Introduction

Cytisine is a quinolizidine alkaloid obtained from plants belonging to the Fabaceae (Leguminosae) family. For centuries, it has been used in traditional medicine in the treatment of, e.g., migraine, insomnia, asthma, or cough. More recently, it has gained more popularity due to its pharmacological action closely resembling that of nicotine [1,2,3]. During the Second World War, leaves of Cytisus Laburnum (known also as “golden chain” or “golden rain”) were used as a tobacco substitute, and Russian soldiers referred to the plant as “fake tobacco” [2, 4].

Nicotinic receptors form a family of acetylcholine (Ach)-gated cation channels made up of different subtypes, localized in the central and peripheral nervous systems [5, 6]. Nicotine, a naturally occurring alkaloid, affects the functioning of peripheral nervous system. It stimulates secretion of catecholamines, leading to increased blood pressure and sugar concentration, which alleviates nicotine withdrawal symptoms and stimulation of the respiratory system [2, 7]. Comparison of the structures of nicotine and cytisine (Scheme 1) showed that the quasi-aromatic ring in cytisine and the pyridine ring in nicotine are associated in a similar way with respect to the nitrogen atom in the bispidine ring system (cytisine) and in the pyrrolidine ring (nicotine) [8]. Cytisine shows an affinity towards the specific subunits of nicotine acetylcholine receptors, nAChRs. It acts as a partial agonist at the α4β2 subunit, and very potently binds to the α3β4, α7 nicotinic subunits [5]. So, cytisine possesses a structure and mechanism of action that are similar to those of nicotine, but it has much lower toxicity [9, 10]. Therefore, cytisine has been applied in nicotine replacement therapy in the form of Tabex®, Desmoxan®, Chantix® etc.

Scheme 1
scheme 1

Comparison of the structures of nicotine and cytisine

Current research focuses on the search of novel compounds—potential therapeutics which are more lipophilic, therefore having improved ability to pass the blood-brain barrier, but still retain the high affinity for α4β2 subtype receptors [2, 9]. Moreover, cytisine derivatives have been explored as potential drugs against Alzheimer’s and Parkinson’s, attention deficit, hyperactivity disorder diseases [9,10,11].

High-resolution X-ray crystallography can be used as a main tool in the analysis of the structure-activity relationships. Here, we present the analysis of a charge density distribution and intermolecular interactions in the cytisine [(−)-1,2,3,4,5,6-Hexahydro-1,5-methano-8H-pyrido(1,2-a)(1,5)diazocin-8-one] (1) and N-methylcytisine [(−)-1,2,3,4,5,6-hexahydro-3-methyl-1,5-methano-8H-pyrido(1,2-a)(1,5)diazocin-8-one] (2) crystals. The results of these studies are supposed to provide high-quality electron density distribution data, which can be used not only in SAR studies but also in docking studies of the active site if the appropriate receptor.

Those two compounds, despite the relatively small changes in their structures, display quite significant differences in affinity towards α4β2 subtype of AChRs [12]. It has been hypothesized that much lower affinity of N-methylcytisine is probably due to the steric hindrance with Tyr-91 [12] or to lack of the possibility of forming hydrogen bonds with hydroxyl group of Tyr-91 or backbone carbonyl of Trp-145, which is formed in case of cysteine.

Experimental

X-ray data

Crystal data and refinement details are presented in Table 1. High-resolution X-ray diffraction data were collected at 100(1)K (under nitrogen flow 5 L/min), on an Oxford Diffraction Supernova Dual diffractometer, equipped with Atlas detector, and microfocus Mova tube with MoKα radiation (λ = 0.71073 Å). The temperature was controlled with an Oxford Instruments Cryosystem device. The data were corrected for Lorentz-polarization effects as well as for absorption (multiscan) [13]. Accurate unit-cell parameters were determined by a least-squares fit of 53,660 (1), and 47,045 (2) reflections of highest intensity, chosen from the whole experiment.

Table 1 Crystallographic measurement and refinement data

IAM model refinement

The calculations were mainly performed within the WinGX [14] and OLEX [15] program systems. The structures were solved with ShelxT [16] and refined (within the Independent Atom Model approximation) with the full-matrix least-squares procedure on F2 by SHELXL97 [16]. Scattering factors incorporated in SHELXL97 were used. All non-hydrogen atoms were refined anisotropically, hydrogen atoms were located by difference Fourier synthesis, and isotropically refined.

