Structural insights into adiponectin receptors suggest ceramidase activity

Abstract

Adiponectin receptors (ADIPORs) are integral membrane proteins that control glucose and lipid metabolism by mediating, at least in part, a cellular ceramidase activity that catalyses the hydrolysis of ceramide to produce sphingosine and a free fatty acid (FFA). The crystal structures of the two receptor subtypes, ADIPOR1 and ADIPOR2, show a similar overall seven-transmembrane-domain architecture with large unoccupied cavities and a zinc binding site within the seven transmembrane domain. However, the molecular mechanisms by which ADIPORs function are not known. Here we describe the crystal structure of ADIPOR2 bound to a FFA molecule and show that ADIPOR2 possesses intrinsic basal ceramidase activity that is enhanced by adiponectin. We also identify a ceramide binding pose and propose a possible mechanism for the hydrolytic activity of ADIPOR2 using computational approaches. In molecular dynamics simulations, the side chains of residues coordinating the zinc rearrange quickly to promote the nucleophilic attack of a zinc-bound hydroxide ion onto the ceramide amide carbonyl. Furthermore, we present a revised ADIPOR1 crystal structure exhibiting a seven-transmembrane-domain architecture that is clearly distinct from that of ADIPOR2. In this structure, no FFA is observed and the ceramide binding pocket and putative zinc catalytic site are exposed to the inner membrane leaflet. ADIPOR1 also possesses intrinsic ceramidase activity, so we suspect that the two distinct structures may represent key steps in the enzymatic activity of ADIPORs. The ceramidase activity is low, however, and further studies will be required to characterize fully the enzymatic parameters and substrate specificity of ADIPORs. These insights into ADIPOR function will enable the structure-based design of potent modulators of these clinically relevant enzymes.

Extended Data Figure 1. Comparison of the three ADIPOR2 structures.

(a) Comparison of the original (top) and revised (bottom) ADIPOR2 crystal structures. The modified sections and the additional molecules modeled in the revised structures are highlighted in red. (b) Overall view of ADIPOR2-scFv crystal structures from within the membrane plane. The heavy and light chain variable region (V_H and V_L) are colored in dark and light grey, respectively. Oleic acid (FA C18:1) is shown as sticks with C atoms displayed as spheres and coloured according to element: carbon, blue; oxygen, red. (c) 2Fo-Fc (dark grey) and Fo-Fc (light grey) density maps used to position oleic acid contoured at 1σ and 2.5σ, respectively. The density provided sufficient features to reliably position a free fatty acid in all structures. However, we could not make a clear distinction between an oleate (C18:1) and a stearate (C18:0) but decided to model an oleate because it is present in greater amount than stearate in insect cells and statistics were marginally better. (d) Hydrophobic binding pocket of the oleic acid within ADIPOR2 displayed as transparent blue surface. Residues forming the pocket are shown as sticks. (e) 2Fo-Fc electron density around the zinc binding site contoured at 1σ in S1, S2 and S3 crystal structures viewed from the intracellular side. The electron density reveals distinct positions of the carboxylic acid moiety and of the tentatively assigned water molecules resulting in the different apparent coordination geometries of the zinc ion. S1, S2 and S3 crystal structures are shown as cartoons and coloured in dark yellow, light blue and pink, respectively. The zinc ion is represented as an orange sphere. Residues participating in zinc coordination and carboxylic acid moiety are shown as sticks with oxygen and nitrogen atoms coloured in red and dark blue, respectively. Oleic acid is shown as sticks with C atoms displayed as spheres and coloured according to element: carbon, blue; oxygen, red. Water molecules are shown as red spheres.

Extended Data Figure 2. Features of the ADIPOR2 continuous cavity.

