Interplay of Noncovalent Interactions in Ribbon-like Guanosine Self-Assembly: An NMR Crystallography Study

An NMR crystallography study shows how intermolecular N–H···O, N–H···N, O–H···N, O–H···O, and CH−π interactions stabilize the ribbon-like supramolecular structures of three different guanosine derivatives: guanosine dihydrate (G), 3′,5′-O-dipropanolyl deoxyguanosine (dG(C3)2), and 3′,5′-O-isopropylideneguanosine hemihydrate (Gace). Experimental solid-state 1H NMR spectra obtained at 20 T using fast magic-angle spinning (MAS), here at 75 kHz, are presented for a dihydrate of G. For each guanosine derivative, the role of specific interactions is probed by means of NMR chemical shifts calculated using the density functional theory (DFT) gauge-including projector-augmented wave (GIPAW) approach for the full crystal and extracted isolated single molecules. Specifically, the isolated molecule to full crystal transformations result in net changes in the GIPAW calculated 1H NMR chemical shifts of up to 8 ppm for O–H···O, up to 6.5 ppm for N–H···N and up to 4.6 ppm for N–H···O hydrogen bonds; notably, the presence...


■ INTRODUCTION
Molecular self-assembly can be exploited to engineer biomimetic and functional materials in aqueous and organic solutions, on surfaces and in the solid state. 1−4 Characterization of noncovalent interactions such as hydrogen bonding and CH−π interactions at the molecular level is crucial to better understand the delicate balance between the forces that hold together supramolecular structure, to establish structure− property relationships and to improve their design and function.
Modified DNA/RNA bases, specifically guanosine (G, 1) derivatives have the versatility to self-assemble into cyclic quartets, continuous helices and ribbon-like structures, as previously characterized by NMR, circular dichroism (CD), and X-ray data. 5−23 Whether the guanosine self-assembly is quartet/ helical or ribbon-like, the intermolecular N−H···N and N−H··· O hydrogen-bonding interactions play an important role, while several other molecular interactions such as electrostatic, π−π, CH−π, Coulombic interactions, van der Waals forces, and cation and anion binding also contribute to stabilizing the overall three-dimensional architectures. Such intermolecular interactions, notably, hydrogen bonding and aromatic ring current effects, have a dramatic effect on NMR chemical shifts. 24−33 By employing an NMR crystallography approach 34−36 using experimental solid-state NMR and GIPAW DFT calculations performed on the full crystal and extracted isolated molecules, we here investigate the role of noncovalent interactions driving the ribbon-like self-assembly in three guanosine derivatives: guanosine dihydrate G, 3′,5′-O-dipropanolyl deoxyguanosine dG(C3) 2 , and 3′,5′-O-isopropylideneguanosine hemihydrate Gace. These guanine derivatives are functionally diverse; for example, lipophilic derivatives of dG(C3) 2 and Gace substituted with longer alkyl chains are well-known to form lyotropic mesophases in organic solvents, 37 and once deposited on surfaces, these molecules exhibit photoconductive 38 and rectifying properties 39 and have potential applications in molecular electronics. 40,41 ■ RESULTS AND DISCUSSION Crystal Packing and Intermolecular Interactions. Together with the chemical structures and atomic numbering, Figure 1 presents an overview of molecular packing in the guanosine derivatives G, dG(C3) 2 , and Gace. Note that in each case, there are two crystallographically independent molecules in the asymmetric unit cell (Z′ = 2), denoted here A and B. For G, the G-ribbons are composed of either purely type A molecules (−A···A···A−) or purely type B molecules (−B···B··· B−), alternatingly stacked on top of each other. By comparison, ribbons of dG(C3) 2 and Gace are composed of both type A and type B molecules self-organized in the form of (−A···B··· A−) with a loose stacking arrangement. The ribbon-like structures presented in Figure 1 are dipolar with the dipoles modulating along the axis of ribbon. Particularly in dG(C3) 2 , the minimal inter-ribbon stacking is due to the substitution by short alkyl chains at the 2′ and 5′ positions causing steric hindrance between successive G-ribbons. Studies also reported that similar steric effects lead to a minimal base-stacking in other guanosine derivatives substituted with triisopropylsilyl and t-butylsilyl groups. 42,43 Guanosine and its derivatives are known to be hygroscopic. 44,45 For G, a humidity-induced crystal transition occurs between hydrous and anhydrous forms passing through an intermediate meta-state. 46,47 The thermal analysis presented in Figure 2 reveals a 12.5% weight loss in G, which corresponds to a dihydrate form G·2H 2 O. The crystal structure of the dihydrate form of G contains two types of water molecules, namely, interlayer water molecules (interconnecting sugar moieties in neighboring ribbon-like structures) and intralayer water molecules (intercalating between guanine frames of the G-ribbons). 47 As previously reported in ref 23, Figure 2 also shows that there is a 2.25% weight loss in Gace that corresponds to a hemihydrate form, Gace·0.5H 2 O.