Multipolar refinement

The charge density was subsequently refined against structure factor amplitudes with the MoPro suite [17] using the multipolar Hansen-Coppens model [18] for pseudoatom electron density:

$$ {\rho}_{\mathrm{atom}}(r)={\rho}_{\mathrm{core}}(r)+{P}_{\mathrm{val}}{\kappa}^3{\rho}_{\mathrm{val}}\left(\kappa r\right)+{\Sigma}_{\mathrm{l}\ }\kappa {\hbox{'}}^3{R}_{\mathrm{l}}\left({\kappa}^{\hbox{'}} r\right)\ {\Sigma}_{\mathrm{m}}{P}_{\mathrm{l}\mathrm{m}\pm }{d}_{\mathrm{l}\mathrm{m}\pm}\kern0.5em \left(\vartheta, \varphi \right) $$
(1)

where the first two terms are the spherically averaged core and valence electron densities of the atom, and the last term corresponds to the non-spherical valence density which is described in terms of real spherical harmonic functions. P val is the valence population, P lm± are the multipole populations and κ and κ’ are the contraction/expansion parameters. R l are radial Slater type functions:

$$ {R}_l(r)=\frac{\xi_l^{n l+3}}{\left({n}_l+2\right)!}{r}^{n l}{e}^{-{\xi}_l r} $$
(2)

The oxygen, carbon, and nitrogen atoms were refined up to octupolar level (l max = 3 and the hydrogen atoms up to dipole level (l max = 1). The n l and ξ l values were set equal to 2, 2, 2, 3, and 4.466 au−1 (O), 2, 2, 2, 3, and 3.176 au−1 (C), 2, 2, 2, 3, and 3.839 au−1 (N), and 1, 1, and 2.00 au−1 (H). The core and valence scattering factors were calculated from Clementi and Roetti wave functions [19]. The structures have been refined against intensities, using several different strategies to obtain the best available model of crystal structures. The refinement was performed for sinθ/λ up to 1.12 Å−1 for cytisine (1) and to 1.0 Å−1 for N-methylcytisine (2). Final refinement strategies for both compounds were similar. Initially, only the scale factor was refined, followed by non-hydrogen, the atomic positions and atomic displacement parameters (ADPs) against all reflections. Subsequently, a high-order (sinθ/λ>0.7 Å−1) refinement for non-H atoms was performed, aiming at the deconvolution of the ADP’s and bonding deformation effects [20]. The ADPs for H atoms were estimated using SHADE server [21]. X-H distances were constrained to average values obtained from the neutron diffraction studies. Then P val, κ, P lm and κ’ were consecutively refined until convergence. The κ an κ’ values for H atoms were restrained to 1.13 (0.02) and 1.29 (0.02), respectively. Chemical equivalent atoms in two symmetry-independent molecules of 1 were restrained to be similar. The details of the data collection and crystallography statistics are summarized in Table 1.

The RDA (residual density analysis) [22] was performed for the final models. This kind of analysis, based on the fractal dimension of the residual electron density map, is used as a method of detection of the systematical errors and quantitative description of the deficiencies of the model. These problems are generally reflected in asymmetric shape of the appropriate plot. Fig. 1 Sup (cf. Supplementary material) shows that in case of both 1 and 2 the plots are quite symmetrical and regular which testifies for the quality of the final models.

Results and discussion

The standard resolution, room temperature crystal structures of cytisine (1) and N-methylcytisine (2) have been reported earlier [8, 23, 24]. Both alkaloids crystallize in non-centrosymmetric space group P 212121. The asymmetric part of the unit cell contains two (almost identical) molecules in case of (−)-cytysine (1) and one for (−)-N-methylcytisine (2). The perspective views of the molecules are presented in Fig. 1. No phase transitions have been observed while lowering temperature down to 100 K.