(a) Extra electron density (2Fo-Fc at 1σ) in the tunnel between TM5 and TM6 assigned to monoolein (rac-Glycerol 1-monooleate) as it is the most concentrated component in the crystallization sample and most likely binds this region with its oleate C18:1 moiety. We cannot rule out that the density originates from another molecule containing a long aliphatic chain. In both cases, the extra density suggests that this opening may play a role in ADIPOR2 function. Occupancy of the glycerol moiety and of the first four carbons from the ester group were however set to 0 during further refinement in absence of a significant electron density as indicated by the 2Fo-Fc map contoured at 1σ. Monoolein and oleic acid are shown as sticks with C atoms displayed as spheres and coloured according to element: carbon, yellow and blue, respectively for monoolein and oleic acid; oxygen, red. Tentatively assigned water molecules (red spheres) in an extra pocket close to the zinc site (b), and in the intracellular cavity and the intracellular opening (c). The extra pocket close to the zinc site is separated from the FFA molecule by F351^TM7 and I223^TM3 and is filled with water molecules. This cavity might be a reservoir of water molecules for the hydrolytic activity but such an hypothesis remains to be demonstrated. The intracellular pocket is split into the intracellular cavity (blue broken line) and the intracellular opening (yellow broken line) by residues D117^Nter and Y205^ICL1. Computational studies suggest that the sphingosine diffuses out from the receptor through the intracellular opening. The role of the intracellular cavity is hard to predict at this time. (d) Residues forming the intracellular cavity (blue) and opening (yellow).

Extended Data Figure 3. Biochemical analyses of ceramide binding to ADIPOR1 and ADIPOR2 and sphingosine formation.

A specific fluorescent signal for ADIPOR2 (a) or ADIPOR1 (b) incubated with NBD-ceramide was observed by FSEC at the elution volume of the receptor (blue broken line). In both cases, the two main peaks in the SEC absorbance traces could correspond to at least two receptor populations but only the one corresponding to the smaller peak on the right is able to bind to NBD-ceramide suggesting that part of the protein is either not functional or in a conformation which cannot accommodate the substrate. The presented figure is representative of two experiments performed with two independent receptor preparations. (c, d) Detected sphingosine in LC-MS analysis revealed ADIPOR1 ceramidase activity and adiponectin stimulation (around twenty five fold increase over basal). Relative sphingosine (c) and detected sphingosine (d) values are represented as the mean ± S.D. of three independent measurements. (e) Mass spectrum for the extracted ion peak (retention time of 2.27 min) of ADIPOR2 with ceramide-C24 sample, and the D-erythro-sphingosine (d18:1) standard sample. (f) Representative LC-MS analysis with an extracted ion chromatogram (m/z from 299.7 to 300.7) of the D-erythro-sphingosine (d18:1) standard sample (left panel) and of ceramide-C24 with N-dodecyl-β-d-maltopyranoside (DDM) and cholesteryl hemisuccinate (CHS) sample (right panel) in which no signals for sphingosine were detected. (g) Representative LC-MS analysis with an extracted ion chromatogram (m/z from 299.7 to 300.7) revealing the formation of the sphingoid base sphingosine (m/z=300.3, retention time of 2.26 min) from the enzymatic reaction between ADIPOR2 and ceramides of different chain length: Ceramide-C6, ceramide-C18 and ceramide-C24. The bottom panel represents the ADIPOR2 untreated samples. (h) Representative LC-MS analysis with an extracted ion chromatogram (m/z from 299.7 to 300.7) revealing the formation of the sphingoid base sphingosine (m/z=300.3, retention time of 2.27 min) from the enzymatic reaction between ADIPOR1 and ceramides of different chain length: ceramide-C6, ceramide-C18 and ceramide-C24. The bottom panel represents the ADIPOR1 untreated samples.

Extended Data Figure 4. Molecular dynamics simulations of ADIPOR2 substrate and products complexes.

