More than Proton Detection—New Avenues for NMR Spectroscopy of RNA

Abstract Ribonucleic acid oligonucleotides (RNAs) play pivotal roles in cellular function (riboswitches), chemical biology applications (SELEX‐derived aptamers), cell biology and biomedical applications (transcriptomics). Furthermore, a growing number of RNA forms (long non‐coding RNAs, circular RNAs) but also RNA modifications are identified, showing the ever increasing functional diversity of RNAs. To describe and understand this functional diversity, structural studies of RNA are increasingly important. However, they are often more challenging than protein structural studies as RNAs are substantially more dynamic and their function is often linked to their structural transitions between alternative conformations. NMR is a prime technique to characterize these structural dynamics with atomic resolution. To extend the NMR size limitation and to characterize large RNAs and their complexes above 200 nucleotides, new NMR techniques have been developed. This Minireview reports on the development of NMR methods that utilize detection on low‐γ nuclei (heteronuclei like 13C or 15N with lower gyromagnetic ratio than 1H) to obtain unique structural and dynamic information for large RNA molecules in solution. Experiments involve through‐bond correlations of nucleobases and the phosphodiester backbone of RNA for chemical shift assignment and make information on hydrogen bonding uniquely accessible. Previously unobservable NMR resonances of amino groups in RNA nucleobases are now detected in experiments involving conformational exchange‐resistant double‐quantum 1H coherences, detected by 13C NMR spectroscopy. Furthermore, 13C and 15N chemical shifts provide valuable information on conformations. All the covered aspects point to the advantages of low‐γ nuclei detection experiments in RNA.


Introduction
Since the developmento fm ultidimensional NMR spectroscopy and the availability of isotope-labeled RNAs, NMR spectroscopy has contributed more than 40 %o fa ll RNA structures in databases. Considerable challenges, however,r emain for the structure determination particularly of large RNAsa nd their complexes by biomolecular NMR spectroscopy.O ften,t he maximum concentration, at which RNA and RNA-proteinc omplexes can be prepared for NMR studies, does not exceed 50 mm,e ither for solubility reasons or for availability of sample, and thus all NMR experimentsh ave to be optimized to maximize the signal-to-noise ratio. The multitude of current approaches to achieve maximal signal-to-noise ratio have been recently summarized. [1] Multidimensional NMR pulse sequences that rely on excitation and detection of the most sensitive nuclei protons( where 'proton" refers to 1 H) have for long been the experimental gold standard in the field of biomolecular NMR spectroscopy. Since 1957, [2] ap lethora of studies have unravelled structure and dynamics of proteins and nucleic acids. Especially for the characterization of RNAs, NMR has led to new fundamentali nsight including transientb ase-pair states modulating replication and transcription, [3] long-lived meta-stable secondary structuresc ontrolling gene-regulation during transcription [4] and translation, [5] and the structureo fav iral packaging signal, [6] to name only af ew.T he advantages of proton-detected experimentsa pparentlys tem from the high natural abundance of the 1 H-isotope and its highest sensitivity of all NMRactive and stable nuclei.
However,a lso proton-detected experiments exhibit difficulties for NMR studies of RNAs. Chemical-shift dispersion is limited for all their resonancesw ith the exception of imino protons. Further,t he proton-density within the nucleobases of RNA is low and many nucleobase-protons are susceptible to conformationala nd solvent exchange. Thus, methods to circumvent these problems are neededa nd we here summarize recent developmentsi nh eteronuclear-detected experiments ( 13 C, 15 N, 19 F, and others). In combination with the traditional proton-detected experiments,t hey will open new possibilities in the structurald escription of RNA dynamicsa nd function at atomic resolution.
