Synthesis and Evaluation of a Rationally Designed Click-Based Library for G-Quadruplex Selective DNA Photocleavage

DNA containing repeating G-rich sequences can adopt higher-order structures known as G-quadruplexes (G4). These structures are believed to form within telomeres and the promoter regions of some genes, particularly in a number of proto-oncogenes, where they may play a role in regulating transcription. Alternatively, G4 DNA may act as a barrier to replication. To investigate these potential biological roles, probes that combine highly selective G4 DNA targeting with photocleavage activity can allow temporal detection of G4 DNA, providing opportunities to obtain novel insights about the biological roles of G4 DNA. We have designed, synthesized, and screened a small library of potential selective G-quadruplex DNA photocleavage agents incorporating the G-quadruplex targeting moiety of 360A with known photocleavage groups linked via “click” chemistry.


Introduction
G4 DNA is highly stable structural conformation that is composed of a series of stacked tetrads composed of four guanines joined through Hoogsteen hydrogen bonds. Monovalent cations such as potassium or sodium occupy the central channel formed by these stacked tetrads. G4 DNA may be OPEN ACCESS formed either by the joining of several strands of single-stranded DNA or intramolecularly within a single-stranded DNA molecule. There are many different G4 conformations that can be adopted, based on the location of loops and the directionality of the strand(s). These conformations are highly dependent on the sequence of the DNA as well as the type of monovalent cation present [1]. Compounds that incorporate planar aromatic systems and positive charges have been shown to bind to G4 DNA with varying specificities, most often by stacking on the planar ends of the G4 DNA or, less commonly, through association with the loops [2].
While the formation of G4 structures in vitro has been known for decades [3], direct evidence for their appearance and roles in biological systems has been primarily limited to single celled organisms such as ciliates [4] and bacteria [5]. However, there is a significant amount of indirect evidence in support of biologically relevant roles in multicellular organisms, including humans. Helicases that have been shown to resolve G4 structures faster and more selectively than they do duplex or other DNA secondary structures suggest that G4 structures do form in vivo and need to be resolved, likely because they pose barriers to replication and transcription. Mutations in some of these helicases are associated with disease states such as Bloom's Syndrome and Werner's Syndrome [6]. Further, sequences that lend themselves to G4 formation are more prominent in many (including human) genomes than would be expected to be present due to chance, suggesting selection for such sequences [7]. These sequences are also disproportionally associated with the promoters of proto-oncogenes [8], suggesting roles in gene regulation and making them interesting potential targets for anti-cancer therapy. The ability of the human telomeric DNA sequence to form G4 structures has also made it one of the most targeted sequences for potential cancer treatments [9].
In order to better understand the biological roles of G4 structures, tools are needed that are capable of directly detecting them within biologically relevant contexts. Small molecules capable of binding selectively to G4 DNA, and which incorporate an inducible sequence are promising candidates to function as such tools. There have been several compounds developed that show reasonable binding selectivity to G4 DNA [10,11]. These can serve as a framework for the development of not only potential drugs, but also for the development of G4-selective probes.
In this study, we developed a library designed around the principle of combining a "targeting" scaffold moiety based on an established G4 DNA ligand with various "warhead" photocleavage moieties. This strategy has given insight into the structure-activity relationship (SAR) associated with G4 DNA photocleavage ligands that will facilitate the design of other small molecular G4 DNA probes.

Library Design
The strategy that was employed to design a library of potential G-quadruplex photocleavage agents is shown in Figure 1. A G-quadruplex DNA targeting moiety is linked via variable-length linkers to a variety of established DNA photoreactive groups. The known G-quadruplex binding di(quinolin-3-yl)pyridine-2,6-dicarboxamide 360A was selected as the G4-targeting ligand based upon its selectivity and the relative ease of incorporation of functionality at the core pyridine 4-position [12]. To facilitate library construction, the linker was designed to incorporate a triazole unit, which could be assembled from copper-catalyzed coupling of appropriate terminal alkynes with different azides [13]. In implementing this approach, the terminal alkyne partners are G-quadruplex targeting ligands appended with different length ω-alkynylalkyl chains and the azides are derived from different photoreactive groups. These photoreactive groups were selected to explore different potential DNA cleavage modes. Photoexcited benzophenones undergo very efficient intersystem crossing, leading to DNA cleavage pathways proceding exclusively vie the triplet excited state [14]. Naphthalenediimides are known to photocleave duplex DNA via both singlet oxygen generation and electron-transfer processes [15]. Anthraquinones affect DNA photocleavage by electron transfer [16].