(a) Comparison of the S1 crystal structure (5lx9) with the starting model used for MDS of the fatty acid and sphingosine system. A zoom of the active site is shown in inset. Calculated RMSD and distances between indicated residues and zinc during MDS performed with the sphingosine and the FA (C16:0) (b) or with the sphingosine and FA (C18:1) (d). In both cases, the sphingosine leaves the active site within the time scale of the MDS and moves towards the cytoplasm. The zinc coordination sphere remains as observed in the crystal structures during the C16:0 MDS (i.e. H202, H348, and H352 interact with the zinc ion), while in the C18:1 MDS some differences are observed. In the first 150ns, S198 interacts with the zinc along with H202, H348, and H352, while D219 is involved in a salt bridge with the sphingosine amine. After 150 ns, the D219/sphingosine interaction is broken and D219 replaces S198 in the zinc coordination sphere. These differences likely arise from the destabilization of the active site by the FA and sphingosine, which most probably requires longer time scales to return to equilibrium, as well as the strong tendency of MDS to remain stuck in local energy minima. (c) Snapshots of the active site extracted from MDS in the presence of the sphingosine and the FA (C16:0) at different times showing that the zinc binding site remained in the configuration observed in the crystal structures (except that the zinc adopted an octahedral geometry by interacting with 3 water molecules). (e) Snapshots of the active site extracted from MDS in the presence of the sphingosine and the FA (C18:1) at different times. In both C16:0 and C18:1 trajectories, the FA carboxylate forms a salt bridge with R278 side chain, which is also observed in the S2 structure. (f) Snapshots of the fatty acid and sphingosine taken every 10 ns along of 460 ns MDS trajectory, highlighting the movements of the sphingosine inside the receptor. The fatty acid and sphingosine are represented as blue and green lines, respectively. The zinc atom is shown as an orange sphere and the receptor is shown in cartoon representation. The carboxylic acid carbon (C₁) of the fatty acid and the nitrogen atom of the sphingosine are shown in spheres and colored using blue-to-orange and green-to-red gradients to help visualize their motion over simulation time. (g) C16:0 ceramide top scoring docking pose reminiscent of the C18:1 pose. Calculated RMSD and distances between indicated residues and zinc (h) during MDS performed with the C16:0 ceramide revealing a behavior similar to C18:1 ceramide. Inset in (h) highlights changes happening at the very beginning of the MDS. Because only the ligand is flexible during docking, and ADIPOR2 receptor was crystallized in a state that corresponds to the product state of the reaction (step four in the proposed mechanism), our interpretation is that the initial relaxation of the system represents the structural adaptation of the receptor to the presence of the substrate (induced fit back to step 1). We suspect that the movements of the receptor are fast because of the presence of the substrate in the binding pocket which is in the product state conformation thus constituting a perturbation of the system. (i) Snapshots of the active site extracted from MDS in the presence of C16:0 ceramide at different simulation times. The C16:0 ceramide and zinc-coordinating residues sampled conformations similar to what was observed for C18:1 ceramide. At late time points in the simulation, the ceramide moved slightly away from the zinc binding site, a likely consequence of the inability of MDS to simulate the destruction and creation of covalent bonds.

Extended Data Fig. 5. Proposed catalytic mechanism for ADIPOR2 ceramidase activity and docking calculations.

(a) Based on the interactions with ceramide and conformational changes of the zinc active site observed in MDS, as well as the zinc-coordinated FFA carboxyl oxygen seen in the crystal structures, we propose a general acid-base catalysis mechanism for the hydrolysis of the amide bond by ADIPOR2. In this mechanism, which is similar to what was proposed for neutral ceramidase (ref. in the main text), the zinc ion activates a water molecule for nucleophilic attack of the amide carbon (1). Y220^TM3, R278^TM5 and Y328^TM6 side chains polarize the amide carbonyl and stabilize the oxyanion formed in the tetrahedral transition state (2). H348^TM7 serves as a general base for proton extraction from water (1) and subsequently acts as a general acid to transfer this proton to the nitrogen of ceramide during, or immediately after amide bond cleavage (3). The active site rearranges following the hydrolysis reaction to yield the product state-associated zinc coordination sphere observed in the crystal structures and MDS (4). In this study, we decided to perform docking and simulations with ceramides and FFA presenting two different acyl chain lengths (C16:0, C18:1) as (i) we anticipated that chain length may not have a major impact on the observed mechanism and (ii) to compare our results with the study of ceramide binding to neutral ceramidase (ref . in the main text), in which docking calculations were performed with C16:0. (b) The top scoring C18:1 FA docking pose obtained using PLANTS (shown as green sticks) is superimposed to the crystallographically observed FA taken from the revised ADIPOR2 structure (5lwy) (blue balls and sticks representation). (c) Comparison of the top scoring C18:1 ceramide docking poses obtained using three different docking programs (PLANTS, Patchdock/Firedock and Glide). The ligands are shown as sticks with hydrogens omitted for clarity. The top scoring pose from PLANTS and Patchdock/Firedock are very similar, while the glide pose is slightly shifted towards the cytoplasm and a significant portion of the sphingosine moiety is exposed to the cytoplasm. In all three cases, the ceramide carbonyl contacts Y328 side chain. (d) Comparison of the C18:1 and C16:0 ceramide top scoring docking poses obtained using PLANTS. The ADIPOR2 receptor is shown as semi-transparent cartoon and surface, and the insets highlight the position of the sphingosine moiety relative to the intracellular surface of the receptor.

