Binding kinetics drive G protein subtype selectivity at the β1-adrenergic receptor

G protein-coupled receptors (GPCRs) bind to different G protein α-subtypes with varying degrees of selectivity. The mechanism by which GPCRs achieve this selectivity is still unclear. Using 13C methyl methionine and 19F NMR, we investigate the agonist-bound active state of β1AR and its ternary complexes with different G proteins in solution. We find the receptor in the ternary complexes adopts very similar conformations. In contrast, the full agonist-bound receptor active state assumes a conformation differing from previously characterised activation intermediates or from β1AR in ternary complexes. Assessing the kinetics of binding for the agonist-bound receptor with different G proteins, we find the increased affinity of β1AR for Gs results from its much faster association with the receptor. Consequently, we suggest a kinetic-driven selectivity gate between canonical and secondary coupling which arises from differential favourability of G protein binding to the agonist-bound receptor active state.

In general the manuscript is well-wriften and the experiments are carried out in a highly diligent manner (as always from this group).I only felt that I had to read through a lot of detail, where the authors describe their system, before gefting to the meat of the paper.I wonder if part of the basic characterizafion of their system can be moved to the supplementary informafion and only a short summary is presented in the main manuscript.Apart from this, I only see minor modificafions required before publicafion (it's rare that I have so few comments to a manuscript).

These points may require further aftenfion:
Which binding partner was used to produce the data shown in figure 1d?The KD, kon and koff values that were experimentally determined only appear as column diagrams in the main manuscript.The logarithmic scale makes it very difficult to read the exact values.Please include the found values in the figure in the main arficle.It's difficult to find the corresponding values in the supplement.
The authors make several statements on whether a binding event progresses via an induced fit or a conformafional selecfion mechanism.To me the data is not sufficient to exclude one or the other mechanism, although one might be more likely than the other.For example, the data indicates that the cytoplasmic binding cavity is not fully formed in the acfive state.The conclusion of the authors is that Gprotein binds to the parfially formed cavity through conformafional selecfion and then TM6 is driven outwards via an induced fit mechanism.However, there is no clear indicafion that TM6 might not visit the outward-oriented conformafion for a fracfion of the fime in the acfive state, rendering binding a mechanism enfirely based on conformafional selecfion.
A direct binding of G-proteins to the pre-acfive state is suggested, i.e. to β1AR-W bound to a parfial agonist.Couldn't the lowered on-rate equally be interpreted in such a way that binding sfill proceeds through the acfive state, since this state has a low populafion in the β1AR-W situafion?
A typo: In line 251, it should probably say …peak locafions THAT were reminiscent…, or …peak locafions reminiscent…

Reviewer #2 (Remarks to the Author):
This manuscript focuses on GPCRs, which are a significant class of membrane proteins involved in transmifting various signals across cell membranes, leading to diverse downstream signaling pathways affecfing physiological processes.GPCRs interact with heterotrimeric G proteins and are known to exhibit selecfive coupling with specific G protein subtypes.Understanding the mechanisms behind this selecfivity is essenfial for comprehending cellular signaling and developing GPCR-targeted therapeufics.
The study specifically focuses on the β1-adrenergic receptor (β1AR), which predominantly couples to Gs but also interacts with Gi.The well-structured design ufilized a set of complementary biophysical methods, allowing it to shed light on the structure, dynamics, and kinefics of β1AR.The study employs spectroscopic techniques, including NMR spectroscopy, to invesfigate the structural dynamics of GPCRs.NMR studies of GPCRs in their acfive state, coupled to different G proteins, are limited but crucial for a comprehensive understanding of G protein selecfivity.It is well-established that agonist binding to GPCRs induces structural changes in the receptor, influencing its acfivafion pathway.However, the exact conformafion of the acfive state is challenging to determine directly due to its inherent instability and dynamic nature.Previous research idenfified ligand-dependent changes in the equilibrium between inacfive and acfivafion-intermediate receptor states.
In this work, the researchers aimed to restore accessibility to the agonist-bound acfive state of β1AR by reverfing a stabilizing mutafion.Using this modified construct, they examined β1AR acfivafion and G protein selecfivity through NMR spectroscopy.The findings indicate that the acfive state resembles the conformafion found in ternary structures but requires addifional conformafional rearrangements to reach the solufion Gs ternary state.Furthermore, subtle conformafional differences between ternary complexes with different G proteins suggest receptor adaptability.
