Heteroleptic Coordination Environments in Metal-Mediated DNA G-Quadruplexes

The presence of metal centers with often highly conserved coordination environments is crucial for roughly half of all proteins, having structural, regulatory, or enzymatic function. To understand and mimic the function of metallo-enzymes, bioinorganic chemists pursue the challenge of synthesizing model compounds with well-defined, often heteroleptic metal sites. Recently, we reported the design of tailored homoleptic coordination environments for various transition metal cations based on unimolecular DNA G-quadruplex structures, templating the regioselective positioning of imidazole ligandosides LI. Here, we expand this modular system to more complex, heteroleptic coordination environments by combining LI with a new benzoate ligandoside LB within the same oligonucleotide. The modifications still allow the correct folding of parallel tetramolecular and antiparallel unimolecular G-quadruplexes. Interestingly, the incorporation of LB results in strong destabilization expressed in lower thermal denaturation temperatures Tm. While no transition metal cations could be bound by G-quadruplexes containing only LB, heteroleptic derivatives containing both LI and LB were found to complex CuII, NiII, and ZnII. Especially in case of CuII we found strong stabilizations of up to ΔTm = +34°C. The here shown system represents an important step toward the design of more complex coordination environments inside DNA scaffolds, promising to culminate in the preparation of functional metallo-DNAzymes.


INTRODUCTION
Proteins are involved in a vast number of processes ranging from structural and regulatory functions to enzymatic reactions. Roughly half of all proteins depend on metal cations helping to maintain a desired folding or serving as catalytic centers or redox cofactors (Raven et al., 1999;Lu et al., 2009;Rubino and Franz, 2012). Which function the respective metal ion adopts is strongly dependent on its properties, including accessible spin states, oxidation potential, Lewis-acidity, and bioavailability (Holm et al., 1996;Waldron et al., 2009). These properties are further fine-tuned by a well-defined first and second coordination sphere. The former is directly involved in metal coordination and usually consists of mixtures of different donor functionalities. Typically involved in coordination are the amino acids histidine, glutamic/aspartic acid, methionine, cysteine, or the backbone amide groups (Holm et al., 1996;Degtyarenko, 2000;Shook and Borovik, 2010;Valdez et al., 2014). In contrast, the second coordination sphere is not directly involved in metal binding but regulates catalytic processes, proton or electron shuttling, substrate transport, and effects selectivity (Colquhoun et al., 1986;Degtyarenko, 2000;Waldron et al., 2009;Shook and Borovik, 2010;Zhao et al., 2013;Valdez et al., 2014;Cornish et al., 2016).
The design of artificial metallo-enzyme mimics with improved or novel properties is attracting increasing interest, but remains challenging. In the area of preparative bioinorganic chemistry, focus is set on small, multidentate chelate complexes, often requiring tedious multistep syntheses and only covering effects of the first coordination sphere (Samuel et al., 2010;Kanady et al., 2011;Anderson et al., 2013;Dicke et al., 2018). More biologically oriented approaches involve the replacement of natural metal cofactors with metal centers not known in nature. An example is the replacement of hemin in myoglobin with an iridium or rhodium porphyrin complex for enantioselective cyclopropanation reactions (Key et al., 2016;Litman et al., 2018). Another approach is embedding metal cofactors by covalent or non-covalent interactions into empty cavities of usually metalfree proteins. This was successfully applied in a series of examples enabling catalysis of the asymmetric transfer hydrogenation of imines (Wu et al., 2019), ring-closing metathesis (Jeschek et al., 2016), oxime (Drienovská et al., 2018), and hydrazine (Drienovská et al., 2018;Mayer et al., 2019) formation and hydration of alkenes (Drienovská et al., 2017). In contrast to the aforementioned examples, a more bottom up approach is the de novo design of new metallo-proteins by the precise arrangement of certain structural motifs to create a metal binding site (Raven et al., 1999;Lu et al., 2009;Rubino and Franz, 2012). In recent years, a more efficient alternative was developed based on small artificial peptoid structures. Due to their simple accessibility by solid phase synthesis and their capability to form well-ordered secondary structures, many examples were shown for selective metal binding and catalytic applications (Baskin and Maayan, 2016;Knight et al., 2017;Baskin et al., 2018;Ghosh et al., 2018).
