Divalent Naphthalene Diimide Ligands Display High Selectivity for the Human Telomeric G‐quadruplex in K+ Buffer

Abstract Selective G‐quadruplex ligands offer great promise for the development of anti‐cancer therapies. A novel series of divalent cationic naphthalene diimide ligands that selectively bind to the hybrid form of the human telomeric G‐quadruplex in K+ buffer are described herein. We demonstrate that an imidazolium‐bearing mannoside‐conjugate is the most selective ligand to date for this quadruplex against several other quadruplex and duplex structures. We also show that a similarly selective methylpiperazine‐bearing ligand was more toxic to HeLa cancer cells than doxorubicin, whilst exhibiting three times less toxicity towards fetal lung fibroblasts WI‐38.

Abstract: Selective G-quadruplex ligands offer great promise for the development of anti-cancer therapies. A novel series of divalent cationic naphthalene diimide ligands that selectively bind to the hybrid form of the human telomeric G-quadruplex in K + buffer are described herein. We demonstrate that an imidazolium-bearing mannoside-conjugate is the most selective ligand to datef or this quadruplex against several other quadruplex and duplex structures.W ea lso show that as imilarly selective methylpiperazine-bearing ligand was more toxic to HeLa cancer cells than doxorubicin, whilste xhibiting three times less toxicity towards fetal lung fibroblasts WI-38.
G-quadruplexes are four-stranded secondary structures that occur in guanine-richr egions of DNA and are over-represented in telomeres and gene promoter regionsi ncluding oncogenes and tumor suppressors. [1] It has been proposed that stabilization of these polynucleotide sequences by smallm olecules could lead to novel anticancer treatments. [2] However,t he development of anti-cancer,d rug-like, bioavailable ligands to selectively stabilizeas pecific G-quadruplex structure is still am ajor challenge, given the approximately 716,310 putative G-quadruplex forming sequences in the human genome. [3] Another hurdle to G-quadruplex ligandsreaching clinical development is their insufficient drug-like character. [4] Although some progress has been made,none of the G-quadruplex ligandsentering clinicalt rials have succeeded to date. [5] Several molecules have been reported that can selectively stabilize one quadruplex structure or topologyo ver another. [6] For instance, telomestatin analogue TOxaPy [6c] is one of the most selectiveligands discoveredt od ate for the human telomeric G-quadruplex in Na + buffer.A nother ligand, N-methyl mesoporphyrin (NMM), can stabilize the human telomeric G-quadruplex in K + bufferw ith no observed stabilization in Na + buffer, although it binds to other parallel quadruplexes. [6a] Work within the Morales group has shown that carbohydratesc an stack onto the G-C base pair,and affect the stability of the G-quadruplex structure as well as hydrogen bond to purines. [7] Indeed, af ew examples of carbohydrate-basedG -quadruplexl igands point towards sugars able to bind to grooves and/or loops of the quadruplex. [8] We thus proposed that glycosides should be both tolerated andu seful binding motifs in quadruplex ligand design.
To test our hypothesis, we decided to replace classical chargeda mine-deriveds ide chains with charged sugars on the quadruplex ligands. We chose the naphthalene diimide( NDI) scaffold since NDI-based compounds are some of the most active and highly studied G-quadruplexl igands to date, such as tetra-substituted NDI MM41 developed by the Neidle group. [9] Despite tri-andt etra-substituted NDI ligands having been extensively researched, [10,11] di-substituted NDI scaffolds have not yet been fully explored. [11a, 12] Herein we report ah ighly modulars ynthesis of as malll ibrary of di-substituted NDI ligands bearing charged carbohydrate moieties alongside classical and non-classical charged groups, and their evaluation as selective G-quadruplex ligands( Figure 1) and their in vitro cell cytotoxic profile in model systems.
