PIP2: Another Player in Biased Signaling of G Protein-coupled Receptors?

Principal Investigator:
Wolfgang Wenzel

Institute of Nanotechnology, Karlsruhe Institute of Technology (KIT)

Local Project ID:

HPC Platform used:
SuperMUC-NG of LRZ

Date published:


The ability of a cell to react to extracellular stimuli is enabled by a complex protein machinery attached to the cell membrane. The core of this machinery consists of transmembrane proteins termed G protein-coupled receptors (GPCRs) and their associated G proteins and arrestins localized inside the cell. While G proteins mostly trigger only signaling pathways, arrestin can either desensitize and recycle the receptor or trigger diverse signaling pathways on its own. Due to their important roles in healthy and disease states, GPCRs are targeted by one third of all current drugs.

Over the last years it was discovered that GPCR signaling is modulated by diverse mechanisms. This modularity is termed biased signaling. One modulator of GPCR signaling is the membrane composition. Thereby, the biased signaling can result from different (co)localization of GPCRs, G proteins, and arrestin, from changes in the oligomerization state of the receptor and from differences in the GPCR conformational flexibility. The most important lipidic modulators are cholesterol [1] and acidic lipids, mainly phosphatidylinositol phosphates (PIPs). The latter were found to be essential for complex formation of β2-adrenoreceptor (AR) (a prominent GPCR responsive to adrenaline) with the kinase (enzyme responsible for GPCR phosphorylation) GRK5.

Moreover, acidic lipids stabilize the active state of the receptor and can influence signaling pathways by determining which G protein will preferentially bind the given GPCR. Moreover, arrestin is only able to internalize GPCRs in presence of PIPs[2]. Although binding sites on both, arrestins and on a range of GPCRs have been determined in the past, no molecular resolution information is available on PIP binding to the complex of a GPCR with arrestin so far.

In order to obtain this information and its possible impacts on the biased signaling of GPCRs by modulation of the relative conformation of the arrestin and the GPCR, here we have performed multiscaling molecular dynamics simulations of β-arrestin2 bound to β2-AR in membranes with and without PIP2.

Results and Methods

Sequential multiscaling molecular dynamics (MD) simulations spare computational resources by studying time consuming processes (like lipid binding to the protein complex or protein-protein association) using coarse-grained (CG) resolution. After conversion back to all-atom (AA) resolution using the tool backward [3] the simulations were continued atomistically, refining the lipid-protein and the specific protein-protein interactions.

In CG MD simulations group of atoms are bundled into so called beads or particles having average properties of the atoms they represent. The resulting speed-up amounts to the factor of 350 for the here applied methodology and simulation system. However, CG simulations cannot be as well parallelized as AA systems of the same size. In detail, we have used 2 SuperMUC-NG nodes (48 cores each) for CG and 13 nodes for AA simulations, producing ~340 ns and ~6.25 ns simulation time per hour, respectively.

All CG simulations were run for 10 µs and all AA simulations for at least 1 µs. All simulations have been performed using the well-established, high performing simulation engine GROMACS ( in single precision. GROMACS has only low requirements concerning I/O. The trajectories and other output files are written regularly (every ~20–30 minutes) by appending data to a few binary and text files. All I/O is done by the first MPI rank only. Typical simulations require 2 GB and 12 GB storage, for 10 µs of a CG and 1 µs of an AA simulation, respectively. However, the storage requirements can increase significantly if the output is written out more often, or if the storage of forces and velocities is needed. Additionally, we have performed extensive steered-MD simulations in which the arrestin was pulled away from the receptor at 0.2 m/s. The necessary increased box size lead to reduction of production to 3.5 ns per hour on 12 SuperMUC-NG nodes.

On-going Research / Outlook

By steered-MD, we have pulled diverse complexes (including different activation states of the receptor and diverse membrane composition) of β2-AR and arrestin apart, and recorded force-time curves and combined them with atomic-force-microscopy. The acquired results build a solid basis for further investigations of the regulation mechanisms of extracellular signal transmission to the cell. We were able to pinpoint residues that are especially important for both stabilisation of the bound states and during unbinding. This knowledge is of special pharmacological importance, because currently only the activation state of the receptor is targeted by pharmaceuticals. However, arrestin binding to the receptor competes with binding of its cognate G protein and causes receptor desensitization, thus opening new pathways for pharmacological intervention.

References and Links

[1] Pluhackova, K., et al., PLOS Computational Biology, 2016. 12(11): p. e1005169.
[2] Gaidarov, I., et al., The EMBO Journal, 1999. 18(4): p. 871-881.
[3] Wassenaar, T.A., et al., J. Chem. Theory Comput., 2014. 10(2): p. 676-690.


Kristyna Pluhackova2, Wolfgang Wenzel1 (PI)

1Institute of Nanotechnology, KIT, Karlsruhe
2Department of Biosystems Science and Engineering, ETH Zürich, Basel

Scientific Contact

Dr. Kristyna Pluhackova
now: Cluster of Excellence SimTech
Stuttgart Center for Simulation Science
University of Stuttgart
Universitätsstraße 32, D-70569 Stuttgart
e-mail: kristyna.pluhackova [@]

Local project ID: pr27wi

March 2021

Tags: LRZ KIT Life Sciences ETH Zürich