The Folding Landscapes of Human Telomeric RNA and DNA G‐Quadruplexes are Markedly Different

Abstract We investigated the folding kinetics of G‐quadruplex (G4) structures by comparing the K+‐induced folding of an RNA G4 derived from the human telomeric repeat‐containing RNA (TERRA25) with a sequence homologous DNA G4 (wtTel25) using CD spectroscopy and real‐time NMR spectroscopy. While DNA G4 folding is biphasic, reveals kinetic partitioning and involves kinetically favoured off‐pathway intermediates, RNA G4 folding is faster and monophasic. The differences in kinetics are correlated to the differences in the folded conformations of RNA vs. DNA G4s, in particular with regard to the conformation around the glycosidic torsion angle χ that uniformly adopts anti conformations for RNA G4s and both, syn and anti conformation for DNA G4s. Modified DNA G4s with 19F bound to C2′ in arabino configuration adopt exclusively anti conformations for χ. These fluoro‐modified DNA (antiTel25) reveal faster folding kinetics and monomorphic conformations similar to RNA G4s, suggesting the correlation between folding kinetics and pathways with differences in χ angle preferences in DNA and RNA, respectively.


Introduction
G-quadruplexes (G4s) are four-stranded oligonucleotides formed both by RNAa nd DNAG -rich sequences that are stabilised through Hoogsteen hydrogen-bonds and binding to monovalent cations,i np articular K + .G 4-forming sequences are abundant in the human genome and their roles in gene regulation have been widely demonstrated. [1] Due to their location in telomeres and in promoter-regions of oncogenes, they have become increasingly important as potential drug targets. [2] DNAG 4s are polymorphic and their structures differ in terms of strand orientation, conformation of the glycosidic torsion angle c (syn, c = 40-808 8,o ranti, c = 180-2408 8) (Scheme 1) and geometry of loops connecting the G-rich tracts. [3] Previously,wehave shown that the K + -induced folding of human telomeric DNAG 4s undergoes kinetic partitioning and long-lived intermediates are populated. [4] Insight from these experimentally observed K + -induced folding investigations is supported by MD simulation. [5] Forakinetic partitioning folding mechanism, abifurcation of the folding pathway is characteristic for DNAG 4. A kinetically favoured long-lived intermediate is formed that first has to unfold before adopting the thermodynamically most stable state.K inetically favoured and multiple thermodynamically stable states are populated, in line with the structural polymorphism of DNAG4s at equilibrium. [4] Comparatively little is known about the kinetics and folding pathway of RNAG4, even if RNAG4s have first been observed already in 1991. [6] RNAG4s have been identified in 5'-UTRs, [7] 3'-UTRs, [8] in introns [9] as well as transcriptome of the telomeres (telomeric repeat-containing RNA [10] short: TERRA). In vivo studies linked translation regulation, [11] splicing regulation [12] and chromatin remodelling [13] to the existence of RNAG4s.However,their existence in vivo is still matter of debate,b ut growing evidence has recently been reported for various G4 RNAf unctional roles. [8,14] In contrast to DNAG 4s,R NA G4s are monomorphic irrespective of salt-condition, molecular crowding or capping structures.M ost structures determined thus far (pdb-codes: 3MIJ,2 M18, 2KBP,2 RQJ,6 GE1, 6K84 6JJH, 4XK0) reveal RNAG 4s to adopt an all-anti all-parallel structure (Scheme 1B). [15] Ther eason has been debated but not finally understood. [16] Only in case of G-rich aptamers as spinach and mango,exceptions have also been published. [17] They contain discontinuous G-stretches and long loops that can influence the G4 structure by formation of further stabilising secondary structures. Thei nvestigation reported here were motivated by the hypothesis that RNAG 4f olding should be faster, since the change from the more stable anti conformation for unfolded guanosine nucleotides to the less stable syn conformation adopted in antiparallel or hybrid G4 architectures would not be required. Further, the folding should be monophasic,since then intermediates on the RNAG4f olding pathway are not expected.
By application of time-resolved NMR following K +induced folding of RNAG 4, we indeed show here that the folding landscapes of human telomeric DNAa nd RNAG 4 are fundamentally different. Thekinetics of RNAG4f olding is in fact faster than DNAG4f olding kinetics.Bycomparison with 19 F-modified DNAi na rabino configuration that also adopts anti conformations,w ep rovide evidence that these altered folding kinetics are linked to the conformation around the glycosidic bond angle c.

