Genetic Code Expansion Facilitates Position‐Selective Labeling of RNA for Biophysical Studies

Abstract Nature relies on reading and synthesizing the genetic code with high fidelity. Nucleic acid building blocks that are orthogonal to the canonical A‐T and G‐C base‐pairs are therefore uniquely suitable to facilitate position‐specific labeling of nucleic acids. Here, we employ the orthogonal kappa‐xanthosine‐base‐pair for in vitro transcription of labeled RNA. We devised an improved synthetic route to obtain the phosphoramidite of the deoxy‐version of the kappa nucleoside in solid phase synthesis. From this DNA template, we demonstrate the reliable incorporation of xanthosine during in vitro transcription. Using NMR spectroscopy, we show that xanthosine introduces only minor structural changes in an RNA helix. We furthermore synthesized a clickable 7‐deaza‐xanthosine, which allows to site‐specifically modify transcribed RNA molecules with fluorophores or other labels.


Introduction
Over the last decade, an increasing number of functional roles has been identified for ribonucleic acids and their biophysical investigation is increasingly pursued. For an umber of reasons, methods developed for proteins cannote asily be transferred for studies of RNAs. Thus, novel methods have to be developed, for example to attach labels that allow for detailed spectroscopic studies of RNA.S tructural studies additionally require labels to be inserted in as ite-selective manner to obtain position-specific readouts.
By chemical solid-phase synthesis, RNA with such positionspecific label can be obtained. Av ariety of chemically modifiable building blocks has been made availableo ver the last years. However, the use of solid phase synthesis bringsa long the well-knownl imitations and drawbacks of this technique; most importantly the limited product yield that decreases with the length of the RNA and the concomitant increase of byproducts make purification of the target RNA difficult.
Severals trategies to circumvent these limitations have been devised and successfully established in recent years. Many of these techniquesu se the standard approach of in vitro transcriptiont og enerate RNA, and employ an additional base pair that is orthogonal to the two Watson-Crick-like base-pairs that establish sequence specificity during transcription. [1][2][3][4] Twom ain strategies have been previously reported:e ither larger hydrophobic moieties establish an entirely novel complementary system or the hydrogen bonding pattern of the standardp urine and pyrimidine-based scaffolds are expanded. Both these strategies have generated additional, orthogonal base-pairs that can be employed to different degrees in in vitro applications or for genetic code expansion in vivo. Some of the designed bases have also been elegantly used for the incorporation of modifiers, that is, for nitroxide spinl abels or fluorophores. [9][10][11] Our focus for the current study wast ou se the capabilities of al abeling strategy employing one orthogonal DNA nucleotide together with the complementary ribonucleoside triphosphate to generate site-specifically modified RNA by in vitro transcription.
To introduce only minimal structural perturbationo ft he target RNA, we therefore opted to follow the strategy of modifying the hydrogen bond arrangemento ft he near canonical kappa nucleobase, which would also be sterically similart ot he natural nucleobases and provide comparable stackingi nterfaces. Based on the work initiated by Piccirillie tal., [4] we first developed an ovel synthesis route for the kappa nucleoside, and incorporated the kappa nucleoside phosphoramidite into a DNA used as at emplate for RNA transcription. We successfully incorporated xanthosine at this positionw ith high specificity and good efficiency.W ei ntroduced xanthosine into as izeable RNA riboswitch and could site-specifically detect the introduced NMR-activel abel. We furthermore implemented novel covalentl abeling options by attaching am odifier to the xanthosine-based nucleobase, focusing on ac lick chemistry compatible derivative.

Results and Discussion
The steps required here to obtain as ite-specifically labeled RNA can be described as follows:( i) Synthesis of the kappa base (and DNA synthesis), (ii)transcription, (iii)quality control, and (iv) labeling.

