High-resolution structural study on pyridin-3-yl ebselen and its N-methylated tosylate and iodide derivatives

The crystal structure of the pyridine analog of the selenium pharmaceutical ebselen is characterized by one-dimensional N—Se chalcogen-bonded chains where the pyridine N atom is the chalcogen-bond acceptor. Charge density analysis using high-resolution Mo Kα X-radiation and subsequent multipole refinement reveals a clear region of positive electrostatic potential at the antipode to the Se—N bond of the isoselenazole moiety. In the N-methylated iodide and tosylate salts, the iodide/tosylate counter-ions are strongly chalcogen bonded to the Se atom.


Introduction
The benzisoselenazolinone scaffold of the drug ebselen (1) is a potent chalcogen-bond donor, due to the presence of a polarizable Se atom covalently bonded to an electron-withdrawing amide/aniline N atom. The propensity for this system to form chalcogen bonds has been studied within our group and by others, with a view to exploiting it in the context of medicinal chemistry (Thomas et al., 2015;Fellowes & White, 2019;Fellowes et al., 2020. However, the concept of chalcogen bonding has also received much attention concerning applications in materials chemistry (Eckstein et al., 2021). In addition to the chalcogen-bond donor, ebselen also contains an amide carbonyl group which nicely fulfils the role of chalcogen-bond acceptor in directing the crystal packing of in difference Fourier maps but were introduced in calculated positions and treated as riding on their parent atoms (C atoms). The H atoms of the water molecules were located in difference Fourier maps and refined isotropically.
Refinements for charge-density analysis of 2 were performed against F 2 , up to a maximum reciprocal resolution of 0.95 Å À 1 for a total of 7174 independent reflections using the MoPro software (Guillot et al., 2001). Beamstop-affected reflections were identified and excluded at the data reduction stage, and disagreable frames were removed. The independent atom model (IAM) structure was first refined using NoSpherA2 in OLEX2 (Kleemiss et al., 2021). This procedure generates aspherical scattering factors for the atoms in the crystal based on a density functional theory (DFT) calculation. The PBE0/def2TZVP level was used for this calculation, and R1/wR2 values of 0.0281/0.0471 were obtained after convergence of the wavefunction calculation and crystallographic refinement. This model was used as a starting point for the multipole refinement in the MoPro Suite (Guillot et al., 2001;Jelsch et al., 2005). As the atomic anisotropic displacement parameters (ADPs) had been adequately determined by refinement using calculated scattering factors, charge density parameters were refined from the beginning, without an initial high-order refinement as is usual in charge density investigations. Statistical weights were used throughout the multipole refinement, and 2% of all reflections were marked as free. Multipole parameters were initialized from the ELMAM2 database for all atoms except for selenium, for which parameters were not available. The multipole expansion was limited to a 32-pole level for the Se atom and to an octupole level for the other heavy atoms. H atoms were modelled at the quadrupole level. Default Slater-type functions were used for all atoms. Charge density symmetry constraints were applied, and kappas were constrained to be equal for chemically equivalent atoms. C-H bonds were constrained to neutron distances and idealized geometries, but U iso values were refined freely. Initially, an overall scale factor was refined, and this was included in all subsequent refinements. Valence and multipole population parameters were then refined, followed by their respective kappas, and this cycle was repeated. When this had converged (shift/<0.001), xyz and U ij were refined. This procedure was repeated to convergence. All heavy atoms were then refined anharmonically (maximum order 3) until convergence, and the Gram-Charlier coefficients of each atom were compared with their estimated uncertainty. If no coefficient exceeded 3�, the atom was removed from the anharmonic refinement. Atoms Se1, O1, N2, C2, C3, C5, C7, C8, C9, C10, C11 and C12 displayed appreciable anharmonic motion, and were thus refined as such. An isotropic extinction parameter was introduced, which substantially reduced residual electron density around the Se atom. Kappa constraints were lifted gradually, followed by multipole symmetry constraints, then all parameters were refined together initially with heavy damping, which was reduced to zero in the final cycles. The final R1/wR2 values were 0.016/0.023 and the goodness-of-fit (GoF) was 1.07. R free remained comparable to R1 throughout the refinement, so we do not believe the model suffers from overfitting. The total number of refined parameters in the final cycle was 496, to give a data/parameter ratio of 14.4. The estimated average error in the electron density was 0.0970 e Å À 3 , with a maximum and minimum residual density of 0.43/ À 0.41 e Å À 3 , which was randomly distributed through the asymmetric unit (Fig. 2).