One of the main drawbacks of NMR spectroscopy of RNA is the narrow chemical shift dispersion of the corresponding resonances due to the limited chemical diversity in building blocks( Figure 1A). This limited dispersion leads to severe spectral overlap, which is particularly problematic in 1 H-spectra ( Figure 1B)a nd, therefore, puts al imit in molecular size of around5 0nts for NMR spectroscopy of RNA. However,w hen site-selective labeling schemes are applied, the current limit is extendedt o1 50-200 nts. [7] Furthermore,c ompared to heteronuclei, the protont ransverse relaxation rates are high and thus their line widths are large and increase fast with moleculars ize again impeding the characterization of larger RNA molecules. The nucleobases as heteroaromatic moieties possess few protons. Thus, only af ew long-range NOE contacts can be detected. Further,e xperiments that correlate NMR-activen uclei by J-transfer steps often require multiple and sometimes long magnetization transfer steps, in which sensitivity is lost due to relaxation processes. Examples for such experimentsa re the 3D TROSY-relayed HCCH-COSY experiment for the correlation of H2C2 and H8C8 in the adenine nucleobase [8] as well as 3D HCCNH experiments for the correlation of the H6C6 (pyrimidines) or H8C8 (purines) with the imino protons in the nucleobases. [9] Both of the experiments are crucial as they offer unique information for the sequential resonance assignment process. However, their long and multiple magnetization transfers are especially challenging for larger RNAs as relaxation is enhanced due to a larger rotational correlation time t c .
The nucleobases are not only protonp oor but the iminoand amino-protons are often involved in different exchange processes.T he iminop roton is in fast exchange with the solvent water if not protected from exchange, mostly through hydrogen bonding or,i nr are cases, other steric protection from exchange. [11] This feature enables the fast determination of stable secondary structures, as only exchange protected imino protons can be detected. But it also leads to loss of information on dynamic regionso ft he RNA, as nucleobases in these regions are not involved in stable hydrogen-bond networks and the imino proton resonances are broadened beyond detectability.U nfortunately,o ften the dynamic regionso fa nR NA are involved in functional processes including ligand-o rp rotein-binding. [5,12] Amino groupse xhibit restricted rotationa roundt he CÀNH 2 bond. The rates of rotation are similar to the chemical shift differenceo ft he two amino protons, andt he signals are thus broadened beyondd etectability in this intermediate exchange regime. [13] Ta ken together,b othe xchange processes of imino and amino protons severelyh amper the collection of information on the orientation of the nucleobase, itsp otentiali nteractions at the exchanging sites and their dynamics.
All four difficulties-1) resonanceo verlap, 2) low 1 H-density, 3) chemical exchange, and 4) relaxation-can be circumvented in heteronuclear-detected NMR experiments. The disadvantage due to low-g detection can be minimizedt hanks to new cryogenic probesw ith inner NMR coils optimizedf or 13 C-, 19 F-, or 15 N-detection. [14] Thus, despite their lower fundamental signalto-noise ratio, heteronuclear-detection schemes have become feasible. These heteronuclear-detected experiments benefit from the larger chemical shift dispersion, coupled to sharper line widths of the heteronuclei. For example, 13 Cn uclei in RNA have chemical shifts from d = 65 to 170 ppm. If the chemical shifts of the heteronucleia re detected during directa cquisi-tion, high resolution due to long FID sampling can be achieved without lengthening the experiments,a sr elaxation delays can be shortened.
The heteroaromatic nucleobases represent ac yclized chain of CÀNf ragments. This particularf eature can be exploited for the direct magnetization transfer in NMR experiments in (multiple) INEPT steps without being dependento n 1 H-excitation or Robbin Schnieders, born in 1991, studied chemistry at the University of Frankfurt and finished studying in 2016 with her master's degree. Since then she has been working on her PhD in the group of Harald Schwalbe and is focused on the development of NMR spectroscopic methods for the characterization of RNAs. H-detection. Information on quarternary carbon or nitrogen atoms then becomes feasible. Additionally,i nh eteronucleardetected experiments, the introductiono fd euterium-labeled nucleotides in large RNAs exploits their favorable relaxation properties in deuterium-decoupled spectra [15] but does not introducet he disadvantage of losing the observable nucleus. In addition, heteronucleiare not affected by solvent exchange.
