Lipids and Phosphorylation Conjointly Modulate Complex Formation of β2-Adrenergic Receptor and β-arrestin2

G protein-coupled receptors (GPCRs) are the largest class of human membrane proteins that bind extracellular ligands at their orthosteric binding pocket to transmit signals to the cell interior. Ligand binding evokes conformational changes in GPCRs that trigger the binding of intracellular interaction partners (G proteins, G protein kinases, and arrestins), which initiate diverse cellular responses. It has become increasingly evident that the preference of a GPCR for a certain intracellular interaction partner is modulated by a diverse range of factors, e.g., ligands or lipids embedding the transmembrane receptor. Here, by means of molecular dynamics simulations of the β2-adrenergic receptor and β-arrestin2, we study how membrane lipids and receptor phosphorylation regulate GPCR-arrestin complex conformation and dynamics. We find that phosphorylation drives the receptor’s intracellular loop 3 (ICL3) away from a native negatively charged membrane surface to interact with arrestin. If the receptor is embedded in a neutral membrane, the phosphorylated ICL3 attaches to the membrane surface, which widely opens the receptor core. This opening, which is similar to the opening in the G protein-bound state, weakens the binding of arrestin. The loss of binding specificity is manifested by shallower arrestin insertion into the receptor core and higher dynamics of the receptor-arrestin complex. Our results show that receptor phosphorylation and the local membrane composition cooperatively fine-tune GPCR-mediated signal transduction. Moreover, the results suggest that deeper understanding of complex GPCR regulation mechanisms is necessary to discover novel pathways of pharmacological intervention.


Simulation system
Simulation type and length Notes pβ2AR* in

Simulation conditions
In coarse-grained (CG) simulations, the simulation input parameters for the polarizable Martini2.2p 1 force field, including the PW water model 2 , were used as recommended for GROMACS5 and GROMACS2016 3 . In detail, the temperature of 310 K was controlled by the velocity rescale thermostat 4 . The pressure was coupled in a semiisotropic manner to 1 bar by the Parrinello-Rahman barostat 5, 6 , using 12 ps time constant and compressibility of 3e-4 bar -1 . The integration step amounted to 20 fs. The Verlet cut-off scheme 7 was used for neighbor search and the van der Waals forces were switched to zero at 1.1 nm over the full interaction length by the Potential-shift-Verlet modifier. PME 8 was utilized to describe the electrostatic interactions above the 1.1-nm cutoff, and the relative permittivity, ! , was 2.5.
In the atomistic simulations performed with the CHARMM36m force field 9-11 , a 2-fs integration step was used. The Verlet cut-off scheme 7 accomplished neighbor search with buffer tolerance of 0.005 kJ/mol/ps per atom. PME 8 was utilized to describe electrostatics above a 1.2-nm cutoff and the Potential-switch function switched the 12-6 Lennard-Jones potential to zero between 0.8 and 1.2 nm. The temperature of 310 K was controlled by the Nosé-Hoover 12 thermostat with the time constant of 0.5 ps. Parrinello-Rahman barostat 5, 6 with the time constant of 5 ps controlled the pressure of 1 bar in a semiisotropic manner using the compressibility of 4.5e-5 bar -1 . The center of mass motion of the system was linearly removed every 100 steps.
Parametrization of phosphorylated amino acids within the framework of Martini2.2 and Martini2.2p.
As atomistic reference data, 500-ns-long all-atom simulations using CHARMM36 force field of 27 individual phosphorylated residues including a backbone with neutral termini were performed separately for phosphoserine and phosphothreonine, each carrying two negative charges on the phosphate. The phosphorylated residues were solvated (1:100) by TIP3p water and neutralized by Na+ counterions.

