Halogen and Hydrogen Bonded Complexes of 5-Iodouracil

.

INTRODUCTION 5-Iodouracil (5IU) is a simple and commercially available pyrimidine nucleobase derivative in which the alkenyl hydrogen closest to the carbonyl group of the uracil moiety is substituted by iodine. This system, along with other 5-halouracils and their derivatives, has been of particular interest due to its biological, viz. antitumor and antiviral, activity. 1 The first crystal structure with a 5IU unit (5-iodo-2'-deoxyuridine) was reported already in 1965, 2 and the online portal to the Cambridge Structural Database (WebCSD) 3 contains two separate reports of the same monoclinic structure. 4 The iodine atom in 5IU is bonded to a sp 2 hybridized carbon and becomes polarized due to the adjacent electron withdrawing carbonyl group. Accordingly, 5IU should manifest itself as a halogen bond (XB) donor, but, to our knowledge, the experimental evidence is limited to some biological systems. 5,6 Of late, halogen bonding has attained increasing interest by many research groups, as discussed in several reviews dealing with XBs in the solid state, 7 in solution, 8 and in biological systems. 9 Recently, the first cases of improved pharmaceutical properties effected/induced by halogen bonding were demonstrated. 10 Despite the potential of 5IU framework for crystal engineering, the number of structural entries in the CSD that contain this unit is surprisingly small: only 26 entries can be found and most of them are derivatives of uridine which is a nucleoside with a ribose ring attached to the uracil ring via a β-N1-glycosidic bond. However, even when examining this limited set of data, it became clearly evident from the range of donor-acceptor bond lengths observed that 5IU could act as a XB donor in different systems. Inspired by this, and due to our own interest on alternative XB donors (i.e. donors in which the halogen atom is not polarized by fluorine), 11 we decided to conduct a thorough investigation of halogen bonding with 5IU. Consequently, crystallization experiments involving 5IU and its three N-substituted derivatives were performed in suitable solvent systems containing electron donors such as N, N-dimethylformamide (DMF), N,Ndiethylformamide (DEF), N-methylformamide (MeF), formamide (FA), dimethylsulfoxide (DMSO), and water. We also performed quantum chemical calculations for simplified model systems as well as co-crystallization attempts with 5IU and its derivatives using selected organic molecules and salts having capabilities as XB acceptors. While the focus of the current work was on the XBs exhibited by the 5IU unit, the examined systems also display a number of hydrogen bonds (HBs) that are included in the discussion.

Scheme 1.
Schematic structures of compounds 1-4 including the atom numbering schemes.
After another 2 h of continuous heating, a second portion of benzyl bromide (0.72 g, 0.5 ml, 4.20 mmol) was added. The heating was continued for 1h, after which the reaction mixture was let to stir overnight at room temperature. This yielded a white precipitate (K2CO3) and a transparent liquid layer. The precipitate was filtered by suction and the solvent removed from the filtrate under reduced pressure. The obtained oily yellow layer was extracted with ethyl acetate (40 ml) and the extract washed twice with water (15 ml). The yellow organic layer was dried over Compound 3 (1-hexyl-5-iodopyrimidine-2,4(1H,3H)-dione): 5IU (0.92 g, 3.86 mmol) and K2CO3 (0.54 g, 3.86 mmol) was dissolved in dry DMF (10 ml) in a round-bottomed flask. The reaction mixture was stirred and heated under N2 atmosphere at 80ºC. After 30 min, 1bromohexane (1.27 g, 1.08 ml, 7.72 mmol) was added into the reaction mixture and the heating was continued overnight. This gave a brownish-yellow oil that was concentrated under reduced pressure. The residue was extracted with ethyl acetate (40 ml) and the extract washed twice with water (15 ml). The organic layer was dried over sodium sulfate and concentrated by rotary evaporator under vacuum. The oily precipitate was crystallized from CHCl3 and the obtained white crystals were collected by filtration, washed with CHCl3, and dried under vacuum. Yield: 0.092 g (7.4 %). 1  Compound 4 (2,2'-(5-iodo-2,4-dioxopyrimidine-1,3(2H,4H)-diyl)

Single crystal X-ray diffraction studies
Crystallographic data were collected at 123K with a Bruker-Nonius Kappa CCD diffractometer with APEX II detector using graphite-monochromatized Mo-Kα radiation (λ = 0.71073 Å) except for 4a, for which with an Agilent SuperNova dual wavelength diffractometer equipped with Atlas CCD area detector with Cu-Kα radiation (λ = 1.54184 Å) were used.

Spectroscopic measurements
All NMR spectra were recorded with a Bruker Avance DRX 500 FT-NMR spectrometer at 303K. The chemical shifts are reported in ppm and referenced internally using the residual polar solvent resonances relative to tetramethylsilane (DMSO-D6 δ = 2.50, D2O δ = 4.80), except for the 15 N data for which an external standard (CH3NO2) was used. Mass spectrometric measurements were performed with a Micromass LCT time of flight (TOF) mass spectrometer with electrospray ionization (ESI).