Moving towards slower relaxing nucleii ncluding 13 Co respecially 15 Nc an bring potentiala dvantages to extendm olecular size limitation as line widths increase slower with molecular size when compared to 1 H. On at echnical side, further advantages of heteronuclear-detectiona re their insensitivity to certain experimental conditions including pH value, temperature or salt concentration andt he non-necessity of water suppression.
The loss in sensitivity due to the lower gyromagnetic ratio is fundamental and remains, however,t he major disadvantagei n heteronuclear-detected NMR experiments. Further,i nu niformly isotope-labeled samples, homonuclear J(C,C) couplings decrease the chemical shift resolution in 13 C-detected experiments.I np articular,t he sizeableh omonuclear 1 J CC couplings lead to splittings, as they are larger than the carbon line widths.H owever,d ecoupling schemes like IP/AP [16] and S3E [17] as well as selective homonuclear decoupling during acquisition are available.

Requirements:F rom NMR probes to sample preparation
Due to the reduced sensitivity of low-g-detected NMR experiments,p robesw ith cryogenically cooled detectionc oils and preamplifiers, so called cryogenic probes, are needed as they increases ensitivity about af actor of 3-4 when compared to room temperature probes. [14] Probes that are optimized for 13 Cor 15 N-detection are even better suited for recording heteronuclear-detected experiments.A so pposed to the so-called inverse probes, which are used as standard probesinb iomolecular NMR,t hey connect the channels of those heteronuclei to the inner coil of the probe enhancing the sensitivity for 13 C-or 15 N-detection due to al arger filling factor.W hen workingw ith fluorinated nucleotides,atwo channel probe for 19 F-detection with simultaneous 1 H-decoupling is needed as J HF scalar couplings tend to be large.
In biomacromolecular NMR of RNA, 13 C-and/or 15 N-isotope labeled samples are indispensable. [18] The simple and fast methodo fi nv itro transcriptionu sing uniformly 13 C-and/or 15 N-labeled rNTPs yields milligram quantities of RNA and is well established. For the characterization of larger RNAs, however, selectivel abeling schemes or the incorporation of modified nucleotides are often necessary.
Using mutants of the T7 RNA polymerase during in vitro transcription allows incorporation of severalm odifications including rNTPs modified with fluorine or amino groups at the 2'-position. [19] While this methodi si ndependent of the RNA size, it is not specific as the modified nucleotide is incorporated uniformly in the RNA of interest. Position-selective labeling of RNA (PLOR) overcomes this limitation and allows the auto-mated enzymatic synthesis of position-specific isotope-labeled RNA by transcription. [20] This methodu tilizes the possibility to pause and restart the RNA polymerase by omitting one nucleoside 5'-triphosphate required for the transcription beyond a desired position. Although with this methodm illigram quantities of RNA with ad esired isotope-labeling scheme are obtained, [20] the method is not commonly used because of its complexity and the lack of the commercial availability of the special apparatus. [21] Solid-phase synthesis is one of the most commonly used methodsf or preparationo fR NAs carrying aw ide range of modifications. [22] To overcome the NMR resolution problems position-specific isotope-labeled nucleosidep hosphoramidites like 6-13 C-pyrimidine, [23] 2'-13 C-methoxy nucleoside, [24] 13 C 5'sugar labeled nucleoside [25] and 