Catalysis of a 1,3-dipolar reaction by distorted DNA incorporating a heterobimetallic platinum(ii) and copper(ii) complex

A novel catalytic system based on covalently modified DNA is described.


Introduction
The main role of DNA is the storage of genomic information leading to the biosynthesis of proteins via diverse forms of RNA. 1 In turn, proteins play multiple roles in living systems, catalysis being among the most important ones. This general pattern has been rened as a consequence of the discovery of catalytic RNAs (ribozymes) 2 and DNAs (deoxyribozymes). 3 However, since DNA is structurally less versatile than RNA and proteins, additional molecules and functional groups are required to expand the catalytic space of DNA. Roelfes and Feringa 4 brilliantly demonstrated the feasibility of this concept by intercalation of aza-chalcones in the double helix and subsequent coordination to copper salts. Further work 5 demonstrated that various intercalating heterocycles and metallic salts in the presence of DNA are able to promote, among other transformations, Diels-Alder reactions, 4,6 Friedel-Cras alkylations 7 and Michael 8,9 (including oxa-Michael) 10,11 additions. It is remarkable that all these reactions were carried out in water, although in several cases the system tolerated organic co-solvents. 7 Other DNA activation methods include covalent attachment of active catalytic sites such as proline organocatalysts, 12 as well as Ir(I) 13 and Pt(II) 14 complexes.
Despite the relevance of (3 + 2) cycloadditions in the chemical synthesis of ve-membered rings, 15 the enzymatic version of this reaction has not been identied in living systems. 16,17 Only a very recent example of a possible enzymatic 1,3-dipolar reaction has been reported to date. 18 In addition, nonenzymatic 1,3dipolar reactions have been postulated in the biosynthesis of several alkaloids 19 and natural products such as furanocembranoids 20,21 and santolin Y. 22 Therefore, to the best of our knowledge a biomolecule-assisted bona de (3 + 2) cycloaddition between azomethine ylides and alkenes to produce unnatural proline derivatives has not been reported to date. Within this context, we herein describe the rst example of a DNA-assisted 1,3-dipolar reaction in water.

