Naphthalene Diimides Carrying Two β-Cyclodextrins Prefer Telomere RNA G-Quadruplex Recognition

Newly synthesized naphthalene diimide carrying two β-cyclodextrins (NDI-β-CyDs) showed improved specificity for the parallel G-quadruplex structure alongside the hybrid G-quadruplex structure. Specifically, the highest binding affinity of NDI-β-CyDs for the telomere RNA G-quadruplex was observed. The binding simulation indicated that β-cyclodextrins might be available for loop nucleobase inclusion under its complex.


Introduction
G-quadruplex structures that comprehensively exist in the gene promoter region have been reported, which might regulate gene transcription and expression, and in the telomere region that is involved in inhibiting telomerase's activity, further inducing cancer cell death [1][2][3]. Therefore, G-quadruplex-specific targeting has been considered a promising anticancer therapy with less side effects due to its biological roles [4,5]. Within the progress of NMR and crystal structure analysis techniques, plenty of G-quadruplex structures with varied sequences have been elucidated in detail [6]. Generally parallel, antiparallel, and hybrid (mixed type) structures were considered as the classical G-quadruplex forms [7]. The telomere DNA G-quadruplex can form a hybrid G-quadruplex with a lateral loop and diagonal loop [8], while the c-myc promoter region forms a flat parallel G-quadruplex with site chain loops [9].
In particular, since it is known that the inhibition of transcription of telomeric repeatcontaining RNA (TERRA), as a telomeric RNA [28], causes the inhibition of telomere length maintenance or ALT activity, pyridostatin [27] and naphthalene diimide derivatives [14,15,[17][18][19][20][21][22][23][24][25][26] have been developed as ligands that bind to TERRA. However, it is well known that telomere RNA G-quadruplex forms a parallel structure from Circular Dichroism (CD) spectra identification, and a parallel G-quadruplex was detailed and identified within two 12nt telomere RNA sequences [29]. There is was detailed and identified within two 12nt telomere RNA sequences [29]. There is still relatively little information about RNA G-quadruplexes, especially longer telomere RNA G-quadruplexes.
Due to the varied preference of molecules for different G-quadruplex topologies, G4 ligands, which can discriminate G-quadruplex patterns, are considered promising G4 ligands with better selectivity. Some promising strategies have been developed for discriminating G-quadruplexes and dsDNA, such as the use of naphthalene diimide, which was designed to interfere with its binding to dsDNA, and therefore enhance its selectivity towards the G-quadruplex [17,18]. Some ligands prefer to recognize the lateral and/or diagonal loop of the hybrid G-quadruplex, and some conjugators even specifically insert into the TTA linker pocket of the two-telomere DNA G-quadruplex [30]. Here, we reported a new strategy that enhances the selectivity of the G4 ligand for the parallel G-quadruplex, especially of the parallel telomere RNA G-quadruplex.
As shown in Figure 1A, we conjugated two β-cyclodextrins with naphthalene diimide to obtain NDI-β-CyDs, 1 and 2. The NDI plane provides recognition and stacking with the G-quartet [19,26], and the bulk body of β-cyclodextrin inhibits the interaction of NDI-β-CyDs with dsDNA due to steric hindrance, and also disturbs its binding to Gquadruplexes with side chain loops. Thus, a flat parallel G-quadruplex was proposed that would be preferable for binding ( Figure 1B). To identify whether this concept is feasible or not, two NDI-β-CyDs, 1 and 2, were synthesized with different lengths of the linker between NDI and β-CyDs, and 3 was adopted as a control ligand for comparison.

Results and Discussion
The synthesis and confirmation of 1 and 2 were listed in supporting information (Figures S1-S6), 3 was prepared according to the previous report [31].