Extended Data Fig. 6. The putative catalytic residues and the substrate binding pocket are highly conserved in the PAQR family.

(a) View of conserved residues around the zinc ion (orange sphere) from the intracellular side. The evolutionary analysis performed by Consurf server reveals that residues H202, H348, H352, and D219 (shown as sticks) coordinating the zinc ion in ADIPOR2 are strictly conserved in the entire PAQR family. S198 potentially involved in ceramide hydrolysis is also strictly conserved within the human members of the PAQR family. The receptor is shown as cartoon and coloured using the Consurf colour scale. Oxygen and nitrogen atoms are coloured in red and dark blue, respectively. (b) The conservation of the internal cavity within PAQR family viewed from within the membrane in two orientations obtained by a 180° rotation. The cavity is represented in surface (cavity mode 1) and coloured using the Consurf colour scale. These data strongly suggest that all members of the PAQR family may have a ceramidase activity. (c) Sequence alignment of the 11 members of the PAQR family coloured using the Consurf colour scale. It is important to note here that ADIPORs also share some homology with alkaline ceramidases further reinforcing the experimental evidence found in this study.

Extended Data Figure 7. Corrected electron density and ADIPOR1 TM5 positions and crystal lattice packing of ADIPOR1-scFv and ADIPOR2-scFv.

2Fo-Fc (blue mesh) and Fo-Fc (green mesh) electron density maps around TM5 are contoured at 1 σ and 2.5 σ, respectively, in the initial ADIPOR1 structure (PBD code 3WXV) fetched from the Electron Density Server (a), the ADIPOR1 structure after modeling in strong positive difference map peaks at ~13 and ~6 σ with two sulfate ions, respectively (b) and the final revised ADIPOR1 structure (c). ADIPOR1 is shown as cartoon and coloured in light grey with TM5 highlighted in black. The zinc ion is shown as an orange sphere. Lattice packing of ADIPOR2-scFv crystals viewed within the membrane plane (d, e) and from the extracellular side (f). ADIPOR2 TM5 (red) does not make any crystal contacts neither with the bound scFv nor with the symmetry related molecules. Lattice packing of ADIPOR1-scFv crystals viewed within the membrane plane (g, h) and from the extracellular side (i). TM5 (red) of ADIPOR1 (R1-a) makes contact with the TM1 and N-terminal short helix (helix 0) (both in blue) from another symmetry related ADIPOR1 molecule (R1-b). (j, k) Closer view of the interaction of ADIPOR1 TM5 with TM1 and helix 0 from symmetry related ADIPOR1 molecule. At the top, TM5 is stabilized by hydrophobic contact between I287 and F125 of the symmetry related helix 0 as well as hydrogen bonds between the main chain carbonyl of A288 in TM5 and R122 in helix 0 as indicated by the black dashed line (j). At the bottom of TM5, Q265 interacts with R158 of the symmetry related receptor molecule intracellular loop 1 as indicated by the black dashed line. In addition, F271, L272 and L276 make hydrophobic contacts with symmetry related TM1 residues L157, I153, F150 and L149. The interacting residues are displayed as sticks and coloured in green for TM5 and yellow for TM1. ADIPOR1, ADIPOR2, V_H and V_L are coloured in pale green, wheat, dark grey and light grey, respectively. In both ADIPOR1-scFv and ADIPOR2-scFv, scFv molecules are contributing the most to the crystal lattice formation. Regarding the contacts between just the receptor molecules, in ADIPOR2 the packing is also mediated by TM4 of two symmetry related molecules.

