Molecular study of interactions of mu opioid receptor with biased and unbiased ligands by molecular dynamic simulation

VIETNAM NATIONAL UNIVERSITY, HANOIVIETNAM JAPAN UNIVERSITY TRAN KY THANH MOLECULAR STUDY OF INTERACTIONS OF MU-OPIOID RECEPTOR IN BINDING WITH BIASED AND UNBIASED LIGANDS BY MOLECULAR DY

VIETNAM NATIONAL UNIVERSITY, HANOI

VIETNAM JAPAN UNIVERSITY

TRAN KY THANH

MOLECULAR STUDY OF INTERACTIONS OF MU-OPIOID

RECEPTOR IN BINDING WITH BIASED

AND UNBIASED LIGANDS BY

MOLECULAR DYNAMIC SIMULATION

MASTER THESIS MASTER PROGRAM IN NANOTECHNOLOGY

Hanoi - 2019

VIETNAM NATIONAL UNIVERSITY, HANOI

VIETNAM JAPAN UNIVERSITY

TRAN KY THANH

MOLECULAR STUDY OF INTERACTIONS OF MU-OPIOID

RECEPTOR IN BINDING WITH BIASED

AND UNBIASED LIGANDS BY

MOLECULAR DYNAMIC SIMULATION

I kindly acknowledge the Japan International Cooperation Agency (JICA) and VNUVietnam Japan University for providing me the financial support to complete mystudy.

Beside, I am so greatful for my friends in VNU Vietnam Japan University and VNUKey Laboratory on Multiscale Simulation of Complex Systems for sharing expertise,valuable guidance and encouragement extended to me

Finally, I would like to express my sincere thanks to Mr Lam for all his patienceand kindness

Hanoi, 10 June 2019

Tran Ky Thanh

Contents i

LIST OF TABLES iii

LIST OF FIGURES iv

LIST OF ABBREVIATIONS vi

OVERVIEW 1

Chapter 1 INTRODUCTION 1

1.1 Opioid painkillers and their side effects 1

1.2 µ-opioid receptor 3

1.3 Biased signaling 6

1.4 Biased and unbiased ligand of mu opioid receptor 8

Chapter 2 METHODOLOGY 10

2.1 Molecular dynamics (MD) simulation 10

2.2 Docking by Autodock 4.2.6 and AutoDockTools 11

2.3 Interaction free energy estimation by MM-PBSA method 12

Chapter 3 RESULTS AND DISCUSSION 15

3.1 Setting up the simulation systems 15

3.1.1 Modeling 15

3.1.2 Docking 18

3.1.3 MD simulation 20

i

3.2 Rout mean square deviation (RMSD) 21

3.3 Cluster analysis 23

3.4 Binding free energy calculated by g_mmpbsa 24

3.5 Root mean square fluctuation (RMSF) 25

3.6 Binding sites 27

3.7 Conformations of mORs 29

3.8 Interaction with Gα protein 32

Chapter 4 CONCLUSIONS 34

REFERENCES 35

ii

LIST OF TABLES

Table 1.1: Properties of Opioid Receptors [34] 4

Table 3.1: Binding free energy in terms of Experiment and g_mmpbsa estimation 25

iii

LIST OF FIGURES

Figure 1.1: Countries consuming the most opioids [33] 2

Figure 1.2: Drug overdose deaths involving any opioid, number among all ages, by gender, 1999-2017 [5] 3

Figure 1.3: Percentage of drugs targeting GPCR family 5

Figure 1.4: Features of class A GPCR [31] 6

Figure 1.5: G proteins signal pathway [18] 7

Figure 1.6: β-arrestin signal pathway [3] 7

Figure 1.7: Illustration for the effects of biased and unbiased ligand when binding to protein [24] 9

Figure 2.1: Illustration for ligand-protein binding 12

Figure 2.2: Thermal dynamics cycle for calculation of ligand-protein binding energy 13 Figure 3.1: µOR (red) and Gα protein (blue) 15

