Precise Distance Measurements in DNA G‐Quadruplex Dimers and Sandwich Complexes by Pulsed Dipolar EPR Spectroscopy

Abstract DNA G‐quadruplexes show a pronounced tendency to form higher‐order structures, such as π‐stacked dimers and aggregates with aromatic binding partners. Reliable methods for determining the structure of these non‐covalent adducts are scarce. Here, we use artificial square‐planar Cu(pyridine)4 complexes, covalently incorporated into tetramolecular G‐quadruplexes, as rigid spin labels for detecting dimeric structures and measuring intermolecular Cu2+–Cu2+ distances via pulsed dipolar EPR spectroscopy. A series of G‐quadruplex dimers of different spatial dimensions, formed in tail‐to‐tail or head‐to‐head stacking mode, were unambiguously distinguished. Measured distances are in full agreement with results of molecular dynamics simulations. Furthermore, intercalation of two well‐known G‐quadruplex binders, PIPER and telomestatin, into G‐quadruplex dimers resulting in sandwich complexes was investigated, and previously unknown binding modes were discovered. Additionally, we present evidence that free G‐tetrads also intercalate into dimers. Our transition metal labeling approach, combined with pulsed EPR spectroscopy, opens new possibilities for examining structures of non‐covalent DNA aggregates.


Introduction
DNAG-quadruplexes,f ormed from p-stacked tetrads of Hoogsteen hydrogen-bonded guanines,p lay important roles in several biological processes,i ncluding regulation of oncogene expression and maintenance of telomeric repeats upon cell division. [1][2][3][4] Beyond their biological role,G-quadruplexes gained al ot of interest as as tructural motif in the field of DNAnanotechnology. [5][6][7] G-quadruplexes tend to form higher-order structures such as dimers, [8][9][10][11] G-wires, [12,13] and other motifs,w hich are thought to influence their function in vivo. [14][15][16][17] It is,for example,not yet understood if and how the multitude of G-quadruplexes in direct neighborhood interact with each other within the telomeric overhangs. [17][18][19][20][21][22] Due to their regulatory function in pathological processes, G-quadruplexes have been identified as interesting drug targets in anticancer research. [23][24][25] Many small molecules [26] and metal complexes, [27] most of them possessing flat psurfaces and positive charges,were found to bind and stabilize the folded secondary structure,t hereby acting as potential anticancer drugs. [28] One well-known G-quadruplex binder is the natural product telomestatin, [29,30] which is ap otent telomerase inhibitor due to its strong interaction with unimolecular G-quadruplexes found in the human telomeric sequence. [31][32][33] Another typical class of G-quadruplex binders is based on perylene diimides,s uch as N,N'-bis [2-(1-piperidino)ethyl]-3,4,9,10-perylenetetracarboxylicd iimide dihydrochloride (PIPER). [34][35][36] An early NMR-based investigation revealed the formation of a1 :2 sandwich complex of PIPER with tetramolecular G-quadruplexes,w here the dye intercalates between two terminal G-quartet sites. [37] Such ligand-mediated or direct non-covalent contacts between terminal Gquartets are anticipated to form and play an important role for the overall stability of quadruplexes,a nd thus their biological function. Therefore,a nalytical methods that provide key structural information about these p-stacked aggregates are required.
In the past years,p ulsed dipolar electron paramagnetic resonance (PDEPR) methods [38,39] have evolved into reliable and versatile tools for structure determination in structural and chemical biology.Pairs of paramagnetic centers based on either organic radicals or open-shell transition metal complexes are required for distance measurements in the nanometer range,providing valuable information on the structure of biomolecules. [40] Thep otential of Cu 2+ ions as spin labels for PDEPR-based studies has been well documented in the literature over the past years. [41][42][43][44][45][46][47][48][49][50][51] While PDEPR is most commonly used to determine structural constraints in protein systems,i ts application to distance measurements within nucleic acids has been steadily expanding. [51][52][53][54][55][56][57][58] So far,f ew reports describe PDEPR-based investigations on G-quadruplexes, [59][60][61] including in-cell spin-labeled G-quadruplexes [62] and quadruplex-metal complex adducts. [63,64] Recently,w ei ncorporated new Cu 2+ -based spin labels into tetramolecular DNAG -quadruplexes of varying Gtetrad count, which allowed us to determine intramolecular distances within the secondary structure with unprecedented accuracy. [65] Key to obtain such precise data was alabel design in which the magnetic orbitals of the square-planar coordinated Cu 2+ cations were fixed in defined spatial orientations by equipping the four-stranded DNAc onstruct with as et of four nitrogen donor ligands,f orming ar igid chelate environment.
