Probing of G-Quadruplex Structures via Ligand-Sensitized Photochemical Reactions in BrU-Substituted DNA

We studied photochemical reactions of BrU-substituted G-quadruplex (G4) DNA substrates with two pyrene-substituted polyazamacrocyclic ligands, M-1PY and M-2PY. Both ligands bind to and stabilize G4-DNA structures without altering their folding topology, as demonstrated by FRET-melting experiments, fluorimetric titrations and CD spectroscopy. Notably, the bis-pyrene derivative (M-2PY) behaves as a significantly more affine and selective G4 ligand, compared with its mono-pyrene counterpart (M-1PY) and control compounds. Upon short UVA irradiation (365 nm) both ligands, in particular M-2PY, efficiently sensitize photoreactions at BrU residues incorporated in G4 structures and give rise to two kinds of photoproducts, namely DNA strand cleavage and covalent ligand–DNA photoadducts. Remarkably, the photoinduced strand cleavage is observed exclusively with G4 structures presenting BrU residues in lateral or diagonal loops, but not with parallel G4-DNA structures presenting only propeller loops. In contrast, the formation of fluorescent photoadducts is observed with all BrU-substituted G4-DNA substrates, with M-2PY giving significantly higher yields (up to 27%) than M-1PY. Both ligand-sensitized photoreactions are specific to BrU-modified G4-DNA structures with respect to double-stranded or stem-loop substrates. Thus, ligand-sensitized photoreactions with BrU-substituted G4-DNA may be exploited (i) as a photochemical probe, allowing “photofootprinting” of G4 folding topologies in vitro and (ii) for covalent trapping of G4 structures as photoadducts with pyrene-substituted ligands.

Guanine-rich DNA and RNA sequences are capable of folding into non-canonic, four-stranded G-quadruplex (G4) structures, composed of several layers of guanine quartets that are stabilized by hydrogen bonds and coordination to central metal cations. G4 structures play important roles as regulatory elements in DNA-and RNA-related processes such as recombination, replication, transcription and genome maintenance 1,2 . Thus, recent bioinformatics 3,4 and high throughput sequencing studies 5,6 gave evidence of formation of numerous G4 structures by genomic DNA, with a particularly higher frequency observed in promoter regions of many genes, as well as in telomeres and in untranslated regions of mRNA 7,8 . However, the debate on the in vivo persistence and the exact roles of these structures continues 9,10 . Of note is the high polymorphism of G-quadruplexes which, unlike duplex DNA, can adopt very different topologies depending on their sequence and conditions. In spite of some attempts, the topology and stability of G4 structures can hardly be predicted purely on the basis of sequence information. In addition, a growing body of high-resolution structural data on G-quadruplexes gives evidence that these structures do not always follow the consensus motif (i.e., G n -N i -G n -N j -G n -N k -G n , where n = 2 to 4, i, j, k = 1 to 7, and N = any base) 11 . For example, snap-back G4 structures feature isolated guanine residues that participate in formation of G-quartets, leading to formation of complex folding topologies that can hardly be predicted by bioinformatics algorithms. In this context, experimental probing of formation of G4 structures in the genome, as well as their topology profiling, are still crucial for elucidation of persistence of these structures at various stages of the cell cycle and understanding their biological roles. Along these lines, reactive and photoreactive probes capable of forming covalent cross-links with G4 structures upon thermal [12][13][14][15][16] or photochemical activation [17][18][19] ,

Results
Binding of ligands to various G4-DNA substrates. The affinity of ligands to various G4-DNA structures and selectivity with respect to double-stranded DNA was initially studied through fluorescence-monitored thermal denaturation ("FRET-melting") experiments 33 . In this method, the extent of ligand-induced stabilization of G4-DNA is assessed by monitoring the denaturation of a G4-forming oligonucleotide terminally labeled with a fluorophore (F, 5′-FAM) and a quencher (T, 3′-TAMRA). Moreover, the drop of ligand-induced stabilization of G4-DNA observed in the presence of unlabeled double-stranded competitor (ds26) provides an additional information on the ligand selectivity. FRET-melting experiments were performed with several quadruplex-forming oligonucleotides representative of different folding topologies (Table 1): F-21G-T (hybrid in K + and antiparallel in Na + conditions), F-22CTA-T (anti-parallel), F-CEB25 WT -T (parallel with a long propeller loop), F-CEB25 L111 -T (simple parallel), and F-Pu24T-T (snap-back parallel).  Table S1). In the case of CEB25 L111 (Fig. 2f), both M-2PY and PhenDC 3 induced a lower stabilization effect (∆T ½ = 11 and 13 °C, respectively; cf. Table S1), which can be attributed to the high intrinsic stability of this G4 structure (T ½ 0 = 68.6 °C in the absence of ligands). In all cases, the stabilizing effect of M-2PY was only slightly reduced by the presence of the double-stranded competitor. Altogether, these data indicate strong and selective binding of this ligand to G4-DNA structures, with a significant preference to the hybrid structure formed by the human telomeric sequence (F-21G-T).
