compared to an LT step). It is probably due to unfavorable stacking of an LBrU step with the overlap of partially negatively charged N1 nitrogen in L and Br atom in BrU, as implicated in our modelling studies (Figures 8 and 9). Incorporating XU residues into PNAs also results in an enhanced binding to ssRNAs but no binding to dsDNAs at near-physiological buffer conditions.
Our work will facilitate rational design of various chemically-modified bases into PNAs for the enhanced and selective recognition of RNA duplexes containing varied base pairs. For example, it is promising that one may design dsRNA-binding PNAs by incorporating BrU residues not immediately downstream of L residues, which are advantageous in selectively binding to dsRNAs over ssRNAs (see Figure 1B, E) (13,33,35,36). Our work provides the foundation for developing modified PNAs as chemical probes and therapeutic ligands for targeting functionally important RNA duplex structures including those found in miRNA hairpin precursors and viral RNAs.