**3. Computational ligand docking**

*Molecular Docking and Molecular Dynamics*

in **Figure 4** was performed using MOE's default settings.

alignment—deletions of positions 361 and 362 in the target sequence and an insertion in position 308. The absence of major gaps in the alignment is favorable for the development of homology models as it reduces the need for loop modeling and grafting, which can be challenging [17]. Model development based on the alignment

*(A) Homology model for the AhR colored by secondary structure. (B) Comparison of backbone traces of homology models obtained by using the MOE modeling suite (template 4F3L) and the automated SWISS-MODEL server (template 5SY7). Coloring is according to RMSD between the two structures (green—yellow—red, in order of increasing deviation), showing very good agreement between the two models. The model obtained from the SWISS-MODEL server had a somewhat longer sequence, resulting in the gray loops at the termini that have no* 

*Ramachandran diagram for the homology model for the AhR. Green ( ) symbols represent torsion angles in favored regions, whereas yellow ( ) symbols represent angles in allowed regions. No entries are present in the* 

**10**

**Figure 6.**

**Figure 5.**

*counterpart in the MOE model.*

*"forbidden" areas.*

Before ligands could be docked into the homology model of the AhR, the exact location of the binding site had to be identified. We subjected the homology model to a binding site search, a feature implemented in MOE that screens the surface of a protein for concave areas capable of binding small molecules. Two areas large enough to accommodate a typical AhR ligand were detected: one on the surface and another one in the core of the receptor. To decide which of these two sites was more realistic, the crystal structures of the ligand/receptor complexes 3F1O [20], 3H7W [21], and 3H82 [21], all of which are proteins related to the AhR, were superimposed onto the homology model. As shown in **Figure 7**, all three ligands were found in an area equivalent to the binding site located at the center of the protein (**Figure 5**). To facilitate a convenient designation of the binding site for the subsequent docking runs, the ligand of 3F1O—N-[2-nitro-4-(trifluoromethyl)phenyl]morpholin-4-amine (**5**)—was copied into the file of the homology model as a point of reference.

#### **Figure 7.**

*(A) Superposition of protein/ligand complexes related to the AhR onto the homology model of the AhR. Spheres delineate the putative binding site predicted by MOE that coincides with the position of the ligands seen in the crystal structures. (B) A closeup view of the ligand 5, overlaid with the spheres depicting the binding site predicted by MOE. (C) The chemical structure of the ligand of 3F1O—N-[2-nitro-4-(trifluoromethyl)phenyl]morpholin-4-amine (5).*

**Figure 8.** *Structures of six dichlorophenylacrylonitriles (4, 6–10) used for docking.*

#### **Figure 9.**

*(A) 3D representation of dichlorophenylacrylonitrile 7 docked into the binding site of the AhR, illustrating the central nature of the site. (B) Interaction diagram of the pyrrole ligand ANI-7 (4), showing π-π -interactions, hydrophobic contacts, and shape complementarity as the driving forces for ligand binding.*

To analyze the utility of this model for further drug development, the structures of a representative ensemble of six dichlorophenylacrylonitriles with known bioactivities (**Figure 8**) were modeled in MOE. Their conformational energies were minimized by molecular mechanics in conjunction with the MMFF94x force field. Docking was performed with the default settings of MOE, utilizing a flexible ligand and a mostly static receptor structure and defining the binding site by a position equivalent to that of the ligand present in 3F1O. The top-scoring pose for each ligand was considered for further analysis.

As shown in **Figure 9**, the docked compounds occupied a narrow and mostly hydrophobic site in the core of the AhR. Almost all ligands in the pool engaged with the AhR in a similar fashion, binding in comparable binding poses and exhibiting similar ligand/receptor interactions. Key hydrophobic contacts were observed between nonpolar regions of the ligands and the side chains (Phe21, Leu34, Phe50, Met66, Leu79, Ala93, Ile105, and Val107). Moreover, the ligand phenyl ring engaged in π-π stacking interactions with the ring of His 17. In addition, the tight fit between the ligands and the site suggested the presence of extensive favorable van der Waals interactions.

**13**

environment.

**Figure 10.**

*standard settings of MOE.*

*Binding of Chlorinated Phenylacrylonitriles to the Aryl Hydrocarbon Receptor…*

In an attempt to attain a quantitative measure of ligand-binding affinity, the docking scores of the compounds were graphed against observed potencies (**Figure 10**). Potencies had been obtained in cell viability assays and the underlying assumption was that ligand binding to the receptor constituted the critical step that would lead to cell toxicity and therefore would correlate with bioactivity. **Figure 10** shows a reasonable correlation between the two quantities with a squared correlation coefficient of 0.79. This data is consistent with the proposed binding mode of AhR ligands, which

*Linear correlation between docking score and bioactivity (−log IC50). Docking was performed with the* 

We complemented our docking-based analysis by MD simulations, whose purpose was twofold. First, we wanted to ensure the stability of a docked pose by monitoring its behavior in a time-resolved system. Second, MD simulations can reveal the role of explicit solvent molecules, something that cannot be accounted for by docking. We selected compound **7** as a representative and subjected it to a simulation time of 100 ns, using the CHARMM36m force field for the protein [22] and the CHARMM general force field for the ligand [23]. The parameters for water were taken from the CHARMM-modified TIP3P water model [24–26] to match those used for the solute. The initial structure of the protein-ligand complex was obtained from docking experiments, and the simulation was performed with the software NAMD [27].