In identical conditions, M-1PY as well as control ligands M-1NH 2 and PY-NH 2 demonstrated much lower stabilization and lower selectivity for G4-DNA structures. Interestingly, the stabilization induced by M-1NH 2 was, in most cases, comparable to that induced by M-1PY, indicative of similar affinity of both ligands for G4-DNA; however, the selectivity of the former was lower, as demonstrated by the significant drop of the ligand-induced stabilization in the presence of double-stranded competitor. This gives evidence that pyrene moiety is crucial for providing high selectivity for G4-DNA, as can be seen from rising selectivity in the series M-1NH 2 < M-1PY < M-2PY.
In parallel, the interaction of M-1PY and M-2PY with oligonucleotides was studied by means of fluorimetric titration experiments. In the case of M-1PY, titrations with a G4-DNA (22AG in K + conditions) revealed a complex evolution of emission spectra, namely a dramatic decrease of intensity of pyrene monomer band (λ em = 422 nm) upon addition of only 0.2 molar equiv. of ligand, accompanied by appearance of a broad, long-wavelength excimer band (λ em ≈ 470 nm). Upon further addition of G4-DNA (up to 4 molar equiv.), the monomer emission band was partly recovered, while the excimer band decreased (Fig. 3a). This intricate behavior precluded the determination of binding constants of the interaction; however, the transient formation of the excimer band provides a strong evidence of formation of species with a 2:1 (ligand : G4-DNA) stoichiometry in the presence of excess of ligand. In the case of M-2PY, the evolution of emission spectra was less complex: the excimer band, initially dominating in fluorescence spectrum of the ligand, was gradually decreased upon addition of 22AG, while the intensity of the minor blue (monomer) band was only marginally affected (Fig. 3b). The fitting of titration isotherms ( Supplementary Fig. S2) to the independent-site model allowed the determination of binding constants with a variety of DNA substrates, assuming a 2:1 binding stoichiometry (fitting to a 1:1 binding model was unsatisfactory in all cases). The results (Table 2) demonstrated the highest affinity of M-2PY to anti-parallel G4-DNA 22CTA (K a = 3.9 × 10 6 M −1 ), followed by hybrid G4-DNA substrates 22AG (in K + conditions) and bcl2Mid (K a = 1.8 × 10 6 M −1 in both cases). The lowest affinity was observed with a parallel G4-DNA CEB25 L111 and the self-complementary duplex ds26 (K a = 0.54 and 0.48 × 10 6 M −1 , respectively). Notably, the results of fluorimetric titrations are in a good agreement with FRET-melting data, which demonstrate lowest stabilization of CEB25 L111 (among all G4-DNA substrates) and good selectivity with respect to the double-stranded competitor, ds26.