Results and discussion
Pyridine-substituted benzisoselenazolinone derivative 2 was synthesized by reaction of diselenide 3 (Scheme 1) with thionyl chloride giving the electrophilic selenium reagent 4 which was immediately coupled with pyridin-3-amine (5). Crystallization of the crude product from hot DMF afforded light-brown plate-like single crystals of 2 suitable for X-ray analysis. Heating compound 2 with methyl iodide or methyl tosylate in DMF gave 2-Me + iodide and 2-Me + tosylate, respectively. Crystals for all samples were obtained from DMF.

Structure analysis for 2
The displacement ellipsoid plot for 2 is presented in Fig. 3, while selected geometrical parameters are given in Table 2. The molecular structure is essentially planar, with an r.m.s. deviation of 0.0357 Å for the non-H atoms. The conformation about the N1-C8 bond sees the pyridine N atom (N2) on the opposite side to the Se atom (Se1); this conformation may be  The molecular structure of compound 2, showing 50% probability displacement ellipsoids.

Figure 4
The chalcogen-bonded chains of compound 2 propagating along the ac diagonal.

Figure 5
The N� � �Se and O� � �Se chalcogen-bonding interactions and �-� stacking interactions in the structure of compound 2.
preferred due to a favourable electrostatic contact between the polarized H atom attached to C9 and amide atom O1, or alternatively this conformation is a consequence of the crystal packing (as discussed below), or both. The structure is characterized by the presence of a strong intermolecular chalcogen-bonding interaction involving the polarized Se1-N1 bond [N2 i � � �Se1 = 2.3831 (6) Å and N2 i � � �Se1-N1 = 177.44 (2) � ; symmetry code: (i) x À 1 2 , À y + 1 2 , z + 1 2 ] which propogates along the ac diagonal (Fig. 4). This strong chalcogen bond combines with a weaker intermolecular chalcogen-bonding interaction involving the less polarized Se1-C1 bond with the amide carbonyl O atom [O1 ii � � �Se1 = 3.3347 (7) Å and O1 ii � � �Se1-C1 = 165.77 (2) � ; symmetry code: (ii) x + 1 2 , À y + 1 2 , z + 1 2 ], and a �-stacking interaction between the pyridine ring [related by the symmetry code (x À 1 2 , À y + 1 2 , z + 1 2 )] and the benzisoselenazolinone ring system [related by the symmetry code (x + 1 2 , À y + 1 2 , z + 1 2 )], having a centroid-centroid distance of 3.412 Å (Fig. 5). These three intermolecular interactions, while not mutually orthogonal, do result in a three-dimensional supramolecular network (Fig. 6). It is worth making a comparison of the structure of 2, which contains N� � �Se1(-N1) chains in the crystal, with the parent ebselen (1) (Thomas et al., 2015), which is characterized by C O� � �Se1-N1 chalcogen-bonded chains. The C O� � �Se1 interaction in 1 is characterized by an Se� � �O distance of 2.533 (1) Å (polymorph 2), which represents a contraction of 0.87 Å compared to the van der Waals radii for Se and O. In comparison, the N2� � �Se1(-N1) distance of 2.3831 (6) Å in 2 is contracted by 1.06 Å from the sum of the van der Waals radii for Se and N of 3.45 Å (Bondi, 1964), suggesting that the N� � �Se(-N) chalcogen-bonding interaction in 2 is significantly stronger than the O� � �Se(-N) interaction in 1. This result is consistent with the pyridine N atom in 2 being a significantly stronger chalcogen-bond acceptor than the amide O atom in 1 and agrees with previous results from cocrystal derivatives of 1 (Fellowes et al., 2019). In addition to dispersion forces, the chalcogen bond has both an electrostatic component (attraction between the positively charged �-hole on the selenium and the electron-rich chalcogen-bond acceptor) and an orbital interaction component [in which the electron-rich chalcogen-bond acceptor (highest occupied molecular orbital, HOMO) donates electron density into the low-lying Se-N �* antibonding orbital (lowest unoccupied molecular orbital, LUMO) on the Se atom] (Pascoe et al., 2017;Kolá ř & Hobza, 2016); this latter interaction results in weakening and lengthening of the internal Se1-N bond distance. Consistent with the apparently stronger N� � �Se interaction in 2 versus the O� � �Se interaction in 1 is the significant lengthening of the Se1-N1 distance [1.9788 (5) Å ] for 2 compared to that [1.905 (1) Å ] for 1, suggesting a significantly increased population of the Se-N �* antibonding orbital in 2.

