NMR chemical exchange measurements reveal that N⁶-methyladenosine slows RNA annealing

Abstract

N⁶-methyladenosine (m⁶A) is an abundant epitranscriptomic modification that plays important roles in many aspects of RNA metabolism. While m⁶A is thought to mainly function by recruiting reader proteins to specific RNA sites, the modification can also reshape RNA-protein and RNA-RNA interactions by altering RNA structure mainly by destabilizing base pairing. Little is known about how m⁶A and other epitranscriptomic modifications might affect the kinetic rates of RNA folding and other conformational transitions that are also important for cellular activity. Here, we used NMR R_1ρ relaxation dispersion and chemical exchange saturation transfer to non-invasively and site-specifically measure nucleic acid hybridization kinetics. The methodology was validated on two DNA duplexes and then applied to examine how a single m⁶A alters the hybridization kinetics in two RNA duplexes. The results show that m⁶A minimally impacts the rate constant for duplex dissociation, changing k_off by ~1-fold but significantly slows the rate of duplex annealing, decreasing k_on by ~7-fold. A reduction in the annealing rate was observed robustly for two different sequence contexts at different temperatures, both in the presence and absence of Mg²⁺. We propose that rotation of the N⁶-methyl group from the preferred syn conformation in the unpaired nucleotide to the energetically disfavored anti conformation required for Watson-Crick pairing is responsible for the reduced annealing rate. The results help explain why in mRNA, m⁶A slows down tRNA selection, and more generally suggest that m⁶A may exert cellular functions by reshaping the kinetics of RNA conformational transitions.

Graphical abstract

N⁶-methyladenosine (m⁶A) is an abundant reversible epitranscriptomic modification found in coding and noncoding RNAs^1–4. It plays important roles in RNA metabolism^5–8 and is implicated in a growing number of cellular processes^9–15. While the modification is thought to primarily exert its function by recruiting reader proteins to specific RNA sites, it can also reshape RNA-RNA and RNA-protein interactions by modulating RNA structure^16–20. A single m⁶A destabilizes RNA duplexes by 0.5–1.7 kcal/mol^21–22, enhancing binding to single-stranded RNA (ssRNA) binding proteins¹⁶. m⁶A destabilizes A-U base pairs (bps) because hydrogen bonding requires that the N⁶-methyl group adopts the energetically unfavorable anti conformation^21–22 (Figure 1).

Figure 1. N⁶-methyl adenosine (m⁶A) destabilizes m^6-A-U pairing and RNA duplexes.

The methyl group has to adopt an anti conformation to form the Watson-Crick H6--O4 hydrogen bond but this leads to unfavorable steric contacts with N7.

The activities of many RNAs also depends on the kinetic rates of folding, protein-RNA, RNA-RNA, and RNA-ligand association/dissociation and conformational transitions^23–29. Surprisingly little is known about how m⁶A and other epitranscriptomic modifications impact these kinetic properties of RNA. Compelling evidence for such a kinetic effect comes from a study showing that in mRNA, m⁶A slows down tRNA selection during translation²⁰. Here, we developed an approach based on NMR spin relaxation dispersion (RD) in the rotating frame (R_1ρ)^30–32 and Chemical Exchange Saturation Transfer (CEST)^33–34 to site-specifically and non-invasively measure hybridization kinetics of nucleic acid duplexes and then used the approach to examine how a single m⁶A impacts RNA duplex hybridization kinetics.

The melting and annealing of RNAs occurs in a wide variety of biochemical reactions^{27, 35}. Relative to other methods for studying hybridization kinetics^36–45, the NMR approach does not require a potentially perturbing label, which could obscure the impact of a small chemical modification, and kinetics can be measured at atomic resolution^{32, 46} to enable characterization of any intermediates that may form at the modified site.

We first evaluated the R_1ρ RD methodology on DNA duplexes whose hybridization kinetics has been extensively characterized previously^{38, 41, 44–45, 47–51}. R_1ρ RD relies on measuring the exchange contribution (R_ex) to transverse spin relaxation (R₂) due to chemical exchange between a major ground-state (GS) and a low-abundance and short-lived ‘excited-state’ (ES)^52–53.

Prior R_1ρ studies on RNA and DNA duplexes were carried out at temperatures below the melting temperature (T_m)^{46, 54–56}. Under these conditions, the population (p_ss) of the single-stranded (ss) species falls below detection (< 0.1%)³¹, enabling studies of bp dynamics. For example, at T=25°C, the R_1ρ profiles measured for various sites in the A₆-DNA duplex^{55, 57} (T_m~51°C and [A₆-DNA] ~ 0.9 mM) reflect exchange between a major Watson-Crick GS and minor Hoogsteen ES⁵⁵ (Figure 2A, 2B, S1). There is no evidence for a transient ss species, which is estimated to have a p_ss ~ 0.1% based on UV melting experiments (Table S1).

Figure 2. Site-specific characterization of A₆-DNA hybridization kinetics using NMR R_1ρ RD.