Thea pproach was based on the concept of metalmediated base pairing,w here canonical nucleotides are replaced by artificial ones,c arrying al igand functionality to coordinate to transition metal cations. [66,67] In earlier studies, we transferred this concept from duplex DNAt oG -quadruplexes.P yridine or imidazole donors were covalently incorporated into defined positions of G-quadruplex structures,e quipping the folded strands with prearranged chelate environments suitable for binding transition metal ions such as Co 2+ ,N i 2+ ,C u 2+ ,o rZ n 2+ . [68][69][70][71][72] Substantial thermal stabilization of the metal ion-bound structures was observed, and the system allows for metal-induced control of its folding topology and protein binding behavior. [73] Here,w ee xtend the use of the Cu(pyridine) 4 tetrad as arigid spin label to detect intermolecular Cu 2+ -Cu 2+ distances in G-quadruplex dimers of different spatial dimensions.T his approach also revealed new binding modes in related sandwich complexes with PIPER, telomestatin, and Gquartets assembled from free guanines.O ur study delivers valuable information on binding stoichiometry and key structural parameters,which are in excellent agreement with MD simulation results.

Results and Discussion
Synthesis and Characterization of Cu 2+ -Binding G-Quadruplexes Fort his study,s ix short modified oligonucleotides were synthesized by solid-phase DNAs ynthesis (oligos A-F, Table 1). Thes equences were chosen based on two known tandem repeat units found in the telomeric regions of different species (TTAGGG and TTG GGG), as these oligonucleotides and related tetramolecular G-quadruplexes are already well investigated. [9,11] Pyridine-modified nucleotides (L,F igure 1A)w ere incorporated next to the 5'-Gquartets,f orming rigid Cu 2+ spin labels at the 5'-ends while leaving the 3'-G-quartets exposed (oligos A and B)o r blocked by additional thymidines (oligo C). Also,i someric sequences of reverse order,carrying an unobstructed G-stack at the 5'-end and the Cu 2+ modification at the 3'-termini, were synthesized (oligos D-F).
We first examined the formation of G-quadruplex structures using CD spectroscopy.For all six oligonucleotides,with ah igh Na + or K + concentration at pH 7.2, formation of ap arallel G-quadruplex topology was indicated in the absence as well as in the presence of Cu 2+ ions ( Figure 1).
In addition, UV-based thermal difference spectra and thermal denaturation profiles confirmed G-quadruplex formation. For all oligonucleotides,t he presence of Cu 2+ ions caused as ignificant increase of the thermal denaturation temperatures,a st he formation of aC u(pyridine) 4 complex was previously shown to raise the overall stability of the hybridized quadruplex structure [68,69] (e.g. DT 1/2 = 17 8 8Cf or [Cu 2+ @B 4 ]i nN aCl-containing solution, Figure 1D;f or further CD and UV spectroscopy results see Figures S3-S23). However,w en ote that the usual set of CD and UV experiments gave no hints on the potential presence of dimeric species.
In the next step,E PR-based investigations were performed. Samples containing 250 mm G-quadruplex monomers (1 mm oligonucleotides) and 375 mm CuSO 4 (1.5 equiv per quadruplex) were prepared in 50 mm potassium phosphate buffer (pH 7.0), then mixed with glycerol (1:1 v/v) and immediately frozen in liquid N 2 .T he final G-quadruplex monomer concentration was 125 mm for all EPR samples investigated in this work.
Atypical field-swept EPR spectrum of the Cu 2+ spin label fixed in its spatial position by the four pyridine ligands is shown in Figure 2A (for [Cu 2+ @A 4 ]). Theb est fit was obtained using the following spin-Hamiltonian parameters: g k = 2.268, g ? = 2.063, A k = 545 MHz. These values are in good agreement with parameters for aC u 2+ ion coordinated by four nitrogen atoms. [74][75][76] We note that the choice of the modified oligonucleotide resulted in slight variations in the EPR line shape of the spin label ( Figure S25). In agreement with previous studies, [51,77] unbound Cu 2+ ions in solution were nearly EPR-silent at pH ! 7a nd provided only an egligible contribution to the overall EPR line shape ( Figure S26). Table 1: Sequences of ligand-modified oligonucleotides used in this work.