Fig. 1
figure 1

Perspective views of (1) and (2) with the numbering schemes, ellipsoids are drawn at the 50% probability level, anisotropic thermal parameters for hydrogen atoms were calculated with SHADE server [21]

Molecular geometries

The Hirshfeld rigid bond test [20] was applied to the non-hydrogen atoms joined by the covalent bonds after the final multipolar refinement. For each pair of non-hydrogen atoms, the values of ΔZAB 2 (differences between the components of the U ij tensor along the bond) do not exceed the acceptable value of 10−3 Å2. The largest differences have been found for N1-C10/N21-C30 (0.00066/0.00076) in 1 and for N1-C6 (0.00091) in 2. The static deformation density maps, presented in Fig. 2, show expected bonding features: accumulation of electron density in the bonds in pyridine ring more significant than in piperidine one, electron lone pairs on oxygen atoms, and subtle change of electron density distribution in the vicinity of the structure modification—in this case, methylation on N12. The dynamic deformation density maps, which show similar features, as well as static deformation density maps for the second molecule of 1, are deposited as a Supplementary material.

Fig. 2
figure 2

The static deformation electron density distribution in plane of a, b pyridine ring c, e N12-C11-C13 plane d, f, C9-C22-C7-C23 plane, the positive (blue) and negative (dashed red) contours are drawn at intervals 0.05 e Å−3

For analysis of the bonds and interactions, we have applied the AIM approach [25]. The bond critical points (BCPs) for carbon-carbon bonds are mainly placed halfway between the atoms involved, with highest values of electron density found at the BCP of C3-C4 bond (pyridine ring), equal to 2.37 e Å−3 in 1, and 2.39 e Å−3 in 2 (Fig. 3). The greatest shift of BCP for C-C bonds has been found for C6-C7 (0.12 Å/0.10 Å towards C7), maybe due to the definitely different character of both carbon atoms. It might be noted, that the distribution of carbon-carbon BCP’s in 2 is apparently less symmetrical: seven out of ten BCP’s are located significantly far from the respective middlepoints of the bonds (in 1 there is only one such example). Carbon-nitrogen and carbon-oxygen BCPs in both structures are—obviously due to the differences in charges—shifted towards the carbon atoms. The C=O bonds have overall largest values of electron density at the BCP: 3.00/2.98 e Å−3 in two molecules of 1, 2.94 e Å−3 in 2. Full characteristics of bond critical points can be found in the Supplementary material (Table 1).

Fig. 3
figure 3

The bond critical points in molecule 1 of 1 (left) and in 2 (right)

The analysis of the basic features, like Laplacian, electron density at the bond critical point, and positive curvature of the Hessian matrix λ 3 (Fig. 4) shows that the significant differences between 1 and 2 are observed in the C and B rings, certainly due to the structural differences: absence/presence of the methyl substituent at N12 (Fig. 4).

Fig. 4
figure 4

Graphical representation of Laplacian [∇2(ρ bcp) (e Å−5)] electron density [ρbcp (e Å−3)] and positive curvature of the Hessian matrix λ 3 in bond critical points for bonds not involving hydrogen atoms

The electrostatic potential plots of (1) and (2) are presented in Fig. 5. In case of 1, the negative potential is located in pyridine ring in both symmetry independent molecules, but the values differ quite significantly, probably as a result of the different interactions in which both molecules are involved. In 2 the values of potential within the molecule are generally smaller than in case of cytisine, the negative potential is observed only around oxygen atom and nitrogen N12 atom (from piperidine ring). Consequently, the dipole moments calculated for cytisine are significantly larger than for N-methylcytisinie (36.52 D/5.67 D).

Fig. 5
figure 5

Electrostatic potential mapped over same electron density iso-surface for cytisine molecules (left), and N-methylcytisine molecule (right). The iso-surfaces are drawn at 0.1 e/Å3

The comparison of atomic charges is not quite straightforward; however, it might be noted that the differences are generally due to the different definitions which can be used, consequently significant discrepancies in values and even in signs that can have been observed [26]. Using consequently, the same definition may allow for the—at least qualitative—comparison of the results. So, to analyze the differences in atomic charges we have compared the results for N valP val differences between the formal and refined population of valence electrons and—on the other hand—the integrated AIM charges [27]. The graphical representation of the values is shown in Fig. 6, all values are collected in Supplementary material, Table. Significantly more coherent results were obtained when calculating atomic charges integrated over the atomic basins (AIM approach). In case of N valP val the differences are much larger. This agrees generally with previous observations (for instance, [26]).

