Screening of minimalist noncanonical sites in duplex DNA and RNA reveals context and motif-selective binding by fluorogenic base probes

Abstract

We hypothesize that programmable hybridization to noncanonical nucleic acid motifs may be achieved by macromolecular display of binders to individual noncanonical pairs (NCPs). As each recognition element may individually have weak binding to an NCP, we developed a semi-rational approach to detect low affinity interactions between selected nitrogenous bases and noncanonical sites in duplex DNA and RNA. A set of fluorogenic probes was synthesized by coupling abiotic (triazines, pyrimidines) and native RNA bases to thiazole orange (TO) dye. This probe library was screened against duplex nucleic acid substrates bearing single abasic, single NCP and tandem NCP sites. Probe engagement with NCP sites was reported by 100–1000X fluorescence enhancement over background. Binding is strongly context-dependent, reflective of both molecular recognition and stability: less stable motifs are more likely to bind a synthetic probe. Further, DNA and RNA substrates exhibit entirely different abasic and single NCP binding profiles. While probe binding in the abasic and single NCP screens was monotonous, much richer binding profiles were observed with the screen of tandem NCP sites in RNA, in part due to increased steric accessibility. In addition to known binding interactions between the triazine melamine (M) and T/U sites, the NCP screens identified new targeting elements for pyrimidine-rich motifs in single NCPs and 2×2 internal bulges. We anticipate that semi-rational approaches of this type will lead to programmable noncanonical hybridization strategies at the macromolecular level.

Graphical Abstract

We have developed a semi-rational approach to detect low affinity binding to noncanonical pair (NCP) sites in duplex DNA and RNA. A set of fluorogenic thiazole orange probes bearing abiotic (triazines, pyrimidines) and native RNA bases was screened against dsDNAs and dsRNAs with single abasic, single NCP and tandem NCP sites. Fluorogenic probe engagement revealed novel and selective base interactions with NCPs, setting the stage for a new strategy of targeting noncanonical motifs in RNA.

Introduction

Nucleic acid recognition^[1] at noncoding^[2] interfaces remains a frontier. Extensive studies have been performed on synthetic binders designed to target the minor and major^[3,4] groove interfaces of duplex DNA and RNA; in particular, the native and designed pyrrole-imidazole polyamide (Pyr-Im) systems^[5] studied by Dervan established rules for sequence selectivity directed by hydrogen bond donor/acceptor matching in the minor groove.^[5–8] Similarly, Benner and others^[1] have shown that functional and orthogonal synthetic base pairs may be prepared by shuffling Watson-Crick interfaces, yielding new pairs that are biophysically similar to native nucleic acids.^[9–11] Recently, high throughput screening (HTS) of small molecule libraries have been applied to noncanonical RNA folds,^[12–16] in analogy to the approach for protein targets. Though selected designed systems^[1,17–24] and RNA HTS^[13,25–27] have yielded many important findings, general principles for targeting non-duplex secondary structures remain elusive. We hypothesized that a semi-rational approach akin to fragment-based strategies^[28,29] that includes both design and screening could be useful to gain insight into targeting of individual and tandem noncanonical pairs (NCPs) within internal bulges.

Noncanonical motifs are high-value targets due to their importance to noncoding function^[30–32] and notably present a richer set of available hydrogen bonding interfaces relative to duplex structures due to their unpaired or weakly paired nature. If it were possible to identify recognition elements for individual NCPs and minimal noncanonical motifs, this could enable programmable noncanonical hybridization to internal bulge structures when binders are displayed together on a macromolecular scaffold. This recognition problem has been solved for the TT/UU noncanonical pair: melamine (M) forms a base-triple with two T/U bases.^[33] Bifacial peptide nucleic acids (bPNAs) exploit this base triple to hybridize with oligo T/U internal bulges,^[34–39] forming hybrid stems that can be compatible^[40–44] or competitive^[45] with native nucleic acid function.