Another type of biopolymers forming well-ordered secondary structures are oligonucleotides. In contrast to peptides, RNA and DNA only consist of four nucleotide building blocks, thus reducing the possibilities to create diverse coordination environments for a range of metal cations. To overcome this limitation, different strategies were developed to covalently or non-covalently anchor metal-chelating ligands inside DNA. Roelfes and co-workers pioneered the design of various oligonucleotides capable of Michael-Additions, Carbene transfer, syn-hydrations of alkenes or Diels-Alder reactions (Roelfes and Feringa, 2005;Coquière et al., 2007;Boersma et al., 2010a,b;Rioz-Martínez et al., 2016). Other groups used modified quadruplexes for sequence-specific DNA cleavage, light controlled thrombin catalysis or peroxidase mimicking DNAzymes (Xu et al., 2009;Ali et al., 2019;Wang et al., 2019). A difficulty of this approach lies in the largely unknown exact position and coordination environment of the catalytic centers. This difficulty could be overcome in the field of metal-mediated base pairing, where the hydrogen bonding interaction of canonical base pairs is replaced by metal coordination, leading to highly stabilized DNA structures (Mandal and Müller, 2017). While first examples included only the involvement of canonical bases (Katz, 1963), the field was later expanded by the incorporation of a variety of artificial nucleobases culminating in the development of programmable metal wires inside DNA duplexes (Tanaka et al., 2006;Clever et al., 2007;Mandal et al., 2016;Sandmann et al., 2019). Later, the concept was expanded from duplex to triplex DNA (Tanaka et al., 2002) and i-motifs (Abdelhamid et al., 2018), while we and others started to focus on G-quadruplexes (Miyoshi et al., 2007;Smith et al., 2012;Engelhard et al., 2013). The latter ones form from guanine-rich sequences where four Gresidues cyclize to planar G-tetrads via Hoogsteen base pairing. Multiple G-tetrads form a G-quadruplex via π-π stacking interactions. Key to their high stability is the incorporation of a central cation-typically Na + or K + (Hänsel-Hertsch et al., 2017;Neidle, 2017). Our group was the first to report Cu IImediated tetramolecular G-quadruplexes based on pyridine and imidazole ligands (Engelhard et al., 2013(Engelhard et al., , 2018bPunt and Clever, 2019a), aimed at a range of applications. For example, dinuclear systems were employed as Cu II -based EPR-rulers for accurate distance measurements (Engelhard et al., 2018a). We later expanded this concept to unimolecular G-quadruplexes, equipped with oligonucleotide loops which form cavities above the G-quadruplex stem in which the metal complexes are embedded (Engelhard et al., 2017). In a recent study, we further showed that these G-quadruplexes can act as robust templates to arrange different numbers of imidazole ligandosides, leading to fine-tuned affinities for a range of transition metal cations with respect to their preferred coordination environments (Punt and Clever, 2019b). While only homoleptic systems were investigated in that study, we herein expand the modular concept to heteroleptic systems with different donor functionalities. We introduce the design of mixed systems with imidazole and benzoate ligands, inspired by metallo-proteins, where the combination of imidazoles and carboxylate is often involved in metal coordination (e.g., in the 2-His-1-carboxylate facial triad) (Greenblatt et al., 1998;Koehntop et al., 2005). We show how this combination affects both, G-quadruplex stability and metal complexation.