Synthesis of a Click-Based Library of G-Quadruplex Photocleavage Agents
The synthesis of the library commenced with the preparation of a series of terminal-alkyne functionalized 360A derivatives (Scheme 1). Chelidamic acid monohydrate was converted to the known diester 2 [17] by reaction with thionyl chloride in methanol. The ester 7 was subjected to Mitsunobu coupling with a series of terminal alkyne alcohols to afford compounds 3a-d, which differ from each other only by the length of the linker carbon chain. The diesters 3a-d were treated with 3-aminoquinoline and trimethylaluminum in 1,2-dichloroethane (DCE) to afford the diamides 4a-d. Next, these diamides were subjected to "click" reaction conditions in DMF with the different azide-functionalized photoactive moieties to afford three different series of compounds: the naphthalenediimides 8a,b, the benzophenones 9a-d, and the anthraquinones 10a,b. DMF proved a more suitable solvent than aqueous t-BuOH [18] for these couplings, although it required the use of copper triflate as a copper source due to the low solubility of copper sulfate in the new solvent system, which was exacerbated by the need for stoichiometric copper. Stoichiometric copper was required due to the copper-chelating capacity of the pyridine-2,6dicarboxamide [19,20] based scaffold. Addition of copper was met by an immediate color change of the solution to a deep green, and the triazole product could only be isolated when employing a slight excess (1.1 equivalents) of copper salt. We presume this is due to the rapid formation of a catalytically inactive 1:1 substrate-copper complex. This copper-chelating capacity also made purification and characterization of the product triazoles more difficult as the complexes needed to be disrupted in order to obtain free compound (see associated experimental). The additional steps required to disrupt the complex contribute to the modest yields for these coupling reactions. Scheme 1. Synthesis of a library of triazole-linked G-quadruplex-photoreactive group molecules.
Finally, methylation of the triazoles 8-10 in the presence of excess methyl triflate in chloroform gave the trimethylated products 11a,b; 12a-d; and 13a,b (Scheme 2). The somewhat unexpected formation of trimethylated products was established by the presence of two different N-methyl resonances for these compounds in the 1 H-NMR spectra and the observation of [M − OTf] + ions in the MADLI MS and [M − 3TfO] 3+ ions in the ESI MS. The location of these methyl groups was established indirectly through the following experiment. A small-scale methylation of 4a with methyl triflate, followed by click reaction with azide 5 in water/t-BuOH afforded a crude dimethylated product that was characterized by 1 H-NMR and LRMS. Comparison of the 1 H-NMR chemical shifts of this product vs. those of the trimethylated 11a showed substantial differences in the linker methylene peak positions best accounted for by methylation on the triazole ring in the case of 11a. It should be noted that attempts to carry out this sequence of reactions on larger scale for the preparation of the dimethylated triazole compounds failed due to the very poor mass recovery from the final copper-catalyzed coupling reactions. This was due to the difficulties in isolating these water-soluble compounds, especially in the presence of the copper salts employed in the final coupling carried out in aqueous t-BuOH. Scheme 2. Preparation of potential G-quadruplex DNA photocleavage agent library.
Photocleavage reactions were carried out in potassium cacodylate buffer with 100 nM fluorescent-labeled oligonucleotides and 500 nM of library compound. After irradiation (UVA lamps) for 30 min, the DNA photoceavage products were either analyzed directly by PAGE or first treated with hot piperidine before PAGE analysis.
As shown in Figure 2, none of these compounds were very effective as photocleavage agents, with the highest apparent cleavage levels around 25%-30% and most at 0%-15%. The photocleavage ability of these compounds was not increased by irradiation with shorter wavelength light (UVB lamps, data not shown). Furthermore, there was no clear pattern of preferred sites of photocleavage for any of the three different photoreactive groups (Figure 2A), indicating that the cleavage may be occurring through a diffusible intermediate. In general, the F-c-MYC22m-T G-quadruplex was more susceptible to photocleavage by these compounds when compared to F21T ( Figure 2B, bars with asterisks). (B) Quantification of G-quadruplex photochemical cleavage from gel electrophoresis analysis after irradiation and piperidine/heat treatment. Unless indicated, F21T was employed as the G-quadruplex substrate. Red, green, and blue bars correspond to compounds incorporating anthraquinone, naphthalimide, and benzophenone respectively. * FcMycT photocleavage data for comparison.
Analysis of the effect of linker length on the efficiency of photocleavage of F-c-MYC22m-T by the benzophenone-containing compounds 12a-d ( Figure 3A) demonstrates little change in photocleavage with linker length. Furthermore, there was not a clear correlation between compound concentration and the extent of photocleavage of FcMycT for either the benzophenone-containing 12b or the analogous anthraquinone compound with the same linker length 13b ( Figure 3B). This latter result was particularly surprising, given that some of the first examples of G-quadruplex DNA ligands were anthraquiones [23]. Therefore, to verify the ability of the anthraquinone moiety to cleave these G4 DNA structures as well as to investigate its binding, compound 14, which incorporates the anthraquinone moiety functionalized through amidation to give a tertiary amine for solubility, was prepared and tested. Concentration-dependent cleavage was observed for 14; however, at levels well below those of the well-established G-quadruplex DNA photocleavage agent TMPyP4 ( Figure 4).

