Naturally occurring quaternary benzo[c]phenanthridine alkaloids selectively stabilize G-quadruplexes

In this work, the interaction of six natural benzo[c]phenanthridine alkaloids (macarpine, sanguilutine, sanguirubine, chelerythrine, sanguinarine and chelirubine) with parallel and antiparallel G-quadruplex DNA structures was studied. HT22 corresponding to the end of human telomere and the modified promoter oncogenes c-kit21 and Pu22 sequences have been used. Spectroscopically-monitored melting experiments and fluorescence titrations, competitive dialysis and nuclear magnetic resonance spectroscopy were used with this purpose. The results showed that these alkaloids stabilized G-quadruplex structures in terms of increments of Tm values (from 15 to 25 C) with high selectivity over duplexes and unfolded DNA. The mode of binding was mainly by stacking on the terminal Gtetrads with stoichiometries 1:2 (DNA:ligand). The presence of non-specific electrostatics interaction was also observed. Overall, the results pointed to a strong stabilization of G-quadruplex structures by these alkaloids.


Introduction
Quaternary benzo[c]phenanthridine alkaloids (QBAs) belong to the group of isoquinoline alkaloids. QBAs are present in plants from families Fumariaceae, Papaveraceae, Ranunculaceae and Rutaceae. In addition to relatively common alkaloids such as sanguinarine and chelerythrine, other less common alkaloids such as sanguilutine, macarpine, sanguirubine or chelirubine have been extracted (Figure 1 and S1) 1 . Some of these have proven antiproliferative effects on skin melanoma cells. 2 Chelerythrine and sanguinarine, as they are commercially available, have been investigated worldwide and their ability to inhibit some important enzymes in cancer cell division has been demonstrated many times. [3][4][5] Macarpine, which is found in plants in very small amounts, was first artificially prepared by T. Ishikawa 6 . This alkaloid and its derivatives also show strong cytotoxic effects in cancer cells. 7 In the case of chelirubine and sanguirubine, antimicrobial, anti-parasitic and anticancer effects have been demonstrated. 8,9 Besides these various biological effects on cells 2, 10 , QBA's in their iminium form were reported to interact with double stranded DNA (dsDNA) with a relatively weak mode 1,11 and comparable to that of ethidium bromide 12 . This interaction leads to changes in their fluorescent properties. Because of these changes, QBAs could also be used as fluorescent DNA probes.
Other DNA secondary structures have gained interest in recent years. One of these structures is the G-quadruplex (GQ) which is present in several protooncogenic-DNA promoters and thus participates in biological processes such as replication, transcription and translation 13,14 . The building blocks of these structures are the G-tetrads: almost planar arrangements of four guanine bases bonded by eight Hoogsteen hydrogen bonds (Figure 1). The G-quadruplex structure can be formed by the intermolecular association of four DNA molecules, by the dimerization of two molecules that contain two G-tracts, or by the intramolecular folding of a single molecule that contains four G-tracts.
The topology of G-quadruplexes may be parallel, antiparallel or hybrid, depending on the spatial orientation of the four G-tracts.
A great interest is observed in the potential of G-quadruplex as anticancer target 15 being the enzymatic activity inhibited by small ligands which stabilize the G-quadruplex. 16 In this work, the ability of natural alkaloids to stabilize GQ structures has been studied considering several relevant DNA sequences that have been shown previously to form antiparallel and parallel GQ structures ( Table 1).
The HT22 sequence, 5'-A(G3T2A)3G3-3', corresponds to the end of the human telomere and may adopt different Gquadruplex structures depending on the environmental conditions. To date, at least five distinct intramolecular Gquadruplex folding topologies have been reported for natural human telomeric repeats 17,18 , four of which were observed in the presence of K + ions 19 . The crystal structure of this sequence in the presence of K + formed a parallel intramolecular G-quadruplex 20 . Subsequent studies suggested that the intramolecular G-quadruplex structure observed in the K + -containing crystal appears unlikely to be the major form in K + -containing solution. Later, studies have shown that the telomeric sequence can form a mixed (3 parallel + 1 antiparallel) structure in K + solution 21 . More recently, another form was observed in K + solution, consisting of a two-G-tetrad basket-type core with extensive base stacking interactions in the loops 17 . Finally, an antiparallel (2+2) structure has been observed in Na + solution 22 .
Several GQ-forming sequences have been identified within the promoter segment of the human c-kit oncogene upstream of its transcription initiation site. The 21-mer sequence (5'-CG3CG3CGCGAG3AG4-3', ckit21) forms polymorphic G-quadruplex structures. [23][24][25] but the mutated sequences (c-kit21T21, Table 1), 5'-CG3CG3CGCGAG3AG3T-3', and c-kit21T12T21, with one G to T mutation at level of 21 and with two G to T mutations at level of 12 and 21 residues, respectively, display more simple conformations. The G21T mutation restrains the length of the third loop to a single nucleotide and the fourth G-tract to three guanines. This modification has a significant effect on the biophysical properties, leading to the stabilization of the parallel-stranded topology. 26,27 The first sequence was used in the present work for CD and fluorescence experiments, the second one for nuclear magnetic resonance (NMR) spectroscopy.
Another important oncogene is c-myc, the overexpression of which is the cause of a wide range of genetic tumors.
Pu22 is a 22-mer sequence mainly responsible for the c-myc transcriptional activity 16 . Pu22-14T23T is the same sequence with two G to T mutations at position 14 and 23. It adopts the single predominant intramolecular parallel GQ conformation under K + physiological concentration, and thus shows better resolved NMR spectra. Recently, it has been reported that Pu22-14T23T gives the same interactions with ligands as wild type Pu22. 28,29 In this work, the interaction of six natural benzo[c]phenanthridine alkaloids (macarpine, sanguilutine, sanguirubine, chelerythrine, sanguinarine and chelirubine) with GQ DNA structures formed by HT22, c-kit21T21, c-kit21T12T21 and Pu22T14T23 sequences was studied. Circular dichroism, spectroscopically-monitored melting experiments and fluorescence titrations, competitive dialysis and nuclear magnetic resonance spectroscopy were used with this purpose.

