The Identification of De-Alkylation Reactions Catalysed by Cytochrome P450 using Pharmacophore Three-dimensional Structure
L Jones, B Howlin, D Povey
cyp, human cytochrome p450, molecular structure, pharmacophore, predictive metabolism, qsar, substrates
L Jones, B Howlin, D Povey. The Identification of De-Alkylation Reactions Catalysed by Cytochrome P450 using Pharmacophore Three-dimensional Structure. The Internet Journal of Pharmacology. 2007 Volume 5 Number 2.
Quantitative structure activity relationships (QSAR) and pharmacophore three-dimensional structure modelling provide possible methods for understanding the first pass metabolism of human cytochrome P450 substrates in the absence of reliable crystal structures of the human enzymes. The increasing need for alternative and objective methods of metabolism prediction has developed into computational approaches to the problem of understanding the enzyme and substrate behaviour. By analysis of the three-dimensional structure known to be catalysed by human P450 and comparison to other substrates involved in similar alkyl removal reactions, along with the alignment of molecular interaction potentials (MIP), a common template for specific de-alkylations is proposed.
Cytochrome P450 is not specifically one entity, but a widely populated family of haem-thiolate enzyme proteins capable of redox (mono-oxygenase) behaviour (G. Smith, M. J. Stubbins, L. W. Harries, 2004). Human P450s are known to remove drugs from the patient efficiently by “first pass metabolism” in the liver (Rang and Dale, 1987; Gibson and Skett 2001).The prediction of the metabolism and specificity of substrates is a difficult yet important task for xenobiotics (Marechal, Sutcliffe, 2006). Various attempts have been made into categorizing reactions according to the active cytochrome engaged in the investigation (Etkins et al. 2001) as well as QSAR (quantitative structure activity relationship) investigations in previous work. Although the number of members per dataset are small, these were limited by the available information through literature (Lewis et al. 2002-2004). For example, the structures displayed in figure 1 appear to show no direct common structural trends with each other, apart from the necessary group for the selected oxygen de-ethylation reaction to occur by P450. By taking the approach of how the cytochrome might “see” the substrates for reception then areas of biological similarity become apparent. This particular area of predictive research is still well within its infancy since the crystal data for human cytochromes with bound ligands are rare and usually of poor resolution. In an attempt to develop the situation, mathematical approaches to metabolism have been growing in popularity since the 1970's (Hansch 1972). Recent research has suggested that Molecular interaction potentials (MIP) may play an important role in the comparisons of several structures for SAR (structure activity relationships) and Quantitative SAR (QSAR)(Rodrigo et al. 2002). By computationally generating MIP, the interaction “probes” of a structure, and thus their interacting properties can be directly compared to other members undergoing a similar reaction or those which are related by catalyst. The most important probe to consider is the Electrostatic Potential (ESP) of a structure. One such program is MOE (Chemical Computing Group Inc.) which generates ESP grids and displays the ESP map by application of the Poisson-Boltzmann Equation (PBE) to the prediction of electrostatically preferred locations of hydrophobic, H-bond acceptor and H-bond donor locations. Comparing two structures from the oxygen de-ethylation dataset in Figures 2 and 3, the structures were first minimised within MOE using the AM1 UHF method, then the ESP map was generated at a –2 kcal/mol contour level. It can be seen that there are areas of attractive ESP within the centres of the aromatic rings in both of the substrates, approximately 2.8Å from the site of reaction, and another ESP site located at a distance of approximately 6.5Å with both sites being located in the plane defined by the aromatic ring system. The precise magnitudes of the ESP points are not critical at this stage of the investigation. With more computational cost, the substrates could be superimposed to further enforce the positions of the ESP points and display the alignment of such “template” points. Also to be taken into consideration are the number of minimised conformations of the more flexible structures. During the calculation of this example, Phenacetin usually resides in the trans conformation, however, the cis formation is equally plausible. From this small example lies the premise for the theory that three dimensional structure and composition is the predominate factor in first pass metabolism prediction over preference for “indirect approaches which although based on the structure, do not directly consider the structure with reference to the corresponding interacting cytochrome. The paper will focus on the similarity between substrates, grouped by reaction type and analysed by their ESP and “template” points designated by non-carbon and non-hydrogen atoms.