15 N-imino/amido nucleoside [26] phosphoramidites are incorporated into RNA by chemical synthesis. However,t his methodi sl imited to RNAs of approximately 50 nucleotides [27] for routine applications, taking the required amountsa nd purity into account.A lternatively,t he genetic alphabet expansion technology allows the incorporation of unnaturalb ase pairs that are compatible with the DNA polymerase and RNA polymerase allowing the amplification of modified nucleic acids by PCR andi nv itro transcription. [28] However,t he unnatural nucleotide is incorporated into the DNA templateb ys olid-phase synthesis, whichh as again as ize limitation. Ar eliable method for synthesis of site-specific modified long RNAs are ligation-based approaches using modified RNA fragments. Methylt ransferases have been used to modify the 5'-endo fR NAs post transcriptionally. [29] Also, the 3'-end can be modified with nucleotidyl transferases that have the capabilityt oi ncorporate modified nucleoside triphosphates [30] or with T4 RNA ligase 1t hat is able to incorporaten ucleoside 3',5'-bisphosphates with modificationsa tt he sugar-,p hosphate-,o rb ase-site. [31] The latter allows the 3'-extension of RNA by as ingle nucleotide, which in af urthere nzymatic step can be dephosphorylated at the 3'-end using ac ommercially availablep hosphatase. Such RNAs carrying ah ydroxyl group and am odified nucleotide at the 3'-end undergo ligationw ith a5 '-phosphorylatedR NA in presence of T4 RNA ligase 2i na n ATP-dependentr eaction. With this method, shown in Figure 2, RNAs up to al ength of 390 nts containing as ingle positionspecific modification have been synthesized. [32] 3. 13 C-detection NMR spectroscopic experiments for RNA The first reports utilizing carbon-directd etection in protein NMR occuredd irectly after the introduction of cryogenic probes. [33] Applications to RNA startedi n2 007 with independent reports by Fiala et al. [34] and FarØse tal. [35] 3.1. Overcoming resonance overlap Due to limited chemical shift resolution, complete assignment of protons in particular for ribose protons is sparse in the BMRB database. [36] The signals of ribose carbon atoms are much better resolved, so that carbond irect detection can con-  Figure 3B). From the obtained carbon chemical shifts, the ribose conformation( C2'endo or C3'-endo)c an be determined. [37] As equential assignment from nucleotidet on ucleotide can then be achieved in the (H)CPC-experiment that correlates the C4' resonance with the 31 Pc hemical shifts in the 5'-a nd 3'-site ( Figure 3C). Additionally,t he phosphorus nuclei can be correlated with C1' and C5' nucleib yi ntroducingaTOCSY-o rC OSYsequences. [38] Compared to analogous 1 H-detectede xperiments, [39] the 13 C-detected experiments have as lightly shorter magnetization transfer pathway.T hey are furthermore also applicablei np artially deuterated samples and in particular C2'/ C3' assignment is facilitated.T he (H)CC-TOCSY experiment is furthermore not only applicable to RNA butt oa ny kind of ribose-containing molecules as shown by Fontana et al.f or carbohydrates. [40] 3.2. Resonance assignment of nucleobase nuclei 13 Cd irect-detected experiments can also significantly contribute to complete resonance assignment of the nucleobases including their quarternaryc arbon or tertiary nitrogena toms. There are two alternative approaches to assign these nuclei:a first suite of experimentse xploits 1 H-excited and 13 C-detected experiments and the second suite exploits 13 C-excited and -detected experiments.