Structure analysis for 2-Me + iodide
The displacement ellipsoid plot for 2-Me + iodide is presented in Fig. 7. The structure is essentially planar, with an r.m.s. deviation of 0.038 Å for the non-H atoms of the cation. The iodide counter-ion, which is strongly associated with the cation, lies close to this plane [deviation 0.131 (1)   The molecular structure of 2-Me + iodide, showing 50% probability displacement ellipsoids.
2.9882 (1) Å and I1� � �Se1-N1 = 178.85 (2) � ], which is perfectly aligned with the antipode of the polarized Se1-N1 bond, is well within the sum of the van der Waals radii for I and Se (3.88 Å ) and is approaching the bond distance for a formal Se-I covalent bond; the Se-I distance in mesityl selenium iodide is 2.536 (1) Å (Jeske et al., 2002) and in 2,4,6tri-tert-butylphenylselenium iodide is 2.529 Å (du Mont et al., 1987). The strength of this chalcogen bond is not only apparent from the short I À � � �Se contact, but also from the significant lengthening of the internal Se-N1 bond distance, which is 2.0053 (6) Å compared to 1.905 Å in the parent molecule 1. Perhaps the I À � � �Se1-N1 moiety is best described as a 3-centre-4-electron bond. The crystal packing of 2-Me + iodide is dominated by strong �-� stacking interactions along the a axis between molecules of the complex, with each planar molecule sandwiched between two parallel molecules related by the symmetry codes (À x + 1, À y + 1, À z + 1), with an interplanar spacing of 3.4146 (8) Å , and (À x + 2, À y + 1, À z + 1), with an interplanar spacing of 3.295 (1) Å (Fig. 8). Selected geometrical parameters are given in Table 3.

Structure analysis for 2-Me + tosylate
The 2-Me + tosylate derivative, represented by the displacement ellipsoid plot in Fig. 9, crystallizes as a trihydrate, which presumably forms as it satisfies the coordination requirements of the tosylate anion, with its three O atoms participating in a number of interactions, including a chal- The �-� stacking interactions in the structure of 2-Me + iodide.