In this paper, the intermolecular hydrogen bonding interactions observed for the G-ribbon crystal structures are classified into (I) sharing donor−acceptor atoms between homologous molecules (A···A or B···B); (II) sharing donor− acceptor atoms between analogous molecules (A···B or B···A); (III) sharing donor−acceptor atoms between nucleobase and water molecules (A···W, B···W). Type (I) interactions are observed in G leading to the ribbon-like structures stacked antiparallel to each other; i.e., the directionality of G-ribbons containing molecules of type A is along the crystallographic axis b, while ribbons made up of type B molecules have an opposite directionality, along the axis −b (Figure 1). Type (II) interactions are observed in dG(C3) 2 and Gace, where the ribbons are modulating parallel to each other. Type (III) interactions reinforce the three-dimensional stacking of ribbons whether they are stacked in a parallel (Gace) or in an antiparallel (G) manner.
Probing Ribbon-like Guanosine Self-Assembly in G by Solid-State NMR Spectroscopy and GIPAW DFT Calculations. In previous studies, solid-state MAS NMR spectroscopy has been used to identify the mode of self-assembly of  guanosine derivatives; notably, 1 H and 15 N double-quantum (DQ) spectral patterns were used to assign quartet-and ribbonlike structures based on distinct intermolecular N−H···N and N−H···O hydrogen bonding interactions, including cases where it was not possible to obtain X-ray diffraction structures. 13,20,23 Figure 3 illustrates how a suite of one-and two-dimensional solid state MAS NMR spectra can be applied to characterize ribbon-like self-assembly in G. Similar NMR spectra have been presented for dG(C3) 2 and Gace in previous studies. 20,23 The 1 H and 13 C cross-polarization (CP) MAS onedimensional NMR spectra of G shown in Figure 3 directly indicate that there are two crystallographically independent molecules in the asymmetric unit cell (Z′ = 2), via a clear observation of a pair of signals for the NH1 and C8 chemical sites. A two-dimensional heteronuclear 1 H− 13 C correlation experiment (Figure 3b) was performed to assign the C−H pairs, i.e., protons directly bonded to carbons. In the 1 H− 13 C refocused INEPT spectrum, 12 C−H correlations were clearly observed, further indicating that there are two crystallographically independent molecules.
Specific correlations observed in the region of the 1 H DQ-SQ (single-quantum) correlation spectrum corresponding to the N−H···N and N−H···O protons are evident in Figure 3d. The full 1 H DQ-SQ correlation spectrum is presented in Figure S2 in the Supporting Information. DQ peaks appear at the sum of the SQ peaks: 31,48,49 the δ DQ peaks at 19.8 ppm (δ SQ , 12.1 + δ SQ , 7.7) and at 20.9 ppm (13.1 + 7.8) are assigned to NH1A− H8A and NH1B−H8B correlations, respectively; similarly, the δ DQ peaks at 18.4 ppm (12.1 + 6.3) and at 19.9 ppm (13.1 + 6.8) are assigned to NH1A−NH 2b A and NH1B−NH 2b B correlations, respectively. In addition, N−H proximities are probed through 1 H− 14 N HMQC spectra ( 14 N, I = 1, 99.6%) recorded using n = 2 rotary resonance recoupling 50 (R 3 ) of the heteronuclear dipolar couplings. 51−56 It is apparent from Figure  3e that one-bond N−H correlations are solely observed for the shorter recoupling time, 106 μs, while longer-range correlations appear for the longer recoupling time, 533 μs (Figure 3f), notably between H8 and N7 and N9.