Initial efforts focused on targeting glucose-and mannosebased NDIs. We hypothesized that the different stereochemical presentation of the OH at C-2 and anomeric linkagea t C-1 should yield useful SAR information on the role the carbohydrate scaffold plays in bindinga nd its ability to pick up additional polar interactions. In addition, am ethylimidazolium cationic group was introduced at C-6 of the glycosides to aid binding. [13] As ac ontrol, uncharged glucoside 1 was also targeted, alongside novel NDIs 4 and 5 bearing only the methylimidazolium group in the absence of the sugar but differing in chain length. Finally,d i-substituted methylpiperazine-a nd di-methylamine-containing 6 and 7 [14] werea lso prepared, as these motifs had been previously reported as components of other NDI ligands. [11a,b] The synthesis of glyco-conjugates 2 and 3 wasa chieved in four steps from azidoalkyl glycosides 8 [15] and 9 [14] (Scheme 1). Selectivem esylation of the C6-hydroxyl using MsCla nd pyridine, followed by S N 2d isplacement with methyl imidazole afforded imidazolium-bearing glycosides 12 and 13 in two steps and 23-33 %o verall yield. Subsequenta zido reduction by hydrogenolysis followedb yc ondensation with 1,4,5,8-naphthalenetetra-carboxylic dianhydride (NTCDA) under basic conditions afforded the desired NDI products 2 and 3.A nalogously,l igands 1 and 4-7 were obtained by reactiono fN TCDA with either aminoethyl glucoside [14] or 1-methylimidazolium amines S5 and S6 in ethanol (see the Supporting Information) and 3dimethylaminopropylamine [16] or 1-(3-aminopropyl)-4-methylpiperazine in toluene, respectively.
With all the compounds in hand, G-quadruplex and duplex DNA stabilization was evaluated in aF RET melting assay at ar ange of concentrations (1-10 mm), (Table1,F igure 2a nd Tables S1-3 in the Supporting Information). Sequences tested were F21T K + (human telomeric G-quadruplex in K + buffer), F21T Na + (human telomeric G-quadruplex in Na + buffer),F -Myc-T (c-Myc Pu-27 G-quadruplex) and F10T (duplex DNA) (see the Supporting Information). Uncharged glucoside 1 did not stabilizea ny of the DNA sequences, which is in accordance with earlier observations that cationic chargei sn eededf or quadruplex DNA stabilization. Excitingly,a ll of the charged ligands 2-6 showed different degrees of stabilization of as ingle G-quadruplex (F21T K + ), with no stabilization of the antiparallel F21T Na + quadruplex observed, whilst exhibitings ignificantly lower stabilization of duplex DNA (F10T). This is somewhat interesting, as it might be expected fors maller,l inear compounds such as 6 to be able to intercalate into duplex DNA. [17] At high concentrationso fl igand (e.g. 10 mm), some binding to both the C-Myc quadruplexa nd to al esser extentd uplex DNA is observed, however as the concentration of ligand decreases the selectivity forF 21T K + increases (see Figure 2). Remarkably, at 1 mm imidazolium-mannoside 3 and methylpiperazine 6 displayed a1 0.1 8Ca nd 9.8 8Cs tabilization for the F21T sequence in K + buffer,r espectively,w ith no observable binding to any other sequence. Furthermore, mannoside 3 showedamuch highers tabilization than glucoside 2.T his could be due to either differences in hydrogen-bonding interactions (OH at C-2 in mannose is axial vs. equatorial in glucose), or the relative orientation of the glycoside due to their anomeric linkage (axial vs. equatorial). These compounds, as far as we are aware, are the mosts electivec ompounds fort he hybrid/parallel type K + human telomeric G-quadruplex over the antiparallel Na + equivalent, and in addition exhibit the highests electivity for    the hybrid/parallel type K + human telomeric G-quadruplex over the parallel type c-Myc promoter quadruplex. Next, we sought to further investigate the quadruplex/ duplex DNA selectivity of the most interesting compounds 2, 3, 5 and 6,b yu sing aF RETcompetition assay with quadruplex forming oligonucleotide F21T K + and unlabeled ds26 competitor duplex DNA (Tables S4-6 and FiguresS2-4 in the Supporting Information). In general, the compounds displayed good quadruplex/duplex selectivity,r etaining 64-39 %s tabilization even at a1 :217 quadruplex:duplex DNA ratio (see Figure S2 in the Supporting Information). It is noteworthy that the quadruplex stabilization effect for glycosides 2 and 3 was less perturbed than for their non-carbohydrate counterparts (5 and 6, Figure S4 in the Supporting Information) suggesting that the former might be more selective towards quadruplex sequences.