Chemicals
Alkaloids were extracted from plant material in Department of Biochemistry, Faculty of Science, Masaryk University (Brno, Czech Rep.). Some of the oligonucleotides used in this work (Table 1) were purchased as dry samples from Thermo Fisher Scientific (USA) at HPLC grade. In other cases, DNA synthesis was performed on an Applied Biosystems DNA/RNA 3400 synthesizer by solid-phase 2-cyanoethylphosphoroamidite chemistry. DNAs were desalted in a Sephadex (NAP-10) G25 column and passed through a DOWEX(Na + ) resin to exchange ammonium to sodium cations. In all cases, DNAs were diluted in re-distilled water with Trizma® base (10 mM) and EDTA (0.1 mM) buffer (pH = 8) to stabilize them during storage. Other chemicals such Trizma® base (C4H11NO3, p.a.) and EDTA (p.a.) were obtained from Sigma-Aldrich (USA). Basic chemicals such as KH2PO4, KCl, NaOH and KOH (all p.a. grade) were purchased from Lach-Ner (Czech Rep.).

Instruments
Absorbance spectra were recorded on an Agilent 8453 diode array spectrophotometer (Agilent Technologies; Waldbronn, Germany). Temperature was controlled by means of an Agilent 89090A Peltier device (Agilent Technologies). CD spectra were recorded on a Jasco J-810 spectropolarimeter equipped with a JULABO F-25-HD temperature control unit (Seelbach, Alemania). Fluorescence spectra were measured with an Aminco-Bowman Series 2 spectrofluorimeter (Thermo-Spectronic, USA), equipped with xenon lamp. Temperature was controlled by means of a water bath. Excitation wavelength was depending on QBA used for titration, ranging from 330 to 350 nm.
Emission wavelength for complex QBA:DNA was set to 600 nm. In all spectroscopic studies, Hellma quartz cells (10 mm path length, and 350, 1500 or 3000 µl volume) were used. The NMR spectra were recorded on a Bruker AV600 spectrometer operating at a frequency of 600.10 MHz, equipped with a 5 mm TXI inverse probe and z -axis gradients. The 1 H spectra were referenced to external DSS (2,2-dimethyl-2-silapentane-5-sulfonate sodium salt) set at 0.00 ppm.

