Fluorobenzene Nucleobase Analogues for Triplex-Forming Peptide Nucleic Acids

Abstract

2,4-Difluorotoluene is a nonpolar isostere of thymidine that has been used as a powerful mechanistic probe to study the role of hydrogen bonding in nucleic acid recognition and interactions with polymerases. In the present study, we evaluated five fluorinated benzenes as nucleobase analogues in peptide nucleic acids designed for triple helical recognition of double helical RNA. We found that analogues having para and ortho fluorine substitution patterns (as in 2,4-difluorotoluene) selectively stabilized Hoogsteen triplets with U-A base pairs. The results were consistent with attractive electrostatic interactions akin to non-canonical F to H-N and C-H to N hydrogen bonding. The fluorinated nucleobases were not able to stabilize Hoogsteen-like triplets with pyrimidines in either G-C or A-U base pairs. Our results illustrate the ability of fluorine to engage in non-canonical base pairing and provide insights into triple helical recognition of RNA.

Graphical Abstract

2,4-Difluorobenzene as a nonpolar isosteric analogue of uridine nucleobase in peptide nucleic acid forms a Hoogsteen base triplet with adenosine-uridine base pair in double helical RNA. The stability and selectivity of modified triplets suggest that 2,4-difluorobenzene uses fluorine and C-H groups in non-canonical hydrogen bonding similar to natural uridine-adenosine-uridine base triplets.

Introduction

Nucleic acid folding and molecular recognition is driven by complex interplay of hydrogen bonding (Watson-Crick, Hoogsteen, etc., Figure 1) and stacking of heterocyclic bases, supplemented by hydrophobic interactions, and interactions with surrounding media (hydration). Chemists have been modifying nucleobases and sugar-phosphate backbone to study these interactions and to develop novel tools for controlling nucleic acid function and designing new therapeutic modalities.^[1] The importance of hydrogen bonding interactions have been studied using hydrophobic nucleobase analogues that cannot form hydrogen bonds.^[2] These studies have developed new tools for studying DNA polymerases^[3] and extended the genetic alphabet beyond the two Watson-Crick base pairs creating unnatural base pairs that function in vivo.^[4]

Figure 1.

Structures and hydrogen bonding patterns of Watson-Crick base pairs and Hoogsteen base triplets involving native nucleobases and fluorinated nucleobase analogues. Blue broken lines indicated potential (not experimentally proven) attractive electrostatic interactions.

Schweitzer and Kool designed 2,4-difluorotoluene (F, Figure 1) as a nonpolar isosteric analogue of thymidine.^[5] The UV melting studies showed that F strongly destabilized DNA duplexes, suggesting that F did not form Watson-Crick hydrogen bonds with adenosine or other natural DNA nucleobases.^[5] This notion was later confirmed in a crystal structure of A-F base pairs in RNA showing 3.92 and 3.78 Å (two independent duplexes) distances between fluorine and NH₂, which were significantly longer than similar distances in Watson-Crick base pairs (~2.9 Å).^[6] On the other hand, Egli and co-workers showed that in G-F wobble pairs in RNA,^[7] and in A-F Watson-Crick pairs in Dickerson-Drew Dodecamer DNA,^[8] the distances between fluorine and NH were 3.03 to 3.12 Å, which was in the range of typical hydrogen bonding distances (2.7 to 3.3 Å).^[7] The former study showed that in RNA G-F base pair was more stable than A-F and concluded that fluorine is engaging in a hydrogen bonding that stabilizes G-F wobble pairs.^[7] Hence, the ability of fluorine to accept hydrogen bond in nucleic acid base pairing appears to depend on the overall structural context. In the present study, we explored if fluorinated benzenes could serve as triplex-forming nucleobases (e.g., F2 in Figure 1) in peptide nucleic acid (PNA).

