Imidazole-modified G-quadruplex DNA as metal-triggered peroxidase† †Electronic supplementary information (ESI) available: Analytical data, native ESI-MS, ABTS assay, synthesis and MD simulation. See DOI: 10.1039/c8sc05020a

Imidazole-modified G-quadruplex DNA binds transition metals Cu(ii), Zn(ii), Ni(ii) and Co(ii). Metal-triggered association with hemin generates a stimuli-responsive peroxidase mimic.


Introduction
Depending on base sequence and solution conditions, DNA can fold into different secondary structures including canonical double strands as well as triplexes, cytosine-rich i-motifs or Gquadruplexes. 1 The latter ones assemble from guanine-rich sequences to form p-stacked G-tetrads, held together via circular Hoogsteen base pairing patterns containing central monovalent cations, oen Na + and K + . Structurally, Gquadruplexes show a higher diversity than one might expect. While the G-tetrads offer only a small structural variety (anti/syn glycosidic angles), the quadruplexes' overall diversity arises from a choice of different strand molecularities (tetra-, tri-, biand unimolecular), orientations (parallel or antiparallel) and loop lengths.
In recent years, G-quadruplexes moved into the focus of research in chemical biology for their ubiquitous presence in the human genome and their potential relevance for diseases such as HIV and cancer. [2][3][4] On a parallel time scale, the eld of DNA nano-technology emerged, including the sub-discipline of metal-mediated base pairing which studies the replacement of the regular Watson-Crick hydrogen bonding scheme by metal coordination. While few examples comprise the bridging of natural DNA bases by metal cations (such as in T-Hg 2+ -T and C-Ag + -C), a large variety of articial base surrogates was introduced, allowing to serve as ligands for a wider range of transition metal cations. [5][6][7] Eventually, this development lead to the design of longer, mixed metal arrays that facilitated a regioselective positioning of different metals to construct nano-wires. 8,9 Until recently, however, most reported examples have featured duplex DNA and only few publications reported on metalmediated base interaction in higher secondary structures such as triplex DNA, three way junctions, i-motifs and G-quadruplexes. [10][11][12][13][14] In 2013, we reported the rst metal-mediated Gquadruplexes where one G-tetrad was replaced by four pyridine modications, allowing the complexation of Cu(II) and Ni(II). 15 This system was exploited as Cu(II)-sensitive topologyswitch in unimolecular G-quadruplexes, as Cu(II)-activated thrombin inhibitor and for the design of EPR-based (Electron Paramagnetic Resonance) molecular rulers. [16][17][18] In a different branch of DNA nanotechnology, preestablished metal complexes were incorporated into chiral DNA scaffolds for the design of articial DNAzymes. 19,20 Examples include systems capable of DNA ligation or hydrolysis, enantioselective syn-hydration of enones, Diels-Alder reactions, Michael additions, and the horseradish peroxidase-mimicking hemin-G-quadruplex DNAzyme, one of the most intensively explored examples. [19][20][21][22][23][24] It was shown that the latter system can catalyse the oxidation of organic substrates to uorophores or chromophores. Also, it catalyses biologically relevant oxidations of thiols and NADH. 25,26 Applications in DNA nanotechnology include the design of logic gates, heavy metal detection and cholesterol diagnostics. [27][28][29] Here we report the rst covalent introduction of an imidazole-modied nucleobase surrogate into tetramolecular G-quadruplexes. Our study is inspired by the ubiquitous involvement of the amino acid histidine in the coordination of transition metal cations in numerous metalloproteins. Different biorelevant metal cations were found to be complexed by these modied G-quadruplex structures resulting in unprecedented high thermal stabilizations and highly accelerated association rates, culminating in the design of a Cu(II)dependent horseradish peroxidase-mimicking DNAzyme.