The difficulty of choosing specific locations within the molecular structure to measure to the presumed point of metabolism arose as either to use a particular atom or the locus of a part of the electrostatic potential distribution. The authors of this document considered the various approaches and decided upon primarily using the actual atoms within the structure followed by areas of high electrostatic potential such as the centre of 6 membered aromatic rings, as the aim was to investigate the three-dimensional substrates specifically before and after the reaction as opposed to the manipulation during the docking to the cytochrome p450 catalyst.
Modelled substrate structures were created and minimised using ArgusLab (ArgusLab 4.0.1, Mark A. Thompson, Planaria Software LLC), by the UFF (Universal force field) method. UFF is a molecular mechanics method introduced by Rappe, 1993 which supports the entire periodic table. This method was chosen over other methods such as AM1 (Austin method) due to the full support of the periodic table and the rough optimisation of geometry, as the computational effort required to locate the global minimum structure of a pharmacophore (if determinable) was assumed to be beyond the scope of this work. The distance between atoms was measured using the ChemIQ software from IDBS Ltd. A C++ program was written specifically for this work and was interfaced to and integrated with the ChemIQ program. The specific distance between the two points:
D2 = (x1-x2)*(x1-x2) +
Where x, y and z are the three-dimensional coordinates for the two points in question. The measurement of the centre of a 6 membered aromatic ring involved measuring from the site of reaction to the atoms which were the farthest, and then from the site of reaction to the ring atom which was nearest, as shown in Figure 4. From this, a respectable point for the centre of the aromatic ring could be established as presented in Figure 5
Results and Discussion
Table 1 summarises the dataset of substrates which undergo oxygen demethylation when catalysed by cytochrome p450 and gives the distance from the site of reaction to other electronegative atoms within the structure, as well as the distance from the site of reaction to the centre of the nearest aromatic ring (CNAR). Tables 1, 2, 3, and 4 display the datasets and distances for substrates for the reactions of oxygen demethylation, oxygen de-ethylation, nitrogen demethylation and nitrogen de-ethylation respectively. The gathered distances from tables 1-4 were evaluated to establish common distances appearing in each member of the reaction dataset to produce the template distances presented in table 5. The Aim is for the molecule to retain the required distances in the three-dimensional minimised form. In order for a structure to undergo oxygen demethylation, the pharmacophore must possess a group of “OCH3” and either 1 of the 2 suggested template distances out of trend distances 1 and 2, or a positive value for the distance from the site atom to the centre of the nearest aromatic ring. For a structure to undergo oxygen de-ethylation, the possession of a group of “OC2H5” and 2 of the 2 suggested template distances out of trend distances 1 and 2. The ring distance from site to nearest centre of aromatic ring does not feature in the suggested template for prediction. In order to undergo a reaction of nitrogen de-ethylation, a structure must possess a group of “NC2H5” and both the noted trend distance and the distance to the centre of the nearest aromatic ring. Finally for a structure to undergo nitrogen de-ethylation, the possession of a group of “NCH3” and 1 of the distances of either the noted trend distance, or the distance to the centre of the nearest aromatic ring. Figures 6, 7, 8 and 9 show for each structure in the nominated reaction dataset, the site of reaction (bold black arrows) and the proposed template points (white arrows). All distances are taken from the minimised structures, so a degree of structural flexibility was considered for the evaluation of plausible template points. For the oxygen demethylation set the possible distances for a template were to be found at 9Å, 6.5Å and 5.25Å. All distances are at a maximum variance of + 0.5Å and may occur more than once in one structure. It can be seen that for Trimethoprim there are 3 available sites of reaction, and the required distances for the template are located on the two heteroatoms in the ring, the centre of the aromatic ring itself and also the joining amine branch. The remaining members of the data set have fairly restricted possibilities for template point opportunities: Naproxen shows valid points using the centre of the second aromatic ring (5.6Å) and the OH at the tail of the structure (9.4Å). 