Supplementary Figure 4:
Interactions of receptor C-terminus residues with all receptor residues evaluated as average distances between C-terminus residues and receptor residues over the simulation time and all simulations of a given type. Top two rows: C-terminus (rightmost) interacts with itself and with helix 8. The seemingly close contacts with TM1 result from their neighborhood in the protein structure and do not result from direct interactions. Bottom row: detailed view on the interactions of helix 8 and C-terminus in the nonphosphorylated state (β2AR*, left), phosphorylated state (pβ2AR*, middle), and their difference (β2AR*-pβ2AR*, right). The rightmost map is colored white to black (white colored residues are further apart in pβ2AR* than in β2AR* and black colored residues are closer in pβ2AR* than in β2AR*). All other maps are rainbow colored red to blue, for distances from 0 to 1 nm and larger. Data obtained from 7 pβ2AR* CM and 5 β2AR* CM independent MD simulations, first 500 ns were excluded from the analysis for equilibration purposes. The residues K251, R237 and K233, which were identified to bind phosphatidylinositol lipids 13 , are shown as large spheres and labeled in purple. The same residues and additionally the residues K327 and K325 (shown as small spheres and labeled in magenta) were experimentally shown to bind hexakisphosphate 14 . Residues at the C-edge which insert into the bilayer are shown as lines and labeled in grey. Additional residues identified here to attract DOPG are shown as sticks and labeled in black. The binding probability of 1 means that at least one DOPG lipid is located within 0.5 nm (a distance reflecting direct hydrogen bonding and hydrogen bonds mediated by a single water molecule) to the given residue for 100% of the analyzed simulation time (first 500 ns were removed from all simulations for equilibration purposes, N=7). intervals. Moreover, both S75 and F76 interact in 3 out of 7 intervals with Q229. In pβ2AR*/βarr2 CM, the following interactions of arrestin fingerloop residues with the receptor were observed: E67-R63 (7/17), D68-R63 (8/17), D70-K267 (9/17) and L74-I135 (6/17), K78-pS262 (10/17), R77-pS262 (6/17), Y64-P138 (6/17). Interestingly, R66 can stretch out to pS355 (9/17). In the pβ2AR*/βarr2 NM complex, residues of the fingerloop formed the following interactions: E67-R63 (5/15), R66-D130 (7/15), D68-R131 (5/15), K78-D234 (7/15), D70-K267 (6/15), D70-K270 (6/15), R66-D331 (5/15), R64-pS355 (6/15).
In accordance with the observed preference of the phosphorylated ICL3 to interact with the neutral membrane surface, the ICL3 of pβ2AR*/βarr2 NM interacts only in 2 out of 15 simulation intervals with arrestin, while in the charged membrane pS262 interacts in 13 out of 17 simulations strongly with arrestin. In the latter case, pS261 and/or pS262 are attracted by K78 and K158, R77, R148, K161, and K313 and repelled by D79, H156 and D70 on arrestin. Moreover, pS246 is in some simulations (5/17) attracted to K313 and K267 of ICL3 builds in 9/17 cases a salt bridge with D70 on arrestin.
The probabilities and the average forces of the individual residue pairs are collected in the Supplementary File Data Sheet 2.xlsx.
The Supplementary Movie Video 1.MP4 shows a self-assembly of the pβ2AR*/βarr2 complex at coarse-grained (CG) resolution using the Martini2.2p force-field. The ICL3 was removed from the CG model in order to avoid possible plugging of the receptor's intracellular binding pocket prior to arrestin's insertion into it. The final conformation resembles very well the complex of rhodopsin/arrestin-1 15 (shown as grey cartoon). Similar results were obtained from three independent spontaneous binding simulations starting from different orientations of arrestin relative to the receptor. The relative orientation of arrestin in the three endpoints of the 10 µs-long self-assembly simulations in terms of insertion depth d and arrestin rotation a amount to 4.05 nm and 17.3º, 3.91 nm and 22.3º, 4.27 nm and 8.3º, respectively. The position of the membrane is visualized in the background as white sticks and spheres. The pβ2AR* is shown as red ribbon, the two main phosphorylation sites on the Cterminus pS355 and pS356 are shown as yellow spheres. The arrestin is shown as blue ribbon with the fingerloop colored green.