Melting experiments
In a typical melting experiment, DNA (final concentration 2 µM) was mixed with QBA (4 µM) together with phosphate buffer (10 mM) and KCl (5 mM). The concentration of KCl was set to 5 mM instead of the most usual 100 or 150 mM concentration in order to reduce the high thermal stability of GQ structures. In this way, the potential stabilization of the GQ by the presence of QBA in terms of Tm could be determined accurately. The sample was heated (96°C) and then allowed to cool slowly. After several hours, the sample was placed to the instrument (Agilent 8453 UV-Vis or Jasco J815 CD spectrometers). Stirred sample was heated to 96°C and cooled down during measurement at a rate of 0.5°C·min -1 . Sample was measured also during heating process starting from 20 °C to 96 °C in the same rate. The absence of hysteresis was checked for some of the QBA:DNA mixtures. Melting temperatures (Tm) were determined as described elsewhere 30 using home-made routines written in Matlab® code.

Fluorescence experiments
Fluorescence measurements were performed to determine binding stochiometries and overall association constants according to equation 1: The constants were determined from fluorescence-monitored titrations of QBAs by GQ at 25 o C. In all experiments, the concentration of QBA was kept constant (3 M) whereas the concentration of the considered GQ was increased.
Binding data analysis was done with the OPIUM program. 31 Alternatively, the Job method was used to determine the binding stoichiometry of the QBA:GQ interaction complex.
In this method, the total molar concentration of the two molecules was kept constant, whereas the molar fraction was varied. The total concentration of QBA and GQ was 3 µM. The stoichiometry of the QBA:DNA interaction complex was estimated from the intersection of two lines fitting those points measured at lowest and highest molar fractions.

Competitive dialysis studies
A 100 l of a 50 M DNA of the different DNA sequences were dissolved in potassium phosphate buffer (185 mM NaCl, 185 mM KCl, 2 mM NaH2PO4, 1 mM Na2EDTA, 6 mM Na2HPO4 at pH 7 and introduced into a separated dialysis unit (Slide-A-Lyzer™ MINI device, 3500MWCO, Thermo Fisher) and a blank sample containing only buffer. All dialysis units were allowed to equilibrate during 24 h at room temperature in a beaker containing the 1 M solution of the appropriate QBA. At the end of the dialysis experiment, the amount of QBA bound to the DNA was quantified by measuring the fluorescence spectra.

Nuclear Magnetic Resonance
The NMR samples of Pu22-T14T23 and c-kit21T12T21 (Table 1) were prepared at concentration 0.34 mM and 0.42 mM. Pu22-T14T23 was dissolved in 25 mM KH2PO4, 70 mM KCl, pH 6.9 and ckit21T12T21 was dissolved in 5mM KH2PO4, 20 mM KCl, pH 6.9. In these salts condition ckit21T12T21 is present as a monomeric form (Form I). 25 The DNA samples were heated to 85°C for 1 min and then cooled at room temperature overnight. with 2048 x 1024 complex FIDs. Mixing times ranged from 100 ms to 300 ms. TOCSY spectra were acquired with the use of a MLEV-17 spin-lock pulse (60 ms total duration). All spectra were transformed and weighted with a 90° shifted sine-bell squared function to 4K x 4K real data points. Proton resonance assignments of GG21-T12T21 and Pu22-T14T23 free and complexed were performed on the basis of previous assignments. 25,32,33 The chemical shift values of the complex of chelerythrine with ckit21T12T21 are reported in Table 2. The chemical shift values of the complexes of Pu22-T14T23 with sanguilutine and chelerythrine are reported in Table S1 and Table S2 respectively.

Results and discussion
Effect of QBAs on G-quadruplex structure by CD and fluorescence experiments First, the overall GQ structures formed by the HT22 and ckit21T21 sequences at the experimental conditions were identified by means of CD spectroscopy ( Figure S2). The shape and position of the bands in the CD spectra reflected the overall antiparallel or parallel nature of the GQ structure. Hence, a positive band around 285 nm indicated the predominance of the antiparallel structure in the case of HT22 sequence, and a negative band around 240 nm were indications of a parallel structure in the case of the ckit21T21 sequence. In general, the addition of QBA to both HT and ckit21T21 GQ structures did not affect dramatically to the shape and intensity of the CD spectra of DNA recorded in the UV region ( Figure S2). This fact was related to the maintenance of the overall GQ structure upon interaction with the ligand. On the other hand, CD measurements in the visible region ( Figure S3) did not show the appearance of any significant signal related to induced CD (ICD) in the ligand. It is known that strong ICD signals are usually related to intercalation 34 . Therefore, the absence of ICD points out to another binding mode, such as endstacking or electrostatics. The absence of intercalation, on the other hand, would agree with the fact that the overall GQ structure is maintained because this mode of interaction probably would alter the DNA structure.