Synthesis of kappa base phosphoramidite
Our synthetic strategy to obtain the kappa deoxynucleoside ( Figure 1) incorporated an umber of novel synthetic developments.M ost importantly,w eo pted to synthesize the CÀCg lycosidic bond using aHeck coupling, which required generation of the glycal intermediate 4 ( Figure 2). [12,13] In addition, formation of the protected amine ring substituents 9 was achieved using apalladium-catalyzed Buchwald-Hartwig coupling. [14] To this end, we started with deoxythymidine 1,w hich was protectedu sing tert-butyldimethylsilyl chloride (TBDMS-Cl) to yield 2.T he elimination of the nucleobase to obtain the double silylated glycal 3 turned out to be experimentally demandingd ue to the instability of 3 and the required purification. The following deprotection was achievedw ith 1.0 m tetrabutylammonium fluoride (TBAF) in THF solution. Here, both reaction monitoring and column purification had to be extensively optimized to reach ay ield of 88 %. We assume that selectivityo ft he deprotection of C5 was limited,t he yield of 4 therefore could not be optimized beyond 45 %; we could, however,n ot fully characterize the side products. As the double silylated glycal presumably cannot undergo the following Heck reaction, [15] it is necessary to block the bottom of the glycal 4 by the sterically demanding TBDMS-groupt oa void the attack from this side which would result in the formation of the wrong a-anomer.
One of the key reactions of our Scheme was the palladiumcatalyzed Heck reaction. The Heck reaction was applied to nucleic acid chemistry by Hocek et al. and Wagenknecht et al. [14][15][16][17] To perform this reaction, the commerciallya vailable 2,4-dichloro-5-iodopyrimidine as the heterocyclich alide and the glycal 4 whicha cted as ac yclic alkene are coupled to yield the protected nucleoside 5.T he b-anomer was obtaineda sa single product.
However,a sc ompound 5 wasu nstable, we opted for an in situ deprotection with 1 m TBAF to yield the ketoriboside 6 with 37 %. Reduction of the ketone to the deoxyribose 7 was achieved using sodium triacetoxyborohydride (62 %y ield). The Figure 1. Examples of unnatural base pairs. To p: The isoC-isoG base pair which was the first developedu nnatural base pair. [5] Z and P as well as S and B are two unnatural base pairs which were developed by Benner et al. [6,7] Middle:H ydrophobic base pairs (NaM and TPT3)w hich was developed by Romesberg et al. [2] and the TPT3 CP and NaM base pair which was developed by Kath-Schorr et al. [8] Bottom:The kappa (k k)xanthosine (X)b ase pair used in this paper.  Buchwald-Hartwig coupling to obtain the benzoyl-protected pyrimidine diamine, which was first appliedb yW agenknecht et al. [14] shows increased selectivity when the C5 of the ribose is silylated. We therefore coupled aT BDMS protection group yieldingc ompound 8.H ere the selectivity was highd ue to the slightly more reactive C5'-hydroxyl group, and the yield was 76 %. The Buchwald-Hartwig coupling is catalyzedb yp alladium, and we increased the amounto fb enzamide to three equivalents to obtain the double protected compound 9 in sufficient yield (37 %). However,p urification of 9 at this point provedd ifficultd ue to the slow migration on the column and the poor separation from the starting benzamide. Removal of TBDMS group using trimethylamine trihydrofluoride yielded compound 10 in 62 %. From here, the protocol for protection to form the phosphoramidite suitable for solidp hase synthesis followed standard procedures:T he DMTrp rotection of C5 to form 11 was achieved by DMAP-catalyzed coupling of 4,4'-dimethoxytritylchloride in dry pyridine. Formation of the phosphoramidite 12 was performed with 2-cyanoethyl N,N-diisopropylethylamine in dry DCM.
As synthesis of the longer DNA especially for transcription of ariboswitchR NA with more than 80 nucleotides is demanding, we purchased DNA from IBA (Gçttingen),p roviding phosphoramidite 12 to be site-specifically incorporated into the desired transcription templates.