In the first suite, the C2 and C4 in pyrimidines and the C2 (A), C4, C5, and C6 atoms in purines are assigned (Figure 4A,B,C). The experiments give rise to CÀHc orrelated spectra and are particularly valuablef or larger RNAs as spectral   H8 to C5 and C4 (blue) or C6 (red) and C HCC-TOCSY in purines. The size of n J(C,C), n J(H,C)a nd n J(N,H) with n = 1,2,3u tilized in the transfer pathways is indicated. Coupling constants, usedd uring IPAP decoupling schemes are depicted in gray.C arbon nuclei at whichm agnetization is detected are marked with ac ircle. Coupling constantsh ave been taken from literature. [34,41] Chem.E ur.J.2020, 26,1 02 -113 www.chemeurj.org 2020 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim overlap can be reduced throught he carbon direct-detection. In the second suite, not only quarternary carbon atoms but also tertiary nitrogen-atoms are assigned using CN-HSQC experiments (magnetization transfer Figure 5A), providedt he J(C,N)-coupling constantsa re sufficiently large. For U, C, and G nucleobases, an ear-to-complete resonance assignmentc an be achieved ( Figure 5B). [34] However, the sequential walk through the CÀNf ragments of adenosinesi sh amperedb yt he low 1 J(C,N) couplings between C6N1, C2N1, C2N3, andC 4N3 (Figure 5A). The requirement of recording atl east three different experiments with matched evolution times for the CN coherence transfer (29.4, 21.7, and 18.5 ms) to observe all nucleobase 13 Ca nd 15 Na toms is the main disadvantage of this approach. [42] Nevertheless, CN-HSQCs have been successfully applied for the complete de novo assignment of aG TP-binding aptamerw ith 39 nts. [43]

Determination of coupling constants
With the chemical shift assignments of 13 C-and 15 N-nuclei in 13 C-direct-detected NMR spectroscopic experiments, also J(C,C)-and J(C,N)-coupling constants can be measured. 1 J(C,C) scalar couplings are determined, for example, in the H6C6C5experiment or in H5C5C4(C5)-experiment ( Figure 6A). [34] The Jcouplings are measured precisely through the deconvolution of the doublet splitting in the direct dimension ( Figure 6B). The obtained scalar couplings are in the range of 1 J(C5,C6) = 67 or 1 J(C4,C5) = 55 Hz. [34] Due to the high resolution in the direct dimension, the precision of the obtained values is high. Although this can be diminished by the inherentl ower sensitivity of carbon detection. These experiments werea pplied in RDC studies of the TAR-RNA. [44]

Detection of exchanging sites
Typically, single-stranded regionsa nd nucleotides within dynamic secondary structuree lements in RNA, such as long loops or bulges, cannot be observed in 1 H-detectede xperiments,b ecause the imino proton reporter signals are broadened beyondd etection by exchange with the hydrogen atoms from the solvent water (Figure7C, D). In ordert oo vercome this blind spot in NMR of RNA, carbon-detection experiments are utilized.
The addition of as pin filter in the CN-HSQC experiment (magnetization transfer Figure 7A) [34,46] allows determination of the status of hydrogen bondinga tt he imino-nitrogen atom. [42] The experiment makes use of the dependence of the scalar 1 J(N,H)-coupling on the proton exchange rate. If the proton is in slow exchange with solventw ater,t he 1 J(N,H)-coupling can evolveu nder an unscaled coupling of % 90 Hz. If the protoni s, however,i nf ast exchange with solvent water,t he scalar coupling is decoupled through scalar relaxation of the second Figure 5. A Magnetization transfer pathway for the 13 C-excited and 13 C-detected 2D CN-HSQC experiment.T he size of 1 J(C,N) utilizedi nthe transfers is indicated. [34] Coupling constants, usedd uring IPAP decoupling schemes are depicted in gray.Carbon nucleia twhich magnetizationi sd etected are markedw ith a circle. B Examples of 2D CN-HSQC spectra for uridiner esidues in a14nts RNA with UUCG tetraloop.The walk through the nucleobase isi ndicated with ag ray dashed line. The figure hasbeen adaptedf rom literature. [42] Figure 6. A Magnetization transfer pathways for 1 H-excited and 13 C-detected 2D experiments thatc orrelate H6 to C6 (black) or H5 to C5 to C4 (red) to measure 1 J(C,C) coupling constants. The size of 1 J(C,H)a nd 1 J(C,C) utilized in the transfers is indicated. Coupling constants,whicha re determined in these experiments, are depicted in gray. [34] Carbon nuclei at whichm agnetization is detected are marked with ac ircle. B 2D H6C6(C5) spectraf or the measurement of the 1 J(C5,C6)c ouplingc onstant in pyrimidines. The figure was adapted from the literature. [34] Chem.E ur.J.2020, 26,102 -113 www.chemeurj.org 2020 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim kind. [47] Due to exchange, the spin state (a or b)o ft he imino proton is not maintained but changes with every H 2 O-imino chemicale xchange process. As ac onsequence, 1 J(N,H) is no longer observable. The spin filter has no effect for nucleotides with fast exchanging imino protons, whereas it inverts the 15 N coherences for imino sites of nucleobases involved in stable interactions ( Figure 8A,B). As the signal intensity is thus modulated by the rate of proton exchange, the underlying exchange rates can easily be evaluated:t he experiment allows quantitative determination of solvente xchange rates between k ex = 10 0 to 10 4 s À1 .T he experiment relies on carbon direct-detection. Therefore, imino exchange rates can also be measured even for samples dissolved in pure D 2 O, and kinetic isotope effects could potentially be determined.