Name
Sequence

Orientation-Selective PDEPR Reveals G-Quadruplex Dimers
In each G-quadruplex, one terminal site was modified to carry the metal complex, while the other end remained unmodified to allow stacking of the terminal G-tetrads as observed in unmodified sequences. [8,9,11] PDEPR experiments were performed to investigate dimer formation by detecting dipole-dipole interactions,a nd thus measuring distances between the two paramagnetic Cu 2+ ions residing in the pstacked G-quadruplex monomers.B oth double electronelectron resonance (DEER, also known as PELDOR) [78,79] and relaxation-induced dipolar modulation enhancement (RIDME) [80,81] techniques were employed, the former providing more robust background correction and the latter resulting in larger modulation depths (D)and being generally less prone to orientation selectivity. [40,50] Forrigid, orientationally correlated spin pairs,orientation selectivity in PDEPR leads to ad eviation of dipolar spectra from aP ake pattern and to the dependence of the dipolar frequency on the selected g-tensor orientations. [82] Therefore, these experiments provide atomic-level structural information on the geometry of the spin system. Some examples include tyrosyl radicals in ribonucleotide reductase, [83] spinlabeled DNA, [57] Co 2+ -a nd Fe 3+ -containing synthetic systems. [84,85] Since the Cu 2+ spin label within aG-quadruplex monomer is fixed in ah ighly rigid fashion, [65] and the total width of aCu 2+ EPR spectrum (148 mT,about 4GHz) is significantly larger than at ypical pulse bandwidth (ca. 50 MHz), orientation selectivity is expected to affect PDEPR dipolar spectra. Indeed, both DEER and RIDME data acquired at different field positions for quadruplexes composed of oligos A, B, D and E showed strong dependence of the dipolar frequency on the excited g-tensor orientation, displaying the rigidity of the systems ( Figures 2B and S28-S33). While there was ac lear difference between RIDME and DEER modulation depths, the two methods produced almost identical dipolar spectra, particularly at orientations that correspond to the distance vector perpendicular and parallel to the magnetic field vector ( Figure S29). In the text below,w ef ocus on the analysis of DEER-derived Cu 2+ -Cu 2+ distances.
PeldorFit, [86] which explicitly takes orientation selectivity into account when deriving distances,was used to analyze the PDEPR data. It allowed us to reproduce DEER time traces including both perpendicular and parallel orientations of the dipolar coupling tensor with one set of parameters (Figure 2B). Importantly,this analysis showed that g z axes of the two Cu 2+ spin labels within aG-quadruplex dimer are aligned collinearly,p erfectly fitting the expected structure for two rigid coplanar Cu 2+ complexes and atight p-stacking interface between the monomers (see Figures S34-S35 and Table S2 for details).
Analysis of DEER data from the [Cu 2+ @A 4 ] 2 sample revealed as ingle Cu 2+ -Cu 2+ mean distance of d A = 2.55 nm with av ery narrow distance distribution of s = 0.02 nm (s is the standard deviation of the distribution, which is assumed Gaussian, Figures 2a nd S30). Thed etected distance was in the expected range for aG-quadruplex dimer formed through tail-to-tail stacking of the 3'-terminal G-tetrads.F or asample of dimer [Cu 2+ @B 4 ] 2 ,alarger mean distance of d B = 3.21 nm (s = 0.02 nm, Figure S31) was obtained, which was about two p-stacking distances longer than that of the shorter [Cu 2+ @A 4 ] 2 dimer (d B Àd A = 0.66 nm). Since each [Cu 2+ @B 4 ] monomer contains one additional G-tetrad, this result is in perfect agreement with the expected distance.