Figure 10
The hydrogen-bonding interactions in the structure of 2-Me + tosylate trihydrate. The undulating chain extends along the a axis.
cogen-bonding interaction with the Se atom, in addition to a number of hydrogen-bonding interactions involving the three water molecules. The water molecules form an undulating hydrogen-bonded tape parallel to the a axis, consisting of alternating six-membered rings fused to four-membered rings, referred to as the T4(2)6(2) motif (Golz & Strohmann, 2015;Custelcean et al., 2000). Each six-membered ring provides four hydrogen bonds to two tosylate anions related by inversion (Fig. 10). The remaining tosylate O atom (O2) forms a chalcogen bond to the Se atom of the cation [O2� � �Se1 = 2.553 (2) Å and O2� � �Se-N1 = 170.57 (10) � ]; the planar cations are approproximately orthogonal to the propagating direction of the water tape and allows for interdigitation from a neighbouring tape by �-� stacking of the benzisoselenazolinone moieties. Each cation is sandwiched between two parallel cations, with interplanar spacings of 3.383 (7) [at (À x + 2, À y + 1, À z + 1)] and 3.405 (7) Å [at (À x + 1, À y + 1, À z + 1)] (Fig. 11), resulting in a two-dimensional network parallel to the (011) plane. Despite the chalcogen-bond interaction in 2-Me + tosylate involving a negatively charged tosylate O atom (quenched to a certain extent by the numerous hydrogen-bonding interactions), the O2� � �Se1 distance of 2.553 (2) Å (a contraction of 0.857 Å from the sum of the van der Waals radii of 3.41 Å ) is clearly weaker than that in the neutral derivative involving the pyridine N atom [N2 i � � �Se1 = 2.3831 (6) Å , a contraction of 1.006 Å from the sum of the van der Waals radii for N and Si of 3.45 Å ]. Selected geometrical and hydrogen-bond parameters are given in Tables 4 and 5, respectively. Consistent with this is the internal Se-N1 bond of 1.926 (2) Å in 2-Me + tosylate (Table 6), which while significantly lengthened compared to the parent ebselen (1.905 Å ), is much less so than in 2 [Se1-N1 = 1.9788 (5) Å ], reflecting the greater extent of the n N -�* Se-N orbital interaction compared to the n O -�* Se-N interaction.

Charge density analysis for 2
We used the experimental electron density from the multipole model to explore the electronic features of the chalcogen bond in 2. Firstly, the electrostatic potential was mapped onto the 0.05 a.u. electron-density isosurface, which revealed a strongly electropositive region along the extension of the Se-N bond, i.e. the hole. Also visible were the lone Hydrogen bonding, chalcogen bonding and �-� stacking in the structure of 2-Me + tosylate trihydrate.

Figure 12
Experimentally determined electrostatic potential for compound 2 mapped onto the 0.05 a.u. isosurface.
pairs of the O and pyridyl N atom, and electron density above and below the �-system (Fig. 12).
The topology of the electron density was also analysed within the QTAIM framework (Bader, 1991), and bond paths corresponding to the Se� � �N and Se� � �O chalcogen bonds were found, along with associated bond critical points (BCPs; Fig. 13). The electron density at the BCP for the shorter Se� � �N chalcogen bond of 0.340 e Å À 3 is significantly larger than that for the Se� � �O chalcogen bond of 0.042 e Å À 3 . The topological parameters associated with these CPs are given in Table 7. The electron density and Laplacian at the critical point (� CP and r 2 � CP for CP Se1-N2 ) are consistent with a closed-shell interaction, but we were intrigued by a number of observations which indicate that this may not be the case. Firstly, the endocyclic Se1-N1 bond is lengthened appreci-ably compared to a gas-phase optimized structure [1.9801 (4) versus 1.8585 Å ; , suggestive of an n N -�* delocalization, leading to partial occupation of the antibonding orbital and thus a lengthening of this bond. Secondly, this same bond has similar topological parameters at the CP to those of the chalcogen bond, which may indicate that the Se atom is participating in a 3-centre-4-electron bond between the two N atoms. Notably, the electronic energy density at the critical point H CP = G CP + V CP is less than zero, corresponding to a dominant potential energy term (V CP ), which is strongly indicative of electron sharing (Cremer & Kraka, 1984;Bone & Bader, 1996). This can be contrasted with the much weaker Se1-O1 chalcogen bond, where the kinetic energy (G CP ) dominates.
The electron localization function (ELF) (Becke & Edgecombe, 1990) is a measure of the probability of finding an  Critical points (CPs) in the vicinity of the Se atom for compound 2. (3,À 1) CPs are shown in red (intramolecular) and green (intermolecular), and (3,+1) CPs are shown in blue.