As opposedt oi mino groups, solvent exchange is negligible in amino groups. [49] Here, ar estricted rotation around the CÀ NH 2 bond is often in intermediate exchange regime and renders the amino protonr esonances undetectable ( Figure 7E,F). This is particularly prominent for adenosines and guanosines as seen in the 15 N-HSQC spectrumi nF igure 9A.F ollowing an approachd eveloped to detect nitrogen-sitesi nt he arginine side chains of proteins, [50] new experiments to detecta ll NH 2 groups in RNA have been developed. [45] In these experiments, 1 H-doublequantum (DQ) coherences are excited in the indirect dimension. This magnetization is transferred to the neighboring carbon atom, where it is detected (magnetization transfer Figure 7B). Evolution of 1 H-DQ coherences is unaffected by chemicale xchange and thus their line width is independento f bond rotation. With the 13 C-detected C(N)H-HDQC experiment aC ÀHc orrelated spectrumi so btained,i nw hicht he 1 H-double quantum signals resonate at the mean protonc hemical shift ( Figure 9B). This experiment enablest he detection of af ull set of sharp resonances for all amino groups independento fa ny kind of exchange.
Te chnicald etails require recording two independent experiments.D uring detection, the signalo ft he C6 in adenosines and the C4 in cytidines are doublets due to 1 J(C6,C5)v alues of Figure 7. A Magnetization transfer pathways for the 13 C-excited and 13 C-detected 2D CN-spin filter HSQC experiment [42] and B the 13 C-excited and 13 C-detected 2D C(N)H-HDQC experiment. [45] The size of 1 J(C,N) and 1 J(N,H) utilized in the transfer pathways is indicated. Coupling constants,u sed during IPAP decoupling schemes are depictedi ng ray.Carbon nucleia twhich magnetization is detected are marked with ac ircle. C 15 N-HSQC and D 13 C-detected CN-HSQC spectra for the imino region of the 14 nts RNA with UUCG tetraloop. 1 J(C,N)-coupling constants utilizedf or magnetization transfer were 17 (yellow) and 27 Hz (green). E 15 N-HSQCa nd F 13 C-detected CN-HSQCs pectra for the amino region of the 14 nts RNA withU UCGtetraloop(right).The figureshave been adapted from Fürtig et al. [42] and Schnieders et al. [45] Figure 8. A 2D CN-spin filter spectrum for the uridines of the 14nts hairpin RNA with UUCGtetraloop. B Structural context of the respective uridines (U11W Cb ase pair,H -bonded imino proton;U6s heared GU base pair,s terically shielded imino proton;U7u npaired nucleotide,f ully solvent exposed iminop roton) are indicated. [48] The figurehas been adapted from the literature. [42] Chem.E ur.J.2020, 26,102 -113 www.chemeurj.org 2020 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim 55 and 1 J(C4,C5) of 75 Hz. Virtual decoupling schemes coined IPAP sequences remove these couplings that cannot else be decoupled for example, by homodecoupling. Thus, either two experiments are recorded with optimized values that match the respective J-coupling or one experiment with an IPAP filter tuned to the averaged coupling of 65 Hz is recorded. The C(N)H-HDQC represents a 13 C-detected experiment that allows detection of NMR signals that completelye vade observation in proton-based experiments.