RESULTS
In this study we report the incorporation of a new benzoate ligandoside L B in combination with the known imidazole ligandoside L I . Both were incorporated in the (S) configuration as GNA (glycol nucleic acid) building blocks (Zhang et al., 2005(Zhang et al., , 2006 by solid phase synthesis into tetramolecular and unimolecular G-quadruplexes. The phosphoramidite of L I was synthesized as previously described (Punt and Clever, 2019b). For the new benzoate ligand L B , a literature procedure was adopted (Engelhard et al., 2017). Accordingly, an initial nucleophilic attack of deprotonated solketal to methyl 4-(bromomethyl)benzoate followed by acidic deprotection reaction led to protected benzoate ligandoside (R)-4. Its structure and absolute configuration were confirmed by singlecrystal X-ray diffraction (Figure 1). The primary hydroxyl group was DMT-protected (DMT = dimethoxytrityl) followed by a phosphorylation reaction yielding phosphoramidite building block (S)-6. DNA solid phase synthesis was then performed according to standard literature procedures with  extended coupling times for the ligandosides L I and L B (see Supplementary Material for details). Coupling efficiencies for L B and L I were typically >99% per step. After solid phase synthesis, oligonucleotides were cleaved from the solid support and deprotected in 0.4 M NaOH in methanol/water (4:1) at 55 • C for 16 h. Standard deprotection with concentrated ammonium hydroxide was avoided due to the risk of forming amides instead of carboxylates from the benzoate esters. After reversed-phase HPLC purification, oligonucleotides were desalted and DMTgroups removed using C18 SepPak cartridges and aq. TFA (2%). The oligonucleotides were then lyophilized at stored at −20 • C.
Since L I had already been established in tetramolecular and unimolecular G-quadruplexes, we first investigated the influence of L B in the tetramolecular G-quadruplex (L B G 4 ) 4 . Clear formation of a parallel G-quadruplex was observed by CD spectroscopy with a positive Cotton effect around ∼260 nm (see Figure S25). Thermal denaturation experiments showed a melting temperature T m of 27 • C which was significantly lower compared to previously reported (L I G 4 ) 4 (T m = 36 • C; Punt and Clever, 2019b). Since L B and L I are sharing the same backbone modification, we ascribe this destabilization to a repulsive effect between the negatively charged benzoates and phosphates (Figure 2). Next, the interaction of (L B G 4 ) 4 with a series of transition metal cations was investigated. In contrast to (L I G 4 ) 4 which was shown to complex Cu II , Ni II , Co II , and Zn II , no signs for metal complexation in (L B G 4 ) 4 were observed (see Figures S3, S4). This may be explained by the harder character of the benzoate ligand, competing with hard ligands such as the contained chloride, cacodylate buffer or phosphate backbones. However, even for hard and oxophilic transition metal cations, including Gd III and Ce III , no interactions were found.
Mixing ligands in tetramolecular G-quadruplexes leads to statistical mixtures, which makes it challenging to design distinct heteroleptic coordination environments (see Supplementary Material for details). On the other hand, the folding of unimolecular G-quadruplexes into discrete topologies enables programmable ligand arrangements. Consequently, we moved forward to incorporate L B in unimolecular Gquadruplexes. At first, L B was incorporated four times in htelL B 4 . Similar to (L B G 4 ) 4 , incorporation of L B caused strong destabilization (T m = 12 • C) compared to htelL I 4 (T m = 33 • C). Successive replacement of L B with L I was accompanied with a linear increase in stabilization for each replacement (htelL B 3 L I T m = 17 • C, htelL B 2 L I 2 T m = 23 • C, htelL B L I 3 T m = 28 • C), highlighting the additive destabilizing effect of L B (Figure 3). CD spectroscopy of htelL B 4 , htelL B 3 L I , htelL B 2 L I 2 , and htelL B L I 3 showed clear signatures corresponding to an antiparallel G-quadruplex topology with a positive Cotton effect around ∼294 nm in all cases (see Figures S26, S27). This is consistent with the previous observations for homoleptic G-quadruplexes containing only L I . Next, the interaction with different transition metal cations was investigated. As for (L B G 4 ) 4 , for htelL B 4 , htelL B 3 L I , and htelL B 2 L I 2 , thermal denaturation experiments showed no signs for interaction with the examined transition metal cations (Cu II , Ni II , Zn II , Co II , V IV O). Pleasingly, this changed for htelL B L I 3 that showed a weak but distinct stabilization after addition of 1 equiv. of Cu II ( T m = + 4 • C). Additional equivalents resulted in no further stabilization consistent with a specific binding of Cu II . CD spectroscopy further confirmed retention of a clear antiparallel topology (see Figures S6, S7, S11-S16).