We speculate that the sequence scope of bPNA triplex hybridization could be expanded by the use of other synthetic and native bases that bind to NCPs other than TT/UU, independently or in conjunction with melamine. Notably, the melamine base itself only binds weakly to DNA,^[46] but can generate low nanomolar affinity binding when displayed multivalently on bPNAs, or micromolar affinity when coupled to an acridine intercalator.^[47] We have developed herein a fluorescence-based screen to facilitate the discovery of new NCP binding bases that is sufficiently sensitive to detect weak binding interactions, thus avoiding the higher synthetic cost of monomer and bPNA preparation. Synthetic and native bases were transformed into NCP probes by coupling to thiazole orange dye. The thiazole orange derivatives studied herein have low background affinity to dsDNA/RNA and low emission in solution, but we find that pendant nitrogenous bases with weak affinity to NCPs in duplex nucleic acids can trigger intercalation, resulting in increased emission. Similar sequence-directed “forced intercalation” and emission has been reported with TO-coupled Py-Im minor groove binders^[48] and PNAs.^[49–51] With these prior studies in mind, a TO-probe library was prepared with native RNA bases, symmetric triazines, and selected synthetic pyrimidines; each base has been shown^[46,52–56]—or has the potential—to form hydrogen bonding interactions with NCPs. In addition to the known native pairings and melamine triple with T/U, among this library is W1,^[57] an amino pseudoisocytidine shown by McLaughlin^[58,59] to form a base-triple with T/C in DNA. Three classes of duplex substrates of increasing complexity were designed and tested: 1) single abasic sites, 2) single NCP sites, 3) tandem NCP sites. Single NCP (1×1 internal bulge) libraries in DNA and RNA were prepared with all 12 non Watson-Crick (WC) pairs. The larger space of tandem NCPs was scanned using a dsRNA library of 43 unique tandem NCPs (2×2 internal bulge); though not comprehensive, this set provides good coverage of purine/pyrimidine usage, excluding WC pairs. Each group was designed to reveal the step-wise impact of changing context on molecular recognition and point the way to a general design strategy for binding NCPs. Importantly, probe screens identified the important role of NCP stability in the inherent probe-reactivity of particular bulge sequences and highlighted stark differences in DNA vs RNA binding. Further, purine probes and adenine-containing sites were conspicuously unreactive, suggesting the sensitivity of this approach to steric demand. Though the context-dependent nature of NCP targeting complicates a difficult problem, the RNA-substrate driven binding trends revealed herein indicate that simple calculations of fold energetics can be useful when choosing an RNA target (Figure 4). Most importantly, the screens have identified ammeline, W1 and cytidine as promising recognition elements for RNA NCP-targeting that could work in conjunction with the known U-site targeting function of melamine.

Figure 4.

(Top) Heat map of tandem NCP RNA library screen. Tandem NCPs are ordered by calculated relative stability, with the most stable on the left. Probes are ordered with the highest integrated fluorescence (most reactive) in the top row and lowest in the bottom row. Fluorescence indicated is relative enhancement over background TO emission. (Bottom) Probe emission data grouped by XQ/WZ size ratio in the 2×2 internal bulge. Size ratio is calculated using purine or pyrimidine molecular weight as an approximation for each base in the tandem NCP. Schematic illustration of the size ratio is indicated above the green bars highlighting the most reactive probe in each grouping. The most reactive probes in each group are labeled accordingly. Samples were prepared in HEPES buffer (pH 7.5, 50 mM, 100 mM NaCl) with 4 μM probe, 2 μM RNA duplex. Data is the average of triplicate measurement, with histograms and error bars available in Supporting Information.

Results and Discussion

Probe synthesis.

We synthesized a library of 11 NCP thiazole orange (TO) probes bearing adenine (A), guanine (G), cytidine (C), uridine (U), melamine (M), ammeline (N), ammelide (D), cyanuric acid (CA), 6-aminopseudouridine (APU), 6-aminopseudoisocytidine (W1) and two linker variations (R1, R2) (Figure 1). The 4 native bases (A, G, C, U),^[60–62] cyanuric acid (CA), and the synthetic pyrimidines (APU, W1)^[58] were alkylated directly to form the acetate, then amidated with Boc-ethylenediamine. The symmetric triazine probes (M, N, D) were all prepared from alkylation of mono-Boc 1,4-diaminobutane with cyanuric chloride, followed by partial reaction with ammonium hydroxide or hydrolysis of the remaining chloride sites. Following Boc cleavage, the amines were acylated with thiazole orange^[63] to yield the R1 (A, G, C, U, APU, W1, CA) or R2 (M, N, D) linker family of probes; R1 and R2 linkers are identical in length (Figure 1). An ammeline probe with the R1 linker (N1) was also prepared to examine linker effects. All probes exhibited negligible emission in solution without RNA (Supporting Information).

Figure 1.

Structures of the fluorogenic probes used in this study. All probes made based on literature procedures to install the R₁ and R₂ linkers indicated. Ammeline was synthesized with both R₁ and R₂ linkers to produce N1 and N2, respectively.