Using 13 C-direct detection in tailorede xperimentsa llows also observation of NOE cross peaks from amino groups. The "amino"-NOESY experiment ( Figure 9C)c orrelates protons and amino groups in NOE close proximityu nobservable in conventional 1 H-detected NOESY experiment. [51] The newly obtained NOE contacts often stem from H1'-to-amino-group correlations.
In the refinement of RNA structure, they are of special value as they describe sequential and cross-strand inter-residual contacts. They significantly improve structure determination, in particularfor dynamic RNAs.

Chemical-shift-to-structure relations
Carbon-direct-detection experiments make chemical shift information accessible, whicho ften is unavailable using proton-detection. Chemical shifts are highly sensitive to the electronic environment of the respective nucleus and, therefore, they can potentially be used for ac hemical-shift-to-structure relation. In proteins this is already as tandard methoda ppliedf or the determination of secondary and even tertiarys tructure. [53] Carbon-chemical shifts of the ribose atoms are used for the determination of the ribose conformation in the RNA'sb ackbone. Based on an empirical calculation of the so-called 'canonical coordinates' the ring pucker and the conformation of the exocyclica ngle (O5'-C5'-C4'-C3': g)c an be extracted. [37] Chemical shifts from 13 Cn uclei in the nucleobase do not depend on specific torsions but are sensitive to hydrogen bondinga nd stacking. With the first complete assignment of all carbon chemical shifts in nucleobases recorded by carbondetection experiments, as tatistical analysis of the so far deposited chemical shifts in the BMRB database was undertaken in order to assign chemical shifts to structural context. [35] The main effects that modulate the carbon chemical shift in the nucleobases are p-stacking and hydrogen-bonding interac-  [45] Experimental details can be takenf rom Schniederse tal. [45] Chem.E ur.J.2020, 26,102 -113 www.chemeurj.org 2020 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim tions. Nucleotides can be classified into three different interaction type categories:h elical (Watson-Crick base pairing and two site p-stacking), terminal (Watson-Crick base pairing and one site p-stacking), and disordered. All of the average chemical shifts of the atoms of the different nucleobases were referenced to the average chemical shift of the helicalr egion and a clear trend can be observed for several carbon atoms in different structural elements (Figure 10 A). [35] Also for the 1 H-DQ chemicals hifts of the amino groups ac hemical-shift-to-structure relation was conducted. [45] Similarly,a sf or carbon chemical shifts, am eanv alue was calculated for all nucleotides involved in Watson-Crick interactions. Therefore, nucleotidese xhibiting ad ifference to this mean must be involved in ad ifferent interaction network. Analysis for five differentR NAs allowed ac lear discrimination between canonical and noncanonical base interactions (Figure 10 B).

15 N-detection NMR experiments for RNA
15 N-direct-detected multidimensional NMR experiments for RNA has only recently been introduced. [54] Here, several 15 N-detected HN-correlation experiments were applied to RNAs of increasingm olecular size. This study was motivated by the development of 15 N-detection TROSYexperiments for the analysis of proteins. [55] In the field of proteins the experiments are particularly interesting for intrinsically disordered proteins (IDPs) as signals are usually very well dispersed in the 15 N-dimension, whereas the 1 H-dimensiono nly covers less than d % 2ppm.
Similarly as for proteins, predictions for RNA show that line width of 15 N-resonances increases much slowerw ith molecular size (rotational correlation time) than their protonc ounterparts. (Figure 11 A).