Figure 3.4: The area of membrane during 50ns MD simulation 16

Figure 3.3: the membrane after 50ns MD simulation from top view (left) (red:cholesterol, purple: DPPC, lime: DOPE) and side view (right) 17

Figure 3.2: the membrane downloaded from MEMBUILDER server from top view (left) (red:cholesterol, purple: DPPC, lime: DOPE) and side view (right) 17

Figure 3.5: Structure of M6G (left) and TRV130 (right) 18

Figure 3.6: Histogram of clustering analysis of M6G 19

Figure 3.7: Histogram of cluster 1 (left) and cluster 5 (right) with respect to binding energy The red line represents the mean value of binding energy of that cluster 19 Figure 3.8: Histogram of clustering analysis of TRV130 20

Figure 3.9: Histogram of cluster 1 (left) and cluster 2 (right) with respect to binding energy The red line represents the mean value of binding energy of that cluster 20 Figure 3.10: The system including proteins, membrane, ligand, ions from top view (left) and side view (right) 21

Figure 3.11: RMSD of backbone µOR 22

Figure 3.12: RMSD of ligands 23

Figure 3.13: The cluster size diagrams of TRV130 (left) and M6G (right) system 24 Figure 3.14: Cluster ID during simulation time of TRV130 (left) and M6G (right) complex 24

Figure 3.15: RMSF of µOR 26

Figure 3.16: RMSF of Gα 26

Figure 3.17: The interaction between M6G (left) and TRV130 (right) with µOR (plot by LigPlot+ [21]) 27

Figure 3.18: Map of residues interacting with M6G (pink) and TRV130 (green) 28

Figure 3.19: Number of hydrogen bonds between M6G and µOR 29

iv

Figure 3.20: Number of hydrogen bonds between TRV130 and µOR 29Figure 3.21: The difference in structure of TM1 (a), TM5 (b), TM6 (c), TM7 (d) 30Figure 3.22: Secondary structure of µOR in M6G (left) and TRV130 (right)

complexes analyzed by DSSP [33] 31Figure 3.23: Conformation of µORs in binding with M6G (left) and TRV130 (right)(forcusing on TM6 and TM7, green-colored residue is PRO295 and GLY325) 32Figure 3.24: Difference in Helix 5 of Gα (other parts of Gα were hiden) 33

v

G protein coupled receptor kinaseMolecular dynamics

Lamarckian genetic algorithmMolecular mechanics Poisson−Boltzmann surface areaMorphine-6-glucuronide

Morphine-3-glucuronideRoot mean square deviationRoot mean square fluctuation

vi

Every year, millions of pain relief drugs prescriptions are written, and many of themare opioids Opioids are among the most strong pain relief in clinical use, but theiranalgesic effect is accompanied with many serious adverse effects, such asconstipation, nausea, vomiting, respiratory depression, and addiction Opioidsoverdose has been resposible for thousands of deaths every year These severeissues have been the driving force behind the development new effective painkillerswhich create less side effects

Opioids creates their effects mainly by binding to mu-opioid receptors (µOR) Theyare considered as unbiased µOR ligands which non-selectively activate µOR in boththe β-arrestin signaling pathway inducing side effects and the G-protein signalingpathway responsible for analgesia A novel drug, TRV130, is a biased µOR ligand

so activates G-protein signal transduction with less β-arrestin recruitment.Consequently, TRV130 provides higher pain relief and reduces side effects

Due to interaction with morphine and TRV130, µOR adopt different conformations,this lead to the different performance of these two drugs To elusidate themechanism of biased signaling, we discovered the conformational difference ofµOR in binding with morphine (unbiased ligand) and TRV130 (biased ligand) byperforming MD simulation

This research calculated the binding free energy of ligands and protein, revealed theinteraction of µOR with biased and unbiased ligands These results would bebeneficial for future research, the design of painkillers targeting µOR

Chapter 1 INTRODUCTION1.1 Opioid painkillers and their side effects.

Opioids have been used to relieve pain for thousands of years Opium is extracted

from the dried milky juice of a species of poppy, called Papaver somniferum.