Results and discussion
The design of the covalent modication of DNA was based on the ability of Pt(II) chemotherapeutic drugs to bind mainly 1,2intrastrand GpG units, 23,24 thus providing a concave-convex distortion of the double helix that could mimic the active sites of metalloenzymes. In previous work 25 we reported new chiral ligands that can bind Cu(II) salts 26 and efficiently catalyze (3 + 2) cycloadditions involving azomethine ylides. Therefore, we reasoned that a DNA-Pt(II)-Cu(II) heterobimetallic complex similar to that depicted in Scheme 1 could catalyze this reaction.
First, we tested the feasibility of this basic design using computational methods. As starting point we took the crystal structure reported by Takahara et al. 27 for a cisplatin-doublestranded oligodeoxynucleotide complex (pdb code: 1aio), where G*G* denotes a 1,2-intrastrand cis-[(H3N)2Pt-d(GpG)] crosslink. To this structure 2,2 0 -bipyrimidin (bipym) was added and the azomethine ylide derived from methyl (E)-2-(benzylidene-amino)acetate 4a was coordinated to a copper(II) metallic centre. The resulting structure was stabilized by incorporating 509 water molecules and 22 sodium cations (Fig. 1a). The whole ensemble was optimized using a hybrid QM/MM ONIOM 28-30 scheme (ESI †). The full structure thus optimized was found to keep the folded geometry of the distorted double helix, where the Cu(II)-bipym-Pt(II)-G*G* system generated a cavity to which the azomethine ylide was coordinated.
As it can be seen by inspection of Fig. 1b, the resulting ensemble closely resembles the active site of a metalloenzyme. As a proof-of-concept experiment, we next examined the ability of (bipym)PtCl 2 complex 1a to bind two equivalents of guanosine. Since 1a and its derivatives posed solubility problems, we monitored the different species in the solid state using Cross Polarization-Magic Angle Spinning (CP-MAS) spectroscopy. 31 The NMR spectrum of bipym showed only one 15 N-NMR signal, as expected from its D 2h symmetry (Fig. 2). In contrast, the two nitrogen atoms of 1a coordinated to Pt(II) could be readily distinguished. Then, we used 5 0 -GMP as a suitable equivalent of G units in DNA and analysed the 1 H-15 N CP-MAS spectra at different mixing times to assign the ve nitrogen atoms of the G-unit. Combination of 1a and 5 0 -GMP resulted in the formation of an adduct whose 15 N chemical shi associated with the N7 atom of the purine system was considerably deshielded with respect to the signal recorded for 5 0 -GMP (Fig. 2).
We concluded that Pt(II) complex interacts with guanine units yielding a square planar complex, in which two equivalent G units bind the Pt(II) centre by means of the respective N7 atoms of the purine bicycle. We next studied the interaction between 1a and oligodeoxynucleotides in order to assess the binding abilities of different DNA sequences.
To this end, we used a quartz-crystal microbalance with dissipation monitoring (QCM-D) device. 32 These QCM-D experiments measured changes in frequency, which correlate with mass adsorption on the sensor surface, and the dissipation of energy of the adsorbed lm, which in turn correlates with its viscoelasticity. 33,34 The dissipation vs. frequency points were collected for various times and ordered graphically from le to right and from bottom to top (Fig. 3). In a reference experiment denoted as "exp. 1" in Fig. 3a, the Pt(II) complex 1a was injected on bare gold (no DNA immobilized) in PBS buffer. Under these conditions, dissipation increased linearly with frequency, thus resulting in an incremental dissipation entirely correlated with the mass adsorbed on the surface. To carry out the experiments in the presence of DNA, we selected two different model oligodeoxynucleotides on the basis of the well-known binding ability of GpG pairs with platinum drugs. 23,24 We tested in duplicate the oligomer containing 5 0 -thiol-AAAAATTAAATTAAA-3 0 binding sequence (Fig. 3a, experiments 2 and 3). An overlay with the reference experiment 1 revealed that this response was indeed non-specic as both experiments showed practically identical proles with respect to the reference. We interpreted these results as negative experiments in which there was no signicant interaction between Pt(II) complex 1a and G-free deoxyoligonucleotide (Fig. 3b). We next examined the behavior of the oligomer containing 5 0 -thiol-AAAAAGGAAAGGAAA-3 0 binding sequence (Fig. 3c, experiments 4 and 5). In this case, upon  addition of 1a both plots showed an identical increase in dissipation while simultaneously no frequency change was observed, in sharp contrast with the previous blank and negative experiments 1-3. This indicated that rst, there was no measurable non-specic binding occurring as it would be detected as a change in frequency (see the dotted lines in Fig. 3c); second, injection of 1a induced a strong change in dissipation, which can be only attributed to conformational changes taking place in the GpG pairs-containing immobilized and hybridized deoxyoligo-nucleotide. From these results, we concluded that GpG-containing DNA interacts specically with Pt(II) complex 1a resulting in an increasingly more dissipative lm, which in turn correlates with additive conformational changes of DNA upon addition of 1a (Fig. 3d).
Once we veried that complex 1a binds selectively G-containing oligonucleotides, we tested its binding ability to DNA strands. As a control experiment, we mixed 1a with 5 0 -GMP and the resulting complex was analyzed by MALDI-TOF mass spectrometry. 35,36 We observed an ensemble of high mass-to-charge (m/z) signals distributed around a value of m/z z 1042 a.u., as shown in Fig. 4a. These signals correlate satisfactorily with complex 7 (Fig. 4b), closely related to 6 ( Fig. 2), in which two ketone units have been generated via dehydration-tautomerization of one hydroxy group of each ribose unit. Aer this control experiment, the same protocol was followed, but instead of 5 0 -GMP we used DNA sodium salt from salmon sperm (salmon sperm DNA in Fig. 4) with a % G-C content of 41.2% and a molecular mass of 1.3 Â 10 6 Da (ca. 2000 bp). In this case, the same prole was observed in the corresponding MALDI-TOF mass spectrum (Fig. 4c), with  a high m/z ensemble centered at ca. 1042 a.u. This response was interpreted in terms of ion 7 or, for two consecutive GpG units in the starting salmon sperm DNA, as the keto-enol phosphoric ester depicted as ion 8 in Fig. 4d. On the basis of these experiments, we concluded that the 1a-G 2 complexes observed in monomeric and oligomeric G-containing species can be extended to double strain DNA.