The 15 N-detection BEST-TROSY experiment was identified as the most sensitive 15 N-detection HN-correlation experiment (Figure 11 C, D) and was thus applied to as et of RNAs ranging in size from 14 to 329 nts. The experimentally determined line width at half heightc onfirmed the theoretical predictionsc oncerningt he trend in increasing molecular size and the differences in line widths for AU-and GC-Watson-Crick base pairs (Figure 11 B). However,t here was no improvementi ns ensitivity when compared to the 1 H-detection BEST-TROSY experiment. It might, however,b ei nteresting to employt he 15 N-detection BEST-TROSYexperiment for the characterization of even larger RNAs due to the favorable relaxation properties.

Other nuclei
Besides the well-establishedn uclei for heteronuclear-detection schemes ( 13 C, 15 N), several experimentsh ave been developed that detect magnetization of either non-native nuclei ( 19 F) or on rather exotic nuclei.
Although naturallyo ccurring RNA nucleotides do not contain 19 F-nuclei,t hey can be introducedb ym eans of chemical or biochemical synthesis at various positions( see the discussion above). It has previously been shown that fluorinem odifications do not necessarily disturbt he structure of the RNA, with the exception of fluorine labels at the 2'-position. [57] In the context of structuredR NAs, the fluorinen uclei are then used as spy nuclei that show chemical shift perturbations overt heir wide chemical shift range in dependence of conformational changes. [58] The great advantage of 19 Fe xperiments is that in biomolecular samples no background signals arise. However, 19 Fl abelling comesa lso with two major disadvantages.T he large chemical shift anisotropy (CSA) rendersd etection of 19 F in large RNAs difficult. Further, the incorporated nucleotides are only 19 Fb ut not 13 Cl abeled, so that only correlation experimentsw ith proton-nuclei can be recorded by exploiting the heteronuclear overhauser effect. Recently,t hese problems have been overcome by the introduction of nucleotides containing ap air of 13 C- 19 Fl abels in the heteroaromatic nucleobases that allow recording of 19 F-13 CTROSY spectra. [59] As an egatively chargedb iopolymer,i nteractions with cations are crucial forR NA/DNA folding and function. [60] Those interactions can be characterized by NMR as relevantc ations including Na + ,L i + ,a nd K + are NMR-active. NMR hase xtensively been conducted for DNA for which particular relaxation measurements have contributed towards understanding the nature of the interaction for Na + and Li + with double-stranded DNA. [61] In addition, G-quadruplexes,f or which cation binding is vital for formation,h ave been characterizedu sing 23 Na, 39 K, 87 Rb, and 205 Tl NMR spectroscopy. [62] The latter is used as as ubstitute for K + in the interaction with G-quadruplexesi nw hich Figure 10. Absolute difference in the chemical shifts of A nucleobase 13 Cnuclei and B nucleobase 1 Ha mino group nuclei in terminal and disordered nucleotides compared to nucleotides involved in helical interactions. The raw data was taken from FarØsetal. [35] and from Schnieders et al. [45] Chem. Eur.J.2020, 26,1 02 -113 www.chemeurj.org 2020 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim differentb inding sites for the G-quartets can be detected [63] and even J(H,Tl) scalar couplings were measured. [64]

Conclusions and Outlook
The development of low-g-detection schemes in NMR spectroscopyh as been an active field over the last 15 years. Now, these possibilities are also exploited for NMR spectroscopy of RNA. Given the increasingly recognizedb iological relevance of RNA and the powero fN MR spectroscopy to characterize its functional dynamics, the application of low-g-detection schemes now allows forwarding NMR spectroscopy to larger RNA molecules. In order to reach this goal, development of ad-vancedN MR methods runs hand-in-hand with improved methods in RNA sample preparation. The low-g-detection schemes are further very compatible with solid-state NMR experiments for RNA, as pioneered in the group of Carlomagno. [65] The advantageso ft hesen ovel direct detection methods for 13 C-and 15 N-nuclei will become even strongera th igher magnetic fields that are now on the horizon. [66]