As controls,G -quadruplexes [Cu 2+ @C 4 ]a nd [Cu 2+ @F 4 ], both carrying obstructing thymidines next to the terminal Gquartets,were probed. DEER time traces showed no dipolar modulation of the Cu 2+ EPR signal ( Figure S36), indicating that no dimers were present in these samples.T his result confirms that extra 3'-o r5 '-terminal thymidines prevent Gquadruplex aggregation in solution. [8,9,11] Our PDEPR experiments demonstrate that av ariety of Cu 2+ -binding G-quadruplex monomers of different lengths, carrying an exposed terminal G-tetrad at either 3'-or5'-end, readily assemble dimers.M ean Cu 2+ -Cu 2+ distances and corresponding distributions for all dimers are listed in Table 2. Thed istance distributions achieved in this work were approximately 5t o1 0t imes narrower than those obtained for DNAa nd RNAs tructures labeled either with nitroxide- [53,55,[57][58][59] or other, less structurally confined Cu 2+ -based [51,63] spin labels.S uch narrow distance distributions not only highlight the pronounced rigidity of our Cu(pyridine) 4 spin label within its G-quadruplex environment, but also demonstrate the overall defined structure adopted by the Gquadruplex dimers investigated in this work.

Sandwich Complexes Based on G-Quadruplex Dimers
Theability to easily distinguish between dimeric species of different lengths with high resolution and accuracy prompted us to employ DEER to investigate intercalation of two wellknown quadruplex binders.
Upon addition of 0.5 equiv of PIPER per G-quadruplex, CD and UV/Vis spectroscopy again confirmed formation of ap arallel G-quadruplex topology for all samples in the absence and presence of Cu 2+ ions.However,the addition of PIPER increased the thermal stability of the secondary structures ( Figure 3). Furthermore,i nt he absence of Gquadruplex DNA, PIPER was not soluble at pH 7.2 and precipitated ( Figure S1). In the presence of folded G-quadruplex DNA, red-colored PIPER stayed in solution, giving rise to ad istinct absorbance signature at 450-600 nm (Figure 3). After thermal denaturation of the secondary structure, PIPER precipitated, and the absorbance signal in the visible region vanished. This observation supported as elective interaction of PIPER with G-quadruplex DNAa sc ompared to single-stranded DNA [87] (see Figures S3-S24 for further CD and UV/Vis spectroscopy results).
EPR samples containing [Cu 2+ @A 4 ] 2 dimers and stoichiometric PIPER concentration showed anew,lower modulation frequency in the DEER data and revealed aC u 2+ -Cu 2+ distance of d P = 2.82 nm, larger than that of pure [Cu 2+ @A 4 ] 2 dimers (d P Àd A = 0.27 nm, Figures S37-S38 and Figure 4A). As imilar result was obtained with samples containing PIPER and [Cu 2+ @B 4 ] 2 dimers ( Figure S39). The increase in the spatial separation between the two Cu 2+ ions demonstrated the formation of sandwich complexes (PI-PER@[Cu 2+ @A 4 ] 2 and PIPER@[Cu 2+ @B 4 ] 2 ), in which the flat organic molecule intercalates between the 3'-faces of the two G-quadruplex monomers. [37] However,the addition of PIPER to both [Cu 2+ @D 4 ] 2 and [Cu 2+ @E 4 ] 2 dimers did not affect the obtained distance distributions,d emonstrating that PIPER does not intercalate into 5'-5'-stacked dimers ( Figure S41), but prefers ad ifferent binding mode.T his result is in agreement with the observed differences in CD signals of PIPER induced by the 3'-a nd 5'-modified quadruplexes ( Figure S24).