Folding Kinetics of wtTel25 and TERRA25 G4s
G4s are not only stabilised by Hoogsteen hydrogen bonds but also by monovalent cations including K + and Na + which are essential for G4-formation. Therefore,G4f olding can be induced by addition of KCl and the build-up of NMR-signals resonating at % 11.5 ppm characteristic for Hoogsteen basepairs can be analysed. Ac areful sample preparation is required to avoid G4-promoting cations beforehand. Arapid mixing device [18] was used to inject aKCl-solution in situ into the NMR tube and folding experiments were conducted in atemperature range between 283 Kand 298 Kfor three of the four systems.D ue to ap oor signal-to-noise (S/N) of TER-RA25skinetic traces at 298 K( Figure S1), however, only the measurements at 283 Kcan be faithfully analysed. Previously, we investigated the folding kinetics of DNAG 4s Te l24 (TT(GGGTTA) 3 GGGA) and other related DNAs equences with different flanking nucleotides.T he DNAG 4f olding kinetics is always biphasic and involves the formation of along-lived kinetic conformation. Thenature of the flanking nucleotides determines the folding rate and the population ratio of major to minor conformation at equilibrium. [4] Here,weinvestigated folding of TERRA25 RNAG4and the homologous wtTel25 DNAG 4, as they have identical sequence but for the RNA(U) to DNA(T) substitution. wtTel25 undergoes biphasic folding kinetics resulting in the formation of at hermodynamically favoured major and ak inetically more rapidly formed, but less stable minor conformation with population ratios 0.6:0.4 at 283 K ( Figure 1A,B and Figure S2, Table S1), in line with previous studies for wtTel26. [4] ThePhan group has identified the major conformation of wtTel25 as hybrid-2 (HT2, Figure 4B). In analogy to our previous investigations,t he minor population of wtTel25 most likely adopts ahybrid-type structure but this has not been assigned so far. [4,19] Forthe RNAG4TERRA25 ( Figure 1C,D and Figure S3, Table S1), we observed 3-4 times faster folding (k 1 = 1.45 AE 0.40 min À1 at 283 K) than for the DNAG 4w tTel25 (k 1 = 0.41 AE 0.32 min À1 at 283 K). Thek inetic traces at T = 283 K show ar apid kinetic phase and as low signal equilibration towards the equilibrium population over several hours.B y NMR, we observe asingle folded state for TERRA RNAG4s. As econd slow phase with amplitude change of % 5% is apparent in the kinetic trace ( Figure 1D). We attribute this slow second phase of low amplitude observed at high NMR concentrations to be caused by dimerization of TERRA25 at low temperature.I nn on-denaturing polyacrylamide gel electrophoreses (PAGE) ( Figure S4) and CD-melting curves ( Figure S14), we observed such dimerization tendencies for TERRA25 with amelting point of dimer around 18 8 8C. Thus, this second slow phase for TERRA G4 reports on dimerization that is not observed in the DNAwtTel25.  [20] DSS was used as internal reference.
Titration with KCl to wtTel25 and TERRA25 We performed 1 H-NMR titration of KCl to wtTel25 and TERRA25 to determine their K + -binding cooperativity. Already the unfolded K + -free states reveal differences for wtTel25 and TERRA25. In the absence of K + ,only unspecific Hoogsten base pairs (broad signal between 10.5-11.2 ppm) can be detected for DNAw tTel25 while TERRA25 shows signals of the parallel folded G4 as well as signals stemming from another conformation visible in the region of 11.65-11.85 ppm even in the absence of K + revealing partial prefolding of quadruplex-like structure (indicated with stars in Figure 2). ForT ERRA25, unfolded conformations are converted into the folded state already upon addition of only 2.0 equiv KCl. In general, TERRA25 has ah igher binding cooperativity towards K + as wtTel25. Complete folding for TERRA25 was reached at 4equiv KCl, and only at 32 equiv for wtTel25.
As imilar behaviour was previously observed in CDtitration on comparable human telomeric DNAa nd RNA sequences. [21] We thus propose that TERRA25 rapidly undergoes hydrophobic collapse as no syn/anti-rotation is required to adopt the stable parallel G4 structure.