Computational details
The geometries of the hydrogen and halogen bonded dimers 5-12 (see below) were optimized using the second-order local Møller-Plesset perturbation theory (LMP2) 14 in conjunction with spin-component scaling (SCS) 15 and density fitting (DF), 16 yielding the DF-SCS-LMP2 method.
This method was also used to calculate the interaction energies of the hydrogen and halogen bonded model dimers 5-12 and the potential energy surface scans of 6 and 11. Starting geometries for the model dimers were obtained from the acquired X-ray crystallographic data.
In all DF-SCS-LMP2 calculations, localized molecular orbitals were constructed using the Pipek-Mezey localization approach, 17 whereas their corresponding domains were defined employing the Boughton and Pulay procedure. 17b Domains were determined at large intermolecular distance and individual monomers were identified automatically in order to ensure interaction energies free of basis set superposition error (BSSE) and to obtain smooth potential energy surfaces. In the SCS correction, scaling factors 6/5 and 1/3 were used for the antiparallel and parallel spin components, respectively. namely aug-cc-pVTZ, were used for all other nuclei except for iodine for which the small core ECP basis set aug-cc-pVTZ-PP was used. 18 Auxiliary basis sets of triple-ζ valence quality were employed in the DF approximation to speed up all calculations. 19 All quantum chemical calculations were done with the Molpro 2010.1 20 program package; for visualization of optimized geometries, the program Mercury 3.1 was employed. 13g

CSD study
As already mentioned in the introduction, only 26 structures containing a 5IU unit were found in the CSD database and only 14 of them manifest XB interactions in which iodine functions as a XB donor. The analysis was restricted to I···A (A = acceptor) contacts that are shorter than 0.9 times the sum of van der Waals (vdW) radii and have close to linear C−I···A arrangement (> 155°). The XB interactions found are listed in the Supporting Information along with the relevant contact parameters. As a whole, the CSD study clearly shows that the 5-iodouracil moiety can act as an XB donor, at least for carbonyl acceptors, although the presence of two carbonyls in the uracil ring often leads to other interactions such as hydrogen bonding.

Structures of 5IU complexes
During the design of the synthetic procedures, the solubility of 1 in different solvents was tested and DMF was found to be a good solvent for this particular system. Consequently, when a solution of 1 in DMF was left standing on an evaporation dish, well-formed colorless plate-like single crystals were obtained. The crystallographic analysis showed the crystals to be that of a DMF solvate of 1 (1a) in orthorhombic space group Pbca. The structural data shows that, in this instance, a short and linear CI···O XB interaction binds the DMF molecules to 1 via the carbonyl oxygen (Figure 1). The I···O contact distance in 1a is 18 % smaller than the sum of vdW radii for I and O atoms, which is of the same magnitude as the shortest XBs involving 1 reported in the CSD. 21 The structure of 1a inspired us to test other formamide solvents under the same conditions. To our surprise, a new polymorph of 1 was obtained when using DEF as the solvent. However, since this structure does not show any XB interactions, it is not discussed herein (see Supporting Information).
The crystallization of 1 from MeF yielded colorless plates which were shown to be a MeF solvate of 1 (1b) by X-ray diffraction. The structure of 1b has a triclinic P-1 space group and it shows similar but weaker CI···O XB interactions than 1a. The I···O distance is rather long and only a 10 % reduction of the sum of vdW radii is observed. The apparently weaker XB in 1b can be explained with the influence of the "interfering" strong N−H···O HB which affects the linearity of the C−I···O unit (Figure 2). The hydrogen bonding pattern in 1b differs from that in 1a, which is expected due to the presence of a strong HB donor in MeF.  Almost all of our attempts to co-crystallize 5IU with different molecules failed (see Supporting Information) and crystalline material (colorless prisms, space group P-1) was only obtained from a mixture of 1 and DABCO in DMF. Subsequent crystallographic analysis gave a somewhat unexpected result: instead of a simple adduct 1·DABCO, the structure contains both neutral and deprotonated (N1) molecules of 1 (1d). A part of the sheet-like motif with HBs and XBs is shown in Figure 4. The CI···O XB interaction involves the anionic 5IU unit (donor) and the neutral molecule 1 (acceptor). A second XB is observed between the CI of 1 and the nitrogen O2. A comparison of the observed XBs with the sum of vdW radii for the corresponding atoms shows ca. 15% reduction, indicative of moderate interaction strength.

