Anita Plazinska, Michal Kolinski, Krzysztof Jozwiak, Irving W. Wainer

Molecular interactions between fenoterol stereoisomers and derivatives and the β2-adrenergic receptor binding site studied by docking and molecular dynamics simulations

Anita Plazinska, Michal Kolinski, Krzysztof Jozwiak, Irving W. Wainer

Anita Plazinska, Michal Kolinski, Krzysztof Jozwiak

Laboratory of Medicinal Chemistry and Neuroengineering, Medical University of Lublin, ul. Chodzki 4a; 20-093 Lublin, Poland. e-mail:

Michal Kolinski

Bioinformatics Laboratory, ul. Dworkowa 3, 00-783 Warszawa, Poland

Irving W. Wainer,

Laboratory of Clinical Investigations, National Institute on Aging, Baltimore, MD.

Online Resource Fig. 1 The dependencies of the experimentally determined pKi values on the MolDock Score function values for all stereoisomers (N = 32) of fenoterol derivatives (A) and for (R,R’)-, (R,S’)- (S,R’)- and (S,S’)-stereoisomers of fenoterol derivatives (B) docked to In_ β2-AR.

Online Resource Fig. 2 The dependencies of the experimentally determined pKi values on the MolDock Score function values for all stereoisomers (N = 32) of fenoterol derivatives docked to Ac_β2AR(A). The dependencies of the experimentally determined pKi values on the MolDock Score function values for each of the stereoisomer groups of fenoterol derivatives (B).

Online ResourceTab.1 The parameters describing the correlation between experimentally determined pKi values and the MolDock Score function values.

Subgroup / R2 / Slope (+/-) / Intercept / N
In_β2ARmodel
(R,R’) / 0.637 / -0.0594 / -1.9188 / 10
(R,S’) / 0.8939 / -0.0802 / -4.7703 / 9
(S,R’) / 0.8168 / -0.8168 / -0.0828 / 6
(S,S’) / 0.6198 / -0.0515 / -1.8233 / 7
(R,*) / 0.7621 / -0.0707 / -3.484 / 19
(S,*) / 0.7053 / -0.0667 / -3.7243 / 13
(*,R’) / 0.6013 / -0.0801 / -4.9328 / 16
(*,S’) / 0.6317 / -0.0793 / -4.9692 / 16
(*,*) (all stereoisomers) / 0.6251 / -0.0812 / -5.1443 / 32
Ac_β2ARmodel
(R,R’) / 0.4890 / -0.0604 / -2.4239 / 10
(R,S’) / 0.7161 / -0.1327 / -12.546 / 9
(S,R’) / 0.6778 / -0.0792 / -6.2858 / 6
(S,S’) / 0.6304 / -0.0636 / -3.9068 / 7
(R,*) / 0.5347 / -0.0815 / -5.4336 / 19
(S,*) / 0.6159 / -0.0637 / -3.9994 / 13
(*,R’) / 0.1868 * / -0.0518 / -1.6799 / 16
(*,S’) / 0.4898 / -0.109 / -9.6266 / 16
(*,*) (all stereoisomers) / 0.3116 / -0.073 / -4.6802 / 32

Sign * stands for the non-statistically significant data, p > 0.05.

Online ResourceTab.2The trends found in the pKi values depending on the type of R2 substituent. The main type of the R2-β2ARinteraction is also given.

FENOTEROLS
ID / 1 / 2 / 4
R2 / / /
pKi / 6.46 / 6.33 / 5.53
Trend in the pKi values /
Type of the R2-β2ARinteraction / HB
NAPHTHYLFENOTEROLS
ID / 52 / 54 / xxx
R2 / / /
pKi / 6.60 / 6.32 / 6.29
Trend in the pKi values /
Type of the R2-β2ARinteraction / Π-Π/ HB

Modeling of the ligand-receptor complexes

We made an attempt to predict N and C terminal domains and third intracellular loop of β2 -AR to preserve whole receptor structure. Although we used one of the best performing methods for ab initio structure prediction: I-TASSER server (residues Met1 to Glu30 and Cys341 to Leu413 respectively) and CABS method (residues Leu230 to Leu266) resulting conformations may significantly differ from those observed in the native form of the receptor protein. The accurate structure prediction is extremely difficult because of high conformational instability of those regions. Observed RMSF values indicated notable conformational flexibility of modeled fragments (Fig. 3 bottom) during 20 ns MD simulation. Nevertheless the predicted fragments did not altered overall receptor structure (see RMSD and RMSF plots, Fig. 3) during conducted molecular dynamics simulations. Moreover, predicted fragments did not affect docking procedure because they were not localized in the vicinity of the receptor binding site.

