Allosteric modulation of dimeric GPR3 by ligands in the dimerization interface

Reviewer #1 (Public review):

Summary:

GPR3 is an orphan receptor that plays a crucial role in central nervous system development and cold-induced thermogenesis, with potential implications for treating neurodegenerative and metabolic diseases. Although previous structural studies of GPR3 have been reported, Qiu et al. presented both active and inactive structures of GPR3 in its dimeric form. Notably, they identified AF64394 as a negative allosteric modulator that binds at the dimerization interface. This interface, primarily formed by transmembrane helices TM5 and TM6, is significantly larger than the dimerization interfaces previously reported for class A GPCRs. The authors further elucidate GPR3's activation mechanism and propose that dimerization may serve as a regulatory feature of GPR3 function. Overall, the study is well-executed, and the conclusions are sound.

Strengths:

Reported a unique dimerization interface of GPR3 and identified AF64394 as a negative allosteric modulator that binds at the dimerization interface.

Weaknesses:

There are some minor issues in the figure presentation.

Reviewer #2 (Public review):

Qiu et al. present active and inactive state dimeric structures of GPR3 with and without the previously identified inverse agonist AF64394. The manuscript combines cryo-EM processing, mutagenesis studies, and live-cell cAMP measurements to provide insights into the mechanism of action of AF64394 as a negative allosteric modulator of GPR3. All resolved structures show the density of a presumably hydrophobic endogenous, co-purified ligand in the orthosteric receptor binding pocket, supporting previous publications by this and other groups that endogenous lipids are endogenous ligands of GPR3. However, the authors also show that none of the proposed endogenous lipids (e.g., oleoylethanolamide) are able to further increase cAMP in living cells in a GPR3-dependent manner when applied exogenously. These data are in contrast to previous studies, but are of interest to the field as they may suggest that GPR3 expressed in different cell types is already saturated by endogenous lipids.

The overall findings are novel and exciting. GPR3 has not previously been proposed to assemble into a homodimeric complex, and no information has been published on where AF64394 binds to the receptor. Several comparative analyses between GPR3 and its close relatives, GPR6 and GPR12, including live cell experiments with GPR3/6 chimera, provide intriguing mechanistic explanations for the different dimerisation behaviour and activity of AF64394 at this GPCR cluster.

The only weakness of the study is that the population shift towards homodimer induced by AF6439, as suggested by 2D classifications of purified GPR3, is not supported by live cell experiments. The fact that AF64394 reduces GPR3-mediated cAMP production in a concentration-dependent manner may also be due to mechanisms independent of homodimerisation. Therefore, a live cell assay that directly detects dimer formation and/or dissociation upon different stimuli would significantly strengthen the findings of Qiu et al.

Reviewer #3 (Public review):

Summary:

The manuscript by Qiu and co-workers describes the single-particle cryo-electron microscopy structures of various oligomeric states of the orphan GPCR, GPR3. It describes the monomeric and dimeric structure of a mutant of GPR3 with a modified G-protein complex (miniGs) and then builds on this work to attempt an inactive 'apo' dimer and an allosteric modulator (AF bound dimer structure, by using an ICL3 insertion and stabilizing FAB fragments.

In general, I'm supportive of the work done in this study, and it does indeed provide valuable insight into GPR3 function. It may be that dimerization of certain class A GPCRs may be a means of signalling regulation or perhaps even amplification. However, some of the interpretation of the single particle data needs some extra attention to strengthen the hypothesis presented in the manuscript.

Firstly, I want to thank the authors for providing the unfiltered half-maps and PDB models for careful assessment. During this review, I did my own post-processing of the half-maps and used the resultant maps for careful analysis of models.

So to begin, I understand that the authors didn't model any lipid in the binding orthosteric binding site in any of the maps, but it may be worthwhile to model something in there, as many readers only download coordinates and not the maps.

A more general point about all the maps. In no case were any focussed refinements carried out. As the point of this paper are some of the finer details between active and intermediate states and the effect of an allosteric modulator, masking out hypervariable portions of the structure and doing local Euler searches would most certainly provide richer insights of the details in GPR3 (especially as the BRIL:Fab structures are not of interest). And also, generally, no 3D-variability studies were performed to see if minor differences in, say, TM4/5/6 positions were due to large variation in the single particles or were a stable consensus position.

As for the PFK dimeric structure. It appears to be refined with C2 point group symmetry (which is not mentioned anywhere except in a tiny bit of text in a supplemental figure). Was this also calculated in C1 to assess if there is any difference in either GPR3 protomer? Also, how certain are the authors of the cholesterol positions at the bottom of TM4/5? At lower map thresholds in the PFK dimer structure, one of them appears to be continuous with the orthosteric lipid. It also appears that there are many unmodelled lipids in this structure, and only two were assigned as cholesterol. It appears that many of the unmodelled lipids are forming bridging connections between the GPR3 protomers. Also, it may be worthwhile to provide a table of the key interactions between the protomers (although I note that there was a figure highlighting them).

With the PFK monomer structure, there was weak density for the same cholesterol, which was not modelled in this one; perhaps some commentary on the authors' approach for deciding how to assign density would be helpful. It also appears that the refinement mask was probably a bit tight in this one (something that cryoSPARC is notorious for), and rerefining with a much looser mask around the TM domain may be helpful in resolving the inner lipid leaflet positions.

The Apo structure, I think, I have the most issues with. Firstly, it is not 'apo'. There is definitely unaccounted for density in the orthosteric site. Also, the structure definitely needs a bit more attention. Firstly, masking out the BRIL and FABs would be a good start in helping better resolve the TMD regions, and then even focussing on a single monomer to increase the map interpretability. My major problem here is that, if this is being called 'apo' and inactive, the map doesn't reflect this; also, the TM5/6 does not look to be in a fully inactive position. The map density (at least around one of the protomers) in this region looks to be poorly resolved, most likely due to averaging due to internal motion. I think some 3DVA is certainly warranted here to strengthen the hypothesis that they have solved an 'apo' inactive.

The AF (allosteric modulator) bound structure is of significantly better quality. But again, only AF is modelled, and no lipids are. How are the authors sure? Perhaps some focussed refinements (and changing the Euler Origin to centre it on the AF molecule could be a good start). To this reviewer, at least in one of the protomers, adjacent to the AF position, there is a density that looks very much like the allosteric modulator, so it could even be forming a bridging dimer. Also, some potential assignments of the lipids may enlighten some of the structure-activity relationship of this modulator, as it seems to make as many contacts with surrounding lipids as it does with TM4/5. Also, it may be worthwhile exploring carefully the 3DVA of this data. In our studies (Russel et al.), we noted that the orthosteric lipid appears to ratchet back-and-forth in concert with TM4/5 twisting. Perhaps in the AF bound structure, as it binds at the 'exit' site of the lipid, perhaps it is locking in a specific conformation.