During human history, opium is considered as “God‟s own medicine” and its tradeand use have been involved in many discreditable commercial, social, moral andpolitical events, for example, the Opium War Opium is the mixture alkaloids whosemajor components are morphine, codeine, and papaverine Whereas, the analgesiceffect of opium is mainly caused by morphine [26]

Nowadays, the use of opioids is different in each country, the United State andCanada are the two countries consuming the most opioids (Figure 1.1) [33] It isworth noting that there is a dramatic increase in prescribing opioids in manycountries According to WHO, in the year of 2016, nearly 34 million people usedopioids and that number for opiates is 19 millions [36] Furthermore, around 90% ofpatients have chronic pain use opioids The proportion of the population sufferingsubstance abuse disorder is 8% which is even more than the percentage of patientshaving chronic pain [1], [8]

Figure 1.1: Countries consuming the most opioids [33].

Opioids are highly effective analgesics used to alleviate acute, surgical and cancerpains, however, they have many side effects The side effects like nausea, vomiting,constipation, sedation, respiratory depression lead to the limitation of dose andeffectiveness of opioids [4] In addition, another common effect of opioids istolerance, the diminish of analgesic response to drug when opioids are usedrepeatedly and patients‟ body adapt with their presence [5] This side effect causes aneed of increasing dose, and then, higher dose results in more serious side effectsand supports the addiction A quarter of people taking opioids long-term becomeaddicted Drug deaths from opioids tend to rapidly increase From 1999 to 2017, inthe United State, the number of death due to opioid analgesic increased significantly

in all gender (Figure 1.2) [5] Besides, opioids may also create several less commonside effects, such as immunologic and hormonal dysfunction, increased painsensitivity, myoclonus, muscle rigidity and so on

Figure 1.2: Drug overdose deaths involving any opioid, number among all ages, by

gender, 1999-2017 [5]

In conclusion, despite their numerous side effects, opioids are very important foranalgesics Consequently, it is necessary to develop new opioid painkillers whichhave diminished side effects

1.2 µ-opioid receptor

The three major types of opioid receptors (OR) (µ, δ and κ) are members of seventransmembrane spanning receptors or G-protein coupled receptor (GPCRs) Theyare present throughout the body but they are in high concentration in the PAG, thelimbic system, the thalamus, the hypothalamus, medulla oblongata and thesubstaintia gelatinosa of the spinal cord Each type of OR are responsible fordifferent function (Table 1.1) [34] Whereas, µOR is the main target of manyopioids The binding of painkillers to µOR leads to clinical analgesics

Table 1.1: Properties of Opioid Receptors [34]

GPCR is one of the biggest family of protein It is a class of transmembrane

receptor coupling with G-protein The GCPRs have seven transmembrane helices

(TM), three intracellular loops (ICLs), three extracellular loops (ECLs), an

extracellular N-terminal and an intracellular C-terminal domain (Figure 4) so they

are also called 7 transmembrane domain receptors (7TMR) G proteins are

membrane protein binding to GDP (guanosine diphosphate) or GTP (guanosine

triphosphate); including three distinct subunits, Gα, Gβ, and Gγ Depending on the

nature of the Gα subunit, G proteins are divided into three major families, Gi, Gq,

and Gs, and each of them shows specific functions by influencing on different

intracellular effectors [22]

GPCR is divided into 6 classes, from A to F ORs are classified as class A which is

the largest class of GPCR family and is targeted by 94% of drugs of GPCR (Figure

1.3)

Figure 1.3: Percentage of drugs targeting GPCR familyMost of Class A of GPCRs shows these features (Figure 1.4) [31]:

• A disulfide bridge between the ECL2 and the upper part of TM3

• A palmitoylated cysteine in the C-terminus

• A highly conserved sequence homology of an Asp-Arg-Tyr motif on the

ICL2

• A sodium ions in the center of seven TMs

• Binding site of small ligands like morphine and TRV130 is between the

transmembrane domains of the receptor In contrast, the binding site ofpeptide and glycoprotein hormone receptors is located between the N-terminus, the extracellular loops and the upper part of the transmembranedomains