To furtheru nderstand the binding and selectivity observed for compounds 3, 5 and 6 with the human telomeric G-quadruplex, circular dichroism (CD) titrationsw ere performed (Figure 3a nd Figures S5-S10 in the Supporting Information) with unlabeled human telomeric oligonucleotide telo23 in pH 7.2 and either 100 mm Na + or K + phosphate buffer.A ll three ligands appeared to cause significant perturbation to the observedC Ds pectrum,i ndicating that these compounds are able to induce changes to the conformationo ft he oligonucleotide.A ss hown in Figure 3, in K + buffer and at low ligand concentrations (1-3 equiv.o f3)s tabilization of the hybrid-type topology was observed, as evidenced by the negative band at 235 nm, ar eduction in the shoulder at 270 nm and an increase in the maximum at 290 nm. [18] As the concentrationo fl igand increases (upt o3 0equiv.), the maximum at 290 nm and the shoulder at 270 nm reverses direction. This effect could imply that these ligandsa ppear to stabilizeahybrid-type topology of telomeric DNA in K + buffer at low concentrations, yet shift to stabilize ap arallel-type topology at higher concentrations, with mannoside 3 exertingt he biggest shift in conformation. CD titrations in Na + buffer,w here the DNA sequence displays an antiparallel type topologys howedashift towards ah ybridtype topology (decrease at 240 nm, increase at 260 nm and 295 nm, Supporting Information Figure S5). This would indicate that ligands in question stillh ave some form of interaction with the antiparallel G-quadruplex, though this does not appear to result in stabilization of the structure in the FRET meltinga ssay.E ven in the presence of telo23 without any buffer, 3 was able to inducet he formationo fahybrid type Gquadruplex (decrease in band at 258 nm, increase in band at 288 nm, Supporting Information Figure S10). These results suggest that these NDI ligandsc an induce significant topological changes to the quadruplexu pon binding, anda re possibly interacting with ah ybrid/parallel-typec onformation of the humant elomeric G-quadruplex.
Isothermal titration calorimetry( ITC) was used to obtain quantitative thermodynamic information on the binding of 3, 5 & 6 to the telo23 K + human telomeric G-quadruplex (Table 2 and Figures S11-S13 in the Supporting Information. Thermodynamic analysis revealed that the ligandsd iffer in their binding affinity with a K a of the order of 10 6 m À1 ,w ith imidazolium 5 having the largest K a ,f ollowed by mannoside 3 and then 6. Further analysiso ft he data revealed that enthalpy was the main driving force for binding for all three compounds tested. Entropicc ontributionsw ere negative in each case, although they were much less for mannoside 3 than for imidazolium 5 or methylpiperazine 6.C ationic mannoside 3 showed the highest stabilizationeffect for F21T K + by FRET analysis and also exhibited the second highest DG values, with the lowest entropic penalty of all three ligandsi nvestigated. This suggestst hat the sugar side chain in the ligand can better displace water molecules from the complex surfacet han the otherm otifs. The ITCderived binding stoichiometry for 5 and 6 showeda pproximately two ligands interactingw ith the quadruplex, whilst the stoichiometry for 3 was 2.7. These differences in bindingt hermodynamicsb etween the different ligands indicatet hat be- Table 2. Thermodynamics of binding of 3, 5 and 6 to telo23 K + measured by ITC. K a = associationc onstant, N = stoichiometry, DG = Gibbs free energy change, DH = enthalpy change, DS = entropy change and T = temperature. Error (in brackets) represents ac onfidence interval of 68.3 %. For morei nformation, seethe SupportingI nformation.  