Binding Behaviors of 1 and 2 with DNA or RNA G-Quadruplex
Classical G-quadruplex DNA and RNA sequences were adopted to conduct the investigation (Table 1). CD spectra of these sequences were determined by adding ligands; for all these DNA sequences, 3 showed influence with CD spectra shift (292 nm up shift of Telomere G4, 267 nm down shift of c-myc and c-kit, 282 nm up shift of ds oligo), 1 and 2 displayed less impact on telomere G4, c-kit and ds oligo, while similarly, a 267 nm shift of c-myc's CD spectra was obtained, which suggested that 1 and 2 might prefer c-myc (forming parallel G-quadruplex structure) recognition ( Figure S7). Although c-kit was also 1 R:

Results and Discussion
The synthesis and confirmation of 1 and 2 were listed in supporting information (Figures S1-S6), 3 was prepared according to the previous report [31].

Binding Behaviors of 1 and 2 with DNA or RNA G-Quadruplex
Classical G-quadruplex DNA and RNA sequences were adopted to conduct the investigation (Table 1). CD spectra of these sequences were determined by adding ligands; for all these DNA sequences, 3 showed influence with CD spectra shift (292 nm up shift of Telomere G4, 267 nm down shift of c-myc and c-kit, 282 nm up shift of ds oligo), 1 and 2 displayed less impact on telomere G4, c-kit and ds oligo, while similarly, a 267 nm shift of c-myc's CD spectra was obtained, which suggested that 1 and 2 might prefer c-myc (forming parallel G-quadruplex structure) recognition ( Figure S7). Although c-kit was also proposed, which forms a parallel G-quadruplex, it was identified with a unique structural scaffold with four loops that interfere with ligand recognition [31]. For telomere RNA G-quadruplexes (23nt, 12nt), adding 1 and 2 could induce the CD spectra to shift both the telomere RNA G-quadruplex, which further indicated the recognition and binding of 1 and 2 for the parallel G-quadruplex (Figures 2, S7 and S8). proposed, which forms a parallel G-quadruplex, it was identified with a unique structural scaffold with four loops that interfere with ligand recognition [31]. For telomere RNA Gquadruplexes (23nt, 12nt), adding 1 and 2 could induce the CD spectra to shift both the telomere RNA G-quadruplex, which further indicated the recognition and binding of 1 and 2 for the parallel G-quadruplex (Figures 2, S7 and S8).   UV-Vis measurements were then performed at 25 °C when Telomere G1, c-myc, and ds-oligo were added to a solution of 50 mM Tris-HCl (pH 7.4), and 100 mM KCl containing 5 μM 1-3, respectively (Figures 3 and S9). The UV-Vis spectra were saturated when c-myc was added in 1 or when Telomere G1, c-myc, and ds-oligo were added in 3. The binding constants are summarized in Table 2 and were obtained using a Scatchard analysis of these measured data.
In other cases, however, the change in absorbance with the addition of DNA was small and the UV-Vis spectrum was not saturated. In this case, the Benesi-Hildebrand analysis was performed ( Figure S10). It was found that cyclodextrin-free 3 could bind to Telomere G1, c-myc, and ds-oligo, but the binding constant for c-myc was lower than that of cyclodextrin-containing 1 (8.6 × 10 5 M −1 ). This indicates that cyclodextrin is highly selective for 1 to c-myc. Similarly, cyclodextrin-containing 2 showed higher affinity for c-myc (1.0 × 10 5 M −1 ) than Telomere G1 and ds-oligo. Hypochromicity (%) was generally corre- UV-Vis measurements were then performed at 25 • C when Telomere G1, c-myc, and ds-oligo were added to a solution of 50 mM Tris-HCl (pH 7.4), and 100 mM KCl containing 5 µM 1-3, respectively ( Figure 3 and Figure S9). The UV-Vis spectra were saturated when c-myc was added in 1 or when Telomere G1, c-myc, and ds-oligo were added in 3. The binding constants are summarized in Table 2 and were obtained using a Scatchard analysis of these measured data.
In other cases, however, the change in absorbance with the addition of DNA was small and the UV-Vis spectrum was not saturated. In this case, the Benesi-Hildebrand analysis was performed ( Figure S10). It was found that cyclodextrin-free 3 could bind to Telomere G1, c-myc, and ds-oligo, but the binding constant for c-myc was lower than that of cyclodextrincontaining 1 (8.6 × 10 5 M −1 ). This indicates that cyclodextrin is highly selective for 1 to c-myc. Similarly, cyclodextrin-containing 2 showed higher affinity for c-myc (1.0 × 10 5 M −1 ) than Telomere G1 and ds-oligo. Hypochromicity (%) was generally correlated with the binding constant, and when the binding constant was high, the hypochromicity was also large, suggesting good stacking between naphthalene diimide and the base. lated with the binding constant, and when the binding constant was high, the hypochromicity was also large, suggesting good stacking between naphthalene diimide and the base.  Telomere G4 c-myc ds-oligo 1 2 To identify whether 1 and 2 showed improved specificity for parallel G-quadruplex binding, isothermal titration calorimetry (ITC) was adopted to evaluate the accurate binding properties (Figures 4 and S11, Table 3). Some substituted naphthalene diimides were designed with enhanced selectivity and showed marked biological performance [20][21][22]25,26]. In particular, sugar-modified trisubstituted naphthalene diimides showed high G-quartet DNA and RNA selectivity [25], while the naphthalene diimide plane could bind to the G-quadruplex structure with no discrimination of different sequences, and demonstrated a comparable affinity for interacting with ds-oligo, which limited the application of the unmodified naphthalene diimide plane as a G4-specific ligand. Similarly, the newly constructed naphthalene diimide substitutions 1 and 2 barely bound to ds-oligo (undetectable in ITC measurement, and Ka was calculated using UV-Vis with Scatchard plot analysis), and both showed a decreased response for the hybrid telomere G4 sequence, especially 2, which was almost completely shut off from telomere G4 recognition. However, it is noticeable that although 1 and 2 were able to recognize the c-myc G-quadruplex, their binding affinity was weaker than 3 (Table 3).