Binding of ligands or Br U modifications do not perturb the folding topology of G4-DNA. Binding of ligands to G-quadruplex structures may induce conformation changes of the latter, as a consequence of a particularly high affinity of the ligand to one or another quadruplex fold. The most prominent example of this transformation is the shift from the hybrid conformation adopted by the human telomeric sequence (22AG) to an antiparallel form, induced by several potent ligands belonging to bisquinolinium series [35][36][37] . As circular dichroism (CD) spectra of G-quadruplexes are characteristic of their conformation 38 , we recorded CD spectra of several quadruplex-forming sequences used in this work in the absence and in the presence of two molar equivalents of ligands (Fig. 4). At the employed conditions (100 mM K + or Na + ), all sequences showed CD spectra typical of their topology (cf. Table 1). In the case of 22AG, the addition of M-1NH 2 and PY-NH 2 induced small changes in CD spectra obtained in K + conditions, namely a decrease of CD signal at 260 nm, indicative of partial conversion from hybrid to antiparallel form (Fig. 4a); however, this change was not observed with the ligands of interest (M-1PY and M-2PY). In the case of hras-1, addition of all ligands led to a small increase of the negative band observed at 260 nm (Fig. 4e), giving evidence of further stiffening of the antiparallel fold, characteristic for this sequence. In the case of other sequences, none of the studied ligands induced significant changes in CD spectra   in the spectral region characteristic of G4 folding topology ( Fig. 4b-d,f-h). Thus, it can be assumed that the pyrene-substituted ligands M-1PY and M-2PY do not perturb the intrinsic conformations of G4 structures, and can be safely used as G4 topology probes. We also assessed the impact of the T > Br U substitution, required for photochemical studies, on the conformation of G4-DNA substrates. It has been previously observed that partial Br U substitution in the human telomeric sequence (22AG) does not lead to conformation changes of the latter 25,27 ; however, the consequences of a total substitution, as well as the impact of this modification on other G4-DNA structures, have not yet been examined. We recorded CD spectra of a number of Br U-modified quadruplex-forming sequences, belonging to different folding topologies (Table 1), and compared them with the ones of their unmodified counterparts. In all cases, CD spectra of Br U-modified and unmodified substrates were nearly identical ( Supplementary Fig. S2), confirming that Br U modification, as a rule, did not perturb the G4 conformation. Hence, Br U-modified sequences represent an adequate model for studies of ligand-sensitized photoreactions with G4-DNA.
Photoinduced reactions with Br U-substituted G4-DNA: human telomeric sequence. We performed photoreactions by irradiating mixtures of Br U-22AG with ligands (5 molar equiv.) using UVA light (365 nm, 3 or 5 min) in K + conditions (100 mM K + ). After irradiation, samples were briefly heated to induce DNA cleavage at the presumably formed heat-labile sites and analyzed by denaturing polyacrylamide gel electrophoresis (Fig. 5a). Along with pyrene-containing ligands M-1PY and M-2PY, control compounds PY-NH 2 and M-1NH 2 were examined in order to elucidate the importance of each ligand moiety (i.e., polyazamacrocycle and pyrene) for the photochemical reaction. Importantly, in the absence of ligands, the Br U-modified substrate was not affected by UVA irradiation (Fig. 5, lanes 1-3). Likewise, a sample of unmodified 22AG, irradiated in the presence of M-1PY or M-2PY in identical conditions, did not give rise to any photoproducts ( Supplementary Fig. S3), giving a clear evidence that Br U residues were indispensable for ligand-sensitized photoreactions.