We also studied the effect of Pt(II) complex 1a on the structure of DNA strands. Thus, samples of DNA (ca. 48 kb), from l phage-infected E. coli were analyzed by atomic force microscopy (AFM). 37 The corresponding DNA strands were unambiguously identied on oxidized silicon 38 by the corresponding AFM images ( Fig. 5a and b). When 1a was added, the AFM scans showed large morphological changes consisting of local kinks and crosslinks ( Fig. 5c and d). On the basis of the previously presented experiments, these changes were attributed to the formation of intra-and interstrand cis-{(bipym)Pt(d[GpG + GpXpG + GpA])} adducts. 39 These advanced analytical studies permitted us to establish the main features of the Pt(II)-mediated binding between molecule 1a and DNA. We concluded that the thus generated hybrid system could catalyse 1,3-dipolar reactions provided that (i) the heterobimetallic Cu(II)-Pt(II) centre is able to generate in situ the required N-metallated azomethine ylide derived from the corresponding imine 4; and (ii) the active site can accommodate the dipolarophile 3.
In order to test the catalytic ability in (3 + 2) cycloadditions of the DNA-heterobimetallic complex, we prepared catalyst 2 (Scheme 1) by using salmon sperm DNA in a buffered solution of (N-morpholino)propane-sulfonic acid (MOPS), to which 1a  was added, followed by copper(II) triate, triethylamine and the corresponding maleimide 3 and imine 4 (Scheme 2). Aer six days of reaction at room temperature, nine distinct endo-(3 + 2)cycloadducts (5aa-5bf, Table 1) could be generated. This endo stereochemistry was secured on the basis of the spectroscopic data and, in two cases, by X-ray diffraction analysis (ESI †).
In order to assess the relevance of each component of the catalytic system we tested 17 possibilities resulting from the combination of all the reagents except at least one (see Table S1 of the ESI †). In all these control experiments no 1,3-dipolar reaction was observed. In particular, when different combinations of triethylamine and the Cu(II) and Pt(II) salts were tested in the absence of salmon sperm DNA, the reaction did not proceed. Similarly, different combinations in the presence of salmon sperm DNA but in the absence of base, 2,2 0 -bipyrimidine and/or one of the metals did not produce any (3 + 2) cycloadduct. It is interesting to note that the combination of DNA, 2,2 0 -bipyrimidine and Cu(OTf) 2 in the absence of Pt(II) was also unproductive (see Table S1 of the ESI, † entry 13), thus indicating that the powerful Roelfes-Feringa method consisting of a metallated intercalating heterocycle cannot catalyse this challenging 1,3-dipolar cycloaddition. In addition, these experiments demonstrate that there is no background reaction (see in particular entry 1 of Table S1 of the ESI †). In summary, these control experiments demonstrated that combination of 2,2 0 -bipyrimidine, Pt(II), Cu(II) and DNA, most likely by bonding to consecutive GG units, is required to achieve moderate yields of (3 + 2) racemic endo-cycloadducts 5 (see Table S1 of the ESI, † entry 10).
We interpreted our results as follows: Previously formed DNA-1a adducts bound Cu(OTf) 2 and the resulting heterobimetallic complex 2a (Fig. 5a) coordinated imine 4 to form intermediate species INT1, from which the corresponding N-metallated azomethine ylide INT2 was formed via triethylamine-assisted deprotonation. This 1,3-dipole interacted with dipolarophile 3 to form the corresponding (3 + 2) cycloadduct and regenerating INT1 via interaction with another equivalent of imine, thus completing the catalytic cycle. In order to understand the origins of the endo selectivity we optimized the possible endo-and exotransition structures under the same computational framework used to optimize the structure of INT2 (Fig. 5b and c). Both saddle points were found to be quite asynchronous but still associated with a concerted [p4s + p2s] symmetry allowed mechanism, as indicated by the bond distances corresponding to the formation of the two new s-bonds. We also found that endo-TS was stabilized by a strong electrostatic interaction between the nitrogen atom of the imide moiety and the metallic centre (Fig. 5b). As a result, exo-TS was calculated to be ca. 13 kcal mol À1 higher in energy than its endo congener, thus predicting the preferential formation of cycloadduct endo-5aa under kinetic control, in nice agreement with the experimental data.
We performed similar calculations in the absence of DNA. In these studies, the features of the computational model (including surrounding water molecules, see the ESI †) remained identical. The results are gathered in Fig. 6d and e. We observed that in the presence of Cu(II) and bipyrimidine, the chief features of the transition structure leading to endo-5aa are similar to those found for the 3a + 4a / 5aa reaction in the presence of DNA. However, the activation energy was found to be ca. 12 kcal mol À1 higher, which corresponds to a k(DNA-Cu-Pt)/k(Cu) ratio of ca. 4.8 Â 10 8 (Fig. 6d). When the square planar diaqua-Pt(II) moiety was incorporated to the reaction coordinate (Fig. 6e), the shape of the corresponding saddle point did not change signicantly and the activation energy was slightly lower than in the previous case, with a calculated Scheme 2 Synthesis of (3 + 2) endo-cycloadducts 5aa-5bf catalysed by the heterobimetallic complex formed by salmon sperm DNA and 1a in the presence of copper(II) triflate. k(DNA-Cu-Pt)/k(Cu-Pt) ratio of ca. 2.7 Â 10 7 . It is interesting to note that the DNA-free simulations also predict the preferential formation of the endo-cycloadduct (see Fig. S3 and Table S3 of the ESI †). These results are in agreement with the absence of reactivity observed when DNA was not present. We interpret the lower activation energies in the presence of DNA in terms of the destabilization of the substrate by restriction of the conformational freedom and by the electrostatic repulsion between the anionic polyphosphate environment and the anionic part of the starting 1,3-dipole, which is alleviated along the reaction coordinate leading to the non-zwitterionic cycloadduct.