Interestingly,f urther chemical shift perturbations (CSPs) can be observed by higher K +-concentration for signals assigned to the lower and upper tetrads (around 11.3 ppm) ( Figure S5-7). 2D data revealed the existence of asecond and probably at hird long-lived conformation at lower K +concentration presumably due to differences in the capping structures ( Figure S8,9).
c Angle Conformation of Guanosine Residues in the Unfolded States of G4 DNAG 4a nd RNAG 4a dopt different quadruplex structures.R NA G4 exhibit parallel G4 strand orientations (Scheme 1B). In this strand arrangement, all glycosidic torsion angles c adopt an anti conformation. If the conformation of c angles for guanosine residues in the K + -free unfolded state of G4 RNAs was also anti,then slow anti/syn conformational transitions do not have to take place during the K +induced folding of G4 RNAs.
Very early on, conformational preferences of phosphorylated guanosines and 2'-deoxyguanosines in mononucleotides and in unfolded states of RNAa nd DNAh ave been investigated in theoretical studies by Olsen [22] proposing the presence of both syn and anti conformations,b ut systematic experimental investigations are missing. Thus,w eh ere determined vicinal 3 J(C8,H1')-and 3 J(C4,H1')-coupling constants that depend on c ( Figure 3A). 3 J(C8,H1')is4.5 Hz for both, syn and anti conformation, but 3 J(C4,H1')i nsyn conformation is 6.0 Hz, and thus larger than the 3 J(C4,H1')  [20] Ac oncentration of 100 mMD NA or RNA was used in 25 mM BisTris pH 7.0 and KCl was added stepwise. Equivalents correspond to strand concentration. 2hfor DNA G4 and 25 min for RNA G4 have been applied at room temperaturef or equilibration, wtTel25 and TERRA25 show differencesi nthe K + -free state and in the cooperativity upon addition of K + .S ignals arising from another conformation than the parallel G4 of TERRA25 were marked with stars in B). The stable E-configuration of the N=N-azo bridge (AzoI) forms an antiparallel tetramolecular G4 with four tetrads that switches into the unfolded Z-configuration upon UV-irradiation. [25] G1* is 13 C, 15 N-labelled C) Regionso flong-range 13 C-HMQC-spectra showing the G1C8H1' and G1C4H1' cross peak region of unfolded AzoG4 and folded AzoG4. Residue G1 adopts a syn conformation (pdb-code: 2N9Q) in the G4 DNA which results in astrong C4H1' cross peak and aweak C8H1' cross peak. The reverse is observed in the unfolded conformation. Experimental conditions:260 mMAzoG4 in 25 mM dTris·HCl (pH 7) containing 15 mM KCl in 100 %D 2 O, DSS was used as reference substance.

Angewandte Chemie
Forschungsartikel of 2.0 Hz reported for anti conformation. [23] Qualitatively,i f the C8H1' coupling is larger than the C4H1' coupling,t he nucleotide adopts (predominantly) an anti conformation while if the 3 J(H1',C8) coupling is smaller than 3 J(H1',C4) coupling,itadopts (predominantly) a syn-conformation (Figure 3A). Fort he mononucleotides rGMP [24] and dGMP ( Figure S10,11, Table S2), the anti conformation is the predominant conformation. We further measured these 3 J(C,H) couplings for aD NA G4 with as ingle 13 C, 15 N-labelled guanosine whose conformation can be triggered by light between the folded G4 DNAc onformation and an extended unfolded conformation in presence of KCl. Thephotoswitching behaviour of this model G4 DNAr elies on the configuration of an azo-group that switches between the stable (E) conformation that enables G4 folding and the (Z)conformation upon UV irradiation that unfolds G4. [25] TheG4structure has been determined by NMR (pdb-code:2 N9Q) and it is known that the 13 C, 15 N-labelled G1 (red color code in Figure 3B)a dopts syn-conformation in the folded state. [25] Theobserved significant change of the signal intensities in the long-range HMBC that report on 3 J(C4,H1')a nd 3 J(C8,H1') shows that not only nucleotide monophosphates have an antipreference but also nucleotides in unfolded G4 ( Figure 3C). Since both dGMP and rGMP have ap reference for the antiglycosidic conformation, we assume that RNAnucleotides in unfolded G4 adopt anti conformation as well.
Further,w ep erformed at emperature jump experiment on wtTel25. These experiments reveal no differences between low K + and high K + conditions ( Figure S12,13, Table S3). These observations provide ab asis for discussing our biophysical findings to conditions where folding was induced at constant K + concentrations.