Results and discussion
Inspired by the amino acid histidine and its omnipresent involvement in the coordination environments of metalloenzymes, we aimed to incorporate imidazole derivatives into G-quadruplex DNA structures. The imidazole-based ligand L was introduced as GNA (glycol nucleic acid) building block enabling internal as well as terminal incorporation (Fig. 1). To access the phosphoramidites needed for solid phase synthesis, literature procedures were modied. 30,31 An initial nucleophilic ring-opening of DMT-protected (R/S)-glycidol was followed by a phosphorylation reaction to access the required phosphoramidite building blocks, each as R-(L R ) and S-(L S ) enantiomers. DNA solid phase synthesis was performed according to standard protocols. 15 Coupling times for L were extended to maximize reaction efficiencies, which were usually > 99% per coupling. Aer solid phase synthesis, DNA samples were deprotected in aqueous ammonia at 55 C and puried 'DMTon' using reversed-phase HPLC. Aer purication, the DMTgroup was removed using C 18 cartridges and aqueous TFA.
Properties of the imidazole-modied G-quadruplexes were investigated using UV-Vis-based melting experiments, thermal difference spectra (TDS), circular dichroism (CD) spectroscopy and high-resolution ESI mass spectrometry. In case of melting experiments, the temperature-dependent change of absorption at 295 nm was observed, characteristic for G-quadruplex denaturation/renaturation. 32,33 To prove whether the imidazole modication would still allow G-quadruplex formation, both diastereomers of the tetramolecular G-quadruplexes G 4 L R/S and G 5 L R/S , containing four and ve consecutive guanine residues followed by the 5 0 -imidazole modication, were investigated. Gquadruplex formation could be observed in the CD spectra ( Fig. 2b, ESI Fig. 13-15 †), with a characteristic positive Cotton effect at $262 nm consistent with a parallel topology. 34 In thermal denaturation experiments for G 4 L R/S , signicantly lower melting temperatures ( were observed, owing to the one G-tetrad shorter sequence in G 4 L R/S (Fig. 2a, Table 1).
Aer having established the formation of parallel Gquadruplexes, metal complexation was investigated. When 1 equiv. of Cu(II) was added to G 4 L R , an unprecedented high stabilization towards thermal denaturation of DT 1/2 ¼ +51 C was observed, while retaining the parallel topology as indicated by CD spectroscopy (Fig. 2b). In addition, a new maximum at $295 nm appeared which was attributed to a previously reported ipping of the 3 0 -guanine residues. 17 In case of already much more stable G 5 L R , Cu(II) addition resulted in thermal stabilities > 95 C and no denaturation could be observed anymore. For the isomeric sequence G 4 L S , thermal stabilization was signicantly lower (DT 1/2 ¼ +48) as compared to its diastereomer G 4 L R , indicating that the conguration of L plays an important role when it comes to metal complexation. This is in contrast to the metal-free G-quadruplex, where the modication's stereochemistry had only a neglectable inuence. UV-Vis spectroscopy-based titrations of the ligand-modied quadruplex with Cu(II) cations showed to be in agreement with a proposed 1 : 1 stoichiometry (Fig. 2c) of Cu(II) and Gquadruplex. Interestingly, when thermal denaturation studies of G 4 L R were performed with Cu(II)-unsaturated systems, two clear transitions were observed at T 1/2 ¼ 32 C and T 1/2 ¼ 83 C, corresponding to Cu(II)-free and -bound G-quadruplex,  respectively. Finally, evidence for formation of the proposed complex was obtained by native ESI mass spectrometry. To determine whether a G-quadruplex retains its native state in the gas phase, two observables can be followed. First, if the secondary structure is destroyed, single stranded DNA instead of a tetramer would be observed. Second, in their native state G n -quadruplexes bind n-1 monovalent cations in their central pockets (with n ¼ number of G-tetrads). On top, ESI mass spectrometry from electrolyte-containing solutions always gives rise to series of unspecic adducts with sodium or potassium cations. For fully denatured species, a statistical distribution of adducts starting with zero cations would be observed while for a native, folded species a distribution is observed starting with n-1 explicitly bound cations. 35 In case of Cu(II)-bound G 4 L R , one main species [(G 4 L R ) 4 K 3 Cu] 4À corresponding to a folded Gquadruplex with three potassium ions and a single Cu(II) was observed as the most abundant and rst peak of a cluster series, followed by further unspecic K + -adducts (Fig. 2d).
