We used the high resolution and accuracy of the Cambridge Structural Database (CSD) to provide detailed information regarding base pairing interactions of selected nucleobases. We searched for base pairs in which nucleobases interact with each other through two or more hydrogen bonds and form more or less planar structures. The investigated compounds were either free forms or derivatives of adenine, guanine, hypoxanthine, thymine, uracil and cytosine. We divided our findings into categories including types of pairs, protonation patterns and whether they are formed by free bases or substituted ones. We found base pair types that are exclusive to small molecule crystal structures, some that can be found only in RNA containing crystal structures and many that are native to both environments. With a few exceptions, nucleobase protonation generally followed a standard pattern governed by pK_a values. The lengths of hydrogen bonds did not depend on whether the nucleobases forming a base pair were charged or not. The reasons why particular nucleobases formed base pairs in a certain way varied significantly.

Understanding intermolecular interactions in crystals of molecular ions continues to be difficult. On the one hand, the analysis of interactions from the point of view of formal charges of molecules, similarly as it is commonly done for inorganic ionic crystals, should be performed. On the other hand, when various functional groups are present in the crystal, it becomes natural to look at the interactions from the point of view of hydrogen bonding, π…π stacking and many other kinds of non-covalent atom–atom bonding. Often, these two approaches seem to lead to conflicting conclusions. On the basis of experimental charge densities of cytosinium chloride, adeninium chloride hemihydrate, and guanine dichloride crystals, with the help of theoretical simulations, we have deeply analysed intermolecular interactions among protonated nucleobases, chloride anions and water molecules. Here, in the second paper of the series of the two (Kumar et al., 2018, IUCrJ 5, 449–469), we focus on applying the above two approaches to the large set of dimers identified in analysed crystals. To understand electrostatic interactions, we analysed electrostatic interaction energies (E_es) computed directly from molecular charge densities and contrasted them with energies computed only from net molecular charges, or from a sum of electric multipolar moments, to find the charge penetration contribution to E_es. To characterize non-covalent interactions we performed topological analyses of crystal electron densities and estimated their interaction energies (E_EML) from properties of intermolecular bond critical points. We show that the overall crystal architecture of the studied compounds is governed by the tight packing principle and strong electrostatic attractions and repulsions between ions. Many ions are oriented to each other in a way to strengthen attractive electrostatic interactions or weaken strong repulsion, but not all of them. Numerous bond critical points and bond paths were found between ions, including nucleobase cations despite their overall repulsive interactions. It is clear there is no correlation between E_EML and E_es. However, strong relation between E_EML and the charge penetration component of E_es is observed. The relation holds regardless of interaction types or whether or not interacting molecules bear the same or opposite charges. Thus, a charge density-based approach for computing intermolecular interaction energies and the atom–atom approach to analyse non-covalent interactions do complement each other, even in ionic systems.

This paper presents experimental charge-density studies of cytosinium chloride, adeninium chloride hemihydrate and guaninium dichloride crystals based on ultra-high-resolution X-ray diffraction data and extensive theoretical calculations. The results confirm that the cohesive energies of the studied systems are dominated by contributions from intermolecular electrostatic interactions, as expected for ionic crystals. Electrostatic interaction energies (E_es) usually constitute 95% of the total interaction energy. TheE_esenergies in this study were several times larger in absolute value when compared, for example, with dimers of neutral nucleobases. However, they were not as large as some theoretical calculations have predicted. This was because the molecules appeared not to be fully ionized in the studied crystals. Apart from charge transfer from chlorine to the protonated nucleobases, small but visible charge redistribution within the nucleobase cations was observed. Some dimers of singly protonated bases in the studied crystals, namely a cytosinium–cytosiniumtranssugar/sugar edge pair and an adeninium–adeniniumtransHoogsteen/Hoogsteen edge pair, exhibited attractive interactions (negative values ofE_es) or unusually low repulsion despite identical molecular charges. The pairs are metastable as a result of strong hydrogen bonding between bases which overcompensates the overall cation–cation repulsion, the latter being weakened due to charge transfer and molecular charge-density polarization.

Conventional refinement strategies used for three-dimensional electron diffraction (3D ED) data disregard the bonding effects between the atoms in a molecule by assuming a pure spherical model called the Independent Atom model (IAM) and may lead to an inaccurate or biased structure. Here we show that it is possible to perform a refinement going beyond the IAM with electron diffraction data. We perform kappa refinement which models charge transfers between atoms while assuming a spherical model. We demonstrate the procedure by analysing five inorganic samples; quartz, natrolite, borane, lutecium aluminium garnet, and caesium lead bromide. Implementation of kappa refinement improved the structure model obtained over conventional IAM refinements and provided information on the ionisation of atoms. The results were validated against periodic DFT calculations. The work presents an extension of the conventional refinement of 3D ED data for a more accurate structure model which enables charge density information to be extracted.