Structure Modulation of DNA G4 by Introduction of 2' 'F-ANA Guanosine Residues
Thec omparison of DNAa nd RNAG 4f olding kinetics reveals faster folding kinetics and no intermediates on the folding pathway of RNAG4. 2'-deoxy-2'F-arabino-modified guanosines (2'F-ANA) forces the nucleobase into anti conformation ( Figure 4A). [26] By studying structure and folding kinetics of this DNAd erivatives,w ec an force the DNA sequence into adopting ap arallel-type G4 with the same sequence as wtTel25 ( Figure 4B). [27] 2'F-ANAGadopts as outheast conformation of the sugar pucker [28] and an antiglycosidic torsion angle to reduce the steric clash between the fluorine atom and the nucleobase. [29] In af irst construct (antiTel25-5F), we substituted those five 2'-deoxyguanosines that adopt syn conformation in the HT2 structure of wtTel25 (G3, G9, G10, G11 and G21) by 2'F-ANAg uanosine ( Figure 4B and D) following previous reports that substitution of guanosines in syn conformation only is sufficient to change the overall structure. [27,30] In as econd construct, we substituted all 2'-deoxyguanosines by 2'F-ANAg uanosines (antiTel25-12F) ( Figure 4E). Thep reviously reported structural change was confirmed by CD spectroscopy (Figure 4C-F) as G4s with hybrid and parallel strand orientation show substantially different CD profiles.
Both modified constructs show the CD-signal characteristics for the parallel conformation with amaximum at 265 nm and aminimum at 245 nm, different to hybrid conformations with maxima at 290 nm and 265 nm and am inimum around 240 nm. [31] Thermal Stability Themelting temperatures (T m )for the four G4 constructs range between 48 8 8Cfor wtTel25 and 91 8 8Cfor antiTel25-12F. Even though two conformations have been determined for wtTel25, only as ingle transition can be observed in CD melting ( Figure S14), arguing for as mall delta T m between minor and major conformation, which is further supported by the near 1:1population of both conformations.AntiTel25-5F has as tabilised major conformation with am elting temperature at 61 8 8C. CD melting curves showed two transitions at 18 8 8Cand at 64 8 8Cfor TERRA25 ( Figure S14) as well as two transitions for antiTel25-12F at 43 8 8Ca nd at 91 8 8C ( Figure S14). Theh igh thermal stability of antiTel25-12F is not surprising.Ithas been previously observed that the introduction of 2'FANA guanosine has astabilising effect on the G4 structure by around 1 8 8Cp er modification but also at hermal stabilisation of 12 8 8Cf or as ingle Gt o2 'F-ANAGmodification was observed. [30,32] Thea dditional stabilisation is determined by an F-H8 pseudo-hydrogen-bond and F-CH-O4' electrostatic interaction. [32] Figure 4. A) Chemical constitution of 2'F-arabino (ANA) guanosine (displayed as red filled circles) in southeast sugar pucker conformation with preference of the anti conformation. Guanosine in syn conformation are displayed as black filled rectangles and in anti conformation as grey filled rectangles. B) Left:t he hybrid-2 (HT2) conformation adopted by wtTel25;right:the proposed parallel conformation of antiTel25-5F that results from the replacementofthe 5g uanosines in syn conformation in wtTel25 by 2'FA NA guanosines. C) Sequence and CD spectra of wtTel25, D) antiTel25-5F, E) antiTel25-12F and F) TER-RA25. The adopted major conformation is shown on the right of each spectrum. Experimentalconditions: 10 mMs ample, 25 mM potassium phosphate buffer pH 7.0, the sample was prepared aday prior to the measurement.
Kinetics of antiTel25-5F and antiTel25-12F We recorded the K + -induced folding kinetics of antiTel25-5F and antiTel25-12F at 283 K ( Figure 5a nd Figure S15,16, Table S1). While the kinetics of antiTel25-5F folding is similar to DNAG 4w tTel25, folding kinetics of antiTel25-12F resembles the kinetics of RNAG 4T ERRA25. Fora nti-Te l25-5F we observe kinetic partitioning of the folding and the rate constants of both conformations (k 1 = 0.38 AE 0.09 min À1 and 0.68 AE 0.16 min À1 at 283 Kf or major and minor (0.75:0.25) conformation, respectively) are not significantly increased compared to the kinetics of DNAG 4 wtTel25. Them inor conformation of antiTel25-5F is less stable and has been fully converted into the major conformation ( Figure S17). By contrast, in antiTel25-12F the formation of any intermediate with Ginsyn conformation is completely supressed and fast folding into one distinct conformation with acomparable folding rate (k 1 = 1.69 AE 0.43 min À1 at 283 K) to the folding rate of RNAG4TERRA25 was indeed observed. Thefolding kinetics of antiTel25-12F show biphasic behavior. CD melting curve ( Figure S14) suggest that antiTel25-12F has similar dimerization tendencya sT ERRA25. In comparison, the stabilised parallel c-MYC22 (PDB:1 XAV) DNAG4has af ast folding rate but the kinetic of its folding still reveal kinetic partitioning and acomparison of tetrameric DNAand RNAG4showed faster assembling of RNAG4. [33] Altogether, the data support the hypothesis of acomplete lack of any guanosine with syn-glycosidic conformation on the TERRA G4 folding landscape.