A GIPAW DFT calculation of NMR chemical shieldings was performed for G (see Table S1 in the Supporting Information for a listing of the GIPAW DFT calculated and experimental 1 H and 13 C chemical shifts, as well as Figure S1, whereby the experimentally observed NMR chemical shifts are plotted against GIPAW calculated NMR chemical shieldings). A stick  Table S1 in the Supporting Information) for G (filled and hollow circles correspond to the A and B molecules in the asymmetric unit cell, respectively). Left panel: (a) A stick spectrum of GIPAW calculated 13 C NMR chemical shifts is plotted on top of an experimental 13

Crystal Growth & Design
Article spectrum of GIPAW calculated 13 C chemical shifts is plotted on top of the experimental 13 C CP MAS spectrum (Figure 3a), and the GIPAW calculated chemical shifts for C−H pairs as overlaid using red crosses in the 13 C− 1 H spectrum ( Figure 3b) show a reasonably good agreement between experimental and DFT calculated chemical shifts. For the DFT geometry optimized structure of G shown in Figure 3, the H−H distances are indicated together with the intermolecular NH··· N and NH···O hydrogen bonding distances. GIPAW calculated chemical shieldings have been reported for dG(C3) 2 and Gace in previous studies. 20,23 For all molecules, the GIPAW calculated 14 N chemical shifts together with the quadrupolar NMR parameters are also listed in Table S2 of the Supporting Information.
Molecule to Full Crystal: GIPAW DFT Calculations Probe the Role of Noncovalent Interactions in Ribbonlike Guanosine Self-Assembly. The above section has shown that there is a good agreement between the GIPAW calculated and experimental NMR chemical shifts for G. This section shows how the effect of intermolecular interactions on NMR chemical shifts can be analyzed by comparing NMR chemical shieldings calculations performed for the full crystal (i.e., wherein all intermolecular interactions are present) versus those for extracted isolated molecules (i.e., in the absence of intermolecular interactions). Similar calculations have been performed on a wide range of organic molecules including compounds exhibiting weak CH···O and CH···N hydrogen bonding interactions, 24,26 amino acids, 25 a camphor derivative, 57 carbazole functionalized isocyanides, 28 and pharmaceuticals. 29,30,32,33 Relative changes in the GIPAW calculated NMR chemical shifts between the full crystal and extracted molecules Δδ iso C−M are investigated here. For 1 H nuclei in G, dG(C3) 2 , and Gace, the cases where Δδ iso C−M values are ≥2 ppm are plotted in Figure 4 (see Table S3 in the Supporting Information for a full listing of the calculated chemical shielding tensors for the full crystal and extracted molecules): Considering Figure 4, in all cases, the larger deviations in the Δδ iso C−M values for H1 and H2b are accounted by intermolecular N−H1···N7 and N−H2b···O6 hydrogen bonding interactions. This analysis reveals that the imino (H1) protons are involved in strong N−H···N hydrogen bonds for which the Δδ iso C−M values are typically larger, 4.3−6.5 ppm, followed by those for the amino (H2b) protons that form relatively weaker N−H···O hydrogen bonds resulting in changes in Δδ iso C−M values between 3 and 4.6 ppm. This is also mirrored in Table 1 wherein longer donor−acceptor distances and deviations up to 27°from linearity of hydrogen bonding angles are observed for the N− H···O interactions. Note that the Δδ iso C−M values of H2b protons are superimposed in dG(C3) 2 due to the similar strengths of N−H2b···O6 interactions for both A and B molecules, which is consistent with their identical donor−acceptor distances, 2.86 Å, as presented in Table 1.
In G, changes observed in the Δδ iso C−M values for the OH2′, OH3′, and OH5′ protons (Figure 4) can be explained by means of intermolecular interactions between guanosine molecules and with water molecules. As illustrated in Figure  5 The symbol → represents the directionality of the hydrogen bonding interaction, with the arrow pointing toward an acceptor site. An overview of molecular packing in G, dG(C3) 2 and Gace is shown in Figure 1. Hydrogen bonding donor and acceptor moieties in A and B ribbons and water (W) molecules are identified. b Inter-ribbon hydrogen bonding interactions between guanosine molecules. c Intramolecular hydrogen bonds within the same ribbon.  Table 1. Table 2 presents an analysis of the hydrogenbonding interactions adopted by the four water molecules. Table 2 also lists the GIPAW calculated 1 H chemical shifts; it is clear that the anomalous low values for one of the OH groups in both a W3 and a W4 water molecule are a consequence of weaker OH···O hydrogen bonding (distances greater than 2.8 Å and bond angles less than 160°). Note that there is insufficient resolution in this region of the 1 H DQ-SQ MAS spectrum (see Figure S2) to identify these chemical shifts experimentally. In conclusion, it is evident that the hydrogen bonding networks formed by the inter-and intralayer water molecules shown in Figure 5 are crucial for holding together the three-dimensional ribbons in G.