According to our initial expectations, samples irradiated in the presence of M-1PY and M-2PY yielded a number of DNA strand cleavage products, evidenced as accelerated bands in PAGE analysis (Fig. 5a, lanes 11-12 and 14-15, respectively); in contrast, control compounds PY-NH 2 and M-1NH 2 did not yield detectable amount of cleavage products, clearly demonstrating the necessity of both parts of the ligand for photoinduced reactions. All photo-induced cleavage sites were located at 5′ ends of nucleotides flanking 5′ of the respective Br U residues, as demonstrated through DMS/piperidine sequencing of a control sample of Br U-22AG (Fig. 5, lane 16). This is in agreement with the cleavage mechanism involving pyrene-sensitized generation of uracil-5-yl radicals, followed by the formation of heat-labile 2′-deoxyribonolactone residues 5′ to Br U residues (cf. Supplementary  Fig. S4) 29,30,39 . Although a total yield of strand cleavage products obtained with M-1PY and M-2PY was comparable (5-7%), the distribution of cleavage sites revealed significant differences between these two ligands: for M-1PY, the principal cleavage sites originated from Br U17 and Br U18 (loop 3 of the G4 structure), whereas in the case of M-2PY most cleavage was due to Br U residues of loop 1 and loop 2.
Surprisingly, along with photoinduced DNA cleavage, we observed formation of a second kind of photoproducts, namely slower-migrating bands (** in Fig. 5), presumably corresponding to covalent adducts formed . However, small amount of the slower-migrating product could be detected with PY-NH 2 when used at a higher concentration (10 molar equiv., Supplementary Fig. S5); this gives an evidence that pyrene residues are involved in its formation. In a separate experiment, we assessed the influence of the ligand-to-DNA ratio and irradiation time on the yield of photoproducts. In the case of M-2PY, the formation of the slow-migrating photoadduct was observed already in the presence of stoichiometric amount of the ligand (~9%), and reached maximal yield (~20%, upon 5 min of irradiation) with ≥2 molar equiv. of M-2PY (Supplementary Figs S6 and S7). However, in the case of M-1PY, much higher amount of the ligand (10 equiv.) was needed to achieve a comparable yield of the photoadduct. This behavior is fully in line with the much higher G4-DNA affinity of M-2PY compared with M-1PY, observed in FRET-melting experiments, and speaks in favor of almost quantitative formation of a complex of the DNA substrate with M-2PY in our conditions. We also observed that optimal yield of the photoproducts was obtained using short irradiation times (M-2PY: 3-5 min, M-1PY: 10-20 min). Upon longer UV exposure (30 min), the photoadduct was degraded, as demonstrated by the disappearance of the corresponding band in PAGE analysis ( Supplementary Figs S6 and S7). This is consistent with the fact that, even in optimal conditions, the total yield of photoproducts does not exceed 30% (photoadduct: ~20%, strand cleavage: ~10%), presumably due to secondary photoprocesses such as photo-oxidation of the pyrene chromophore or light-induced degradation of primary photoproducts.
In the quest of our understanding of ligand interaction with different isoforms 22AG, we also assessed the photoreaction of M-1PY and M-2PY with the antiparallel form of Br U-22AG, predominantly formed in Na + conditions (100 mM Na + ). Similarly to what was observed in K + conditions, both ligands gave rise to two kinds of photoproducts ( Fig. 5b; quantification data: cf. Figure 6b). In the case of M-2PY, the yield of the covalent adduct was slightly higher in Na + conditions (22%) comparing with K + conditions (16%), while the one of cleavage products was not significantly affected (~7%). In the case of M-1PY, the yield of the photoadduct (~4%) was not significantly affected; however, the yield of cleavage products was reduced from 5% (in K + ) to 2% (in Na + ).