Conclusions
In this work we have shown that (2,2 0 -bipym)PtCl 2 complex 1a binds 5 0 -GMP by its N7 atom. In addition, oligomers containing GpG pairs can interact selectively with 1a, in line with the major formation of cis-[Pt(II){GpG}] complexes in platinum drugs. Moreover, 1a promotes signicant structural changes in l-DNA, resulting in highly distorted microenvironments around the metallic centres, most likely because of the formation of cis-[Pt(II){GpG} + {GpXpG} + {GpA}] adducts. On the basis of these experiments, we conclude that a suitable bimetallic complex based on Pt(II) can distort the double helix of DNA to generate an active site similar to those found in well-known metalloenzymes, as supported by QM/MM calculations. Finally, we have shown that the system formed by 1a, salmon sperm DNA and Cu(OTf) 2 catalyses (3 + 2) cycloadditions between azomethine ylides (formed in situ from the corresponding imines) and maleimides. Only the racemic endo-cycloadducts are obtained, probably because of competing geometries of the intermediate N-metallated azomethine ylides linked to the 1a-DNA hybrid system. This result demonstrates that modied biomolecules can catalyse chemical reactions in water, for which there is no equivalent in living systems.

Computational methods
All the stationary points were fully optimized and characterized (harmonic analysis) within the QM/MM scheme using B3LYP hybrid functional. 29 LANL2DZ basis set and effective core potential, 24 as well as UFF force eld 25 were used within the ONIOM 28 framework as implemented in the Gaussian09 (ref. 30) suite of programs.