In addifion to NMR spectroscopy, the researchers employed bio-layer interferometry (BLI) to invesfigate the in vitro binding of G proteins to β1AR.This complementary approach provided insights into affinity and binding kinefics under condifions akin to NMR experiments.These binding assays confirmed that complex formafion with Gs occurs more rapidly, suggesfing a kinefic-driven selecfivity gate between canonical and secondary coupling, driven by the favorability of G protein binding to the receptor's acfive state.There are a couple of weaknessess in the experimental design as descripfion as well as the resulfing analysis and interpretafion of the data that requires parficular aftenfion to strenghthen the conclusions of the paper: (1) Regarding the in vitro biofinylafion step, have you conducted an analysis to assess the level of biofinylafion?This would provide a more accurate esfimate of the extent of non-specific binding of nonbiofinylated material during the BLI experiments.
(2) After capturing the receptor with biofin, did you implement a step to block the streptavidin (SA) surface to reduce the binding of the analyte (= G-proteins)?A clarificafion on the blocking procedure would be helpful to understand how you approached the reducfion of non-specific interacfions to allow for proper data interpretafion (3) Could you please provide more clarity on the preparafion of the reference channel?The experimental descripfion menfions the use of un-biofinylated avi-tagged receptor for the reference channel.However, it appears that a dissociafion phase in BLI buffer was employed to eliminate non-specific binding of unbiofinylated avi-tagged receptor.This leaves some confusion about the reference channel's role.Could you explain how it was prepared and its specific purpose in the experiment?
(4) In light of Figure 2e in the supplementary material, where the dissociafion phase drops below the zero response level, it seems indicafive of deposifion on the reference surface.Considering this, do you agree with my assessment that a properly prepared reference surface with an unrelated but biofinylated protein might be necessary to comprehensively assess the specificity of G-protein binding in your experiments?
(5) The experiments appear to have been conducted at a single concentrafion, which deviates from the standard procedure where dose-response experiments using mulfiple concentrafions are performed.Using mulfiple and varying concentrafions of G-proteins in your experiments would enable the esfimafion of the maximum binding signal Rmax, allowing for the determinafion of stoichiometry and a more accurate assessment of the level of non-specific binding.Could you discuss the rafionale for using a single concentrafion and whether future experiments should include dose-response studies ?(6) In the data analysis, you assumed monophasic associafion and dissociafion curves with local parallel fifting.Could you please elaborate on the basis for this assumpfion and whether there are other data or analyses that support this choice?It would be helpful to understand the reasoning behind this specific modeling approach and if there is supporfing evidence.
(7) How did you evaluate the structural integrity, folding, and funcfional acfivity of the analyte (Gprotein)?In other words, how did you ascertain the funcfional concentrafion of the analyte, or did you solely rely on the nominal protein concentrafion for your experiments ?What problems would you forsee with this approach and how would that impact the analysis?(8) In Figure 2d of the supplementary material, it's evident that there are notable discrepancies in the binding levels of the same G-protein to β1AR-E and β1AR-W constructs.These differences raise quesfions about the capture levels achieved for both constructs and the nominal Rmax values expected for a 1:1 interacfion in your BLI experiments assuming 100% ligand-binding competent material.Could you please provide a detailed explanafion of the capture levels for each construct and the nominal Rmax values?Addifionally, could you elaborate on how these values are reflected in the binding isotherm, shedding light on the observed binding behavior?What Rmax values have been computed as part of the data fifting approach and how do those values relate to the theorefical Rmax values ?