Next, we attempted to identify preferential reaction sites for photoinduced strand cleavage and covalent adduct formation with both ligands. For this purpose, the reactivity of each loop was systematically assessed by selectively substituting both thymine residues with Br U, and the photoreactions were performed as described above. The results demonstrated no exclusive reactivity of any loop towards one or another photoreaction: in all cases, both kinds of photoproducts were observed (Supplementary Fig. S8). In the case of M-2PY, the loop 1-modified sequence (22AG L1 ) gave the highest amount of cleavage products both in K + and in Na + conditions, which is consistent with the distribution of cleavage products obtained with the all-Br U-substituted sequence (cf.  Figure 5a). In the case of M-1PY, loop 2-modified sequence (22AG L2 ) gave the smallest yield of cleavage products (≤1%), with 22AG L1 and 22AG L3 giving comparable yields (3-5%). The formation of the covalent adduct was observed with all three Br U-modified loops, with M-2PY giving systematically higher yields than M-1PY. Altogether, these results suggest that the topology of G4 substrate has a certain influence on the formation and distribution of photoproducts; however, the polymorphism of the human telomeric G4 structures makes the results difficult to interpret. Therefore, subsequent experiments were performed with other well-characterized G4 structures.
Ligand-sensitized strand cleavage is specific with respect to loop topology in G4-DNA. In order to evaluate the impact of G4 topology on ligand-sensitized photoreactions, we exploited a panel of Br U-modified analogues of other well-characterized G4-DNA structures (Table 1). Among these substrates, hybrid (bcl2Mid) and antiparallel (22CTA, hras-1) G4 folds gave rise to strand cleavage products upon photoreactions with M-2PY, with the cleavage sites corresponding to the distribution of Br U residues in G4 loops (Fig. 6a). A lower total yield of ligand-sensitized strand cleavage (<5%, Fig. 6b), compared with 22AG, is likely due to a smaller number of reactive Br U residues per substrate (two in bcl2Mid and hras-1, three in 22CTA, six in 22AG). In the case of M-1PY, strand cleavage could only be observed with 22CTA (~4%). In stark contrast, parallel (CEB25 WT , CEB25 L111 ) and snap-back parallel (Pu24T) G4 structures did not undergo strand cleavage in similar conditions (Fig. 6a). At the same time, the formation of slow-migrating covalent photoadducts was observed with all substrates. Quantification of gel electrophoresis data indicated that the adduct yield was systematically higher with M-2PY, compared with M-1PY, and with parallel G4-DNA substrates (up to 27% in the case of CEB25 WT ). Taking into account the structural features of the G4 substrates (Fig. 6c), our observations may be rationalized upon assumption that ligand-sensitized DNA photocleavage is specific for Br U residues of lateral and diagonal quadruplex loops. This model is consistent with the results obtained with 22AG, which forms two interconverting hybrid structures in K + conditions: thus, photoinduced cleavage occurs in the lateral loops 2 and 3 in the hybrid-1 structure, as well as in the lateral loops 1 and 2 in the hybrid-2 structure; as a consequence, all three loops are cleaved in a sample containing a mixture of both forms. In Na + conditions, the photocleavage affects the lateral loops 1 and 3 as well as the diagonal loop 2. Likewise, 22CTA presenting three Br U-containing lateral loops also gives rise to three distinct strand cleavage products. In the case of bcl2Mid and hras-1, each presenting two Br U-residues in one of lateral loops, the cleavage is localized to these residues. In contrast, parallel folds CEB25 WT and CEB25 L111 (which possess only propeller loops) as well as Pu24T (in which three Br U residues are located in three propeller loops), do not yield any photocleavage products.