Next, the sample was subjected to trapped ion mobility timeof-ight mass spectrometry (timsTOF). With this method, the gas-phase conformational distribution of ionic molecules or supramolecular objects can be determined by measuring their relative mobility K 0 when pushed against a variable electric eld by a constant stream of collision gas. Large ions elute faster, while smaller ions remain longer trapped. Mobilities can be converted to collisional cross sections (CCS) by applying the Mason-Schamp equation (details see ESI †). For Cu(II)-bound G 4 L R , very sharp ion-mobility distributions could be measured corresponding to a CCS of 793Å 2 , being in the same range as similar, but unmodied tetramolecular G-quadruplexes and indicating a compact, well-dened 3-dimensional structure (Fig. 5, ESI Fig. 26 †). 36 To evaluate the scope of metal complexation, further transition metal cations were screened. Ni(II), Zn(II) and Co(II) were found to bind to the modied G-quadruplexes as shown by strong thermal stabilizations (see Table 1) while retaining the parallel topology (ESI Fig. 14 and 15 †). Noteworthy is the successful binding of Zn(II), for which only this year a rst example for a specic metal-mediated base pair in a DNA double strand was reported by Müller et al. 37 To probe if the thermal stabilization was a reversible process, EDTA was added to the G-quadruplex-metal complexes. For Zn(II) and Co(II), EDTA addition led to an immediate reversal of the metal-induced thermal stability but in case of Cu(II) and Ni(II) no change was observed, indicating that EDTA is hindered of readily removing the bound Ni(II) and Cu(II) cations. This is attributed to a kinetically trapped state since no metal-mediated stabilisation was observed for Cu(II) and Ni(II) if EDTA was added prior to G-quadruplex formation and metal addition.
Next, we followed the association of G 4 L R/S from its single strand components over time, indicated by the CD spectral change at 262 nm, at 7 C (Fig. 3, ESI Fig. 22 †). In absence of transition metals, the typical slow association was observed, stagnating without ever reaching full conversion aer days. Association rate constant k on was calculated using eqn (1) and was in a similar range to unmodied tetramolecular DNA Gquadruplexes reported in the literature (k on (G 4 L R ) ¼ 2.5 Â 10 9 M À3 s À1 ; k on (G 4 L S ) ¼ 1.4 Â 10 9 M À3 s À1 ). 38,39 Aer addition of 1 equiv. Zn(II), for G 4 L R a 10-fold accelerated association rate (k on (G 4 L R -Zn) ¼ 1.9 Â 10 10 M À3 s À1 ) was determined while for G 4 L S only a weak acceleration could be observed (k on (G 4 L s -Zn) ¼ 3.3 Â 10 9 M À3 s À1 ). When Cu(II) was added, however, Gquadruplex association was enormously accelerated with a half-association time of only 10 min and full G-quadruplex formation was observed within hours. To calculate k on for the metal containing G-quadruplexes, fourth-order kinetics were assumed as plausible model, 38 based on the encounter of four individual DNA strands of which one, at a given Cu : DNA ratio of 1 : 4, brings in a rapidly precoordinated Cu(II) cation. The correspondingly calculated values k on (G 4 L R -Cu) ¼ 4.2 Â 10 11 M À3 s À1 and K on (G 4 L S -Cu) ¼ 3.2 Â 10 11 M À3 s À1 reect a 100fold acceleration as compared to the metal free sample (Fig. 3a  and b). Our hypothesis for the accelerated association is based on a very fast coordination of the imidazoles of four strands to Cu(II), yielding an intermediate complex. This precoordinated species then quickly folds to the corresponding G-quadruplex under formation of Hoogsteen hydrogen bonds and incorporation of three sodium ions (Fig. 3c). Interestingly, when Table 1 Thermal stabilities expressed as melting temperature T 1/2 for G 3 L R/S , G 4 L R/S and G 5 L R/S in absence and presence of Cu(II), Ni(II), Zn(II) and Co(II) (in brackets, thermal stabilizations DT 1/2 after metal addition are given) association kinetics were determined in presence of a tenfold excess of Cu(II) cations, the reaction was becoming slower again (k on (G 4 L R -Cu) ¼ 1.8 Â 10 11 M À3 s À1 and K on (G 4 L S -Cu) ¼ 1.3 Â 10 11 M À3 s À1 ), as observed in CD (Fig. 3a) as well as UV-Vis spectroscopy experiments (ESI †). In light of the proposed model, we explain this with saturation of a majority of single strand imidazoles with coordinated Cu(II) cations, rendering an encounter of four DNA strands around a single Cu(II) less likely.