G-Quadruplex DNA Binding
It was postulated that the presence of the triazolium moiety introduced in the final step of the synthesis might interfere with compound binding and/or photoactive group orientation, resulting in the lowered levels of photocleavage of these G-quadruplex DNA structures. In order to test this, the ability of a subset of the library compounds to bind to F21T and FcMycT was determined by FRET-based ΔTm experiments. The degree of binding is assumed to be proportional to the increase of the Tm of the complex relative to the Tm of the untreated DNA [24]. Most of the compounds tested exhibited minimal binding as evidenced by ΔTm shifts of only a few degrees Celsius, especially when compared to the large stabilization exhibited by 360A ( Figure 5). Interestingly, G-quadruplex DNA binding was partially restored in the case of 12d, the benzophenone-containing compound with the longest linker length. This suggests that shorter linkers are not well tolerated for G-quadruplex DNA binding by these compounds, perhaps due to the positioning of the triazolium group.
The results from the G-quadruplex photocleavage and binding studies reported here suggest several potential areas for improvement in the design of photoactive probes targeting G-quadruplex DNA. While longer linkers were stabilizing to the DNA-compound complex, clearly allowing binding, this did not appear to improve photocleavage performance. This may be because of the nature of the triazolium linker used. While increasing linker length would alleviate destabilization of the DNA-compound complex by increasing the distance between the primary DNA-binding moiety of the compound and the triazolium group, the distance between the triazolium group and the photoactive moiety remains unchanged. If the triazolium group is responsible for poor orientation of the photoactive group, then poor photocleavage performance would not necessarily be improved by increased linker lengths.
All reactions were carried out under argon in oven-dried glassware with magnetic stirring. Unless otherwise noted, all materials were obtained from commercial suppliers and were used without further purification. THF was distilled from sodium/benzophenone prior to use. Dichloromethane and 1,2-dichloroethane were distilled from CaH2 prior to use. Unless otherwise noted, organic extracts were dried with Na2SO4, filtered through a fritted glass funnel, and concentrated with a rotary evaporator (20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30). Rf values are reported for analytical thin-layer chromatography (TLC) performed on EM Reagent 0.25 mm silica gel 60-F plates with UV light or KMnO4 visualization. Flash chromatography was performed with EM Reagent silica gel (230-400 mesh) using the mobile phase indicated. Melting points (open capillary) are uncorrected. Unless otherwise noted, 1 H-and 13 C-NMR spectra were determined in CDCl3 or d6-benzene on a spectrometer operating at 400 and 100 MHz, respectively, and are reported in ppm using solvent as internal standard (7.26 ppm for 1 H and 77.0 ppm for 13 C in CDCl3. 7.15 ppm for 1 H and 128.0 ppm for 13 C in d6-benzene). All mass spectra were obtained in the positive mode either by chemical ionization using methane as the ionizing gas or by electrospray ionization.