Computational studies of 5IU complexes
Even though the crystallographic evidence conclusively highlights the ability of 1 to act as a XB donor in 1a-d, high-level quantum chemical calculations for five simplified model dimers (5-9, Figure 5) were carried out in order to get insight into the strength of the observed interactions.
For comparison purposes, hydrogen bonding interactions in four different model dimers involving 1 were also examined (10-12, Figure 5).
Because both XBs and HBs involve a varying degree of dispersion, electrostatics and charge transfer, not all quantum chemical methods can be used to describe them. 24 One computational method that has been widely used for this purpose is the second-order Møller-Plesset perturbation theory (MP2). 25 The MP2 method offers reasonable accuracy with relatively low computational cost, but unfortunately its standard implementation tends to overestimate the strength of both XB and HB. 26,27 However, it was recently shown that the performance of MP2 with respect to XBs can be significantly improved by employing spin component scaling (SCS) 15 in conjunction with large basis sets, and by removing the basis set superposition error (BSSE) using counterpoise correction. 26a,28 The BSSE can also be removed with the help of localized orbital basis (LMP2), 14 and the required CPU time can be significantly reduced with density fitting (DF). 16 Consequently, the combined SCS-DF-LMP2 method was used in the current work in together with the aug-cc-pVTZ-(PP) basis sets. 18 The geometries of the dimers 5-12 were optimized in the gas phase at the SCS-DF-LMP2/augcc-pVTZ-(PP) level of theory; the XBs and N···O distances in the optimized structures are listed in Table 2. As seen from the optimized structures, the calculated data are in good agreement with the experimental values, with the exception of 9 and 11 for which the deviation between the two sets of numbers is more than 0.15 Å. The differences between the experimental (solid state) and computational (gas phase) results originate most likely from the flatness of the potential energy surface with respect to both halogen and hydrogen bonds. 11g,29 For instance, a ±0.20 Å displacement of the bonds from their calculated equilibrium distance weakens them by 1-2 kJ mol 1 (see Supporting Information). Such small energy differences are easily offset by crystal packing forces which the gas phase calculations naturally do not take into account. However, despite the small deviations between the experimental and optimized geometries, the calculated data can be safely used to evaluate the interaction energies of the HBs and XBs in 1a-d. The calculated interaction energies (at 0K temperature, Table 2) show that the strength of the XBs in the model dimers 5-9 varies between 14.0 and 20.0 kJ mol 1 , which supports the statement that XBs involving 1 range from weak to moderate. For comparison, the ability of 1 to act as a XB donor has been briefly mentioned in another computational study in which the authors concluded that the interaction energies involving 1 settle between 5.0 and 16.0 kJ mol 1 . 6 The calculated interaction energies in Table 2 also show that XBs do not solely determine the solid state packing of 1a-d since all HBs are calculated to be of equal strength to the XBs (or stronger). Thus, it can be concluded that the calculated data shows the potential of 5IU unit as a XB donor of medium strength, and that the crystal packing observed for 1a-d results from delicate interplay between XBs and HBs.

Structures of compounds 2-4
Compounds 2 and 3 crystallized out only from DMF and DMSO solutions, respectively; X-ray diffraction quality crystals were also obtained from co-crystallization of 3 with DABCO in ethyl acetate. However, since none of these structures shows any XB interactions, they are not discussed further (see Supporting Information).
The crystallization of 4 from FA yielded colorless plates which were shown to be a dihydrate of 4 (4a) with a triclinic P-1 space group, though dry solvents were used throughout. The 5IU unit binds to a bromine anion via long CI···Br  XB interaction ( Figure 6). The contact is approximately linear, but the reduction of the sum of vdW radii is only 11 %. The observed I···Br  interaction is perturbed by two O−H···Br  hydrogen bonds involving the co-crystallized water molecules. Apparently these HBs withdraw the anion away from the iodine and weaken the XB contact. It seems that 4, due to its several potential HB acceptors, absorbs water easily, which naturally influences the crystallization process and plays a dominant role in the packing of 4a. Interestingly, neither of the two carbonyl units on the uracil moiety functions as a HB acceptor, while four molecules of water and two Br  anions form a ring-like assembly connected via HBs, and the cations and anions bind to his assembly. Unfortunately, 4a does not show XB interactions involving oxygen or nitrogen acceptors, possibly due to the weak XB donor nature of 5IU. Figure 6. Non-covalent assembly in the crystal structure of 4a, formed by two dications, four anions, and four water molecules, highlighting the relevant XB (blue) and HB (black) interactions.
In the present study, we have investigated the halogen and hydrogen bonding properties of 5iodouracil (5IU, 1) and its derivatives (2)(3)(4) in polar solvents including formamides, DMSO, and water. In addition, co-crystallization experiments with 5IU and potential halogen bond acceptors were carried out. Nine new crystal structures were obtained of which five showed halogen bonding based on subsequent structural and computational analyses. Three of these structures contained solvated compounds (1a-c) while the remaining two were co-crystals (1d) or not solvated (4).
The obtained crystal structures, along with an analysis of the existing crystallographic data, demonstrate that the 5IU unit can act as a halogen bond donor, albeit a rather weak one. Halogen bonding interactions of the type CI···O were seen in four of the structures, while CI···N or CI···Br  interactions were observed in only instance each. The strongest interaction was seen in the structure of 1a which displayed the largest (18 %) reduction of the sum of van der Waals radii for the contact atoms. Typically, the observed reduction varied from 10 to 17 %, indicating weak to moderate interaction strength. Quantum chemical calculations on simplified model systems confirmed the attractive nature of the observed halogen bonds and showed that their interaction energy varies between 15 and 20 kJ mol 1 . The CI···N type interaction observed for 1d is significantly shorter than the one currently reported in the literature, while the I···Br  interaction seen in 4a has not previously been demonstrated for 5IU or its derivatives.