In dG(C3) 2 , the substitution with short alkyl chains causes a steric hindrance between successive ribbons with each ribbon being surrounded by alkyl chains as shown in Figure 6 Table 5 of ref 20). For G, dG(C3) 2 , and Gace, the 13 C chemical shift values for the two crystallographically independent molecules are presented in Table 3. It is thus evident that a 13 C C8 chemical shift above or below 135 ppm is characteristic of a syn-or an anti-conformation, respectively. In addition, torsion within the furanose ring is characterized by means of torsional angles measured across specific dihedrals for the DFT optimized structures and for X-ray diffraction structures. These torsional angles are reported in Table S4 in the Supporting Information.

■ CONCLUSIONS
This study has probed similarities and differences in the ribbonlike supramolecular structures for three guanosine derivatives in the solid state. A NMR crystallography approach has been employed to quantitatively unravel the role of specific intermolecular interaction by means of GIPAW DFT calculated NMR chemical shifts on full crystal vs isolated molecules for the two crystallographically independent molecules in each guanosine derivative. The NMR crystallography analysis reveals that intermolecular interactions are differently experienced for the two crystallographically independent molecules. In particular, GIPAW calculations performed on full crystal versus extracted isolated molecules show that intermolecular N−H···N hydrogen bonds are stronger than the N−H···O hydrogen bonds, corresponding also to shorter donor−acceptor distances and an increased linearity of the hydrogen bonding angles. In addition to the N−H···N and N−H···O hydrogen bonds which are interconnecting guanosine units, several other noncovalent interactions such as O−H···N, O−H···O, and CH−π interactions contribute to stabilizing the overall three-dimensional structures in a cooperative manner. For the ribbon-like structures studied here, a diversity in hydrogen bonding interactions can be seen.
For G and Gace, NMR crystallography is complemented by thermal analysis that shows that G is a dihydrate and Gace is a hemihydrate. The presence of lattice water reinforces stacking of two-dimensional molecular sheets in G and Gace. In G, two interlayer water molecules are interconnecting sugar rings and two intralayer water molecules are connecting guanine frames in the neighboring ribbons by means of an extended hydrogen bonding network. For the three guanosine derivatives studied here, the base-sugar conformations are both syn and anti in G, all anti in dG(C3) 2 and all syn in Gace. It is shown that the 13 C C8 chemical shift (above or below 135 ppm) is diagnostic of a syn or an anti-conformation.
To conclude, this study has demonstrated the power of a NMR crystallography approach for a detailed study of the interplay of noncovalent interactions governing supramolecular self-assembly; further applications within materials chemistry are to be envisaged. 58 ■ EXPERIMENTAL AND COMPUTATIONAL DETAILS G was purchased from Sigma-Aldrich (Gillingham, UK), recrystallized from water, and vacuum-dried prior to thermal analysis. Recrystallized G was subjected to solid-state NMR characterization.   Table S4 in the Supporting Information, and the experimental C8 NMR chemical shifts are listed in Table 3.  22 , the homonuclear decoupling cycles were applied with a duration of 32 μs (320 steps of 100 ns each). Prepulses of duration 0.8 μs were used before and after the eDUMBO-1 22 pulses. The SPINAL64 1 H heteronuclear decoupling scheme was used during acquisition with a pulse duration, 5 μs. The indirect dimension was acquired using 48 t 1 FIDs, each with 512 coadded transients, by using the States-TPPI method to achieve sign discrimination with an increment of 80 μs. The total experimental time was 17 h, with a recycle delay of 3 s. The 1 H chemical shifts are scaled in the F 1 dimension by a factor of 0.63.