Addressing these quesfions should help to enhance the overall depth and clarity of your study's conclusions.I strongly believe that this would not only bolster the accuracy of your experimental data but also offer a more robust foundafion for your findings.This, in turn, would strengthen the validity of the conclusions drawn from your well-designed study, ulfimately advancing our understanding of the mechanisms underlying G-protein binding to β1AR.Such precision and clarity in your analysis would not only benefit the scienfific community but also potenfially guide further research in this crucial area of study

Reviewer #3 (Remarks to the Author):
Jones et al. present a thought-provoking study hypothesizing that the primary determinant for G protein subtype selecfivity at β<sub>1</sub>AR is the on-rate (k<sub>on</sub>) of G protein binding to the fullagonist acfivated receptor.Gs is the canonical G protein for the β<sub>1</sub>AR and thus likely to outcompete non-canonical G proteins in a cellular environment due to its faster on-rate.The authors show that sequence variafions of the C-terminal α5 helix of Gα are sufficient to modulate on-rates, while the off-rates (k<sub>off</sub>) of the tested G proteins were not significantly different.The authors further show that G protein coupling to the full agonist bound β<sub>1</sub>AR mainly follows a conformafional selecfion mechanism, followed by a somewhat minor induced fit component.G protein coupling to pre-acfive (e.g.parfial-agonist bound) β<sub>1</sub>AR, however, followed an induced fit mechanism leading to similar global conformafions of the receptor as seen in presence of the fullagonist.The global conformafion of the receptor was highly similar when in complex with different G protein subtypes with the most substanfial differences found in the TM7 and ICL4 regions, hinfing at a more producfive Gs:β<sub>1</sub>AR interacfion.For this study, the authors reverted the thermostabilizing mutafion E130W<sup>3.41</sup> in the in the previously used (pre-acfive) β<sub>1</sub>AR-W construct to generate the more acfive β<sub>1</sub>AR-E variant.Ligand-binding to both variants was characterised using a BRET-based in-cell ligand-binding assay, while G protein coupling was assessed using the TRUPATH assay.Biolayer interferometry (BLI) was employed to measure the affinity of mini-G variants of canonical and noncanonical Gα subunits for both β<sub>1</sub>AR-W and β<sub>1</sub>AR-E.Conformafional fingerprints of the different β<sub>1</sub>AR-W and β<sub>1</sub>AR-E complexes (apo, ligandbound, and ternary-complexes with mini-G's and a G protein mimicking nanobody Nb6B9) were obtained by <sup>13</sup>C-methyl methionine and <sup>19</sup>F NMR using endogenous or introduced probes.This study is logically structured and excepfionally well wriften.The NMR data presented is of high quality and sufficiently supports the claims made in this study.The combinafion of <sup>13</sup>Cmethyl methionine and <sup>19</sup>F NMR is highly complementary, allowing the authors to address specific quesfions either technique alone would not be able to cover.To my knowledge, this is the first fime the kinefics of interacfions between a GPCR and different G proteins have been studied using BLI while the KD of mini-Gs binding to β1AR seems somewhat comparable to previously published data (Nehme et al., 2017;Ref 48).The conclusions made seem to be supported by the data and the discussion sufficiently addresses the limitafions of the study.The research presented by Jones et al. is highly relevant and furthers the understanding of G protein subtype selecfivity and ternary complex formafion.I highly recommend this manuscript for publicafion.General comments/quesfions: 1.I would suggest a fitle that befter reflects the main findings of the study such as "Binding kinefics drive G protein subtype selecfivity at the β1-adrenergic receptor" 2. This study relies a great deal on kinefic data generated using biolayer interferometry (BLI).Raw data (i.e.BLI binding curves) is only shown for mini-Gs, and for a single concentrafion of the binding partner (2 μM) in Extended Figures 2e/f.However, these binding curves are not shown for either Nb6B9 or any of the other mini-G proteins analyzed.Given that the BLI results are crucial for this study, I would have expected to find all binding curves in a supplementary Figure .Furthermore, it is unclear if the BLI experiments were carried out with different binding partner concentrafions.3. I understand that Apyrase was added in NMR experiments to "mifigate variafions in GDP binding" but BLI experiments suggest it had pracfically no effect and Apyrase was thus omifted from BLI experiments.Why the discrepancy?Specific comments & suggesfions: Lines 157 -159: I fully agree with the authors assessment of M223 adopfing a pre-acfive conformafion in β<sub>1</sub>AR-W, compared to the acfive conformafion in β<sub>1</sub>AR-E.It appears that the Iso-bound M223-W resonance almost sits on a linear trajectory between the Apo M223-E and the Isobound M223-E