Figure 1.4: Features of class A GPCR [31]

1.3 Biased signaling

G proteins and β-arrestins are the most recognized signaling pathways of GPCRs.They present different biochemical and physiological functions

The G protein signaling pathway is displayed in Figure 1.5 When an agonist binds

to GPCR and changes the conformation of the receptor, G-protein is activated, Gasubunit releases guanosine diphosphate (GDP) and associates with guanosinetriphosphate (GTP) This leads to the dissociation of Gα from Gβγ subunits.Dissociated Gα and Gβγ subunit modulate downstream effector pathways To stopsignal, G-protein will be inactivated by the hydrolysis of Gα-GTP complex byGTPase, which convert GTP into GDP

β-arrestin signal pathway is a downstream signal pathway (Figure 1.6) After ligandbinding and G protein activation, G protein coupled receptor kinase (GRKs)phosphorylates the receptor, typically on its cytoplasmic tail β-arrestin recognizedthe phosphorylated sites and binds to the receptor β-arrestins mediate manyreceptor activities, including desensitization, downregulation, trafficking, andsignaling

Figure 1.5: G proteins signal pathway [18]

Figure 1.6: β-arrestin signal pathway [3]

After binding the GPCRs, most agonists are thought to equally activate both Gprotein and β-arrestins signaling pathways However, recently, a novel concept,

biased agonism, has emerged, in which biased agonists are able to selectively

activate the signaling pathway leading to the desired effects but not the signalingpathway causing adverse effects This concept enables to develop selective drugswith higher efficacy and reduced side effects [29] Several biased ligands havedemonstrated their efficient treatment and safety in clinical trials [28] Therefore,studying GPCR biased signaling may create a new generation of drugs

1.4 Biased and unbiased ligand of mu opioid receptor

Morphine is an unbiased ligand, signaling both the GPCR signaling pathway tocreate analgesia, and the β-arrestin signaling pathway responsible for side effects.[24] TRV130 is a biased ligand of µOR, it activates the GPCR signal transductionwith less β-arrestin recruitment [6], [24] TRV130 has been evaluated in clinical trialfor severe acute pain treatment Compared with morphine, TRV130 provides similaranalgesia but causes less adverse effects [6], [30], [35] (Figure 1.7) In addition,another compound, PZM21, was discovered by computational modeling andstructure-based screening Similar to TRV130, PZM21 showed higher analgesia,reduced adverse effects than morphine in preclinical trial

The different performance between unbiased and biased ligands might result fromdifferences in binding conformations of ligands and receptor Therefore, it isimportant to understand the binding conformations of µOR with both biased andunbiased ligands to determine the important differences in µOR and ligandsstructures leading to the different signal pathways activations

Figure 1.7: Illustration for the effects of biased and unbiased ligand when binding to

protein [24]

Chapter 2 METHODOLOGY2.1 Molecular dynamics (MD) simulation

Newton‟s equation of motion is solved for a system of N atoms in MD simulation:

(2.1)Where mi and ri are mass and coordinate of ith atom

The formula for the forces are the negative derivatives of a potential function V (r1;

r2;…; rN):

(2.2)

In each small time step, the equations are solved simultaneously Following the time,the system remains at a required temperature and pressure, the outputs (coordinates,force, velocity, …) are written regularly at a specific time The trajectory of thesystem is represented as the coordinates, which is a function of time Generally, thesystem will become equilibrium after a period of time Many macroscopic propertiescan be extracted from the average values of the equilibrium trajectory in the outputfile

Here is some limitation of MD simulation method [2]:

• The simulations are classical: Using Newton‟s equation of motion

automatically means that the motion of atoms is described by classicalmechanics For most atoms at normal temperatures, this is acceptable, except

in some cases For example, hydrogen atoms' motion may be similar toprotons motion which can have quantum mechanical (QM) characteristic andclassical mechanics can not treat properly this case