Synthesis of functionalized 7-deazaxanthosine
Severalo ptions are availablet op rovide am odifiable xanthosine derivative that maintains its Watson-Crick-faceb ase-pairing interactions. Introduction of chemical modifications at C8 would be the mosti mmediate choice. However,i ntroduction of as terically demanding modification at C8 introduces steric clashes in the major groove of an A-form RNA helix. [18] Thus, attachment of modifications at N7 promises to be structurally less perturbing, but their synthesesa re difficult. However,s everal nucleosided erivatives have been reported that introduce a7 -deazapurine backbonea nd an umber of synthetic strategies [19][20][21][22] as well as suitable starting compounds are available. Major advantages of this approach are the conservation of the hydrogen bondingp attern at the Watson-Crick side, the advantageouss ubstrate properties of such modified nucleotides for RNA polymerases [23] as well as the possibility to specifically couple extended linkersusing cross-coupling reactions.
In ap reviously published synthesis, [21] the methyl ether at the C4-position was removed with trimethylsilyl (TMS) chloride and sodium iodide. Under these conditions, it was reported that a3:2 ratio of the 7-deazaxanthosine without iodine and 7deazaxanthosin with iodine was recovered and attempts to remove the methyle ther using sodium hydroxide were unsuccessful. In our case, the terminal alkyne linker is sufficiently stable during Sonogashira coupling conditions. We therefore directly conducted the Sonogashira coupling of the 1,7-octadiyne linker to obtain 19.W eu sed 10 equivalents of 1,7-octadiyne and a2 :1 ratio of copper/palladium. [26] The methylp rotectiong roup at O6 was only then removed using TMSc hloride ands odium iodide at room temperature with a6 1% yield of 20.
Triphosphorylation of the modified xanthosine derivativei s necessary to make it amenable for T7 RNA polymerase-based transcription. Therefore, the final step is the coupling of the triphosphate specifically to C5 of the ribose. Here, protecting strategies have been used previously. [27] We found that the conditions used by Yoshikawa andL udwig [28][29][30] were compatible with our modification, and provided the advantage that protection of the other two ribose hydroxyl moieties was not requiredt oo btain the desired specificity for C5. We adapted the protocola nd extended the hydrolysis step with 0,1 m TEAB buffer to eight hours. After purification by RP-HPLC, the final nucleoside triphosphate 21 was obtained in 16 %y ield. With this, both orthogonal base-pairing building blocks were ready to be testedi nt ranscription andl abeling.