(A) The A₆-DNA duplex. Δω = ω_ES – ω_GS obtained from global fitting of the R_1ρ RD profiles is color-coded on each atom. Sites which are not colored indicates that no measurements were done. (B) Off-resonance R_1ρ (¹³C) RD profiles measured in A₆-DNA at 25°C (left) and 45°C (right). T9(C1′) RD at 25°C were reprinted by permission from⁵⁸. Buffer conditions were 25 mM NaCl, 15 mM sodium phosphate, 0.1 mM EDTA and 10% D₂O at pH 6.8. (C) The site-specific k_off values obtained from 2-state fitting of the R_1ρ RD profiles measured for A₆-DNA at 45°C. (D) Comparison of Δω_ES-GS = ω_ES – ω_GS measured by RD with Δω_ss-ds = ω_ss – ω_ds values obtained from the major and minor resonance observed in 2D [¹³C,¹H], [¹⁵N,¹H] and [¹⁵N, ¹³C] HSQC spectra of A₆-DNA at 45°C.

Based on simulations⁵⁰, increasing the temperature so that p_ss>1.0% should bring hybridization kinetics within R_1ρ detection (Figure S2). Indeed, the R_1ρ profiles for A₆-DNA changed when increasing the temperature to T=45 °C (p_ss~10%). RD is now apparent at A16(C2) and T9(C1′), which are otherwise flat at T=25°C (Figure 2B). A single peak was observed in all cases consistent with two-state exchange (GS⇌ES). Fitting the R_1ρ data to a 2-state exchange model yielded very similar k₁ = k_off (differences < 2-fold; k_off is the rate constant for dissociation) for different sites as expected for concerted melting and annealing of the duplex (Figure 2C). This is in stark contrast to Hoogsteen exchange at T=25°C, in which k₁ varies 50-fold across sites reflecting sequence-specific differences in bp dynamics⁵⁹. The ES chemical shifts measured for various sites were also in excellent agreement with those measured for the isolated ss, confirming that the ES is the ss species (Figure 2D, S3).

In the ‘zip-up’ model^{48, 60}, DNA annealing proceeds through a slow nucleation step followed by a fast zipping step occurring on the ns-μs timescale which is too fast for RD detection. Since the Hoogsteen exchange at higher temperatures is likely too fast for RD detection, ‘all-or-nothing’ behavior is observed with strands either being fully annealed or fully unzipped. These results establish the utility of R_1ρ RD to measure hybridization kinetics in DNA duplexes with site-specific resolution.

The backward rate constant k_-1 = k_on × [ss] (k_on is the rate constant for duplex annealing) was ill-defined when fitting the R_1ρ RD data (Figure S4). Such a degeneracy is expected when the exchange is slow on the NMR timescale and when using spin lock powers (ω₁) in the R_1ρ experiment that exceed the exchange rate (k_ex = k₁ + k_-1)^61–62. Indeed, in the slow exchange limit, the line broadening of the GS resonance only depends on the forward rate. To address this degeneracy, we used CEST experiments which can employ much lower spin locking fields more suitable for characterizing systems in slow exchange^33–34. CEST relies on measuring the resonance intensity of the GS as a function of the power and offset of an applied weak radio frequency (rf) field. At T=45°C, the CEST profiles for A₆-DNA revealed a dip at the chemical shift of the ss ES (Figure 3A, 3C, S5). Fitting the CEST profiles allowed the reliable determination of all exchange parameters including k_on (Figure S4), resulting in values (Figure 3B) that are in good agreement with those previously reported values for similar DNA duplexes^{45, 50}.

Figure 3. Site-specific characterization of hybridization kinetics using CEST.

(A) ¹³C CEST profile for G11(C8) measured in A₆-DNA at 45°C. (B) Comparison of Δω_ES-GS, k_off and k_on values obtained from R_1ρ and CEST (fits of the R_1ρ profiles were preformed fixing p_ss to the value measured using CEST). Buffer conditions were 25 mM NaCl, 15 mM sodium phosphate, 0.1 mM EDTA and 10% D₂O at pH 6.8. (C) The sequence of A₂-DNA and A₆-DNA. Δω = ω_ES – ω_GS obtained from CEST fitting is color-coded on each atom. (D) ¹³C CEST profiles for G11(C8) measured in A₂-DNA and A₆-DNA at 50°C. (E) Comparison of k_on and k_off values measured for A₂-DNA (red) and A₆-DNA (green).

Fixing p_ss to the CEST determined value, the R_1ρ RD profiles could be satisfactorily globally fitted (Figure S5), yielding exchange parameters (k₁ = k_off, k_-1 = k_on × [ss] and Δω_ES-GS that are in excellent agreement with the CEST derived values (Figure 3B, Table S3, S4). This mutual consistency further supports the validity of the approach. Finally, we further evaluated the CEST methodology by comparing the hybridization kinetics of A₆-DNA with another A₂-DNA duplex, which has higher stability (T_m~60°C and [A₂-DNA] ~ 0.8 mM) (Figure 3C, 3D). Consistent with prior studies^47–49, the two duplexes have similar k_on values but k_off is 20-fold faster for the less stable A₆-DNA duplex (Figure 3E).