Abasic sites.

Single T and C abasic sites in DNA exhibited the expected binding with complementary A and G probes, albeit with modest signals (Figure 2). Much stronger binding signals were observed with M and W1 to T-abasic and T/C-abasic sites, respectively, confirming prior selectivity studies on these synthetic bases. Among the synthetic bases not previously examined, APU showed weak binding to T/C abasic sites and D showed a modest binding to G-abasic; all binding patterns can be rationalized through complementary hydrogen bonding patterns. Of the new bases, ammeline showed the strongest binding as well as a significant linker effect, with N2 binding to G and C abasic sites while N1 bound to G, C and T abasic. Though N and W1 have tautomers with the same hydrogen bonding patterns, their binding profiles are distinct, even when the same linker (R1) is used for both (N1, W1). Notably, the A-abasic site did not accept any probe, and pyrimidine-abasic sites were overall much more probe-reactive. Remarkably, though there were general similarities, RNA binding profiles were significantly different to those in DNA. Like in DNA, A-abasic sites refused all probes and strong M, N selectivity for U/C abasic sites was observed. However, other selectivity patterns were diminished and blurred: while W1 elicited the most intense integrated binding signals in DNA, it was not a stand-out binder in RNA. Furthermore, while probe turn-on reached ~1000X over background in DNA, the range in RNA was about half, suggesting that NCPs in dsRNA are generally less probe-accessible than in dsDNA.

Figure 2.

Heat map of fluorogenic binding for the abasic (Left) DNA and (Right) RNA) libraries. Fluorescence is reported as average intensity from triplicate measurement relative to unbound probe. Samples were prepared in HEPES buffer (pH 7.5, 50 mM, 100 mM NaCl) with 2 μM probe, 2 μM duplex. See histogram analysis with error bars in Supporting Information.

Single NCP sites.

The general trends seen in the abasic library played out the single NCP libraries: probe emission with DNA was ~7X more intense than in RNA and triazine probes M-R2 and N-R2 were again the best binders, targeting T/U and C containing sites. Pyrimidine-pyrimidine (YY) NCPs were clearly more accepting of base probes while purine-purine (RR) and pyrimidine-purine (YR/RY) sites were generally unresponsive. Single mismatch stability likely plays a role, with GU the most stable due to wobble pairing and CU/UC/CC mismatches are among the least stable; among YY pairs, UU is the most stabilized by wobble pairing.^[64] Probe W1 again showed affinity for CC, TT and TC/CT sites, consistent with McLaughlin’s prior observations; binding was muted in RNA as seen in the abasic library. The affinity of melamine and ammeline for T/U sites was affirmed, with the affinity for mixed CT and TC sites perhaps indicative of single base pairing rather than base-triple formation. Again, W1 and N1 exhibited distinct behavior, particularly in DNA, where W1 was a strong binder to all YY sites, while N1 did not exhibit clear binding to any site. However, N2 showed good binding to all YY sites except CC. These observations underscore a significant linker effect as well as a fundamental difference between pyrimidine and triazine bases despite similar hydrogen bonding tautomers. In addition, the ~7X fold advantage in probe brightness in DNA over RNA is likely reflective of steric barriers to probe insertion presented by A-form dsRNA, which is more tightly wound than B-form dsDNA. The abasic DNA and RNA libraries are more similar in probe emission intensity, possibly because the abasic site is more inherently accessible than a site with two bases; further, the lack of probe reactivity in purine-abasic, RY and RR sites relative to YY sites is consistent with sterically blocked purine sites. We speculate that probes approach NCPs from the major groove in DNA and from the minor groove in RNA, as these are sterically the most accessible paths in B and A-form helices, respectively. The altered probe binding profiles would thus reflect the distinct hydrogen bonding interfaces offered upon approach from either groove (Figure 3). One would expect similar melamine binding to TT and UU sites due to the symmetry of both the pyrimidine faces and the triazine itself, while W1 and N would encounter significantly different interactions at the major vs minor grooves. Notably, while W1 binding to TC sites in DNA has been studied,^[58,59] binding to RNA has not been previously reported; in our assay, W1 exhibits no binding to the single UC or CU site in dsRNA (Figure 3).

Figure 3.