Site-specific incorporation of xanthosine by in vitro transcription
We first tested incorporation of axanthosine (X) to be decoded by aDNA nucleotide containing the kappa nucleobase by standard T7 run-off transcription.T he kappa nucleobase was incorporated into two DNA-templates,e ncoding a1 4mer RNA hairpin with aU UCG tetraloop motif, and the aptamer domain of the guanine-riboswitch-aptamer RNA (Gsw 73 )f rom Bacilluss ubtilis. Run-off transcription can utilize single-stranded DNA where only the T7 primerr egion of the template DNA is double stranded.U sing this strategy,i ti ss ufficient to synthesize the kappa DNA phosphoramidite. [31] As ap roof of concept, aG 9X mutation was introduced into the 14mer UUCG-tetraloop hairpin RNA, as guanosine hash igh structural similarity to xanthosine. Transcription yielded a highly pure RNA.T he yield upon switching from Gt oXd ropped by 87 %( Figure S1, Table S1). We furtheri nvestigated the impact of introducing xanthosine on RNA structure using NMR spectroscopy.1 D 1 HNMR showedt hat folding into the hairpin structurei sr etained and introduction of xanthosine leads to only small structural deviations ( Figure S2). The UUCG tetraloop is usually highly ordered [33] and was destabilized by the mutationw hich is visible in both 1D 1 Ha nd 1 H, 1 H-NOESY NMR spectra (Figure 4a nd FigureS2). These results showt hat preparative transcription using the modified constructs is feasible, and that replacing ag uanosinew ith ax anthosine yields an interesting atomicm utagenesis approacht oi nvestigate local structural effects by NMR.
We then turned to al arge structured RNA to test whether these findings also hold true for the analysiso ff unctional/biologicallyr elevant RNA. Here, we used the guanine sensing ri-boswitchR NA from B. subtilis. [34][35][36][37] For this construct,w ereplaced G79 forming aG ·U wobble base pair in the P1 stem with X79 ( Figure 5A). We optimized the transcription conditions with regard to the concentrationo fX TP (FigureS3). Under optimized conditions, the yield of the full-length G79X transcript was 71.4 %i nc omparison with transcription of the unmodified Gsw 73 -transcript. Not surprisingly,w eobserved abortion products at the site of the mutation, with ar atio of 77:23 full-length RNA to abortion fragment. In absence of the xanthosine triphosphate, we alsoo bserved 22 %f ull-length product ( Figure 5B,T able S2), which presumably arises from misincorporation of adenosine. [4] CD meltinge xperiments showed as malld ifferenceo f0 .25 K in the melting points (DT m )b etween the Gsw 73 and G79X riboswitch, indicating that the structure is retained( Figure 5C). We furtherp urified the full-length G79X RNA and obtained about 22 nmol of RNA, whichw as sufficient for NMR studies. The NMR data fully supported our proposed strategyt hat introduction of xanthosine induces minimal structural perturbation. Furthermore, the previous G79H 1 ' cross peaks are now visible as X79 H1' cross peaks, indicating that the structure is retained ( Figure 5D).
We prepared ar eversed labeled riboswitch with 13 C, 15 N-labeled ATP, CTP,G TP and UTP together with unlabeled XTP. With this isotopically labeled RNA sample, we performed 13 C and 15 N-filtered 1D 1 H-NMR experiments that suppress signals from hydrogen atoms bound to 13 Ca nd 15 N. In this experimental setup, we detected the single 12 C, 14 N-"labeled" xanthosine, in particular the signals annotated XH1 and XH3 imino-and the XH8 aromatic signal in the background of the other 73 nucleotides of the riboswitch RNA ( Figure 5E). Thus, transcribing RNA from as sDNA template that contains the kappa nucleotide can yield large functional RNAs, as both selectivitya nd specificity are very high. Purification yields ah ighly homogenous RNA sample suitable for NMR spectroscopy.

Post-transcriptional functionalization of RNA via click reaction
Site-specific introduction of an on-natural nucleobase can be exploited forlabeling purposes. [4] We therefore used the 7-deazaxanthosine (7dX) derivative containing at erminal alkyne in the form of at riphosphate during in vitro transcription to introduce aposition-selective spectroscopic label.
Transcription of the kappa-modified Gsw 73 DNA-template with the alkyne-XTP 13 resulted in a1 :1 ratio of full-length . Overlaid imino-amino-and imino-sugar region of 1 H, 1 H-NOESY spectra of the 14mer(grey)a nd 14mer X (red) RNA displayed on the left. The assignment of the 14merhas been transferred from Fürtig et al. [32] The signals of the G9X modificationa re significantly shifted and no crosspeaks can be detected for the X9, whereas the cross signalso fG 2, G10 and G12 are mostly retained.T his findingh ints towards ar estructuringi nthe loop region by conservation of the hairpin structure. Experimental condition: 1mm RNA in 25 mm potassium phosphate buffer pH 6.2 and 10 %D 2 Ow ere measured at 278 Ka nd 600 MHz.
G79-7dX to abortion fragment (67 nt), with at otal yield of 34.5 %f ull-length RNA compared to the unmodified riboswitch ( Figure 6A,F igure S4,T able S3). As expected, the rate of misincorporation in absence of 7dXTP remainsc omparable to the misincorporation observed in absence of XTP under the respectiveo ptimized conditions for both transcripts. The modified 73mer RNA wasp urified by extraction from ap olyacrylamide gel. For labeling, ac lick reaction with Cy3-Azide was performed, and the resulting RNA analyzed by PAGE. The signals shown in the gel lanes for RNA stainings howedt hat the RNA remained intact duringl abeling, and no significant shift was observed. For RNA that contained the alkyne-modified xanthosine, afluorescentsignal could be observed comigrating with the unlabeled RNA, demonstrating ah ighly specificl abeling (Figure6B). The fluorescently labeled RNA was further excised from the gel, anda nalyzed using UV/Vis spectroscopy. Based on the intensities at the absorption maxima (DNA: 260 nm, Cy3:5 55 nm), we calculated the labeling yield of the RNA to be 11 %( Figure 6C).