In order to confirm this hypothesis, we performed additional investigations with two hybrid structures offering interesting combinations of loops, namely bcl2Mid (hybrid-2, loop sequence: llp) and Pu24T (snap-back parallel fold presenting a forth diagonal loop in addition to the three propeller loops, pppd). In contrast to 22AG, these two structures, to the best of our knowledge, are not polymorphic and could allow the discrimination of propeller versus lateral/diagonal loops by ligand-sensitized photocleavage, provided the corresponding loops contain Br U residues. However, while the T > Br U substitution had no impact on the G4 topology, we anticipated that substitution of other bases could potentially disturb the intimate base-base interactions which define the G4 folding, and therefore had to be performed with great care. In the case of bcl2Mid, we deliberately replaced the cytosine C20 located in the propeller loop 3 with a T (or a Br U) residue (cf. Table 1 and Fig. 6c). In the case of Pu24T, we mutated the adenine residue A21 belonging to the diagonal loop 4; a detailed inspection of the Pu24T structure reveals that this adenine residue is not involved in hydrogen bonding interactions with other bases, and therefore may be replaced by another base without perturbing the overall folding motif 40 . The CD spectra of the two mutated sequences, bcl2Mid T20 and Pu24T T21 , as well as their Br U analogues, displayed only minimal differences with respect to the parent sequences ( Supplementary Fig. S2c,j), confirming that their folding topology was not affected. Gratifyingly, ligand-sensitized photoreactions of these sequences provided full support of our hypothesis. Thus, the photoreaction pattern of Br U-bcl2Mid T20 with M-1PY and M-2PY was identical to the one observed with Br U-bcl2Mid (Fig. 7a), giving evidence that Br U20, located in the propeller loop, does not give rise to photoinduced strand cleavage. On the other hand, a photoreaction of Br U-Pu24T T21 with M-2PY resulted in a small, but significant yield (1.2%) of a strand cleavage product with a cleavage site localized 5′ to Br U21 (Fig. 7b), confirming that this Br U residue, located in the diagonal loop, is susceptible to the ligand-sensitized photocleavage. Altogether, these results demonstrate that ligand-sensitized photoreactions may discriminate the topology of different Br U-containing loops within the same G4 structure, provided the substrate is not polymorphic in experimental conditions. Ligand-sensitized photoreactions are specific for G4 substrates. To investigate the selectivity of ligand-sensitized photoreactions for G4 structures with respect to double-stranded DNA, we initially performed photoreactions with Br U-22AG in the presence of increasing concentrations of calf thymus DNA (ct DNA) competitor. In the case of M-1PY, we observed a gradual decrease in the photoproduct yield upon increasing concentration of the competitor: thus, in the presence of 20 mM (phosphate) ct DNA, the yield of photoproducts was nearly zero (Fig. 8a and Supplementary Fig. S9). However, in the case of M-2PY, the yield of both photoproducts (i.e., cleaved DNA and covalent adduct) was not affected even at the highest used concentration of ct DNA (i.e., G4 DNA/ct DNA = 1:180). This result (Fig. 8b) suggests an exceptional selectivity of M-2PY for G4-DNA with respect to double-stranded DNA, and is in full agreement with the results of FRET-melting demonstrating high binding selectivity of this ligand.
Next, we assessed the reactivity of both ligands towards Br U residues in various double-stranded and single-stranded DNA context. Taking into account the fact that bis-naphthalene macrocycles preferentially bind to the sites of pairing defects in double-stranded DNA, such as bulges and homopyrimidine mismatches 41,42 , one could reasonably expect enhanced photoreactivity at such sites. We studied the photoreactions of M-1PY and M-2PY with (i) a self-complementary 26-mer duplex DNA substrate ( Br U-ds26, seven Br U residues per strand); (ii) a hairpin structure with Br U residues both in the stem and in the 5-nt loop parts ( Br U-hp29); and (iii) an analogous hairpin bearing two Br U<> Br U mismatches in the stem part ( Br U-hp29 MM , cf. Table 1). In the case of M-2PY, the analysis revealed no photoproduct formation with either of these substrates (Fig. 8c-e), giving a clear evidence that ligand-induced photoreactions are indeed specific for Br U residues located in loops of G4 structures. However, in the case of M-1PY, we observed formation of a small amount (~2%) of strand cleavage products when the photoreaction was performed with the mismatch-containing substrate Br U-hp29 MM , with cleavage sites localizing to one of the mismatch sites (Fig. 8e) 17,19 . Indeed, a cross-link of two DNA strands would have a much lower mobility, as demonstrated by a control experiment with a 44-nt DNA fragment ( Supplementary Fig. S10). Moreover, the fact that this product was systematically detected in samples that underwent thermal treatment (10 min at 95 °C, cf. Materials and Methods) speaks in favor of a stable, irreversible modification. To determine the nature of this modification, the product isolated from the slow-migrating band was subsequently treated with snake venom exonuclease, an enzyme which cleaves the DNA substrate in the 3′-to-5′ direction, stalling at crosslink and lesion sites 17 . While a control sample of Br U-22AG was totally digested, the photoreaction product gave significant amount of unreacted material as well as two incomplete digest products, observed as slow-migrating bands in PAGE analysis ( Supplementary  Fig. S11). These results give additional evidence that the photoproduct represents an irreversible covalent modification (and not, for example, a non-covalent complex), which can be formed in three different sites, presumably corresponding to the Br U-modified loops of the G4 substrate. Further evidence of formation of a covalent ligand-DNA crosslink was obtained through fluorescence analysis. Indeed, only the slow-migrating band demonstrated blue fluorescence under UV light ( Fig. 9 and Supplementary Fig. S12). Moreover, DNA extracted from this band demonstrated a fluorescence spectrum with three well-resolved maxima at λ = 377, 397 and 419 nm, typical for pyrene derivatives (Supplementary Fig. S13). Altogether, these results provide an evidence that the covalent adduct is formed with the ligand (and not between two DNA strands) and contains at least one pyrene residue. Lastly, we assessed the possibility of formation of a covalent adduct by performing a model photoreaction between PY-NH 2 (c = 1 mM) and 5-bromo-2′-deoxyribouridine (BrdU, c = 10 mM) in an aqueous solution. In the absence of the template effect of G4-DNA, the photoreaction appeared notoriously inefficient despite the high concentration of reactants. However, after a prolonged UVA irradiation (365 nm, 3 h 30 min), LC/MS analysis revealed the formation of three novel, minor peaks characterized by UV absorption at 350 nm and identical m/z values of 629.5 (Supplementary Figs S14 and S15), presumably corresponding to covalent adducts forming in a reaction PY-NH 2 with BrdU upon a loss of bromine atom. Although the structure of these adducts cannot be established at the current stage, the formation of three isomeric products is in line with the presence of several putative reactive sites in the structure of PY-NH 2 , most probably corresponding to α-carbons of amino groups. One plausible mechanism could involve hydrogen abstraction by uracil-5-yl radicals from α-carbon atoms of amino groups in the ligand, leading to generation of α-amine radicals and addition of the latter to nucleobases of DNA.

UV Ligand
Lane + + + + + + + + + + 1 2 3 4 5 6 7 8 9 b) Pu24T T21  with any of non-G4, Br U-modified DNA substrates. The high efficiency and G4-specificity of M-2PY are likely due its higher binding affinity to G4-DNA (comparable with best benchmark G4-ligands), as reflected by FRET-melting experiments. Both ligands can induce two alternative photochemical reactions with Br U-modified G4-DNA, namely DNA strand cleavage and formation of covalent photoadducts. Most spectacularly, the first process is observed exclusively with G4 structures presenting Br U residues in diagonal or lateral loops, such as hybrid, antiparallel, or snap-back G4 folds. The loop-topology specificity of ligand-induced strand cleavage may be explained considering the structural differences of propeller and diagonal or lateral G4 loops, similarly to what was proposed by Sugiyama et al. 25 . Thus, Br U residues of lateral or diagonal loops remain in proximity of neighboring nucleotides, allowing hydrogen abstraction from C1′ of 5′-flanking nucleotides upon photogeneration of uracil-5-yl radicals and the subsequent strand cleavage (Fig. 10, left panel). In contrast, Br U residues of short propeller loops point outwards the G4 core and are remote with respect to C1′ of 5′-flanking nucleotides (cf. Fig. 10, right panel); consequently, hydrogen abstraction reactions leading to strand cleavage are highly unfavorable. Thus, ligand-sensitized strand cleavage may be considered as a sensitive photochemical probe for "photofootprinting" of G-quadruplex topology through discrimination of diagonal and/or lateral versus propeller loops. Conversely, the second photochemical process, namely the formation of covalent ligand-DNA adducts, is observed with every studied G4 substrate, including those presenting Br U residues in propeller loops. We may suggest that, in the absence of neighboring sugar residues predisposed for hydrogen abstraction reactions, the uracil-5-yl radicals of propeller loops, generated upon photoreaction with the pyrene-containing ligand, undergo   46 , illustrating the differences in photoreactivity of Br U residues belonging to lateral ( Br U15, Br U16) and propeller ( Br U20) loops. Nucleobase colors: G, blue; A, green, C, violet; Br U, orange; the red arrows illustrate the distances between C5 of Br U residues and H1′ of 5′-upstream nucleotides.