The binding constant of 3 to G1-r23nt was lower than that to Telomere G1. The affinity of 3 for c-myc, a parallel structure, was also somewhat lower than for Telomere G1, a hybrid structure. Nitrogen-bearing 2-substituted NDI is known to hydrogen bond to the phosphate in the lateral loop of telomere G1 [26]. When 3 is stacked in the G quartet, it   To identify whether 1 and 2 showed improved specificity for parallel G-quadruplex binding, isothermal titration calorimetry (ITC) was adopted to evaluate the accurate binding properties (Figures 4 and S11, Table 3). Some substituted naphthalene diimides were designed with enhanced selectivity and showed marked biological performance [20][21][22]25,26]. In particular, sugar-modified trisubstituted naphthalene diimides showed high G-quartet DNA and RNA selectivity [25], while the naphthalene diimide plane could bind to the G-quadruplex structure with no discrimination of different sequences, and demonstrated a comparable affinity for interacting with ds-oligo, which limited the application of the unmodified naphthalene diimide plane as a G4-specific ligand. Similarly, the newly constructed naphthalene diimide substitutions 1 and 2 barely bound to ds-oligo (undetectable in ITC measurement, and K a was calculated using UV-Vis with Scatchard plot analysis), and both showed a decreased response for the hybrid telomere G4 sequence, especially 2, which was almost completely shut off from telomere G4 recognition. However, it is noticeable that although 1 and 2 were able to recognize the c-myc G-quadruplex, their binding affinity was weaker than 3 (Table 3).
In contrast, when testing the affinity of ligands for telomere RNA G-quadruplex formed by G1-r23nt, both NDI-β-CyDs showed a stronger affinity than 3, especially 1, which displayed the K a for G1-r23nt at 19 × 10 5 M −1 .
The binding constant of 3 to G1-r23nt was lower than that to Telomere G1. The affinity of 3 for c-myc, a parallel structure, was also somewhat lower than for Telomere G1, a hybrid structure. Nitrogen-bearing 2-substituted NDI is known to hydrogen bond to the phosphate in the lateral loop of telomere G1 [26]. When 3 is stacked in the G quartet, it can hydrogen bond to the lateral loop in Telomere G1 where the lateral loop is present, but in the parallel structure, hydrogen bonding is not effective and affinity is expected to be slightly lower.
oligo or different G-quadruplex structures, both 1 and 2 showed an enhanced affinity for the G-quadruplex. Furthermore, 1 and 2 could prefer recognizing G1-r23nt by more than 10 fold higher than for telomere G4 (10.7 folds of 1, and 20.5 folds of 2), which could be considered superior specificity for targeting the telomere RNA G-quadruplex. Regarding the structure, the 2′-OH hydroxyl groups in the RNA quadruplex play a significant role in redefining the hydration structure, which may provide a positive environment for NDIβ-CyDs with β-CyD as a side chain to interact with the loops [26].   A 12nt parallel telomere RNA G-quadruplex (G1-r12nt) which has been reported with structure details was also adopted for investigation [17]. When compared to G1-r23nt, both NDI-β-CDs displayed a weaker affinity for the intramolecular parallel RNA G-quadruplex. Based on this result, the telomere RNA G-quadruplex (G1-r23nt) formed a distinct Gquadruplex structure with G1-r12nt, with an unconnected 12U, 1U, and 2A structure, especially the 3 side chain of G1-r23nt, which might be more useful for the recognition of NDI-β-CyDs. Compared to the poor performance of 3 in discriminating ds-oligo or different G-quadruplex structures, both 1 and 2 showed an enhanced affinity for the Gquadruplex. Furthermore, 1 and 2 could prefer recognizing G1-r23nt by more than 10 fold higher than for telomere G4 (10.7 folds of 1, and 20.5 folds of 2), which could be considered superior specificity for targeting the telomere RNA G-quadruplex. Regarding the structure, the 2 -OH hydroxyl groups in the RNA quadruplex play a significant role in redefining the hydration structure, which may provide a positive environment for NDI-β-CyDs with β-CyD as a side chain to interact with the loops [26].
Compared to 3, the binding of 1 and 2 with the G-quadruplex generally produced larger negative ∆S values, which indicated a more stable condition of the NDI-β-CyDs-Gquadruplex complex, and the ∆S value of 1 with G-quadruplex was more negative than 2, which also suggested that 1, with a shorted linker between NDI and β-CyD, may restrict the adjustability more during the binding process. However, except for G1-r23nt, similar thermal parameter values but distinct K a values among 1, 2 and 3 were obtained (Table 3).