(Top) Fluorescence heat maps of NCP-probe binding. Average emission intensities from triplicate measurement are shown according to the scales at right, indicating fold enhancement over unbound probes for DNA (Left) and RNA (Right). Probes and NCPs in both maps are ordered with the highest integrated relative fluorescence signal in DNA at the top and right, respectively. (Bottom) Possible interactions of synthetic bases melamine (M), ammeline (N) and W1 with NCPs approached from the (above) major groove and (below) minor groove. Samples were prepared in HEPES buffer (pH 7.5, 50 mM, 100 mM NaCl) with 2 μM probe, 2 μM duplex.

Tandem NCP sites.

We focused our subsequent studies on tandem NCP sites in RNA only, as RNA is more likely to contain noncanonical motifs. A library of tandem NCPs in dsRNA, forming 2×2 internal bulges (Scheme 1) was annealed. It was anticipated that 2×2 bulges would be more dynamic than 1×1 bulges and would potentially be more accessible for binding at one or both NCPs, thereby providing a new structural context for probe binding. A limited set of 43 unique 2×2 bulges in dsRNAs was prepared to generally assess the reactivity of RR/RR, RR/RY, RY/RY, RR/YY, YY/YR and YY/YY NCP sites towards our probe library. In contrast to the monotonic output from the single NCP substrates, the tandem NCP library yielded an information-rich set of profiles (Figure 4) that indicated context-dependent binding properties (Figure 3). Not only were probes with broad scope binding identified, we found internal bulges that accepted many probes; in particular, AG/CU could be targeted by W1, C, N1, APU, CA and U probes. Notably, C, U, APU and CA did not show any binding in the abasic or single NCP screens, suggesting that tandem NCP motifs are significantly more accessible and/or reactive. We speculate that motifs such as AG/CU are probe-reactive due to the relative instability of the internal bulge such that probe binding provides an energetic benefit. For the specific bulge AG/CU, CU is a particularly destabilizing mismatch while AG is moderately stabilizing; moreover, a size mismatch between RR and YY is destabilizing due to poor stacking overlap.^[65] It is reasonable that this relatively unstable motif would present binding opportunities complementary all four native bases, resulting in broad scope of binding among the probes studied. When the tandem NCPs were scored according to size matching of the stacked pairs, we found the strongest probe binding with the largest size mismatch (RR/YY and YY/RR). Generally, probes did not bind well to size matched pairs (eg-RY/RY, YY/YY, RR/RR) (Figure 3). An exception is N2 binding to UC/CU and UC/CC, which suggests selectivity of N for UC pairs (Figure 4). However, for reasons not fully understood, N2 did not show binding to UC/CU, which has the lowest stability as calculated with bifold.^[66] It appears that fold energetics provide a backdrop to site reactivity, which remains directed by the base recognition element (Figure 4, Table S2.4). For instance, UU/UU did not accept any probes, despite the fact that bPNAs are well-established to bind to UU sites in RNA with M bases; however, M showed a strong fluorescence response with UU/GG. We rationalize this observation by noting that the UU/UU bulge is highly stabilized by wobble pairs and any benefit of melamine binding would be diminished by the loss of existing structure.^[67] In contrast, the YY/RR size mismatch in UU/GG renders the motif unstable, leading to a potential energetic benefit upon expansion of the UU site to a UMU triple; this would correct the size mismatch and increase base-stacking against GG. Interestingly, RR pairs have been observed to bridge the structural transition between G-quadruplex or base-triple domains and duplex domains.^[68,69]

Scheme 1.

Illustration of the three nucleic acid library designs targeted by the fluorogenic probe library. (Left) The abasic library in DNA and RNA pairs a purine (R) or pyrimidine (Y) with an abasic site (Ap) wherein the anomeric carbon is replaced with CH₂; the single noncanonical pair (NCP) library contains all non WC pairings in DNA and RNA; the tandem NCP library in RNA only contains 43 unique motifs of two non WC pairs in tandem. (Right) Illustration of fluorogenic probe binding to the noncanonical sites.

While melamine and ammeline were good binders, overall weak binding was observed with the related triazines ammelide and cyanuric acid, possibly due to deprotonation of ammelide and cyanuric acid (pK_As ~6.9) while ammeline and melamine are neutral at pH 7.^[70] Anionic bases impose a protonation cost to binding^[54] and are particularly detrimental to base recognition;^[71] however, no changes in binding profiles were observed at pH 4–6.5 (Supporting Information). Similarly, while W1 (6-aminopseudoisocytidine) exhibited many NCP interactions, APU (6-aminopseudouridine) showed nearly no binding. It is notable that the APU and W1 are the pyrimidine analogs of ammelide and ammeline, respectively, and their reactivity patterns are nearly identical, suggesting that replacement of the exocyclic amine in ammeline and W1 with a keto group in ammelide and APU could be causative in loss of NCP binding, possibly due to introduction of carbonyl lone pair repulsions at the NCP interface.