To better understand how G 4 L R/S is able to accommodate both Cu(II) and Zn(II), since the former usually prefers squareplanar and the latter tetrahedral coordination, MD simulations were performed using a bonded model for metal interactions (Fig. 4) according to a previously published protocol. 16 In absence of Zn(II) and Cu(II), the modelled structure of G 4 L R reveals that the imidazole ligands interact with the groves (Fig. 4a). This is in contrast to our previous observation for pyridine-carrying tetramolecular quadruplexes, where p-stacking of the aromatic modications with the 3 0 -terminal G-tetrad was proposed based on MD simulations and melting studies. 17 Interestingly, MD simulation of diastereomer G 4 L S indeed found the ligand stacking on top of the 3 0 -G-tetrad. Subsequent MD simulations where performed in complex with Cu(II) and Zn(II), respectively, showing that the ligand was just long enough to facilitate a square-planar coordination for Cu(II) (Fig. 4b and d) and a tetrahedral coordination for Zn(II) (Fig. 4c  and e).
Aer exploring the scope of metal binding, we set out to exploit the tremendous thermal stabilization and accelerated assembly upon Cu(II) addition to develop a switchable DNAzyme capable of mimicking the metalloenzyme horseradish peroxidase. As a crucial prerequisite for this purpose, the modied Gquadruplex has to maintain the capacity for binding the ironporphyrin complex hemin. ESI-MS investigations could show that the modied G-quadruplex was indeed able to bind one molecule of hemin, as two main signals corresponding to [(G 4 L R ) 4 K 3 Cu] 4À and [(G 4 L R ) 4 K 3 Cu-hemin] 4À were observed (Fig. 5). No signal corresponding to a two-hemin adduct could be determined. In addition, ion-mobility experiments showed two well separated signals corresponding to [(G 4 L R ) 4 K 3 Cu] 4À (793Å 2 ) and [(G 4 L R ) 4 K 3 Cu-hemin] 4À with a signicant larger CCS ¼ 860Å 2 . Noteworthy, the CCS increase of the hemin adduct exactly corresponded to the addition of one G-tetrad as observed for [(G 5 L R ) 4 K 4 ] 4À (858,Å 2 , ESI Fig. 27 and 28 †).