Binding Model of 1 and 2 with DNA or RNA G-Quadruplex
The binding model of NDI-β-CyDs with G-quadruplex structures was simulated by computer calculation (Figures 5 and S12). Other than the difficulty observed in fitting 1 and 2 with the hybrid telomere DNA G-quadruplex (unable to get a stable model), for both c-myc and the G1-r12nt parallel G-quadruplex structure, stable binding models were successfully simulated. Interestingly, in both models, except for the stacking of the NDI plane with G-quartet, β-CyDs were observed as a pocket for nucleobase inclusion (T7 and T16 in c-myc; U7 and U18 in G1-r12nt). Since previous reports supported the interaction between β-CyD and nucleobase [32], the c-myc sequence within two AP sites (apyromidinic sites without the base of T7 and T16) was adopted to investigate its influence on NDI-β-CyDs' recognition of the G-quadruplex. Although a slight decreasing affinity of 1 or 2 for the c-myc AP site was observed with increased entropy (Figure 6, Table 3), it is still difficult to directly claim whether such nucleobase inclusion could indeed occur during G4 ligand binding with the G-quadruplex or not. Moreover, varied positioning of β-CyDs for c-myc or G1-r12nt was observed, which indicated that the shape of the G-quadruplex could be an important factor impacting NDI-β-CyDs' G-quadruplex recognition.    Finally, we analyzed the stabilization effect of 1 and 2 for the G-quadruplex structure based on T m measurement ( Figure 7, Table 4). A comparison of 3, 1 and 2 revealed a weaker ability to enhance the thermal stability of the G-quadruplex, which might be because the bulky structure of β-CyDs causes more resistance to maintain the NDI-β-CyDs-G-quadruplex complex, while there was almost no stabilization of NDI-β-CyDs for the telomere DNA G-quadruplex, and a comparatively stronger ability of NDI-β-CyDs for stabilizing the telomere RNA G-quadruplex was observed.