After we could show that at least three imidazole ligands are required to complex Cu II , we moved forward to a new series of sequences with six incorporated ligands (htelL B 4 L I 2 , htelL B 3 L I 3 , htelL B 2 L I 4 ). Again, the formation of G-quadruplexes with a clear antiparallel topology was observed by CD spectroscopy (see Figures S28, S29). Likewise, comparison of the thermal stabilities showed the destabilizing effect of L B (htelL B 4 L I 2 T m = 17 • C, htelL B 3 L I 3 T m = 26 • C, htelL B 2 L I 4 T m = 26 • C), however, not in the linear fashion as observed for the series htelL B 4−n L I n (n = 0-4). For the examined set of sixligand-containing sequences, however, direct T m comparison is not appropriate due to the chosen modification pattern (see Table 1). When investigating the interaction with metal cations, for htelL B 4 L I 2 , a clear stabilization after addition of Cu II ( T m = + 6 • C) was observed. Considering that for htelL B 2 L I 2 almost no stabilization was observed ( T m = + 1 • C), we conclude that in htelL B 4 L I 2 an involvement of one or two ligandosides L B into metal coordination is very likely. When further replacing L B with L I as in htelL B 3 L I 3 and htelL B 2 L I 4 , the Cu II -mediated thermal stabilization successively increased from T m = + 9 • C (htelL B 3 L I 3 ) to T m = + 34 • C (htelL B 2 L I 4 ). This extremely high thermal stabilization  [a] Punt and Clever (2019b). is unprecedented for unimolecular G-quadruplexes and much higher compared to the reported G-quadruplexes htelL I 6 ( T m = + 18 • C) and htelL I 4 A ( T m = + 23 • C) (Punt and Clever, 2019b). The formation of 1:1 complexes for htelL B 2 L I 4 with Cu II and Ni II was further confirmed by native ESI mass spectrometry. To understand whether a G-quadruplex is folded or unfolded in the gas phase, the intrinsic property of G-quadruplexes is exploited that in their folded state they always bind n−1 potassium ions (where n = number of G-tetrads). For a folded G-quadruplex with two G-tetrads, a main signal corresponding to the adduct with one distinct potassium ion would be expected, followed by a statistical distribution of adducts with further unspecifically bound potassium cations. On the other hand, for an unfolded G-quadruplex, the main signal would correspond to the mass of the DNA strand without potassium ions. The mass spectrum shows a main signal corresponding to [htelL B 2 L I 4 +Cu+K-7H] 4− (Figure 4), thus strongly indicating a folded G-quadruplex coordinating to a Cu II or Ni II ion in the gas phase (D'Atri et al., 2015;Lecours et al., 2017).
Jahn-Teller-distorted Cu II usually favors the coordination of four strongly associated ligands in a square planar geometry, with two additional ligands more loosely bound in axial positions (Halcrow, 2012). After proving a 1:1 complex for htelL B 2 L I 4 and Cu II , the question was if all six ligands are participating in metal coordination or if only L I is involved. Therefore, a new sequence htelL I 4 B was synthesized where L B was replaced with thymidines. Addition of Cu II led to a thermal stabilization of T m = + 20 • C, much lower compared to htelL B 2 L I 4 ( T m = + 34 • C). However, when looking at the absolute melting temperature T m in presence of Cu II , one notices that they are the same for both sequences (htelL B 2 L I 4 T m = 60 • C, htelL I 4 B T m = 60 • C). This could mean that Cu II coordination by htelL B 2 L I 4 simply compensates the destabilizing effect of L B and no benzoate ligand was involved in Cu II coordination. Further studies are required to shed light on this question.