Three tandem NCP sites in dsRNA were selected for careful analysis of affinity: UU/GG, UC/CU and AG/CU. These sites have a clear preferences for M, N2 and W1 probes, respectively (Figure 4). Using fluorescence readout, binding curves for all three NCP sites and all three probes (M, N2, W1) were generated in triplicate and fit to a 1:1 binding model (Figure 5). The good fit and low standard deviation supported the 1:1 binding model (Supporting Information), and importantly, substantiated the triplicated single concentration measurement data in the screen. As expected, M, N2 and W1 were the best binders for UU/GG, UC/CU and AG/CU, respectively (Table 1). Notably, probes that gave a weak RNA binding signal over free probe background (0–50X turn-on) could not saturate binding.

Figure 5.

Binding isotherms of M, W1 and N2 probes binding to the indicated tandem NCPs, obtained by fluorescence (RFU) as a function of RNA (duplex) concentration, showing fit to a 1:1 binding model and standard deviation error bars from triplicate measurements. RNA substrates and probes were the same as in Figure 4.

Table 1.

Probe affinity to selected tandem NCPs^a

Probe	Kd (μM)
	UU GG	AG CU	UC CU
M	5.7 (1.3)	>1000	>1000
N2	>1000	>1000	33.5 (9.6)
W1	28.4 (7.2)	8.9 (2.4)	>1000

^aStandard deviation error from triplicated measurements shown in brackets. Non-saturating binding curves shown as >1000 μM K_d. Probe concentration was 500 nM

Conclusions

The observations herein indicate that probe binding is influenced by base recognition, NCP stability and accessibility, and underscore the fact that DNA and RNA targeting are entirely distinct problems. We speculate that differences in labeling profiles stem from greater NCP accessibility from the B-form major groove of DNA helices than the A-form minor groove of RNA helices; the probes thus encounter distinct steric challenges and hydrogen bonding interfaces. The abasic and single NCP screens validated expectations regarding hydrogen bond complementarity and also revealed steric limitations of targeting single base pair anomalies with the current probe design. We were both surprised and gratified to observe a broader range of probe reactivity in the tandem NCP library, which we attribute to increased accessibility and energetic driving force for probe binding. Thus, we find a greater diversity of binding selectivities, even among probes that were silent in the abasic and single NCP screens. In each set of screens, M, N and W1 exhibited the strongest binding signals to pyrimidine NCPs (TT, TC, CT, UU, UC, CU). Though apparently overlapping in substrate, the probes demonstrated selectivity in the tandem NCP screen, with M uniquely targeting the UU/GG motif and N uniquely targeting UC/CU and UC/CC motifs, respectively. Non-overlapping targeting is critical for broadening the scope of noncanonical hybridization and the ammeline base appears to be a useful tool to apply in conjunction with melamine. Though silent in the RNA single NCP screen, synthetic pyrimidine W1 has a wide array of substrates in the tandem NCP screen. The RNA targeting profile of W1 has not been previously reported and was found to be surprisingly rich in the tandem NCP screen; W1 could be applied as a “universal base,” capable of binding many different RNA motifs. Likewise, the AG/CU bulge may be a universally accepting motif due to its reactivity with many different probes. Despite the relative promiscuity of the AG/CU internal bulge, it is notable that neither M nor N bind well to this site, suggesting that macromolecular display of a combination M, N and W1 could potentially bind to an internal bulge containing UU, CU and AG NCPs in tandem. The C-probe exhibits a similar profile to W1 (Figure 4), despite different hydrogen bonding patterns in the neutral base. Interestingly, protonated C (C+) shares the same pattern with W1 so it is possible that this is the active targeting base; however, limited screens under acidic conditions did not enhance or alter binding patterns (Supporting Information). Thus, the synthetic and native base probes studied herein have revealed 4 potentially useful recognition elements (M, N, W1, C) for noncanonical hybridization of macromolecular reagents to RNA internal bulges, and studies to incorporate these bases into bPNAs are ongoing. Taken together, these data show that a step-wise, design-based approach can provide insight and complement other strategies for targeting noncanonical RNA sites with synthetic reagents.