As a readout for peroxidase-activity, the oxidation of ABTS with H 2 O 2 to the strongly green ABTS-radical (A max ¼ 414 nm) was performed. To quantify the catalytic activity, the initial rate (V 0 ) was determined, dened as the concentration of the ABTS radical as function of time. First experiments showed that G 4 L Shemin indeed catalysed the oxidization of ABTS (V 0 ¼ 390 nM s À1 ), however, even in absence of Cu(II) G 4 L S was already stable at 25 C (T 1/2 ¼ 31 C (R), 32 C (S)), hemin binding further increased the thermal stability (G 4 L S -hemin T 1/2 ¼ 45 C, DT 1/2 ¼ +14 C) and thus Cu(II) addition only resulted in a small activity increase (V 0 ¼ 523 nM s À1 , ESI Fig. 24 †). To destabilize the system, the concentration of NaCl was reduced to 1 mM but still G 4 L S -hemin was forming in absence of Cu(II) (ESI Fig. 7 †). However, when the number of G-tetrads was reduced to three, the resulting G-quadruplex G 3 L R/S was extremely unstable, and no G-quadruplex formation could be observed at 7 C by CD (ESI Fig. 16 †). While addition of Zn(II) and Co(II) could not trigger Gquadruplex formation, addition of either Cu(II) or Ni(II) lead to the formation a parallel stranded G-quadruplex. In case of Ni(II), G-quadruplex formation was not complete as indicated by CD spectroscopy. In case of Cu(II), quadruplex formation was smooth, thus enabling the design of a switchable peroxidase mimic.
As expected, when the oxidation of ABTS was performed only with G 3 L R or G 3 L S and hemin, no activity could be observed.  Also aer addition of various transition metal cations such as Fe(II/III), Zn(II), Co(II) or Mn(II) no activity was observed. Only when G 3 L R or G 3 L S was formed in presence of either Cu(II) or Ni(II) and hemin, the active DNAzyme could form, showing a strong catalytic activity (Fig. 6, ESI Fig. 25 †). In presence of Cu(II), a signicant higher initial rate was determined (V 0 (R) ¼ 785 nM s À1 , V 0 (S) ¼ 615 nM s À1 ) than in presence of Ni(II) (V 0 (R) ¼ 387 nM s À1 , V 0 (S) ¼ 369 nM s À1 ) which was attributed to the incomplete G-quadruplex formation with Ni(II). The DNAzyme was also shown to be stable aer addition of up to 10 equiv. of EDTA and incubation at 25 C for 1 h, attributed to a kinetically trapped metal complex, however an activity decrease (V 0 (R) ¼ 281 nM s À1 , V 0 (S) ¼ 353 nM s À1 ) was observed. The decrease was mainly attributed to a general interference of EDTA in this assay which was further supported by time-dependent CD spectroscopy of G 3 L R and G 3 L S in presence of EDTA (ESI Fig. 17 †) that was showing no signs of G-quadruplex decomposition aer 1 h at 7 C. 40 Finally, when the DNAzyme was formed in presence of Cu(II) in mixture with competing transition metals including Fe(II/IIIIII), Mn(II), Zn(II) and Co(II/III), only a modest deceleration was observed for G 3 L R (V 0 ¼ 712 nM s À1 ) while for G 3 L S (V 0 ¼ 641 nM s À1 ; Fig. 6, blue line), even a small rate increase was observed. Hence, our system shows potential for Cu(II) sensory applications in the presence of competing metal ions.

Conclusions
The covalent introduction of imidazole ligands into tetramolecular G-quadruplex structures is reported. Modied Gquadruplexes were shown to complex several transition metal cations such as Cu(II), Zn(II), Co(II) and Ni(II). In case of Cu(II), metal complexation lead to an unprecedented strong thermal stabilization and highly accelerated association rate which could be exploited to design a Cu(II)-switchable peroxidase. Once formed, the peroxidase was shown to be stable in presence of EDTA, attributed to the formation of a kinetically trapped complex. Detailed mass spectrometric investigations could show that only one hemin was binding to the 3 0 -end, while the 5 0 -end was blocked by the formed metal complex. Further, trapped ion-mobility experiments indicated a signicant increase of the collisional cross section (CCS) upon hemin binding from 793 to 860Å 2 which exactly corresponded to the CCS of a one G-tetrad larger G-quadruplex (858Å 2 ). The herein established switchable DNAzyme presents a novel approach to stimuli-responsive, functional oligonucleotide structures with potential application in diagnostic setups, triggered probes for chemical biology studies as well as dynamic DNA nanotechnological devices.

Conflicts of interest
There are no conicts to declare.