Besides Cu II , the addition of Zn II and Ni II to htelL B 2 L I 4 and htelL B 3 L I 3 led to thermal stabilizations. These results were highly intriguing for two reasons. Quadruplex htelL B 2 L I 4 was significantly more stabilized with Ni II ( T m = + 22 • C) compared to Zn II ( T m = + 6 • C). However, in htelL B 3 L I 3 , the opposite effect was observed, showing a higher stabilization after Zn II addition ( T m = + 5 • C), while for Ni II no complexation was observed. This adds to the established variation of ligand number and position a third layer to our system to fine-tune metal affinities by the introduction of heteroleptic systems. As last question, we were interested whether Zn II in htelL B 3 L I 3 is coordinated by one or more benzoates. Interestingly, other sequences shown to complex Zn II (htelL I 4 A T m = + 3 • C, htelL I 6 T m = + 8 • C) always contain at least four counts of L I . Since in htelL B 3 L I 3 only three L I were available, we conclude that an involvement of L B in coordination to the Zn II cation is likely.

CONCLUSION
A new benzoate-based ligandoside L B was established in tetramolecular and unimolecular G-quadruplex structures. Homoleptic G-quadruplex (L B G 4 ) 4 was found to form a clear parallel topology. Its thermal stability indicated a strongly destabilizing effect of L B compared to L I which was attributed to an accumulation of negative charges. Also, no interactions between a series of transition metal cations and (L B G 4 ) 4 were found. Similarly, for the unimolecular G-quadruplex htelL B 4 , a destabilizing effect of L B and no interactions with transition metal cations were observed. The successive replacement of L B with L I in htelL B 3 L I , htelL B 2 L I 2 , htelL B L I 3 , and htelL I 4 resulted in a linear increase of the thermal stability. In addition, for htelL B L I 3 , a weak thermal stabilization after addition of 1 equiv. Cu II indicated specific binding.
When moving to systems with six incorporated ligands, a tremendously high thermal stabilization was observed after addition of Cu II to htelL B 2 L I 4 ( T m = + 34 • C). In comparison, for htelL I 4 B, addition of Cu II resulted in a stabilization of only T m = + 20 • C. However, the absolute melting temperatures T m of htelL B 2 L I 4 (T m = 60 • C) and htelL I 4 B (T m = 60 • C) are the same, indicating that Cu II complexation is rather compensating the destabilizing effect of L B . More interesting were the results for htelL B 2 L I 4 and htelL B 3 L I 3 after addition of Zn II and Ni II , respectively. HtelL B 2 L I 4 was significantly more stabilized by Ni II ( T m = + 22 • C) compared to Zn II ( T m = + 6 • C). However, in htelL B 3 L I 3 , the opposite effect was observed, showing a higher stabilization after Zn II addition ( T m = + 5 • C) while for Ni II no complexation was found. This expands our toolbox to design tailored binding sites for various transition metal cations. Previously, we had shown to fine-tune coordination environments by varying position and number of ligands. Here, we expand this approach by combining two ligandosides, L B and L I , which we regard as an important step for the design of metal-selective G-quadruplexes with application in diagnostics, selective catalysis, and DNA nanotechnology.

DATA AVAILABILITY STATEMENT
The datasets generated for this study can be found in the Cambridge Crystallographic Data Center under the CCDC identifier 1961648.

AUTHOR CONTRIBUTIONS
PP and LS conducted all syntheses and DNA experiments. SS contributed to the tetramolecular systems. LK and CS contributed the X-ray structure of compound 4. PP, LS, and GC designed the study, conceived the experiments, analyzed the data, and authored the manuscript.

ACKNOWLEDGMENTS
Prof. Herbert Waldmann from the MPI Dortmund is thankfully acknowledged for access to the MALDI-TOF spectrometer.