sides the NDI scaffold, the side chains play an important role in binding to the oligonucleotide. Furthermore, binding of imidazolium-glycoside 3,i midazolium 5 and methylpiperazine 6 to telo23 K + is greater than that of cationic porphyrin TmPyP4 as measured by Bončina et al.,a nd of as imilarm agnitude as that of the quadruplex ligand Phen-DC3. [19] Molecular docking simulations were used to try to understand the binding mode of our ligands 2-6 with several of Gquadruplex( g4) structures (parallel, antiparallel and 3 + 1h ybrid human telomeric g4, and the Pu27T c-Myc g4) using Molsoft ICM-Pro [20] (Figure 4, Supporting Information Tables S7-9 and FiguresS14-41). All of the compounds evaluated exhibitedt he NDI end-stackingo nto the quadruplex, with one side chain pointing into ag roove. Models of binding for both 3 and 2 with the parallel human telomeric G-quadruplexr evealed severalp utative hydrogen bonds and imidazolium stacking interactionsw ithin this groove of the quadruplex, with one of the carbohydrate side chains inserted. Compound 6 also displayedahydrogen-bonding interaction between the protonated tertiary amine and thep hosphate in the groove. The additional interaction of the positivelyc harged imidazoli-um with the negatively charged phosphate of 3 in the groove is consistent with its enhanced K a over 6.F urtherw ork is still neededtof ully understand the precise mechanism of cell cytotoxicity for thesec ompounds and whether the observed activity is directly attributed to G-quadruplex based processes. [2a, 21] In order to evaluate the cell cytotoxicity profile of compounds 1-7 and doxorubicin as the benchmark, we compared 72 hi ncubations with WI-38 (embryonich uman lung fibroblasts), HeLa (human cervical cancerc ells), MDA-MB231 and MCF7 (human breast cancer cells) over the range of 10 fm-100 mm ligand, quantifying the number of live cells (Calcein fluorescence, Ta ble 3) and metabolic competence (Alamarblue, measureo fr eductive metabolism, Supporting Information Ta ble S10). Whereasg lyco-conjugates 1-3 did not appear to be toxic at concentrations up to 100 mm for all the cell cultures tested,a ll other compoundsi nt he class appeared to exhibit varying degrees of cytotoxicity,w ith increased toxicity towards cancer cells. HeLa cells were the most susceptible of all the cancer cells screened for example, 6 and 7 displayed 410 and 380 nm IC 50 values, respectively, as lighti mprovement in potency when compared to doxorubicin (530 nm). Methylpiperizine 6 displayed the most potent in vitro cancer-cell killing activity of all the compounds screened (e.g. IC 50 of 410 nm for HeLa and 810 nm for MDA-MB32 vs. 2.28 mm for the healthy WI-38 cells) (Table 3). It is important to highlight that whereas doxorubicin displayed at wo-fold selectivity for HeLa over WI-38, 6 displayed as ix-folds electivity for HeLa over WI-38, making it three times as selective. Compounds 4 and 5,w hich did not show significant G-quadruplex stabilization, were about one order of magnitude less toxic to HeLa, MCF-7 and WI-38 than 6.Anon-monotonic dose response was observed when 4 and 5 were tested against MDA-MB231 cells and IC 50 valuesc ould not be obtained, which suggest ad ifferent mode of action.
Ta king advantage of the inherentf luorescenceo fo ur ligands,confocal microscopy was used to assess the intracellular uptake in HeLa cells for compounds 3 and 6 (100 mm)a fter 30 min. and 16 hi ncubations. As shown in Figure 5, while significant intracellular uptake is seen form ethylpiperazine 6 after 30 min. (Figure 5-B and 5-F), the uptake for 3 is much lower (   selectiveG -quadruplex binding,c ould be attributed to poor cellular uptake. [22] Further work is still needed to fully understand the precise mechanism of cell cytotoxicity for these compounds,a nd whether the observed activity is directly attributed to G-quadruplex based processes. [2a, 21] In conclusion, the modulars ynthesis and evaluation of as eries of di-substituted naphthalened iimideG -quadruplex ligands that display high selectivity towards the hybrid-type topologyo ft he K + human telomeric G-quadruplex is described. Wes howed that imidazolium mannoside 3 is the most selectivel igand towards the F21T K + quadruplex sequence to date. Excitingly,c ompound 6,w hich is easily accessible in one step, is more toxic than doxorubicin towards cancer cells whilst exhibiting three times more selectivity.O ur findings suggest that lead divalent compounds 3, 5 and 6 providea ni nteresting platform for furtherd evelopment and that chargedc arbohydrates can be exploiteda sb inding motifs that can be tuned to interact with G-quadruplexg rooves. These results represent an exciting step towards developing more selective and bioactive G-quadruplex ligands with potentially improved anti-cancer activity.