Library Synthesis
3-Azidopropyl 9,10-dioxo-9,10-dihydroanthracene-2-carboxylate (7). 100 mg (1 eq, 0.37 mmol) of 9,10-dioxo-9,10-dihydroanthracene-2-carbonyl chloride was dissolved in 3 mL freshly distilled dichloromethane in an oven-dried flask under argon. This flask was then cooled to 0 °C. In a separate oven-dried flask, 150 mg (2.55 eq, 0.92 mmol) of 62 wt % 3-azidopropan-1-ol [27] in DMF was dissolved in 1 mL of distilled dichloromethane under argon. 52 µL (1 eq, 0.37 mmol) of freshly distilled triethylamine was then added. The solution containing the alcohol and the amine was then added slowly under argon to the flask containing the acyl chloride at 0 °C. The reaction was then allowed to return to room temperature and was stirred for 19 h when it appeared complete by TLC (75% ethyl acetate in hexanes). After flash chromatography (25% ethyl acetate gradually up to 50% ethyl acetate in hexanes), 74 mg of the light yellow solid was obtained (approximately 80% purity, 45% yield  Dimethyl 4-oxo-1,4-dihydropyridine-2,6-dicarboxylate (2) [28]. Distilled methanol that had been stored under argon over 4 Å molecular sieves (3.1 mL, 31 eq, 77.5 mmol) was added to a oven-dried flask under argon and the flask was then placed in an ice bath. Slowly, 1.1 mL (6.2 eq, 15.5 mmol) of thionyl chloride was added. The solution was allowed to stir for 5-10 min before 500 mg (1 eq, 2.5 mmol) of chelidamic acid was added under increased argon flow. The flask was outfitted with an oven-dried condenser. The mixture was then stirred for 72 h, under argon, allowing the ice bath to slowly warm to room temperature. The faintly yellow solution was then diluted approximately 2× with methanol, transferred to a larger flask, and the solvent removed under reduced pressure to give a white solid residue. The flask was then placed in an ice bath for 15 min before 3 mL of chilled distilled water was added with swirling, followed by 0.75 mL of chilled 10% sodium carbonate solution and 0.75 mL of chilled 50% aqueous methanol. After swirling, the mixture was allowed to stand in the ice bath for 20 min before being filtered under reduced pressure and washed with 3 mL, 3 mL, and 1 mL portions of chilled 50% aqueous methanol, giving 500 mg of crude white product. The crude product was adsorbed to 1 g of silica and purified through chromatography on a silica plug (about 5-6 g SiO2, EtOAc as eluent), giving 417 mg of purified product as a white solid (79% yield). 1 H-NMR (400 MHz, CDCl3) δ 10.0 (1H, s), 7.5 (2H, s), 4.0 (6H, s) (matches lit. [21]); 13 C-NMR (100 MHZ, CDCl3) δ 172.8, 163.3, 144.1, 117.7, 53.3 (matches lit. [21]).

G-Quadruplex DNA Photocleavage Assays
The photocleavage experiments were adapted from the published photocleavage assay [21]. Briefly, solutions containing a single library compound and 100 nmol FRET pair-labeled DNA or blank controls in potassium cacodylate buffer were prepared and then added in triplicate to a 384-well plate (20 µL reaction volume) and irradiated in a Luzchem photoreactor for 30 min before being diluted to 95 µL with dissociation buffer (large excess calf thymus DNA in potassium cacodylate buffer). The plate was then sealed and heated in an oven at 85 °C for 30 min before being slowly cooled to room temperature overnight. After spinning down condensation in the plates, triplicate wells from the plate were combined, then split into two samples, one of which was treated with piperidine and heat while the other remained untreated. These samples were dried down under vacuum and resuspended in formamide denaturing buffer prior to being subjected to 20% denaturing PAGE at 8 W (constant current). Gels were visualized using fluorescein fluorescence on a Typhoon Trio gel reader. Cleavage bands within a lane were normalized against total lane signal and control lane signals subtracted so that the relative cleavage in each lane could be determined to give % cleavage of the DNA strand above background. Band intensities were quantified using GelQuant. NET provided by biochemlabsolutions.com. Percent Cleavage = ((Cx/Totx − FC0)/(1 − FC0)) × 100 (1) where Cx is the sum of the cleavage band intensities, Totx is the total band intensities in a lane, and FC0 is the ratio of cleavage band intensities to total band intensities in a control lane loaded with the untreated control.

G-Quadruplex DNA Melting Assays
200 nM of FRET pair-labeled DNA in 5 mM potassium cacodylate buffer containing 0.25 mM disodium EDTA was first precycled (heated to 100 °C, then cooled slowly back to 10 °C) three times prior to being run to help improve reproducibility between replicate DNA melts. Precycled DNA samples were then treated with 0 nM-1 µM compound and incubated for 15 min prior to being heated in a Varian Eclipse fluorimeter from 10 °C to 80 °C at 1 °C/min and from 80 °C to 100 °C at 0.5 °C /min before resting at 100 °C for 2 min and then cooling to 10 °C at 1 °C/min. The melting traces followed the fluorescence signal of fluorescein (FAM) and data was fit to the logistic equation below through nonlinear regression using SciDavis in order to determine melting points.
where I is fluorescence intensity, I0 is initial fluorescence intensity, If is final fluorescence intensity, s is the signal increase rate (a scaling factor), T is temperature, and Tm is the melting temperature.

Conclusions
The strategy of combining a "targeting" moiety with a "warhead" moiety through "click" chemistry still shows some promise, although the levels of both photocleavage and binding were less than originally expected. This could be due in part to the strong dependence of binding on the length of the linker used as well as the non-ideal photocleavage moiety positioning due to the charge introduced to the linker. Preparation of another library that does not incorporate this charge while continuing the exploration of linker length could improve both the binding and photocleavage activity, giving analogues that are better suited as photoactive probes for G4 DNA.