Conclusion
We prepared two well-structured RNAs-a model 14mer as well as a7 3nta ptamer of ar iboswitch-via run-off transcription and introduced as ingle guanosine-to-xanthosinem utation, in each case yieldinga100 mm NMR sample. We show that the structuralp erturbations introduced into two different RNA constructs by the unmodified xanthosine are minor.W e furthermore devisedamodifiable 7-deazaxanthosine triphosphate carrying am odifiable linker for posttranscriptional attachments of spectroscopic labels for example, for FRET-, EPRor IR-measurements. Due to the demanding synthesis of this compound, the final triphosphorylation currently limits the obtainable yield, motivating the urgent need for novel synthetic routesf or triphosphorylation. [38] We showt hat our approach facilitates labeling of biologically relevant RNAs (i.e. riboswitches) in ap osition-selective manner.W ith the synthesis of both the kappa DNA phosphoramidite and am odifiable RNA-xanthosinetriphosphate derivative, we provide ap roof of concept to employ non-natural, Watson-Crick-like base pairs for sitespecific bioorthogonal labeling of RNA.

Experimental Section
Solid-phase synthesis of the kappa (k k)-containing DNA template The phosphoramidite containing the k-nucleobase has been incorporated into the DNA via solid-phase synthesis by commercial DNA synthesis (IBA Lifesciences;Gçttingen, Germany). . Again, astrongshift for the iminos of the mutated position and minor shifts for the imino signals of the nearby nucleotidescan be observed. Here,t he cross signals for the X79 are detectable showing complete structural conservation. Experimental condition:G79X (100 mm)w as recordeda t9 50 MHz in 25 mm potassium phosphate pH 6.2 containing5 0mm KCl and 10 %D 2 OG sw 73 was recorded by JaninaBuck( unpublished data) in the same buffer at 900 MHz. The partial assignment has been transferred from Buck et al. [36] E Imino and aromatic region of 13 C, 15 NA ,C,G,U-labeled G79X. Bottom:spectraw ithout x-filter, middle: 15 N-/ 13 C-edited spectra, only hydrogens bound to 15 No r 13 Ca re visible, top: 15 N-/ 13 C-filtered spectra, only hydrogens boundt ot he unlabeled xanthosine are visible. Experimental condition:The RNA was in a25mm potassium phosphatep H6.2 containing 50 mm KCl and 10 %D 2 Of or the imino spectra and 100 %D 2 Of or the spectrao ft he aromatic region.S pectra were recorded at 600 MHz. The protein was expressed in BL21(DE3)cells carrying ap olyhistidine (His 6 -Tag). Expression was induced at OD600 = 0.6-0.8 with 0.5 mm isopropyl-d-1-thiogalactopyranoside (IPTG). T7 RNA Polymerase was expressed over night at 37 8Ca nd were purified via Hi-sTrap HP(GE Healthcare) columns using 50 mm Tris/HCl (pH 8.1), 400 mm NaCl, 20 mm imidazole and 5mm b-mercaptoethanol as lysis buffer.The protein was further purified via size exclusion chromatography with ab uffer containing 20 mm sodium phosphate (pH 7.7) 150 mm NaCl, 1mm EDTAa nd 5mm DTT. In vitro transcription 40 mm tris glutamate (pH 8.1), 2mm spermidine, 20 mm DTT and 10 ng mL À1 P266L T7 RNA polymerase were used for all the transcription reactions. The other reaction components have been optimized for the 14mer to 45 mm Mg 2 + ,1m m NTPs (each) and 300 nm DNA template 1a nd for the 14mer X to 10 mm Mg 2 + ,2m m NTPs (each), 1mm XTP,a nd 200 nm DNA template 2. The transcription reactions of Gsw 73 ,G 79X and G79-7dX were performed with the same concentration of tris glutamate, spermidine, DTT and RNA polymerase. Gsw 73 was transcribed from 300 nm DNA template 3i npresence of 35 mm Mg 2 + and 6mm NTPs (each). For the in vitro transcription of G79X 200 nm DNA template 4w ere used and 35 mm Mg 2 + ,6m m NTPs (each) and 3mm XTP (Jena Bioscience) were added. The transcription of G79-7dX was performed in presence of 200 nm DNA template 4, 35 mm Mg 2 + ,6m m NTPs (each), 15 %D MSO and 2.5 mm 7-deazaxanthosine derivative. The relative yield was determined with ImageJ [39] or ImageLab software after polyacrylamide gel electrophoresis (PAGE).