7-methoxyresorufin shows a similar layout but with the points located at the bridging nitrogen (5.6Å) and the ketone (9.2Å). Codeine is more difficult to visualise due to its complex three dimensional structure, but valid points are positioned at the heteroatoms oxygen (6.5Å) and nitrogen (7.3Å). Finally metoprolol has points positioned at the first chain oxygen from the aromatic ring ( 6.4Å) and the OH group ( 8.8Å). This structure, however is less rigid than the rest, and like Trimethoprim, could be minimised in such a way that the nitrogen (currently 9.9Å) could also provide a valid point. For the oxygen de-ethylation set the possible distances for a template were to be found at 6.5Å and 5.25Å. All distances are at a maximum variance of ± 0.5Å and may occur more than once in one structure. The remaining members of the data set have fairly restricted possibilities for template point opportunities: 7-ethoxycoumarin shows valid points using the bridging oxygen (4.9Å) and the ketone at the tail of the structure (6.7Å). 7-ethoxy-4-triflourocoumarin displays a similar layout but with the points again located at the bridging oxygen (6.6Å) and the ketone (4.9Å). Phenacetin too, shows valid points positioned at the nitrogen (5.7Å) and tail OH group (6.3Å). For the nitrogen demethylation set the possible distances for a template were to be found at 2.5Å and 5.75Å. Distances may occur more than once in one structure. Some of the dataset members have fairly flexible possibilities for template point opportunities given that the some of the members may be able to alter their conformations to accommodate a missing template point: Diazepam shows a valid point using the oxygen (2.4Å), and a possible secondary point using the chlorine (6.1Å). Benzphetamine possesses possible points at the centre of both of the aromatic rings (3.6Å and 5.2Å), but due to conformational felexibility, the structure may adopt a more suitable conformation for the template. S-mephenytoin displays a layout with the points located at the ring nitrogen (3.6Å) and the centre of the aromatic ring (5.7Å). Zopiclone shows valid points positioned at the first nitrogen (2.9Å) and centre of the tail aromatic ring (6.6Å). MPTP shows a valid point using the centre of the aromatic ring (2.4Å), although there appears to be no suitable candidate for a secondary point for further template confirmation. MDMA possesses a valid point at the centre of the aromatic ring (5.2Å) and possible secondary points with further conformation possibilities. Finally rosiglitazone has template points positioned at the first chain oxygen(2.4Å) and the nitrogen of the pyridine ring (2.5Å). For the nitrogen de-ethylation set the possible distances for a template were to be found at 3.65Å and a ring distance of 6.25Å. All distances are at a maximum variance of ± 0.5Å and may occur more than once in one structure. The members of the data set have fairly flexible possibilities for template point opportunities given that some of the members may be able to modify their conformations to accommodate a missing template point. Alfentanil shows a valid point using a nitrogen in the centre of the structure (3.8Å), this structure, however is less rigid than the rest, and like Trimethoprim, could be minimised in such a way that the remaining heteroatoms on the tail of the structure such as nitrogen (currently 7.4Å) or either of the two oxygen atoms (7.9Å and 8.5Å) could also provide a valid point. Lidocaine displays a layout with the points located at the bridging nitrogen (3.6Å) and the centre of the aromatic ring (5.7Å). Metoclopramide too, shows valid points positioned at the nitrogen (3.8Å) and centre of the aromatic ring (6.8Å). Finally sulpiride has template points positioned at the first chain nitrogen (6.4Å) and once again the centre of the aromatic ring (5.7Å).
In conclusion, it is proposed that the distribution and location of electronegative “points” within a three-dimensional structure can provide a start towards the prediction of first pass metabolism reactions, and that further studies could provide yet more structure/activity relationships.
The work described above has been carried out with partial funding of ID Business solutions (Guildford, UK) and we acknowledge a PhD studentship (LLJ) from ESPRC.