Reaction Coordinates
Beside the melting point, av antH off analysis of the melting curves was performed to determine DH8 8, DS8 8 and DG8 8 (Table S4). As TERRA25, antiTel-25-5F and antiTel25-12F exhibit ah igh melting point, the baseline might be insufficient for an accurate determination of DH8 8, DS8 8 and DG8 8 and this approach has its limitation when more than two states are involved. So,the data have to be handled with care but give an estimation of the relative free energy of the constructs.Furthermore,the population of the conformations has been determined by integration of well resolved signals in 1 H-NMR spectra ( Figure S18) to calculate DDG of the involved conformations.T ogether with the folding rates,t his complete set of thermodynamic and kinetic data enables us to propose reaction coordinates for the folding landscape of the different G4 constructs ( Figure 6). TheD NA G4 wtTel25 folds into the thermodynamically stable hybrid-2 and asecond kinetically stabilised hybrid structure which have similar free enthalpies (DDG = 0.26 kcal mol À1 ). Thea ntiTel25-5F construct folds into aparallel G4 and forms akinetically trapped intermediate with af inal population ratio major/minor of at least 0.92:0.08 (DDG ! 1.37 kcal mol À1 )a sd erived from the detection limit due to S/N in the 1 H-NMR spectra (Figure S18). wtTel25 and antiTel25-5F follow akinetic partitioning folding mechanism, while the folding landscapes of antiTel25-12F and TERRA25 are funnel-like with ap arallel G4 as final structure.The parallel structure of antiTel25-12F is significantly more stable than the one of TERRA25 due to stabilising-effect of 2'F-ANAguanosine discussed above. [30] Conclusion Human telomeric RNAG4s have afundamentally different folding landscape than human telomeric DNAG4s.F ive anti-directing modifications were introduced in a25mer DNA G4-forming sequence at positions with guanosines in syn conformation allowing us to drive the DNAs tructure from ahybrid towards aparallel G4 conformation as observed for RNAG4s.H owever,r eplacing five deoxyguanosine residues was not sufficient to completely reshape the folding landscape from the typical DNAG 4f olding landscape with kinetic partitioning to the RNAG 4f unnel-like folding landscape. Only if all 12 deoxyguanosine residues are replaced by 2'F-ANAg uanosines the DNAG 4f olding landscape changes from akinetic partitioning to afunnel-like folding landscape. Theexperimental data thus support our hypothesis that RNA G4s have ah igher propensity for the anti-glycosidic conformation both in the unfolded and the folded state and no other than the parallel G4 structure is formed, even not transiently during folding.O nt he contrary,D NA G4 have at hermodynamic preference for the syn-glycosidic conformation. Only if the syn/anti rotation is prevented by constraining all the G-quartet forming nucleotides in anti conformation, the DNAG4f olding behaviour is comparable to the one of RNAG4. Recently,i th as been shown that the folding rate is also influenced by the loop length. [34] Hence, both factors the required syn/anti flipping and the loop length Figure 5. A) Change of the imino signals of antiTel25-5F over time upon KCl addition. B) Change of the signal intensities over time for selected signal of the major conformation( black) and the minor conformation (grey) of antiTel25-5F. The signals are indicated with arrows on top of the 1D 1 HNMR spectrum before (grey) and after (black) KCl addition. The minor conformation of antiTel25-5F is fully converted into the major conformation after at least 1.5 day at room temperature ( Figure S17) C) Change of the imino signal intensities over time of antiTel25-12F after KCl addition. D) Change of the signal intensity over time for aselected imino signal of antiTel25-12F after KCl addition.

Angewandte Chemie
Forschungsartikel have to be taken into account to estimate G4 folding rates and the potential population of long-lived intermediates on the folding landscape.