Experimental Section

Materials and Instrumentation

All chemicals were used without further purification from commercial sources, unless otherwise noted. DNAs and RNAs were purchased from Integrated DNA Technologies (IDT). Nucleic acid strands shorter than 20 nt were used without further purification while longer DNAs/RNAs were purified by TBE-urea denaturing gel. SYBR^Ⓡ gold was purchased from Thermo Fisher Scientific. DNA stock solutions were serially diluted in Milli-Q (18 MΩ) water and concentrations were determined by solution absorbance (260 nm) with a Thermo Fisher Nanodrop 2000. Sample fluorescence was measured on Thermo Fisher Nanodrop 3300. Buffer for fluorescence turn-on (1X) contained: HEPES (pH=7.5) 50 mM, NaCl 100 mM. Analytical and semi-preparative HPLC was carried out with C₁₈ reverse phase columns with the following HPLC solvents: solvent A (99% Milli-Q H₂O, 1% HPLC grade acetonitrile, 0.1% trifluoroacetic acid) and solvent B (10% Milli-Q H₂O, 90% HPLC grade acetonitrile, 0.07% trifluoroacetic acid).

General procedure for R1 probe synthesis.

The procedure for the cytidine probe C is provided as a representative protocol, see Supporting Information detailed synthetic schemes. A Boc-protected thiazole orange derivative (17a, 9.8 mg, 0.02 mmol, 1 equivalent) was added to trifluoroacetic acid (1 ml) and the reaction was stirred for 4 hr at room temperature. Then TFA was removed by N₂ stream and the residue 17b was dissolved in DMF (1.5 ml) and treated with N,N-diisopropylethylamine (20 mg, 0.2 mmol, 10 equivalents) and cytidine acetate (5.1 mg, 0.03 mmol, 1.5 equivalents). After dissolution, HBTU (12 mg, 0.03 mmol, 1.5 equivalents) was added to the mixture and the reaction was stirred overnight. The reaction was concentrated in vacuo and by N₂ stream and the residue was resuspended in solvent A (90% acetonitrile, 10% water, 0.1% TFA), centrifuged and the supernatant was purified by HPLC. The low solubility of the R1 probes (<0.2 mg/ml) limited NMR characterization and HPLC-ESI was used to characterize the TO probes.

General procedure for R2 probe synthesis.

The procedure for melamine probe M is provided as a representative protocol, see Supporting Information detailed synthetic schemes. A acetate thiazole orange derivative (18a, 7 mg, 0.02 mmol, 1 equivalent), N²-(4-aminobutyl)-1,3,5-triazine-2,4,6-triamine (5.9 mg, 0.03 mmol, 1.5 equivalents) and N,N-diisopropylethylamine (10 mg, 0.1 mmol, 5 equivalents) were added to DMF (1.5 ml). Following dissolution, HBTU (12 mg, 0.03 mmol, 1.5 equivalents) was added and the reaction was stirred overnight at room temperature. Solvent was removed by N₂ stream and the residue was resuspended in solvent A, centrifuged and the supernatant was purified by HPLC. The low solubility of the R2 probes (<0.2 mg/ml) limited NMR characterization and HPLC-ESI was used to characterize the TO probes.

Fluorescence measurements

Samples for screening were prepared as solutions with 50 mM HEPES (pH 7.5), 100 mM NaCl, 2 μM dsDNA/dsRNA, 2 μM probe (4 μM probe with tandem mismatch). Prior to use, nucleic acids were annealed (95^oC/5 min, slow cool to RT over 2 h) to pre-form the duplex. Prepared samples were incubated at room temperature for 30 min before measuring fluorescence (Thermo Fisher Nanodrop 3300, excitation/emission=470/522 nm). For each measurement, 2 μl of sample was applied to the center of the nanodrop probe. All fluorescence values are the average of triplicate measurements, starting from fresh sample preparation. Error bars indicate standard deviation. The relative intensity (turn-on) was calculated with the equation:

$$\text{relative}\text{F}=\frac{RFUofprobewithDNA/RNA}{RFUofprobeonly}$$

Binding affinity

Probes (M, N2, W1) were held at a constant concentration of 500 nM and treated with RNA to final concentrations from 0 to 20 μM in buffer (50 mM HEPES, pH 7.5, 100 mM NaCl). The experiments were triplicated from sample preparation and error bars indicate standard deviation. The data were fit with the following equation to obtain dissociation constant K_d:

$$\text{RFU}=\frac{m\text{∗}\left(Kd+x+500\right)-m\text{∗}sqrt\left[sqr\left(Kd+x+500\right)-4\text{∗}x\text{∗}500\right]}{2}$$