Generals
DNA and RNA oligonucleotides (Table 1) were purchased from Hokkaido System Science Co., Ltd. (Sapporo, Japan). Before use, the DNA was annealed under the following conditions: heating to 95˚C for 10 min, then cooling to 25 °C at 0.5 °C/min. A previously

Generals
DNA and RNA oligonucleotides ( Table 1) were purchased from Hokkaido System Science Co., Ltd. (Sapporo, Japan). Before use, the DNA was annealed under the following conditions: heating to 95 • C for 10 min, then cooling to 25 • C at 0.5 • C/min. A previously reported procedure was used to synthesize 3 [33].
UV/Vis and CD spectra were measured with 50 mM Tris-HCl, 100 mM KCl to fix the salt concentration. However, the ITC measurement was difficult with 50 mM Tris-HCl buffer (pH 7.4) containing 100 mM KCl, and the measurement conditions were the recommended conditions of the nano ITC (50 mM KH 2 PO 4 -K 2 HPO 4 buffer, pH 7.0). Before performing the ITC measurement, the CD spectrum measurement of G4 with the addition of KCl was performed in a 50 mM KH 2 PO 4 -K 2 HPO 4 buffer (pH 7.0), following which it was confirmed that the spectrum of the G4 structure did not change and that a G-quadruplex structure was sufficiently formed. In addition, when Tm a measurement could not be performed with 50 mM Tris-HCl, 100 mM KCl, the conditions were changed. Since the Tm curves of G1-r23nt and c-myc at 100 mM KCl did not change completely, the KCl concentration was set to 5 mM. On the contrary, G1-r12 was insufficient for the Tm measurement at a KCl of 100 mM, so it was measured with 50 mM KH 2 PO 4 -K 2 HPO 4 buffer (pH 7.0) and 100 mM KCl.

Synthesis of 2
A total of 50 mL was added to 1,4,5,8-Naphthalenetetracarboxylic acid 1.1 g (4.1 mmol) and Glycine 1.3 g (18 mmol), and reflux was performed with heating for 6 days. Then, the solution was cooled to room temperature, and kept on ice for 1 h. The yellow precipitation was collected by suction filtration, and rinsed with MilliQ several times. After reduced pressure drying, the obtained solid was washed with 100 mL acetonitrile, and the precipitation was dried under reduced pressure drying again. 4 was obtained and confirmed with MALDI-TOF-MS ( Figure S3), and 1 H-NMR ( Figure S4). Yield 1.5 g (3.8 mmol), yield ratio 94%. 1  N, N -Bis(3-methylaminopropyl)naphthalene-1,4,5,8-tetracarboxylic acid diimide 0.2 g (0.5 mmol) and Tosyl-β-CyD 2.7 g (2.1 mmol) were dissolved in 10 mL, and mixed in 70 • C for 48 h. Then, the solvent was removed by reduced pressure evaporation, and 10 mL of 0.1% trifluoroacetic acid (TFA) containing 7% acetonitrile was added and mixed in room temperature, after which the solution was collected by filtration. The target compound 1 was purified by RP-HPLC, and dried by freeze drying to obtain white powder. 1 was confirmed by 1 H-NMR ( Figure S1), HPLC ( Figure S2) and HRMS (EI + ).

Synthesis of 2
A total of 50 mL was added to 1,4,5,8-Naphthalenetetracarboxylic acid 1.1 g (4.1 mmol) and Glycine 1.3 g (18 mmol), and reflux was performed with heating for 6 days. Then, the solution was cooled to room temperature, and kept on ice for 1 h. The yellow precipitation was collected by suction filtration, and rinsed with MilliQ several times. After reduced pressure drying, the obtained solid was washed with 100 mL acetonitrile, and the precipitation was dried under reduced pressure drying again.