1 H Double-Quantum (DQ) Spectroscopy. A DQ-SQ correlation spectrum of G was acquired using the 850 MHz spectrometer with a JEOL 1 mm HX probe. The MAS frequency was 75 kHz. One rotor period of the BABA 62,63 (back to back) recoupling sequence was used for the excitation and reconversion of DQ coherences. A 16-step phase cycle was used in order to select Δp = ±2 on the DQ excitation pulses (4 steps) and Δp = −1 (4 steps) on the z-filter 90°pulse, where p is the coherence order. 256 t 1 FIDs, each with 32 coadded transients, were acquired using the States method to achieve sign discrimination in the F 1 dimension with a rotor-synchronized t 1 increment of 13.3 μs, corresponding to an overall experimental time of 5 h using a 2 s recycle delay. 14 53 was employed to record 14 N− 1 H HMQC spectra by applying a second 1 H 90°pulse (90°out of phase with respect to the first 90°pulse) immediately after the first 1 H 90°pulse and using phase inversion (every rotor period) of the n = 2 (ν 1 = 2 ν R ) rotary-resonance recoupling pulses. 55 A four-step nested phase cycle was used to select changes in coherence order Δp = ± 1 (on the first 1 H pulse, 2 steps) and Δp = −1 (on the last 14 N pulse, 2 steps). The 1 H and 14 N 90°pulse durations were 2 and 5 μs, respectively. For each of 48 t 1 FIDs (using the States method to achieve sign discrimination in F 1 with a rotor synchronized increment of 13.3 μs), 128 transients were coadded with a recycle delay of 2 s corresponding to a total experimental time of 4 h.
All 1 H and 13 C experimental shifts are calibrated with respect to neat TMS using adamantane as an external reference (higher ppm 13 C resonance, 35.8 ppm 64 and the 1 H resonance, 1.85 ppm 65 ). 14 N chemical shifts were referenced to a saturated NH 4 Cl aqueous solution at −352.9 ppm, corresponding to the primary reference, liquid CH 3 NO 2 (0 ppm). To compare to the alternative reference of liquid NH 3 at −50°C as used in protein NMR, it is necessary to add 379.5 ppm. 66 GIPAW DFT Calculations. Calculations were performed at the Warwick Centre for Scientific Computing using plane-wave based DFT approaches 67,68 implemented in the CASTEP code, UK academic release version 6.1. 69 Initial atomic coordinates were taken from published crystal structures: G, Guanosine dihydrate, 5 CSD code GUANSH10, Z = 4, Z′ = 2, space group P2 1 , 156 atoms/cell (including 8 H 2 O); dG(C3) 2 , 3′,5′-O-dipropanolyl deoxyguanosine, 9 CSD code MOFBUE, Z = 2, Z′ = 2, space group P1, 96 atoms/cell; Gace, 2′,3′-O-isopropylidineguanosine, 6 CSD code VUYMIL, Z = 4, Z′ = 2, space group P2 1 , 166 atoms/cell (including two H 2 O). In all cases, the unit cell parameters were fixed, the space group symmetry was imposed, and periodic boundary conditions were applied during the geometry optimization. NMR shielding calculations of G, dG(C3) 2 , and Gace were performed using the gauge-including projector-augmented wave (GIPAW) approach. 67,68 Both geometry optimizations and NMR chemical shift calculations used a plane-wave basis set and the PBE exchange correlation functional 70 at a basis cutoff energy of 600 eV with integrals taken over the Brillouin zone by using a Monkhorst−Pack grid of minimum sample spacing 0.08 × 2π Å −1 . A semi empirical dispersion correction was applied using the TS scheme 71 in both geometry optimization and NMR shielding calculations with on-the-fly (OTF) ultrasoft pseudopotentials. 72 In all cases, forces, stresses on the unit cells, energies, and displacements were converged to better than 0.01 eV Å −1 , 0.1 G Pa, 0.0000004 eV, and 0.001 Å, respectively. From the DFT geometry optimized crystal structures, single molecules (in each case A and B individually) are extracted and placed in a periodically repeating unit cell with dimensions: G, 17.52 × 23.00 × 13.32 Å 3 ; dG(C3) 2 , 23.19 × 13.51 × 19.59 Å 3 and Gace, 17.81 × 20.65 × 21.95 Å 3 , in order to separate the molecules in such a way that no two molecules are within a 8 Å distance. GIPAW calculations were performed on extracted single molecules using the above stated parameters. GIPAW calculated NMR shieldings of full cell and extracted molecules are viewed, processed, and tabulated through the CCP-NC output files visualization tool, MagresView. 73 ■ ASSOCIATED CONTENT

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.cgd.5b01440.