Extended Data Figure 8. A broken N-terminus-TM5 interaction between the close (ADIPOR2) and open (ADIPOR1) structures.

The polar interaction between the R275^TM5 and D117^N-term of ADIPOR2 (a) is broken in the opened ADIPOR1 structure with the corresponding R264^TM5 shifted away and the D106^N-term side chain repositioned to interact with the Y194 (b). ADIPOR1 and ADIPOR2 are shown as cartoons and coloured in light green and dark yellow, respectively. R275, D117, R264, D106 and Y194 are shown as sticks. The zinc ion is represented as an orange sphere. Residues coordinating the zinc ion are shown as lines. Oxygen and nitrogen atoms are coloured in red and dark blue, respectively.

Figure 1. Crystal structure of ADIPOR2-scFv bound to a fatty acid.

(a) Overall view of ADIPOR2-scFv crystal structure at 2.4 Å from within the membrane plane. Oleic acid (FA C18:1) is shown as balls and sticks. (b) 2Fo-Fc (dark grey) and Fo-Fc (light grey) density maps used to position oleic acid. (c) Hydrophobic binding pocket of the oleic acid within ADIPOR2 displayed as transparent blue surface. Side chains of residues forming the pocket are shown as sticks. (d) Arrangement of polar residues and tentatively assigned water molecules, represented as red spheres, around the carboxylic acid moiety and the zinc ion in ADIPOR2 crystal structures viewed from the intracellular side. The black dashed lines indicate polar contacts. In all panels, the zinc ion is represented as an orange sphere.

Figure 2. A continuous cavity in the ADIPOR2 structure.

The large internal cavity is shown as surface (cavity mode 1) and coloured according to the Eisenberg hydrophobicity classification. (a) The extra electron density (2Fo-Fc contoured at 1σ) is shown in dark blue. The intracellular cavity extruding towards the N-terminus domain/TM1/TM2 is contoured in blue. (b) Highlighted in yellow are the two openings accessible to solvent (TM opening/Intracellular opening, shown in insets as surface views). The zinc ion and oleic acid are represented and coloured as in Fig. 1.

Figure 3. Biochemical and computational analyses of ceramide hydrolysis.

(a) Fluorescent spectra revealing the binding of NBD-C18 ceramide to ADIPOR2 (green line), compared to a control GPCR (black line). (b) ADIPOR2 ceramidase specific activity with ceramide substrates presenting different lengths (from C6 to C24 ceramides). Relative sphingosine values are represented as the mean ± SD of three independent measurements. (c) Representative Michaelis Menten analysis of ADIPOR2 ceramidase activity (n of 3). Initial velocity values (μM.min^-1) are represented as the mean ± SD of three measurements. (d) Adiponectin increases twenty fold the basal ADIPOR2 ceramidase activity. Detected sphingosine values are represented as the mean ± SD of three independent measurements. (e) Surface view of the C18:1 ceramide top scoring docking pose in comparison with the FFA (S2 experimental structure). Calculated RMSD (f) and minimum distances between indicated residues and zinc (g) during MDS performed with the C18:1 ceramide. Inset in (g) highlights changes happening at very short time during MDS. (h) Snapshot of the active site extracted at 70 ns.

Figure 4. Comparison of ADIPOR1 and ADIPOR2 crystal structures.

ADIPOR1 and ADIPOR2 are superimposed and shown as light green and dark yellow, respectively, with views from the membrane (a) and the intracellular side (b). TM5 in ADIPOR1 structure is tilted by 20 degrees as indicated by the angle between the two black lines. The distances between α-carbons (spheres) are shown as dashed lines. The molecular surface of ADIPOR1 viewed from the membrane (c) and intracellular side (d) highlighting the accessibility of the zinc catalytic core in stark contrast to ADIPOR2 (e,f). The positions of TM5 and TM6 in ADIPOR1 and ADIPOR2 are highlighted in light green and dark yellow, respectively.