Figure 14
Electron localization function (ELF) in the plane of the ring system for compound 2.

Table 7
Topological parameters at bond critical points (CPs) in the vicinity of the Se atom for 2.

Critical point
Distance electron of like spin in the vicinity of a fictitious reference electron. It recovers the orbital structure of atoms, while not requiring any knowledge of a wavefunction. An ELF of 1 corresponds to complete localization of an electron pair, a value of 1 2 corresponds to a uniform electron gas-like delocalization, while a value of 0 denotes the border between electron pairs. The ELF in the plane of the aromatic system is plotted in Fig. 14, which shows partial electron localization in the Se1-N1 chalcogen bond, lending further support to the hypothesis that this is not a closed-shell interaction. The ELF along both the strong N1-Se1-N2 chalcogen bond and the weak C1-Se1-O1 chalcogen bond is plotted in Fig. 15, clearly showing the difference in ELF at the BCP of these two contrasting cases. In the stronger chalcogen bond, the ELF is approximately 0.2, while in the weak chalcogen bond it is almost zero.

Conclusions
The crystal structure of the pyridine-substituted benzisoselenazolinone 2 is dominated by strong intermolecular N� � � Se(-N) chalcogen bonding, where the N� � �Se distance of 2.3831 (6) Å is well within the sum of the van der Waals radii for N and Se (3.34 Å ). This strong interaction results in significant lengthening of the internal N-Se distance, consistent with a significant orbital interaction component to the N� � �Se chalcogen bond. Much weaker intermolecular O� � �Se chalcogen bonding occurs between the amide-like O atom in 2 and the less polarized C-Se bond in this structure. Charge density analysis of 2 using multipole refinement of highresolution data revealed the presence of a positive electrostatic surface potential at the antipode to the Se-N1 bond corresponding to the �-hole. Topological analysis of the electron-density distribution in 2 within the QTAIM framework revealed bond paths and (3,À 1) BCPs for the N� � �Se-N moiety consistent with a closed-shell interaction. However, the potential energy term suggests a significant contribution from electron sharing. Analysis of the electron localization function (ELF) for the strong N� � �Se and the weak O� � �Se chalcogenbonding interactions in the structure of 2 suggests significant electron sharing in the former interaction and a largely electrostatic interaction in the latter. Conversion of 2 to its N-methylated derivatives by reaction with methyl iodide and methyl tosylate removes the possibility of N� � �Se intermolecular chalcogen bonding and instead structures are obtained where the iodide and tosylate counter-ions fulfill the role of chalcogen-bond acceptor, with a strong I À � � �Se interaction in the iodide salt and a weaker p-Tol-SO 3 À � � �Se interaction in the tosylate salt.

Special details
Geometry. All esds (except the esd in the dihedral angle between two l.s. planes) are estimated using the full covariance matrix. The cell esds are taken into account individually in the estimation of esds in distances, angles and torsion angles; correlations between esds in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell esds is used for estimating esds involving l.s. planes.

Fractional atomic coordinates and isotropic or equivalent isotropic displacement parameters (Å 2 )
x y z U iso */U eq C1 0.71969 (11) 0.39143 (9) 0.29142 (6) 0.01123 (9)    Special details Geometry. All esds (except the esd in the dihedral angle between two l.s. planes) are estimated using the full covariance matrix. The cell esds are taken into account individually in the estimation of esds in distances, angles and torsion angles; correlations between esds in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell esds is used for estimating esds involving l.s. planes.
Fractional atomic coordinates and isotropic or equivalent isotropic displacement parameters (Å 2 )

Special details
Refinement. Refinement of F 2 against reflections. The threshold expression of F 2 > 2sigma(F 2 ) is used for calculating Rfactors(gt) and is not relevant to the choice of reflections for refinement. R-factors based on F 2 are statistically about twice as large as those based on F, and R-factors based on ALL data will be even larger.