Next, we investigated the effect of PIPER concentration on the sandwich complex formation. At aP IPER-to-[Cu 2+ @A 4 ] 2 ratio of 0.5:1, the peak intensities corresponding to pure dimer (d A = 2.55 nm) and PIPER@[Cu 2+ @A 4 ] 2 com-  plex (d P = 2.82 nm) in the dipolar spectrum were equivalent ( Figure 4A), revealing approximately equal concentrations of pure and PIPER-containing dimers.O nce the PIPER-todimer ratio was raised to 1:1, the frequencycomponent of the pure dimer disappeared and only the distance of d P = 2.82 nm was detected, indicating ahigh binding constant for PIPER to the [Cu 2+ @A 4 ] 2 dimer.T hese results suggest am onomerdimer equilibrium far on the side of the dimer,inagreement with literature data for (TTAGGG) 4 at high K + concentrations. [9] Thes ame conclusion may be reached based on the DEER modulation depth parameter D,w hich reflects the number of dimers present in the sample. [88] In our case,since D obtained in the g ? region is not affected by the excess Cu 2+ (Figure S27), it provides information on the dimerization efficiency of G-quadruplexes.U pon addition of one PIPER molecule per dimer, D did not increase,s uggesting that the whole G-quadruplex population was already present in the dimeric form prior to the PIPER addition. Surprisingly,further increase in the PIPER-to-dimer ratio led to the appearance of an ew Cu 2+ -Cu 2+ distance of d 2P = 3.21 nm, about one p-stacking distance longer than that of the PIPER@[Cu 2+ @A 4 ] 2 complex (d 2P Àd P = 0.39 nm). This distance was assigned to as pecies where two PIPER ligands intercalate between the monomers of at ail-to-tail arranged G-quadruplex dimer (2PIPER@[Cu 2+ @A 4 ] 2 ,F igures 4C and S40). To the best of our knowledge,t his binding mode of PIPER to DNAG -quadruplexes has never been described before.Itresembles areported motif found in the solid state, where two naphthalene diimide derivatives intercalate into ah ead-to-head arranged dimer of unimolecular G-quadruplexes. [89] Them odulation depth strongly decreased with an increase in the PIPER-to-dimer ratio beyond 2:1( Figures 4B  and S37). This suggests ad isruption of the 2PIPER@[-Cu 2+ @A 4 ] 2 complex and the formation of am onomeric species,presumably PIPER@[Cu 2+ @A 4 ], which contains only one Cu 2+ center, and thus cannot be detected with PDEPR. There appears to be am arginal decrease in D upon PIPER addition up to 2:1P IPER-to-dimer ratio.S ame has been observed for 5'-5'-stacked dimers,i nw hich PIPER does not intercalate ( Figure S41). These data suggest that up to two equivalents PIPER addition reduces the number of dimers by forming as mall amount of monomeric adducts with quadruplexes.
EPR-based distance measurements were also performed in the presence of telomestatin, [29,30] another well-known Gquadruplex binder.The natural product was described to bind to unimolecular G-quadruplexes,interacting with terminal Gquartets via p-stacking. [31,90] Interestingly,s amples of the [Cu 2+ @A 4 ] 2 dimer containing telomestatin (1 equiv per dimer, Figure S42) revealed as econd peak in the distance distribution with an increased Cu 2+ -Cu 2+ distance of d T = 2.88 nm, with respect to that of the pure dimer (d T Àd A = 0.33 nm). This observation strongly suggests that telomestatin can indeed intercalate into a3 '-3' p-stacked dimer formed from two tetramolecular [Cu 2+ @A 4 ]m onomers with parallel topology to build at elomestatin@[Cu 2+ @A 4 ] 2 sandwich complex. To the best of our knowledge,this is the first time such asandwich binding mode of telomestatin has been observed. Since human telomeric sequences are able to form (unimolecular) parallel G-quadruplexes with exposed terminal G-tetrads, [18,22] the discovered binding mode might play ar ole in the ability of telomestatin to inhibit telomerase.

Intercalation of Free G-Quartets into G-Quadruplex Dimers
Moreover,w ei nvestigated the interaction of derivatives of free guanine with G-quadruplex dimers.F irst, we added guanine (4 equiv per dimer) to samples of [Cu 2+ @A 4 ] 2 and [Cu 2+ @B 4 ] 2 .F or both dimers,t his resulted in the appearance of as econd peak in the distance distribution, accounting for an increase in length of Dd = 0.33 nm as compared to the pure dimers ( Figures S43-S44). Theconcentration ratio of the new extended dimeric species to the original one was approximately 1:1for both samples ( Figure 5, green traces). Thenew distance can be explained by as pecies where aw hole untethered G-tetrad assembled from four free guanines intercalates between the two quadruplex monomers (guanine 4 @[Cu 2+ @A 4 ] 2 and guanine 4 @[Cu 2+ @B 4 ] 2 ,F igure 5E and Table 2).