M-1PY M-2PY
Scientific RePoRTS | (2018) 8:15814 | DOI:10.1038/s41598-018-34141-z a radical addition reaction with the same ligand molecule. Notably, the reactive species (uracil-5-yl radical) is generated on the G4 substrate and, in this sense, the formation of covalent adducts with pyrene-containing ligands described here is fundamentally different with respect to the previously described photochemical probes that covalently target G4-DNA through light-induced generation of ligand-centered reactive species, such as benzophenone diradicals, nitrenes 17,18 , or activated metal complexes 43,44 . Although the mechanistic details of this process could not be elucidated in details at the present stage, a preliminary support of such reaction could be provided by formation of photoadducts in a model photoreaction of BrdU with PY-NH 2 . Beyond the fundamental aspects shedding light on the reactivity of Br U residues in the context of G4-DNA substrates, we may expect that this photochemical process could be exploited for design of novel agents for covalent trapping of G4 structures in longer sequences and potentially applied for identification of such structures in Br U-labeled genomes. Thus, the bifunctional agents described in this work complement the currently available toolbox of photochemical tools for probing DNA structures.  CD spectroscopy. Circular dichroism spectra were recorded with a J-710 spectropolarimeter (JASCO) equipped with a Peltier temperature controller, in quartz cells with a path length of 5 mm. Samples for CD spectroscopy contained 5 µM oligonucleotides in K-100 or Na-100 buffer, in the absence or in the presence of ligands (10 µM) or DMSO control (0.5% v/v). The scans were recorded at 20 °C from 210 to 330 nm using the following parameters: data pitch, 1 nm; bandwidth, 2 nm; response, 2 s; scan speed, 50 nm min −1 ; the reported spectra are the average of four scans.

Materials
Photoreactions. Unless stated otherwise, samples were prepared by annealing a mixture of the 5′-32 P-labeled G4-DNA (100 000 cpm) with 5 µM of the corresponding non-radiolabeled sequence to a total volume of 10 µL in K-100 buffer. The samples were incubated for 1 h at 37 °C in the presence of ligands (25 µM) and subsequently irradiated for 3 to 5 min with UV light (xenon light source MAX-303, 300 W, 365-nm monochromatic light) by keeping a distance of 10 cm between the light source and the sample in a 1.5-mL centrifuge tube placed on ice (irradiance: 5.44 mW cm −2 ). After irradiation, samples were concentrated in vacuo, supplemented with 10 µL of loading dye (bromophenol and xylene blue in a cocktail of 800 µL of formamide and 200 µL of 0.5 M EDTA) 39 , heated for 10 min at 95 °C to allow the cleavage of heat-labile 2′-deoxyribonolactone residues, and loaded onto the gel. Gel electrophoresis. Samples were analyzed by 20% denaturing PAGE (7 M urea, 1X TBE as running buffer) at 1000 V (run duration: 4 h). After exposure to a phosphor-imaging screen overnight, the gels were visualized with Typhoon 2000 scanner and quantified using ImageQuant software. The yield of photoproducts was estimated as the ratio of total counts of the photoproduct (crosslink or cleavage) bands to the sum of counts of photoproduct and unreacted DNA bands in the same lane, after background subtraction.