Online ResourceFig. 3 Plot presenting the RMSD values for TM segments of simulated models (top). Plot presenting the RMSF (RMSD fluctuations from the average residue coordinates over the simulation time) for each residue of simulated receptors (bottom). In_β2-AR - blue, Ac_β2-AR - green and β2ART4 lysozyme fusion protein crystal structure with bond (S)-carazolol - red. Localization of TM helices of the receptor proteins are marked with brown lines.

RMSD plots for TM domains of the In_β2-AR, Ac_β2-AR modelsand β2-AR crystal structure (PDB ID: 2RH1) are very similar. The RMSD value estimated for transmembrane helices of three proteins models becomes stable at the level of 0.14-0.18 nm after 20 ns of MD. The extracted structures of In_β2-AR and β2-ART4 (PDB ID: 2RH1) are very similar to each other. The conformations of TM domains of these two structures are almost identical. The obtained receptor models: In_β2-AR and Ac_β2-AR have one additional helical turn at the intracellular end of TMHV. The plot of RMSF values indicates small fluctuations of TMHII to TMHVII in In_β2-AR and Ac_β2-AR during the MD simulation. RMSF values for TMHI are noticeable smaller for two created receptor models when compared to simulated crystal structure plausibly due to the presence of N terminal domain which stabilized the extracellular part of TMHI. Inspite of the high flexibility of N terminal domain (residues: Met1 to Asp29) of the constructed receptor model In_β2-AR, residues Asp21 to Thr25, close to the beginning of TMHI, formed one helical turn that was well preserved during the whole MD simulation time. The second extracellular loop (ECL2) covering the receptor binding site has the same shape and localization in three compared receptor structures. Little changes can be seen in conformations of all short loops connecting seven TM helices located at the extracellular side of simulated receptors. The C terminal domain (Leu340 to Leu413) and third intracellular loop (ICL3, Leu230 to Leu266), connecting TMHV and TMHVI of both In_β2-AR and Ac_β2-AR, showing high RMSF values, did not adopt the single preferred stable conformation during the MD simulation. Protease susceptibility and intramolecular fluorescence resonance energy transfer experiments [49] indicated that those two receptor regions are the most unstructured ones.No change was observed in the states of well known molecular switches of β2-ARincluding: the “ionic lock” involving amino acid sequence DRY from TMH III, sequential motif NPLIY at the end TMHVII and the “rotamer toggle switch” combined with the structural water molecule (visible in the receptor crystal structure: PDB ID: 2RH1) when compared to the staring structures. The position and conformation of the Ile121 and Phe282 residues involved in recently reported molecular switch [11], were well conserved in the simulated structures of In_β2-AR and the β2-ART4 lysozyme fusion protein presenting inactive state of the receptor. In contrast, conformation of these two residues was not stable during the simulation of Ac_β2-AR representing the receptor active state. High conformational instability of Ile121 and Phe282 may result from the fact that the MD simulation of Ac_β2-AR was conducted using unliganded receptor protein and in absence of nanobody seen in the crystal structure (PDB ID: 3P0G).

MD simulations - binding modes of two fenoterol isomers

Online ResourceFig.4 Starting position of (R,R’)-fenoterol - color red, and (S,S’)-fenoterol - color blue, inside

the In_β2-AR (A) and Ac_β2-AR(B) binding site resulting from 200 ps restrained MD simulation.

Representative conformations of both (R,R’)- and (S,S’)-fenoterol isomers in the complexes with In_β2-AR and Ac_β2-ARwere extracted from 44 MD simulations. Inspection of the obtained 44 MD trajectories of the receptor-ligand complexes revealed much higher conformational stability of the (R,R’)-fenoterol when compared to the (S,S’)-fenoterol isomer. The flexibility of two ligands has been monitored in the form of RMSD autocorrelation function (Fig.S5). The observed averaged RMSD values are lower for (R,R’)fenoterol isomer when interacting with both In_β2-AR and Ac_β2-ARfor all monitored time intervals. The differences in RMSD values were ranging from about 0.01 nm to about 0.016 nm for 1 ns to 2.5 ns MD simulation time for In_β2-AR and from about 0.005 nm to about 0.02 nm for 1 ns to 2.5 ns MD simulation time for Ac_β2-ARrespectively, Fig.S5.

Online Resource Fig. 5Autocorrelation function of RMSD parameter values for two fenoterol isomers calculated for In_β2-AR(top) and Ac_β2-AR(bottom) complexes: black - (S,S’)-fenoterol, red - (R,R’)fenoterol.

The averaged values of measured distances between various ligand-receptor atom pairs are presented in Figures S6.

Online Resource Fig. 6The average measured distances between receptor-ligand atom pairs for In_β2-ARand Ac_β2-ARcomplexed with (R,R’)- and (S,S’)-fenoterol during 5 ns MD simulations with no restraints.