Next, the experiments were repeated with 7-deazaguanine ( Figure 5, red traces), guanosine (blue traces), and guanosine monophosphate (GMP) under the same conditions (Figures S45-S47). 7-Deazaguanine lacks anitrogen atom necessary for Hoogsteen hydrogen bonding,and thus is unable to form tetrads. [91] As anticipated, its addition did not alter the distance distributions of the pure dimers,i ndicating that not asingle nucleobase,but aquartet acts as intercalating species. Thea ddition of GMP did not alter the distance distribution either,since tetrads formed from GMP are highly negatively charged at pH 7, and thus suffer from electrostatic repulsion. In contrast, guanosine (as free guanosine quartet) was found to intercalate even more efficiently than ag uanine tetrad, as judged by the relative intensities in the dipolar spectra, presumably due to its higher solubility in water. Thus,a ll additional results support the hypothesis of an untethered G-quartet assembled from free guanines intercalating into the G-quadruplex dimers.

MD Simulation of Dimeric G-Quadruplex Structures and Sandwich Complexes
In order to relate experimentally determined distances to structural models,M Ds imulations were performed for each dimeric system. Starting structures were created assuming at ail-to-tail or head-to-head stacking of the terminal Gtetrads,a nd 50 ns MD runs in explicit TIP3P water with 100 mm KCl concentration were conducted (for more details, see the Supporting Information). Thedimeric structures were preserved throughout the simulation time (Figures 6and S48-S51) and the obtained intermolecular Cu 2+ -Cu 2+ distances and distance distributions matched the experimental values very well (Table 2).
Sandwich adducts with PIPER, telomestatin or free guanosine quartets as intercalating species were also simulated. Starting structures were created with typical p-stacking distances between the respective intercalator and the Gquadruplex monomers.A gain, the sandwich structures were preserved during the whole MD run, and the Cu 2+ -Cu 2+ distances agreed extremely well with DEER-derived ones ( Table 2, Figures 7and S52-S58). In the special case of adduct 2PIPER@[Cu 2+ @A 4 ] 2 ,the simulation gave information on the Figure 5. A,B) Background-corrected DEER time traces (g eff = 2.061) for [Cu 2+ @A 4 ] 2 and [Cu 2+ @B 4 ] 2 ,r espectively,i nthe absence and presence of 7-deazaguanine, guanine or guanosine (4 equiv per dimer). C,D) Corresponding dipolar spectra, with mean Cu 2+ -Cu 2+ distances shown on the top axis. E) Sandwich complexesguanine 4 @[Cu 2+ @A 4 ] 2 and guanine 4 @[Cu 2+ @B 4 ] 2 with afree G-tetrad intercalatingbetween the two G-quadruplex monomers. relative orientation of the two PIPER molecules with respect to each other. Throughout the MD run, the relative rotation angle was quite flexible at around 40-608 8 ( Figure S57).
Theo verall agreement between experimentally obtained and MD-derived distances confirmed that PDEPR spectroscopy allows measuring intermolecular Cu 2+ -Cu 2+ distances in p-stacked G-quadruplex dimers and related sandwich complexes with high accuracy.

Conclusion
We have shown that paramagnetic Cu(pyridine) 4 spin labels incorporated into DNAG -quadruplexes are suitable for the investigation of higher-order G-quadruplex structures such as dimers,b ym eans of intermolecular Cu 2+ -Cu 2+ distance measurements using PDEPR techniques.D ue to the unprecedented rigidity of the spin label and the concomitant sharp peaks in the distance distributions,t he method provides as imple readout for the unambiguous characterization of dimers of different lengths and composition. Moreover,i ntercalation of G-quadruplex-binding ligands such as PIPER and telomestatin was clearly demonstrated with the new method, revealing previously undescribed binding modes.
Surprisingly,wewere also able to show that untethered Gquartets,composed of free guanines or guanosines,intercalate into G-quadruplex dimers.T his observation may be relevant for applications exploiting the supramolecular interaction of functionalized G-quartet probes with G-quadruplexes of biological origin. [92] Furthermore,t he introduced methodology herein, based on an easy to synthesize DNAm odification and an established EPR protocol, showcases the possibility to investigate more complex G-quadruplex systems,o ther higher-order oligonucleotide architectures,a sw ell as DNA-protein interactions.W eexpect this method to provide valuable contributions to the investigation of structure and dynamics in both biological and DNA-nanotechnological contexts.