resonance, which would further support the assumpfion that M223-W is a pre-acfive state.It would be interesfing to compare NMR spectra of parfial agonist-bound β<sub>1</sub>AR-E to see where M223 would sit with respect to this trajectory.Lines 166 -169: A recent publicafion (Chashmniam et al., 2021; ChemBioChem doi:10.1002/cbic.202000701)hypothesized that neighbour effects contribute more to the methyl <sup>13</sup>C of Methionines than internal side chain dihedral angles.According to their theory, the upfield chemical shift change of iso-bound M223 in the 13C dimension may also be a result of the F<sup>6.44</sup>shielding effect.Line 180: I'm not sure if I agree with "reduced conformafional variability".To me, the single (broadened) peak of M178 in the in the iso-bound β<sub>1</sub>AR-E spectrum suggests faster exchange rates compared to slow exchange in the iso-bound β1AR-W spectrum.What I find interesfing, is that the minor M178 peak in the iso-bound β<sub>1</sub>AR-W spectrum (Supp Fig 4a) shifts to the β<sub>1</sub>AR-E posifion when bound to iso/Nb6B9 (Supp Fig 4b).Furthermore, the iso/Nb6B9 spectrum seems to show a minor M178 peak near the major peak observed in the iso-bound β<sub>1</sub>AR-W spectrum (Supp Fig 4a).It may be helpful to menfion that agonist-bound X-ray structures of GPCRs often show TM6 in an inacfive (or inacfive-like conformafion).Interesfingly, the b36-m23 mutant used for crystallography (Ref 43) did not seem to harbor the E130W mutafion, but Warne et al. argued that their thermostabilizing mutafions resulted in preferenfial adopfion of the inacfive state regardless of the ligand pharmacology.Line 209: I understand that PDB 2Y03 is a "poor choice" for visualizing the acfive state as TM6 largely remains in an inacfive conformafion.It may be helpful to add an acfive (in complex with Gs) structure to visualize the extent of TM6 twisfing/solvent exposure upon acfivafion.Lines 211 -214: Why did the authors choose to do CPMG and not CEST experiments if ms-s fimescale exchange is consistent with large TM6 helical movements?Line 215: Could the authors comment on why BTFA was used for TM6 and TET was used for TM7 labelling and/or add a reference to previous tesfing?Lines 257 -259: Do the authors have a means to esfimate experimental errors of peak intensifies shown in Supp Fig 6a?For example, if one condifion was measured in duplicate, an average error can be determined and extrapolated.Line 265: Reference 59 refers to β<sub>2</sub>AR.Has this also been shown for β<sub>1</sub>AR?Lines 306 -308: How confident are the authors with the deconvolufion?A residual error analysis (as described by Kim et al., 2013;JACS doi:10.1021/ja404305k)may be helpful.Line 532: I would be more specific and use the term GPCRs instead of receptors.Lines 555 -557: The authors could employ a <sup>19</sup>F saturafion transfer experiment as previously carried out by Frei et al., 2022;Nat Commun, doi: 10.1038/s41467-020-14526-3) to probe if a low-populated acfive state is present in the parfial agonist-bound b1AR-E solufion ensemble.Line 572: A space is missing after "conformafion".Lines 572 -575: I suggest a more granular discussion of the authors findings in comparison with reference 27.The cryo-EM structures in reference 27 used heterotrimeric G proteins while the authors used mini-Ga subunits.The simplified system using only the mini-Ga subunits may very well indicate similar koff rates for canonical and non-canonical G proteins, but the cryo-EM structures in reference 27 clearly suggest G protein subtype-specific β<sub>1</sub>AR:G protein interacfion interfaces that extend beyond the α 5 helix.Hence, off rates for heterotrimeric G proteins (or wild type Ga subunits) may differ from those observed for the mini-G's.It may be useful to compare the effect of the mini-G proteins used in this study with that of the wild-type heterotrimeric G proteins on ligand binding (like Han et al., 2023 (Ref 35); Extended Figure 10c).Such an experiment could inform on how well the engineered mini-G proteins reflect their wild type counterparts and potenfially give the authors more confidence in their claim that G protein on rates are the main drivers of G protein subtype selecfivity even it the full heterotrimers were used.This could also be done in the presence and absence of nucleofides (here I'm referring to lines 601 -607).Lines 583 -586: I suggest using "is in parfial agreement" since reference 69 also highlights the Lines 188 -195: The authors claim that the more upfield 1H chemical shift of M153 of iso-bound β<sub>1</sub>AR-E in Fig 2b is indicafive of a more acfive conformafion.However, the M153 1H chemical shift of Apo β<sub>1</sub>AR-E (Fig 2e) is located further upfield compared to iso-bound β1AR-E in Fig 2b.Does this mean that Apo β<sub>1</sub>AR-E is more acfive than iso-bound β1AR-E?Line 194: I suggest labelling Fig 2d clearly as corresponding to the β<sub>1</sub>AR-E L289M mutant.A label in the figure will make it easier for the reader to rafionalize the absence of the L289M resonance in Figs 2b & 2c.