Reconstruction of apo A2A receptor activation pathways reveal ligand-competent intermediates and state-dependent cholesterol hotspots

Abstract

G-protein coupled receptors (GPCRs) play a pivotal role in transmitting signals at the cellular level. Structural insights can be exploited to support GPCR structure-based drug discovery endeavours. Despite advances in GPCR crystallography, active state structures are scarce. Molecular dynamics (MD) simulations have been used to explore the conformational landscape of GPCRs. Efforts have been made to retrieve active state conformations starting from inactive structures, however to date this has not been possible without using an energy bias. Here, we reconstruct the activation pathways of the apo adenosine receptor (A2A), starting from an inactive conformation, by applying adaptive sampling MD combined with a goal-oriented scoring function. The reconstructed pathways reconcile well with experiments and help deepen our understanding of A2A regulatory mechanisms. Exploration of the apo conformational landscape of A2A reveals the existence of ligand-competent states, active intermediates and state-dependent cholesterol hotspots of relevance for drug discovery. To the best of our knowledge this is the first time an activation process has been elucidated for a GPCR starting from an inactive structure only, using a non-biased MD approach, opening avenues for the study of ligand binding to elusive yet pharmacologically relevant GPCR states.

Figure 1

Apo A2A activation landscape. (A) Density map of the apo A2A receptor structures sampled during the adaptive simulation plotted along the two descriptors characterising GPCR activation: the distance between R102^3.50 and E228^6.30 and the RMSD from the inactive position of residue Y288^7.53. The blue dot represents the crystal structure 5uig that was used as starting structure in the MD simulation. Cluster centroids (cl) of the six kinetic macrostates are projected with red markers and the numbering is assigned from the least populated to the more populated cluster. (B) Structures corresponding to the six macrostates identified by the MSM model. Specific residues and structural elements whose change in conformation characterize active, inactive and intermediates states in A2A are highlighted (pink, blue, red, cyan, green and orange for M1, M0, M5, M4, M2 and M3 respectively). These features include: the TM6 helix, the ionic lock residues (R102^3.50 and E228^6.30), Y197^5.58 of TM5 and Y288^7.53 of TM7.

Figure 2

Ligand-competent intermediate states. (A) Density map of the apo A2A receptor plotted along the two descriptors of GPCR activation and sampled during simulation: the distance between R102^3.50 and E228^6.30 and the RMSD from inactive of residue Y288^7.53. The blue dots represent the distances of some of the solved crystal structures of A2A in the Protein Data Bank: 0 = 3eml, 1 = 5g53, 2 = 4eiy, 3 = 2ydv, 4 = 5uig, 5 = 3qak, 6 = 5nm2, 7 = 3pwh, 8 = 3rfm, 9 = 4ug2, 10 = 6gdg, 11 = 5wf5, 12 = 2ydo, 13 = 5olg (see Table S2 in Supplementary Information for details on the considered crystals). Centroids of the six kinetic macrostates have been projected and identified by red markers. (B) Structures corresponding to the six macrostates identified by the MSM model are differently coloured (pink, blue, red, cyan, green and orange for M1, M0, M5, M4, M2 and M3 respectively) and aligned to the Cα atoms of the following crystal structures: M1, M5 aligned with agonist-bound 2ydo; M2 aligned with miniG_s-protein bound 5g53; M0 aligned with antagonist-bound 3pwh, M4, M3 aligned with antagonist-bound 3eml.

Figure 3

Activation pathways of the apo A2A receptor. Each macrostate is schematically represented by a coloured labelled circle mapped on the two descriptors of A2A activation. Percentages for each pathway flux are reported below the corresponding plot with the corresponding standard deviation. The thickness of the arrows correlates with the value of the related percentage. The thicker the arrow, the higher the value, and thus, the relevance of the pathway observed. (A) Net kinetic flux reconstructed from the MSM model built for the apo A2A receptor. The reconstructed activation pathway considers the transition from M4 (inactive-like state) to M2 (active-like state). (B) Net kinetic flux reconstructed from the MSM model built for the apo A2A receptor. The reconstructed activation pathway considers the transition from M5 (active intermediate) to M2 (active-like state).

Figure 4

Cholesterol hotspots. Cartoon representation of the A2A receptor showing cholesterol occupancy for each of the most populated macrostates: M2, M3, M4 and M5. (A) Mesh surfaces represent the hotspots where cholesterol has the higher occupancy. Each surface is colour-coded to correspond to the respective macrostate. (B) Transverse sectional view of the EC aspect of the A2A receptor. The hotspots corresponding to M2, M4 and M5 are shown in green, cyan and red respectively. (C) Transverse sectional view of the IC aspect of the A2A receptor. The hotspots corresponding to M2, M3, M4 and M5 are shown in green, orange, cyan and red respectively.

Figure 5

Allosteric compounds and lipids binding sites. Examples of allosteric compound and lipids binding to areas of the A2A receptor with high cholesterol occupancy. (A) Structure of BPTU allosteric compound (in magenta) bound to the TM2-TM3 EC cleft of P2Y1 (pdb code 4xnv). (B) Two molecules of cholesterol bound to the TM2-TM3-TM4 IC cleft of β2 adrenoreceptor (pdb code 5x7d). (C) Cholesterol hemisuccinate bound to TM6-TM7 EC cleft of P2Y1 (pdb code 4xnv).