Next, we applied the methodology to examine how m⁶A impacts hybridization kinetics in an RNA duplex containing the most abundant m⁶A consensus sequence (GGACU) in eukaryotic mRNA^1–2 (T_m~80°C and [dsGGACU] ~ 0.7 mM with Mg²⁺). In canonical RNA duplexes, there are no contributions from Hoogsteen exchange or any other process as verified for Watson-Crick bps in a variety of sequence and structural contexts⁵⁴. However, since m⁶A could induce local melting of the duplex, it was important to carry out measurements on the m⁶A residue itself. To this end, two dsGGACU duplexes were chemically synthesized containing ¹³C2/C8 labeled m⁶A or A near the center of the duplex (Figure 4A, S1, S6) (see methods). m⁶A destabilized the dsGGACU duplex by ~1 kcal/mol (Table S1), consistent with prior studies^21–22.

Figure 4. Measuring the impact of m⁶A on dsGGACU hybridization kinetics using CEST.

(A) The dsGGACU sequence. Δω = ω_ES – ω_GS obtained from global fitting of CEST is color-coded on each atom. (B) ¹³C CEST profiles measured for A6 in unmodified (left, green) and m⁶A modified (right, red) dsGGACU at 65°C in the presence of 3 mM Mg²⁺ (profiles in the absence of Mg²⁺ are shown in Figure S7). Buffer conditions were 25 mM NaCl, 15 mM sodium phosphate, 3 mM Mg²⁺, 0.1 mM EDTA and 10% D₂O at pH 6.8. (C) Comparison of Δω_ES-GS = ω_ES – ω_GS measured by CEST with Δω_ss-ds = ω_ss – ω_ds values obtained from the major and minor resonance observed in 2D [¹³C,¹H] HSQC spectra of dsGGACU with (red) and without (green) m⁶A at 65°C. (D) Comparison of k_on and k_off measured for unmodified (green) and m⁶A modified (red) dsGGACU.

The CEST and R_1ρ profiles for both unmodified and modified dsGGACU duplex at T=65°C revealed a single peak/dip consistent with 2-state exchange (Figure 4B, S7). However, the profiles for the modified duplex differed markedly from its unmodified counterpart (Figure 4B, S7). In both cases, global fitting of the CEST and R_1ρ data yielded ES chemical shifts that are in excellent agreement with those measured for the isolated ss (Figure 4C, S7, S8). Fitting the CEST data revealed that m⁶A changes k_off by 0.7–1.7 fold but decreases k_on by 4–9 fold (Figure 4D, S7). This m⁶A induced slowdown of annealing was observed robustly with or without Mg²⁺ (Figure 4D, S7), for a different sequence derived from Hepatitis C virus (HCV)¹⁵ (T_m~76°C and [dsHCV] ~ 0.7 mM with Mg²⁺) (Figure S1, S7), at a higher concentration of monovalent ions (Figure S7), and when using the R_1ρ RD data (Figure S7).

When unpaired, the N⁶-methyl group favors the syn conformation, while the anti conformation required for Watson-Crick pairing and duplex annealing is unfavorable with an estimated population of ~5%⁶³. Rotation of the N⁶-methyl group is likely responsible for the reduced annealing rate. Mismatches have also been shown to reduce k_on by up to 50-fold^{27, 64} through mechanisms that are not fully understood. Further studies are needed to dissect the kinetic mechanism by which m⁶A slows the annealing rate and how this varies with position and sequence context⁶⁴.

In conclusion, we have described an NMR strategy for site-specifically resolving duplex hybridization kinetics. The ease and throughput of these experiments can be improved in the future by using longitudinal optimized ¹H-CEST experiments⁶⁵ as well as other approaches for optimal data collection^{34, 66}. The approach can also be applied to mismatch containing duplexes ideally by targeting remote sites that are not involved in any local mismatch dynamics and to use multi-site exchange models as needed to fit data⁵⁶. Our results show that in the middle of a duplex, m⁶A minimally affects the melting rate but substantially decreases the rate of annealing. This may help explain why tRNA selection during translation is slower for mRNAs containing m⁶A²⁰. m⁶A is also found in the seed sequence of microRNAs and in their mRNA target sites⁶⁷ and mismatches that slowdown microRNA:mRNA annealing have substantial effects on gene expression⁶⁴. Thus, m⁶A could similarly affect gene expression by altering the kinetics of annealing. m⁶A may also affect the kinetics of RNA-protein and RNA-ligand association and also reshape co-transcriptional RNA folding pathways^68–71 by prolonging the lifetime of the unpaired conformation^{25, 72–73} perhaps in a manner analogous to cis-trans proline isomerization in proteins^74–75.