RNA purification
The 14mer transcription reaction was purified using the protocol described by Helmling et al. [40] The G79X/-7dX transcription reaction was desalted with ddH 2 O using centrifuge concentrators (5 kDa MWCO cut-off, Satorius), subsequently gel purified with a1 5% denaturing polyacrylamide gel. The desired fraction were extracted from gel by incubating in 0.3 m NaOAc over night at 37 8C. The supernatant was separated and the dissolved RNA was precipitated by addition of 2.5 V9 9.5 % EtOH (Carl Roth) and storage at À20 8Cf or 2h.T he RNA was collected by centrifugation at 8500 gf or 30 min and the pellet was dried and dissolved in an appropriate volume of ddH 2 O. For removal of PAAi mpurities, the RNA was HPLC purified, lyophilized, desalted using centrifuge concentrators (5 kDa MWCO cut-off, Satorius) and first precipitated with EtOH (see above), then with 5V 2% LiClO 4 in Acetone (w/v)f or 3times to remove residual salt from HPLC-buffer.I nt he last step, the sample was desalted again and concentrated to 100-300 mm.

Click-reaction
2 mm RNA, 15X excess of Cy-3-azide, 50 %D MSO, 0.3 mm CuSO 4 , 2mm TBTAa nd 5mm sodium ascorbate were mixed and incubated at 20 8Cf or 1h.Asample was separated using PAGE and analysis of the fluorescence signal was carried out with aT yphoon9400. After further PAGE purification of the RNA, the click efficiency could be determined via UV absorption measurement using e 260 of 796.5 Lmmol À1 cm À1 for the RNA and e 555 of 150.0 Lmmol À1 cm À1 for Cy3.

CD melting analysis
16 mm RNA in 25 mm potassium phosphate buffer pH 6.2 and 50 mm KCl was provided. 400 mm Mg 2 + was added. The melting curves were recorded between 5-95 8Ca nd reverse at 264.2 nm. SigmaPlot 12.5 was used for the analysis. Ab isigmoidal fit was used to determine the melting points.

NMR-spectroscopy
14mer and 14mer X were measured at 278 Ka nd 600 MHz at 1mm RNA concentration in 25 mm potassium phosphate buffer pH 6.2 and 10 %D 2 O.
The 1 H, 1 H-NOESY of G79X was recorded at 950 MHz on an unlabeled sample (100 mm)i n2 5mm potassium phosphate pH 6.2 containing 50 mm KCl and 10 %D 2 O. As equence with jump-returnecho water suppression [41] was used. For the x-filter experiments, [42] a 13 C, 15 N-A, C, G, Ul abeled sample was prepared. For the imino region, jump-return-echo water suppression was used as well. The concentration was 160 mm in the same buffer as described above and 10 %D 2 O. The aromatic region was recorded with transfer using DIPSI2 sequence for mixing with 13 C-and 15 N-filter [42][43][44][45][46] and watergate sequences for water suppression. [47] The concentration was 60 mm in the same buffer as described above in 100 %D 2 O. The spectra were recorded at 600 MHz at 308 K.