As shown in **Figure 11**, the differences between the poses before and after 100 ns of simulation time were minor. The overall position of the ligand did not change significantly and the only notable difference related to a slight rotation around the central axis of the molecule which placed the nitrile group in a somewhat different

Interestingly, analysis of the MD simulation data revealed the presence of several water molecules in close proximity to the ligand. This observation was somewhat unexpected; while polar water molecules have been found in predominately hydrophobic cores of proteins, it is a rare occurrence [28, 29]. Residues exposed to water molecules included Leu315, Thr289, His 291, Gln383, Ser365, and His 337. In some cases, the solvent molecules formed bridged hydrogen bonds between the nitrile group and the ligand. The latter could explain the abovementioned slight twist of

relies predominately on hydrophobic but also on additional π-π interactions.

*DOI: http://dx.doi.org/10.5772/intechopen.84818*

**4. Molecular dynamics simulations**

the nitrile group into a more favorable position.

*Binding of Chlorinated Phenylacrylonitriles to the Aryl Hydrocarbon Receptor… DOI: http://dx.doi.org/10.5772/intechopen.84818*

#### **Figure 10.**

*Molecular Docking and Molecular Dynamics*

*Structures of six dichlorophenylacrylonitriles (4, 6–10) used for docking.*

To analyze the utility of this model for further drug development, the structures of a representative ensemble of six dichlorophenylacrylonitriles with known bioactivities (**Figure 8**) were modeled in MOE. Their conformational energies were minimized by molecular mechanics in conjunction with the MMFF94x force field. Docking was performed with the default settings of MOE, utilizing a flexible ligand and a mostly static receptor structure and defining the binding site by a position equivalent to that of the ligand present in 3F1O. The top-scoring pose for each

*(A) 3D representation of dichlorophenylacrylonitrile 7 docked into the binding site of the AhR, illustrating the central nature of the site. (B) Interaction diagram of the pyrrole ligand ANI-7 (4), showing π-π -interactions,* 

*hydrophobic contacts, and shape complementarity as the driving forces for ligand binding.*

As shown in **Figure 9**, the docked compounds occupied a narrow and mostly hydrophobic site in the core of the AhR. Almost all ligands in the pool engaged with the AhR in a similar fashion, binding in comparable binding poses and exhibiting similar ligand/receptor interactions. Key hydrophobic contacts were observed

between nonpolar regions of the ligands and the side chains (Phe21, Leu34, Phe50, Met66, Leu79, Ala93, Ile105, and Val107). Moreover, the ligand phenyl ring engaged in π-π stacking interactions with the ring of His 17. In addition, the tight fit between the ligands and the site suggested the presence of extensive favorable

ligand was considered for further analysis.

van der Waals interactions.

**12**

**Figure 9.**

**Figure 8.**

*Linear correlation between docking score and bioactivity (−log IC50). Docking was performed with the standard settings of MOE.*

In an attempt to attain a quantitative measure of ligand-binding affinity, the docking scores of the compounds were graphed against observed potencies (**Figure 10**). Potencies had been obtained in cell viability assays and the underlying assumption was that ligand binding to the receptor constituted the critical step that would lead to cell toxicity and therefore would correlate with bioactivity. **Figure 10** shows a reasonable correlation between the two quantities with a squared correlation coefficient of 0.79. This data is consistent with the proposed binding mode of AhR ligands, which relies predominately on hydrophobic but also on additional π-π interactions.

## **4. Molecular dynamics simulations**

We complemented our docking-based analysis by MD simulations, whose purpose was twofold. First, we wanted to ensure the stability of a docked pose by monitoring its behavior in a time-resolved system. Second, MD simulations can reveal the role of explicit solvent molecules, something that cannot be accounted for by docking. We selected compound **7** as a representative and subjected it to a simulation time of 100 ns, using the CHARMM36m force field for the protein [22] and the CHARMM general force field for the ligand [23]. The parameters for water were taken from the CHARMM-modified TIP3P water model [24–26] to match those used for the solute. The initial structure of the protein-ligand complex was obtained from docking experiments, and the simulation was performed with the software NAMD [27].

As shown in **Figure 11**, the differences between the poses before and after 100 ns of simulation time were minor. The overall position of the ligand did not change significantly and the only notable difference related to a slight rotation around the central axis of the molecule which placed the nitrile group in a somewhat different environment.

Interestingly, analysis of the MD simulation data revealed the presence of several water molecules in close proximity to the ligand. This observation was somewhat unexpected; while polar water molecules have been found in predominately hydrophobic cores of proteins, it is a rare occurrence [28, 29]. Residues exposed to water molecules included Leu315, Thr289, His 291, Gln383, Ser365, and His 337. In some cases, the solvent molecules formed bridged hydrogen bonds between the nitrile group and the ligand. The latter could explain the abovementioned slight twist of the nitrile group into a more favorable position.

**Figure 11.**

*MD simulations. (A) Ligand/AhR complex before (gray) and after (blue) 100 ns of simulation time. (B) Density of water molecules in the binding site as highlighted by orange grids. Residues are colored according to their polarity (pink: nonpolar; green: polar uncharged).*