Our interest in novel PNA nucleobases was motivated by our recent discoveries that modified nucleobases, such as 2-aminopyridine (M, Figure 1), enabled strong and sequence selective triple helix formation with double-stranded RNA (dsRNA) at physiological pH and salt conditions.^[9] Combined with other heterocyclic modifications,^[10] triplex-forming M-modified PNAs showed biological activity by inhibiting mRNA translation^[11] and microRNA maturation^[12] in live cells. Others have also reported favorable RNA binding^[13] and promising biological activity of nucleobase modified triplex-forming PNAs.^[14] In addition, novel Janus-wedge triplets^[15] and other innovative heterocycles^[16] have been investigated for molecular recognitions of complex folded RNA and DNA.^[17] However, further development of novel heterocyclic nucleobases, especially for recognition of Hoogsteen faces of pyrimidine nucleosides, is required to expand the applications of PNA-dsRNA triplexes in biological systems.^[18]

In the present study, we tested five fluorinated PNA nucleobases in comparison to unsubstituted benzene. We found that F2, and the closely related F3, as nucleobase surrogates in triplex forming PNA preferred binding the Hoogsteen face of U-A base pair, though with relatively modest affinity and selectivity over other base pairs. Our results are consistent with some electrostatic stabilization possibly involving non-canonical hydrogen bonding interactions in the U-A•F2 and U-A•F3 triplets (Figure 1). We also tested if fluorinated benzenes could form base triplets that engage pyrimidines. We were hoping that fluorine may provide electrostatic stabilization with exocyclic N-H of cytosine (Figure 1); however, no selective stabilization was found for either G-C or A-U base pairs.

Results and Discussion

To study fluorinated benzene derivatives as nucleobase surrogates in PNA, we used the model system of four dsRNA hairpins (HRP1-HRP4, Figure 2) having a variable base pair in the middle of the stem, as in our previous studies on nucleobase modified PNAs.^[9–10] We synthesized six PNAs having modified nucleobases with various fluorine substitution patterns (PNAX, Figure 2) ranging from none (X = F0) to fully substituted pentafluorobeznene (X = F5). PNA monomers were synthesized using our previously reported methods starting from commercially available substituted phenylacetic acids (for details, see Supporting Information).^{[9–10, 19]}

Figure 2.

Sequences of RNA hairpins and PNA, and structures of the nucleobase analogues.

We evaluated the stability and specificity of base triplets involving fluorinated benzenes using UV thermal melting of the triple helices formed between model hairpins and modified PNAs. UV thermal melting is a common technique to measure stability of nucleic acid secondary structures; however, for triple helices the measurements may be complicated by overlap between the triplex to duplex and duplex to single strands transitions. Accurate measurements become difficult, if not impossible, if the base line between the two transitions is not well defined. To overcome this limitation, we took advantage of the unique absorbance of M nucleobase at 300 nm (Figures S6 and S7A). Because natural nucleobases have negligible absorbance at 300 nm (Figures S6 and S7A), the UV melting curves at this wavelength reported only on the melting of triplex with minimal, if any, residual signal from melting of the hairpins (Figure 3).

Figure 3.

Representative UV thermal melting curves of triplexes formed by RNA hairpins and PNAF2 recorded at 300 nm.

Overall, triplets formed by the benzene (F0) or various fluorinated benzene (F1-F5) nucleobase analogues exhibited decreased stability (Figure 4) compared to the canonical U-A•T (T_m ~ 69 °C) or C-G•M+ (T_m ~ 66 °C) observed in analogous sequences in our previous study.^[9c] The melting temperatures of triplexes ranged between 30 and 50 °C while hairpins HRP1-HRP4, as observed in our previous study,^[9a] melted at ~90 °C (see also Figure S7B). Benzene (F0) and fully fluorinated benzene (F5) were not able to selectively stabilize triplets with any of the variable base pairs (T_m ≤ 40.5 °C). This was not unexpected because of the uniform substitution pattern (all H or all F). F1 having a fluorine substituent para to the backbone attachment gave a similar result. Nucleobases F2 and F3 having fluorine substituents at ortho positions were somewhat more stabilizing and showed slight selectivity (T_m ~ 48 °C) for binding adenosine in HRP2. F4 was less stabilizing and less selective (T_m ~ 44 °C with HRP2), which was consistent with the expected ability of para fluorine to interact with the exocyclic amino group of adenosine (Figure 1). Interestingly, binding affinity and selectivity required both para and ortho fluorine substitutions. It is conceivable that the strongly electron withdrawing neighboring fluorines decreased the pK_a of meta hydrogen increasing its ability to participate in attractive electrostatic interactions.

Figure 4.

UV thermal melting results of triplexes formed by RNA hairpins and modified PNAs (averages of three measurements; standard deviations are illustrated with error bars).