• Force fields are approximate: Force fields includes a set of potential

equations and their parameters provide the forces These functions are used

to create the potential energy and the force

• The force field is pair-additive

• Electrons always remain in their ground state

• Long-range interactions are cut off

• Boundary conditions are unnatural: when the system is small, there is a lot of

undesirable interacting area with the vacuum Periodic boundary conditionsare used to avoid real phase boundaries

GROMACS is a free program which performs molecular dynamics simulations andenergy minimization [2]

2.2 Docking by Autodock 4.2.6 and AutoDockTools

Automated docking is the prediction how small molecules, such as drugs andsubstrates, bind to a 3D structure biomolecular [17]

AutoDock is an automated docking tool which has shown an effective ability ofquickly and accurately predicting bound conformations and calculating theirbinding energies by a semiempirical free energy force field Autodock can searchthe large conformational space available to a ligand around a protein by using agrid-based method which rapidly evaluates the binding energy of trialconformations This means that the target area macromolecule is located in a gridand each grid point is a probe atom whose the interaction energy with target protein

is computed During the docking, this grid of energies may be used as a lookup table[17]

Lamarckian genetic algorithm is primarily used in the conformational searchingmethods At first, a number of trial conformations are generated and, from them,successive conformations are chosen and next generations are created by mutatingthese individuals, exchanging conformational parameters, and competing in asimilar way with the biological evolution, finally, individuals with lowest bindingenergy are selected In the “Lamarckian” approach, each conformation is able to

search their local conformational space, find local minima, and then pass thisfeature to next generations AutoDock4 also provides other search methods such asSimulated annealing and a traditional genetic algorithm [17].

AutoDockTools is an effective graphical user interface tool for preparing coordinate,designing experiment and analyzing the results AutoDockTools help users to formatinput molecule files, with a set of selection from protonation, calculating charges tospecifying rotatable bonds in the ligand and the protein In addition, AutoDockToolsusers can simply design and prepare the docking experiments by specifying theactive site and determining visually the volume of space searched in the dockingsimulation, specifying search parameters and launching docking calculations.Finally, AutoDockTools includes a variety of novel methods for clustering,displaying, and analyzing the results of docking experiments [17]

2.3 Interaction free energy estimation by MM-PBSA method

Molecular mechanics Poisson−Boltzmann surface area (MM-PBSA) approach hasbeen widely used to compute interaction energies, especially, for biomolecularcomplexes In combination with molecular dynamics (MD) simulations, this method

is also able to consider conformational fluctuations and entropic contribution intothe binding energy [20]

Generally, the binding free energy is calculated by the equation below [27]:

(2.3)Where Gcomplex is the total free energy of the protein-ligand complex and Gprotein

and Gligand are total free energies of the isolated protein and ligand in solvent,respectively (Figure 2.1)

Figure 2.1: Illustration for ligand-protein binding

However, the solvent-solvent interactions would mainly contribute to the energyand the fluctuations in total energy would be an order of magnitude larger thanbinding energy To avoid the inordinate amount of time to converge, it is moreefficient to divide up the calculation according to the thermodynamic cycle inFigure 2.2 [27].

Figure 2.2: Thermal dynamics cycle for calculation of ligand-protein binding energyBased on the above thermal dynamics cycle, the solvation binding free energy

Gbind,solv is:

Solvation free energies are calculated by either solving the linearized PoissonBoltzmann for each of the three states (this gives the electrostatic contribution to thesolvation free energy) and adding an empirical term for hydrophobic contributions[27]:

(2.5)

Gvacuum is obtained by calculating the average interaction energy between receptorand ligand and taking the entropy change upon binding into account if necessary[27]

temperature,

analysis

In practice, entropy contributions can be neglected in case of a comparison of states

of similar entropy, such as two ligands binding to the same protein The reason forthis is that normal mode analysis calculations are computationally expensivecompared with MM-PBSA and its magnitude of standard error can significantlymake the result uncertain

The average interaction energies of receptor and ligand are usually obtained byperforming calculations on an ensemble of uncorrelated snapshots collected from

MD trajectory These structures have to come from the equilibrated MD simulation[27]