Signaling inside out

G protein-coupled receptors are a very large superfamily of transmembrane proteins that transmit signals into cells. They all have seven membrane-spanning helices, and most of them can activate a so-called G-protein that binds guanine nucleotides. I say “most” because there are some families that are sometimes grouped with GPCRs that seem to signal independently of G-proteins, for instance the Hedgehog receptor Smoothened and the PAQR receptors that are the subject of this post.

The early and mid-2000s saw the first structures of these receptors, at first in the “off” state with antagonists bound, and later in the “on” state with agonists bound. I have posted about some of these structures before. These structures revealed that when an agonist binds to the outside of the receptor, the outer halves of several of the helices in the bundle move slightly closer together, and this triggers, in a kind of “see-saw” manner, their outer halves to splay apart, creating the G-protein binding site. In particular, the outside ends of TM5(blue-green), TM7(red-orange), and to a lesser extent TM6 (light orange) move in, and the outer end of TM6 moves dramatically out, with TM5 and TM8 moving less. Here, the inactive "off" state is in rainbow colors, and the active "on" state is in light blue, and this figure is based on two structures of the beta2 adrenoceptor ((1) and (2)).

There is one peculiar family of “sorta-GPCRs” that are called PAQR receptors, which include receptors for the anti-obesity hormone adiponectin as well as a non-classical type of progesterone receptor. The adiponectin receptors are the weirdest, as they are “inside out”--the end of the bundle where ligands bind in “normal” GPCRs faces INTO the cell, and the end that normally binds the G-protein faces outward. This immediately raises the question of how the signal is transmitted. If the activation mechanism is similar, then the adiponectin would pry the helices apart on the outside, acting like the G-protein, and the intracellular ends would be pushed closer together, but this wouldn’t seem to help something bind inside to pass on the signal.

The first structures of the two adiponectin receptors were released in 2015(3), and confirmed the “inside out” structure. They showed that the equivalent of the normal ligand binding pocket contains a zinc ion and resembles an enzyme active site, and in fact these receptors had been suggested to be ceramidases. Ceramide is a special lipid that acts as a signal, and removing one fatty acid chain deactivates it--this happened when adiponectin was added in one experiment(4), though it is unclear whether other PAQR members signal in the same manner (5).

Now just this week, new structures were published--one being a new structure of AdipoR2, and one being a revised structure of AdipoR1 from the 2015 data, along with new biochemical data once again showing ceramidase activity(6). The new AdipoR2 structure resembles the 2015 structures of BOTH receptors, except for a very slight closure on the inside, and has a fatty acid bound near the zinc ion. The revised AdipoR1 structure is much more open on the inside than any of the other structures.

Comparing them to each other and to class A GPCRs, it seems to strongly imply that the 2015 structures and the new AdipoR2 structure are in fact close to the “on” state, while the revised AdipoR1 structure is “off”. Nearly all of the hallmarks of activation known from true GPCRs are there, but the relative magnitudes of the helix movements differ greatly from class A GPCRs. In particular, the outward movement of TM6 (here yellow) on the outside, “G-protein binding” (really adiponectin-binding) end is barely visible, yet the inward movement of TM5 (half green, half yellow) in the inside is far more dramatic than in any activated GPCR structures.

It’s unclear how much of the former is a family difference and how much is an artifact of the conditions, as the outward movement of TM6 was barely apparent in the first GPCR agonist structures too--it required adding a “nanobody”, a kind of antibody, to play the role of the G-protein and pry it open. So possibly, adiponectin would be required to do the same here. The large movement of TM5 is unlikely to be an artifact, however, as in the “off” state it opens up a path for ceramide to enter the bundle from the membrane.

This suggests a mechanism for how the receptor works: In the “off” state, ceramide enters from the membrane and binds loosely, but is too disordered relative to the zinc ion to be cleaved. When adiponectin binds, it inserts like a G-protein and wedges open the outside of the bundle, in turn contracting the inside of the bundle and preorganizing the active site for attack on the substrate. In this regard it is reminiscent of a set of bolt-cutters, using the lever action of the helices to pinch the substrate against the zinc ion “blade” to cut it. Note that in this highly simplified diagram, I am representing helices 5 and 6 together as the "lever", and neglecting to show that the helix that moves the farthest on one end of the bundle is not necessarily the same helix that moves the farthest on the other.

If the lever action were very rigid, this would mean that the adiponectin would need to at least partially unbind to let the “jaws” open again and release the product/bind new substrate, which means that each binding event triggers a single turnover. I consider this unlikely because one of the main purposes of catalysis is to amplify signals. Given the fact that the two ends of the bundle can be partly uncoupled in structures of class A GPCRs, it is likely that adiponectin binding only biases the active site toward the closed “on” state, such that it’s on part of the time as opposed to nearly always off and allowing multiple turnovers. It’s unclear why AdipoR2 seems to prefer the “on” state in crystals even without adiponectin, and whether this is biologically relevant. Biochemical results published with the new crystal structures actually show contrarily that AdipoR2 is about half as catalytically active as AdipoR1 in cells, both with and without ligand.

These structures suggest how evolution has gone in two directions from a similar starting point. It is likely that the first seven-transmembrane proteins were transporters/channels, as shown by for example by bacteriorhodopsin. These had inner and outer gates that had to be partly synchronized to move substrate across the membrane. Although it seems that the true GPCRs and PAQRs likely evolved convergently, given that there is no known common ancestor for the two (7), they both have ancestors in bacteria and attained a very similar fold. However, they have co-opted a similar conformational change, that may well be an inherent “normal mode” of this fold, to both transmit signals but in completely different ways.

(1) http://www.rcsb.org/pdb/explore/explore.do?structureId=2RH1
(2) http://www.rcsb.org/pdb/explore/explore.do?structureId=4LDE
(3) Tanabe H, Fujii Y, Okada-Iwabu M3, Iwabu M, Nakamura Y, Hosaka T, Motoyama K, Ikeda M, Wakiyama M, Terada T, Ohsawa N, Hato M, Ogasawara S, Hino T, Murata T, Iwata S, Hirata K, Kawano Y, Yamamoto M, Kimura-Someya T, Shirouzu M, Yamauchi T, Kadowaki T, Yokoyama S. Nature. 2015 Apr 16;520(7547):312-6.https://www.ncbi.nlm.nih.gov/pubmed/25855295
(4) Holland WL1, Miller RA, Wang ZV, Sun K, Barth BM, Bui HH, Davis KE, Bikman BT, Halberg N, Rutkowski JM, Wade MR, Tenorio VM, Kuo MS, Brozinick JT, Zhang BB, Birnbaum MJ, Summers SA, Scherer PE. Nat Med. 2011 Jan;17(1):55-63. https://www.ncbi.nlm.nih.gov/pubmed/21186369
(5) Chen JJ, Lin DJ, Liu MS, Chien EJ. Nat Med. 2011 Jan;17(1):55-63 https://www.ncbi.nlm.nih.gov/pubmed/24269742
(6) Vasiliauskaité-Brooks I, Sounier R, Rochaix P, Bellot G, Fortier M, Hoh F, De Colibus L, Bechara C, Saied EM, Arenz C, Leyrat C, Granier S. Nature. 2017 Mar 22. https://www.ncbi.nlm.nih.gov/pubmed/24269742
(7) Tang YT, Hu T, Arterburn M, Boyle B, Bright JM, Emtage PC, Funk WD. J Mol Evol. 2005 Sep;61(3):372-80. https://www.ncbi.nlm.nih.gov/pubmed/16044242