The trends described above were most significant for engaging purine nucleobases in HRP1 and HRP2, while differences in binding uridine in HRP4 were less pronounced. Interestingly, all nucleobase analogues, including the unsubstituted benzene, had approximately the same binding affinity for cytosine in HRP3 (T_m ~ 40 °C). Contrary to our hypothesis, fluorine was not able to engage the exocyclic amino group of cytosine in stabilizing interactions. These observations were consistent with the well-known difficulty of triple helical recognition of pyrimidines.^[10b] The fluorinated nucleobases showed similar selectivity for the C-G base pair – PNAF2-PNAF4 were binding stronger to HRP1 (than other fluorinated nucleobases) albeit with lower affinity than to HRP2 (Figure 4). This result could be explained considering that the C-G•F2 mismatched triplet lacked the F to H-N attractive interaction but would not suffer from significant other repulsive interactions compared to the U-A•F2 triplet (c. f., U-A•F2 and C-G•F2 triplets in Figure 1). This result and explanation are also consistent with higher stability of C-G•T mismatched triplets (having similar H-bonding ability as C-G•F2) compared to other mismatched triplets observed in our previous studies.^{[9b, 9c]}

A previous study examined F1 and F5 as PNA nucleobases in PNA-DNA duplexes and found both acted as universal nucleobases with little discrimination in pairing with native nucleobases.^[20] The range of Tm values for F1 and F5 Watson-Crick base pairs with A, C, G and T was only 1.8 °C.^[20] In contrast, our study showed that fluorinated nucleobases were more discriminative in the context of Hoogsteen base triplets that spanned larger Tm ranges from 7.1 °C for F1 to 12.9 °C for F2 (Figure 4 and Table S2).

Conclusion

The ability of fluorine in nucleobase analogues to engage in non-canonical hydrogen bonding has been a matter of continuous debate. Early results by Kool and co-workers showed that 2,4-difluorotoluene (F) recognized adenosine using steric fit but did not form hydrogen bonds. Egli and co-workers found similar results for A-F pairing in RNA. However, in G-F wobble pairs the distance between fluorine and NH was in the range of hydrogen bond (~3.05 Å). The slight but notable stabilization of U-A•F2 (T_m = 48.9 °C, Figure 4) and U-A•F3 (T_m = 48.1 °C) triplets observed in this study suggested that in the structural context of Hoogsteen triplets, the fluorinated nucleobases may be able to engage in attractive electrostatic interactions akin to non-canonical F to H-N and C-H to N hydrogen bonding. Interestingly, F2 and F3 were also binding stronger (compared to other fluorinated nucleobases) to C-G base pairs, although the stability of these triplets (T_m ~ 42 °C) was significantly lower than those involving U-A base pairs. The fluorinated nucleobases were not able to stabilize Hoogsteen-like triplets with pyrimidines in either G-C or A-U base pairs. Our results provide another example of fluorine engaging in non-canonical electrostatic interactions to stabilize base pairing and provide insights into triple helical recognition of double helical RNA.

Experimental Section

RNA hairpins were purchased from Dharmacon. Prior to use, RNA hairpins were deproteced in accordance with manufacturer instructions and purified using reverse phase HPLC and a gradient of MeCN in 50 mM aqueous triethylammonium acetate buffer (pH 7). The purity of the final products was confirmed by reinjection in HPLC. UV-melting experiments were performed on Shimadzu UV-2600 or UV-1800 spectrophotometers equipped with TMSPC-8 temperature controllers. RNA hairpins were dissolved at 10 μM in a buffer containing 50 mM potassium phosphate, 2 mM MgCl₂, 90 mM KCl, 10 mM NaCl, pH 7.4, vortexed, centrifuged, and heated to 95 °C for 3–5 minutes. The solution was snap cooled to 4 °C and kept at that temperature for 3–5 minutes. The RNA hairpins were allowed to warm to room temperature and incubated for 10 min. The modified PNAs were then added at 10 μM to the sample and the mixture was vortexed and centrifuged again followed by incubation at room temperature for 10 min. Samples were transferred to an eight-cell cuvette. A temperature ramp rate of 1 °C/ min was used, typically from 20°C to 95°C, and the absorbance was monitored at 300 nm. Typical melting curves are shown in Figure 3; the experimental results are listed in Table S2. Three replicates were then used to determine average and standard deviation in melting temperature (Tm).