Fig. 6
figure 6

Graphical representation of atomic charges calculated from N valP val or integrated atomic charges over the atomic basins (AIM)

Intermolecular interactions

Bond critical points (3,−1) have been found for 33 intermolecular contacts in the crystal structure of cytisine (30 of them have electron density at the critical point larger than 0.02 e Å−3), while in the structure of N-methylcytisine there are 20 such contacts with 16 with electron density at the CP above 0.02 e Å−3. The data for all these CPs are deposited as the Supplementary materials (Tables 3 and 4). Few shortest, and the most important interactions are listed in Tables 2 and 3 below.

Table 2 Distances and topological characteristics of the BCPs in 1
Table 3 Distances and topological characteristics of the BCPs in 2 (definitions as in Table 2)

As has been mentioned before, in course of the discussion of intramolecular BCPs, two independent molecules of 1 are involved in similar but not identical molecular interactions. For instance, oxygen atoms from both molecules accept different numbers of significant N-H···O and C-H···O hydrogen bonds, and the characteristics of formally similar interactions are also quite different. For instance, the most important N-H···O hydrogen bond is much stronger for this accepted by molecules A than for that accepted by B.

This difference can be related to the less directional geometry of N12-H12···O22 (Fig. 7) interaction, which can be attributed to the fact that O22 atom is acceptor of more CH∙∙∙O contacts than O2.

Fig. 7
figure 7

The 3D view of static defamation charge density map in the area of a N32-H32···O2 and b N12-H12···O22 interaction. The iso-surfaces are drawn at 0.125 e Å−3. Positive and negative values are represented in blue and red, respectively. Representation of c N32-H32···O2 and e N12-H12···O22 in Laplacian maps, logarithmic contours in e Å−5. Red and blue contours denote positive and negative values, respectively. Representation of d N32-H32···O2 and f N12-H12···O22 gradient lines of the total electron density

Among the most important interactions, besides CH∙∙∙O hydrogen bonds, also C-H···N and C-H···C contacts have been found (Fig. 8). The first one might be an interaction between the hydrogen atom from aromatic ring and the lone pair of piperidine nitrogen atom. The second one is less directional, and—although critical point has been found between hydrogen and carbon atom from aromatic ring—it cannot be classified as C-H···π interaction with the double C3=C4 bond electrons. It is consistent with the abovementioned fact, that C3 and C4 atoms are the area of the most negative potential in the whole molecule, so this interaction can have attractive nature, which can be seen in Fig. 8.

Fig. 8
figure 8

C-H···N and C-H···C interaction: a bond path and associated critical point b 3D view of static deformation charge density map in the area of C-H-N and C-H-C contacts. The iso-surfaces are drawn at 0.125 e Å.3 Positive and negative values are represented in blue and red, respectively

Due to the methylation of atom N12, the molecule 2 lacks strong hydrogen bond donors, and therefore in its structure all intermolecular interactions are much weaker than in case of 1 (Table 3).

One of the strongest contacts is of H∙∙∙H type (Fig. 9), with the highest value of electron density at the critical point (0.050 e/Å3). Such H···H bonding between two similar hydrogen atoms (both connected to carbon atoms) has almost all characteristics of its critical point similar to other weak hydrogen bonds, which can prove the importance of such type of interactions. The presence of H···H bonding has been postulated in other studies (for instance, [28]). It can be noted that in case of N-methylcytisine the strongest H···H interactions are found between slightly different types of hydrogen atoms (for instance one from the aromatic ring, and the other from the methylene group), and in such cases the critical point is shifted towards the atom with higher positive charge (as defined by the integration over the appropriate atomic basin).

Fig. 9
figure 9

H···H interaction: a bond path and associated critical point b 3D view of static defamation charge density map in the area of H···H contact. The iso-surfaces are drawn at 0.125 e/Å3. Positive and negative values are represented in blue and red, respectively

Figure 10 shows the plot of the chosen characteristics of the intermolecular BCPs vs. R ij distance. This plot, after removing the debatable H···H interactions, fits quite well into exp.-type dependence, as postulated, e.g., by Wozniak et al. [29, 30].

Fig. 10
figure 10

A plot of ρ(CP) vs. R ij for intermolecular BCPs in 1 after removing H···H interactions

Conclusions

The experimental charge density studies as well as topological analysis have been performed for two bioactive molecules: cytisine and its N-methyl derivative. The analyzed features show that besides the structural similarities of both alkaloids, the electron density distribution and consequently all its derivatives differ quite significantly, which can be (except the steric factors) the reason of definitely lesser activity of (2) towards α4β2 receptor.