Thermal stabilization
Melting experiments were carried out to observe any positive contribution of QBAs to the thermal stability of GQ structures. It is known that a shift in the unfolding process of GQ in presence of a ligand towards higher temperature values reflects the tendency of the ligand to interact more strongly with the GQ folded species than with the unfolded strand 35 . Table 3 summarizes the determined melting temperature (Tm) values in the absence and presence of QBAs.
The melting experiments with antiparallel structure HT22 were monitored either with UV-Vis or CD spectroscopies.
The determined Tm values in both cases were very similar (within ±0.5°C). Melting temperature of HT22 was found to be 51.0°C (Figure 2). Practically no changes between melting temperatures calculated from data obtained during cooling or heating the studied systems were observed ( Figure S4), which ruled out the presence of hysteresis. In general, all other QBAs produced a clear stabilization of this GQ structure. The highest Tm value in case of HT22 was observed for sanguinarine (ΔTm = 18.1 °C), which reflects a greater tendency of this ligand to interact with the folded GQ structure than with the unfolded strand. On the other hand, macarpine (ΔTm = -0.9 °C) showed a similar tendency to interact with both species (GQ folded and unfolded strand).
Melting experiments on parallel structure ckit21T21 were exclusively made by CD spectrometry because of the low absorbance changes observed at 295 nm during measurements (see an example in Figure S5). The Tm value of ckit21T21 was found to be 50.0°C. In this case, the most conspicuous shifts in case of ckit21T21 were observed for chelirubine (Tm = 24.5 °C) and sanguilutine (Tm = 22.4 °C). Again, macarpine showed the lowest GQ-stabilization properties, as Tm only increased 6.3 °C. To our knowledge, the stabilizing propierties showed by chelirubine is one of the highest observed differences in melting values for parallel GQ structures. 33,36,37 As example, the model ligand TMPyP4 stabilizes telomeric DNA less than selected alkaloids about (ΔTm = 1 -13°C) at the same concentration ratio. [37][38][39] In general, the studied alkaloids show Tm values higher in the case of parallel GQ (ckit21T21) than in the case of antiparallel (HT22) structures. This fact may be related to the different loop geometry in both structures, as the GQ core formed by three G-tetrads is similar. It seems that these alkaloids interact much better with a structure showing double-chain reversal loops, like ckit21T21.  Figures 1 and S1). All the other alkaloids, which do not include R6 group, showed GQ-stabilizing properties.

Selectivity
At this point, it is necessary to mention that melting experiments with dsDNAs and QBAs did not show any significant increase of melting temperature (Tm < 2°C) (Figure 2b). Therefore, QBAs appeared as a potential selective group of ligands to bind GQ structures. To gain more information about the selectivity of QBAs for DNA structures or sequences, competitive dialysis experiments were performed ( Figure 3 and Table S4) 41 . In these, a set of dialysis units containing different DNA sequences, some of them prone to form folded structures, is placed inside a solution of the considered ligand. After an equilibration period, part of the ligand enters to the dialysis units depending on its selectivity towards each sequence or structure, as well on the binding stoichiometry. Quantification is done by measuring emission fluorescence at 600 nm. In this work, a set of different DNA sequences (Table 1) representing several nucleic acid structures was used. 16 for both HT22 and ckit21T21 GQ structures (see below), the fluorescence intensity is higher for the parallel structure than for the antiparallel one.