Circular Dichroism (CD) Measurements
CD spectra of the annealed 1.5 µM DNA or RNA were measured in 50 mM Tris-HCl buffer (pH 7.4) with 100 mM KCl at 25 • C, with a JASCO J-820 spectrophotometer equipped with a temperature controller, in the presence of 0 µM to 4.5 µM of 1-3. The measurement was performed at a scan rate of 50 nm/min, using a Jasco J-820 spectropolarimeter (Tokyo, Japan) with the following conditions: response, 4 s; data interval, 0.2 nm; sensitivity, 100 mdeg; bandwidth, 2 nm; and scan number, 4 times.

Isothermal Titration Calorimetry (ITC) Measurements
ITC measurements were performed using a low volume nano ITC (TA instruments, USA) with a cell volume of 190 µL at 25 • C. The annealed DNA solution and chemicals were degassed for 10 min before loading. The measurement was performed with titrating 1, 2 or 3 (0-100 µM) to the G-quadruplex (telomere G1, c-myc, c-myc AP site, G1-r23nt, G1-r12nt) in 50 mM KH 2 PO 4 -K 2 HPO 4 buffer (pH 7.0) at 25 • C. In each titration, 2 µL of the ligand solution was injected into a quadruplex solution every 120 s up to a total of 25 injections, using a computer-controlled 50 µL microsyringe, with stirring at 300 rpm. The binding curve was fitting on the independent binding model.

UV-Vis Absorption Spectroscopy
The binding affinity of 1-3 to telomere G1, c-myc and ds-oligo were studied with Hitachi U-3310 spectrophotometer (Tokyo, Japan). A total of 120 µL of 150 µM telomere G1, c-myc and ds-oligo were added to 5 µM of 1, 2 or 3 in 50 mM Tris-HCl buffer (pH 7.4) and 100 mM KCl. Absorbance spectra were taken at 25 • C. The observed spectrum changes at 385 nm were rearranged with a Scatchard plot with the following equation in the case of non-cooperative binding, where ν, L, n, and K refer to the saturation fraction as the amount of bound ligand per added DNA, amount of unbound ligand, binding number of ligands per one G4, or double stranded DNA, and binding affinity, respectively. When the absorption spectrum was not saturated, nK values were obtained using the Benesi-Hildebrand method and the following Equation [36],

Modeling Simulations
Molecular Modeling of these complexes was constructed with MOE [37]. Data of NMR in the aqueous solution of K + (PDBID: 1xav (c-myc); 2kbp (G1-r12nt)) were utilized in the structural construction of the binding model simulation. The AMBER10:ETH [38,39] force field was used for this modeling simulation. The following modeling simulation processes were performed with the tether constrained to the three planes of the G-tetrad core in the structure. 1 or 2 was placed on the binding site of Telomere G1 and the energy minimization of these complexes was carried out in order to resolve the 3D obstacle. As a next step, the models interacting between β-CyDs and nucleobase (T7 and T16 in 1xav; U6 and U18 in 2kbp) were created. Molecular dynamics simulations of these complexes with distance restraints between three atoms of the β-CyDs and three atoms of the nucleobases were carried out until 1 or 2 was located in the binding site as a stable condition. Additionally, the distance restraints were then removed from the complexes, and molecular dynamics simulations for 500 ps were carried out. The complexes were observed to maintain the binding form in the simulation. Finally, the complex molecular models by energy minimization were obtained as shown in Figures 5 and S12.

Conclusions
Newly constructed NDI-β-CyDs, 1 and 2, were identified with significant specificity for recognizing parallel G-quadruplex, especially for a 23nt telomere RNA G-quadruplex. β-CyD might be an interesting medium to enhance the ability to perform this discrimination, whereas due to the bulky structure of β-CyD, the stabilization effect of NDI-β-CyDs for the G-quadruplex weakened. Some β-CyDs analogs with smaller volumes might be good candidates when conjugating with NDI to improve their stabilization performance.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/molecules27134053/s1, Figure S1 Funding: This study was supported in part by the Nakatani Foundation for the advancement of measuring technologies in biomedical engineering, Japan (S.T.) and by Grants-in-Aid from the Ministry of Education, Culture, Sports, Science, and Technology, Japan (19H02748, S.T.).