NMR experiments
Liquid NMR spectra were recorded on either a Bruker Avance 500 MHz or 400 MHz spectrometer at standard temperature and pressure, equipped with an z gradient BBO probe. The 15N spectra were recorded using an inverse gated sequence at

QCM-D experiments
The experiments were carried out using a quartz-crystal microbalance with dissipation monitoring from Biotin Scien-tic and respective quartz sensors QSX 301 Gold (Biolin Scien-tic). For each experiment, a new sensor was used. Each experiment started with obtaining a stable baseline by passing PBS-buffer (Sigma Aldrich) at a ow rate of 100 mL min À1 through the sensor module (E1. Biolin Scientic) using a peristaltic pump (Ismatec, Reglo Digital). Frequency and dissipation data were recorded using the acquisition soware QSo 401 (Biolin Scientic). The 5th harmonic of the frequency was used for further data processing as this harmonic has previously proven to yield most reliable signals. 34 Stability was assumed when the average frequency signal was less than 0.05 Hz during ve min. Subsequently, each thiol-containing oligomer (Biomers) was injected (5 mM) at the same feed ow velocity for 10 min, followed by rinsing with PBS buffer during additional 10 min in order to wash-off any loosely bound DNA. The complementary strand (5 mM) was then injected at the same ow rate and duration, followed again by injection of PBS buffer. Finally, the Pt(II) complex 1a was injected at a concentration of 0.1 mg mL À1 , followed by rinsing with buffer. For better visualization, in Fig. 3 the frequency data were normalized for the steady-state frequency (F max ) measured when the DNA duplex had formed aer injection of the complementary strand and before injection of the Pt(II) complex. In this way, the effect of the Pt(II) complex on the dissipation of the DNA was comparable between measurements irrespective of the absolute DNA coverage of the sensor.

Reagents and catalysis
[Pt(DMSO) 2 Cl 2 ] and [(bipym)PtCl 2 ] were prepared following the procedure described in the literature. 42 The a-imino glycinate esters R 2 HC ¼ NCH 2 CO 2 Me 4, with R 2 ¼ Ph (4a), p-Me(C 6 H 4 ) (4b), cyclohexyl (4e), 3 0 -thienyl (4f), p-F(C 6 H 4 ) (4d), and p-OMe(C 6 H 5 ) (4c), were prepared using the synthetic procedures described in the literature. 25 Synthesis of endo-methyl 4,6-dioxo-3-aryloctahydro-pyrrolo [3,4-c]pyrrole-1-carboxylates (5). Low molecular weight (ca. 200 bp) DNA from salmon sperm (Sigma-Aldrich) was used as purchased. All cycloaddition experiments were carried out in an orbital stirrer at room temperature. In a typical reaction procedure, salmon sperm DNA (1 mmol) was added to 20 mL of a buffered solution of MOPS (20 mM, pH 6.5) and the resulting mixture was stirred for 24 h. Then, 1a (16 mg, 0.015 mmol) was added and stirring was resumed for additional 12 h. To this mixture copper(II) triate (16.27 mg, 0.015 mmol) was added and aer 30 min of stirring triethylamine (2.27 mL, 0.015 mmol), the corresponding maleimide 2 (0.05 mmol) and imine 3 (0.05 mmol) were consecutively added. The resulting mixture was stirred for 6 d. Then, the reaction mixture was extracted with dichloromethane (3 Â 10 mL) by means of an ultrasound bath. The three organic layers were collected and washed with brine (saturated solution) and water. The resulting organic layer was dried with anhydrous magnesium sulfate. Evaporation of the solvent under reduced pressure furnished the crude cycloadduct 5 as white solid or colourless oil, which was puried by trituration with diethyl ether.

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