Determination of DNA:ligand stoichiometry
The interaction of GQ with the QBAs causes an increase of their intrinsic fluorescence that can be used to determine   Figure S6). The 1:3 stoichiometry was the best fit for both studied GQ structures with almost all alkaloids. At neutral pH QBAs are mostly in iminium (positively charged) form as their pKR+ values lie between 7.7 and 9 43 . As previously reported, QBAs in iminium form interact with dsDNA forming highly luminescent complexes. 44 Although there is no clear evidence for the mechanism of interaction with dsDNA, it is supposed that planar positively charged alkaloids are intercalated between base pairs of DNA and due to this incorporation, the luminescence is enhanced (except for sanguinarine). A similar enhancement of the luminescence emission was observed also for mixtures of QBAs with GQ. However, the intercalation mechanism does not seem a very plausible possibility, as pointed out by CD measurements, because of insufficient space for the ligands between tetrads of GQ which, moreover, are probably occupied by cations 45,46 . In this sense, the stoichiometries determined in this work are far from the 1:1 (GQ:QBA) stoichiometry described by Bhadra 47 for sanguinarine, coralyne, palmatine and berberine with 5'-AG3(T2AG3)3-3'.
However, it should be taken into account that the experimental conditions were different, and these could not only affect to the GQ structure (which was an hybrid parallel/antiparallel in that work) but also to the presence of additional interactions.
Xiong et al. 48 (Table S5), b) the titration experiments and c) the inter-residue NOE connectivities between the H1 imino and aromatic H8 of guanine residues.
A large change in the chemical shifts of the H1 imino protons (Δδ ≥0.80 ppm) was observed both for the H1 imino protons belonging to outer G-tetrads and for the internal one (Δδ ≥0.40 ppm) ( Table 2). Several NOE contacts were found: the aromatic proton H6 of chelerythrine with H1 imino protons of G18, G7 and G19, the NCH3 protons with G20, and the 8-ethylene-dioxy with G18, suggesting strong interactions of chelerythrine with the oligonucleotide (Table 4) To better clarify the mode of binding of these alkaloids with the parallel G-quadruplex structure and to extend our investigation to the sequence responsible for the c-myc transcription activity we performed the NMR experiments with the Pu22-T14T23 sequence.
The titration with chelerythrine and sanguilutine induced, even at low R=[ligand]/[DNA] ratio, a broadening of all the signals of DNA and of the ligand. At R ≥ 1.5 a new set of imino protons signals appeared at up-field shift, and the signals sharpened at R=2.0 ( Figure 6). This suggested the formation of a defined complex with two ligand molecules interacting with the GQ structure. A further addition of ligands to Pu22-T14T23 caused only small changes in the H1 imino protons until the R = 3.0 was reached.
The analysis of the spectra at R=3.0 was performed starting from the attribution of the three tetrads by inter-residue NOE connectivities between the H1 imino and the aromatic protons H8 of guanine residues (following the procedure used for the study of other ligands. 28,52 The results reported in Table S6 show that the quadruplex structure is conserved. Also, for these complexes a significant shielding was observed for the H1 imino protons of the internal tetrad (δ = -0.30/-0.60 ppm) although lower than the values of the external tetrads (δ ≥-0.60 ppm). In particular, it is relevant the δ =-1.36 ppm observed for the G16H1 of the chelerythrine complex (Tables S1 and S2).
Sanguilutine showed NOE contacts of H6 with G22 H1. Other NOEs were found between some aromatic protons and DOSY experiment, performed on the complex with chelerythrine, showed a diffusion coefficient indicating that the stoichiometry of the complex may be more than two ligands for G-quadruplex and excludes a higher aggregation of the nucleotide. The significant upfield chemical shifts of the imino protons belonging to the internal tetrad are difficult to be explained but they can suggest an interaction also at the level of this tetrad.

Conclusions
The interaction between human telomeric DNA (HT22) and modified c-kit structure (ckit21T21) and six plant The NOESY experiments, performed with c-kit21T12T21 and Pu22T14T23 sequences, show that sanguilutine and chelerythrine have NOEs contacts with units at 3'and 5' end, with two molecules being located respectively over the outer tetrads.
The possibility that the ligand molecules may interact also at the level of the internal tetrad is suggested by the strong shelding of the imino protons signals of these units. This appears more probable for the complex with ckit21T21 sequence, where the ligand positioned in different binding sites appears in chemical exchange.
Also, other studied alkaloids, except macarpine, have shown strong stabilizing effect, forming 1:3 or 1:4 complexes with GQ. These compounds seem to be potential selective GQ ligands as they practically do not stabilize the B-DNA duplex structure.
Overall, these results suggest the potential use of these minor, non-commercial QBAs as G-quadruplex stabilizers in vivo, or for the development of analytical methods based on fluorescence spectroscopy.