Structural Flexibility of the Gαs Α-Helical Domain in the Β2

Home , Heterotrimeric G protein, Ras GTPase

Structural ﬂexibility of the Gαs α-helical domain in the β2-adrenoceptor Gs complex

Gerwin H. Westﬁelda,1, Søren G. F. Rasmussenb,c,1, Min Sua,1, Somnath Duttaa,1, Brian T. DeVreed, Ka Young Chungb, Diane Calinskid, Gisselle Velez-Ruizd, Austin N. Oleskiea, Els Pardone,f, Pil Seok Chaeg, Tong Liuh, Sheng Lih, Virgil L. Woods, Jr.h, Jan Steyaerte,f, Brian K. Kobilkab,2, Roger K. Sunaharad,2, and Georgios Skiniotisa,2

aLife Sciences Institute and Department of Biological Chemistry, dDepartment of Pharmacology, University of Michigan Medical School, Ann Arbor, MI 48109; bDepartment of Molecular and Cellular Physiology, Stanford University School of Medicine, Stanford, CA 94305; cDepartment of Neuroscience and Pharmacology, The Panum Institute, University of Copenhagen, 2200 Copenhagen N, Denmark; eStructural Biology Brussels and fVIB Department of Structural Biology, Vrije Universiteit Brussels, 1050 Brussels, Belgium; gDepartment of Chemistry, University of Wisconsin, Madison, WI 53706; and hDepartment of Chemistry, University of California at San Diego, La Jolla, CA 92093

Contributed by Brian K. Kobilka, August 19, 2011 (sent for review August 9, 2011) The active-state complex between an agonist-bound receptor and Results and Discussion a guanine nucleotide-free G protein represents the fundamental In a first step, we sought to examine the architecture of com- signaling assembly for the majority of hormone and neurotrans- plexes in the nucleotide-free state of Gαs. Before coupling with mitter signaling. We applied single-particle electron microscopy an agonist-bound receptor, the nucleotide binding pocket of the β (EM) analysis to examine the architecture of agonist-occupied 2- α-subunit of the Gαsβγ heterotrimer is occupied by GDP. Upon adrenoceptor (β AR) in complex with the heterotrimeric G protein 2 forming a complex with the β2AR, GDP dissociates, and the Gs (Gαsβγ). EM 2D averages and 3D reconstructions of the deter- resulting nucleotide-free β2AR-Gs complex is highly stable (2). gent-solubilized complex reveal an overall architecture that is in EM visualization of the nucleotide-free complex showed a very good agreement with the crystal structure of the active-state monodisperse particle population (Fig. 1A, and SI Appendix, Fig. α α ternary complex. Strikingly however, the -helical domain of G s S1). Reference-free alignment and classification of ∼17,000 fl appears highly exible in the absence of nucleotide. In contrast, particle projections revealed characteristic class averages with an the presence of the pyrophosphate mimic foscarnet (phosphono-

overall density that is in very good agreement with the crystal PHARMACOLOGY formate), and also the presence of GDP, favor the stabilization of structure of the complex (2). Because of its shape, the complex α α the -helical domain on the Ras-like domain of G s. Molecular adsorbs on the carbon support with small variations (± 20°) of α modeling of the -helical domain in the 3D EM maps suggests mainly two diametrically opposite preferred orientations that that in its stabilized form it assumes a conformation reminiscent generate practically identical, mirror-related 2D projections (SI α γ to the one observed in the crystal structure of G s-GTP S. These Appendix, Figs. S2 and S3). α data argue that the -helical domain undergoes a nucleotide- The distinct features of the class averages in these preferred fl dependent transition from a exible to a conformationally orientations allowed us to assign the negative stain projection stabilized state. profiles from specific components of the complex (Fig. 1 B and C). A central oval density represents the β2AR in a detergent G protein-coupled receptor | negative stain electron microscopy | random micelle, with a small protruding density corresponding to T4 conical tilt lysozyme (T4L) that replaces the unstructured extracellular N terminus of the receptor and serves as an orienting landmark. he majority of hormones and neurotransmitters communi- This interpretation was confirmed by EM analysis of complexes Tcate information to cells via G protein-coupled receptors lacking T4L (SI Appendix, Fig. S4). Some class averages of the (GPCRs), which instigate intracellular signaling by activating T4L-β2AR-Gs complex do not reveal a density corresponding to their cognate heterotrimeric G proteins on the cytoplasmic side. the T4L. Besides the presence of a relatively flexible linker GPCRs constitute the largest family of membrane proteins and connecting T4L and the β2AR, this effect is mostly because T4L play essential roles in regulating every aspect of normal physi- lies at an angle to the longitudinal axis of the complex, as shown ology, thereby representing major pharmacological targets. De- in the X-ray structure (2). Because of this geometry, even a 10° spite a wealth of biochemical and biophysical studies on inactive variation in the way the particle adsorbs on the carbon support and active conformations of several heterotrimeric G proteins, drastically reduces the visibility of the T4L projection profile, as the molecular underpinnings of G-protein activation remain demonstrated by projection simulation experiments (SI Appen- fi elusive. The β2-adrenergic receptor (β2AR) and its complex with dix, Fig. S5). Thus, the visibility of the T4L projection pro le is heterotrimeric stimulatory G-protein Gs (Gαsβγ) represent an very sensitive to even limited out-of-plane particle tilting (e.g., ideal model system for the large family of GPCRs activated by because of particle “rock” and “roll” or because of variations in fl diffusible ligands. Agonist binding to the β2AR promotes inter- the atness of the carbon support). Because we observe a single actions with GDP-bound Gsαβγ heterotrimer, leading to the density corresponding to T4L, the detergent micelle contains β exchange of GDP for GTP, and the functional dissociation of Gs only a single copy of the 2AR, in agreement with the crystal fi into Gα-GTP and Gβγ subunits. To examine the architecture of structure. Therefore, the signi cant additional density around agonist occupied β2AR in complex with Gαsβγ under different conditions, we used electron microscopy (EM) and single-particle analysis. Because of the limited size of the protein complex Author contributions: V.L.W., J.S., B.K.K., R.K.S., and G.S. designed research; G.H.W., S.G.F.R., ∼ M.S., S.D., B.T.D., K.Y.C., D.C., G.V.-R., A.N.O., E.P., P.S.C., T.L., S.L., and G.S. performed ( 148 kDa), we visualized specimens embedded in negative research; and B.K.K., R.K.S., and G.S. wrote the paper. fi stain, which provides suf cient contrast from relatively small The authors declare no conflict of interest. protein assemblies (1). This approach allowed us to obtain 2D 1G.H.W., S.G.F.R., M.S., and S.D. contributed equally to this work. projection averages and 3D reconstructions that provided new 2 β To whom correspondence may be addressed. E-mail: [email protected], sunahara@ insights into dynamic features of the 2AR-Gs complex, and umich.edu, or [email protected]. helped guide a successful approach to crystallize the complex This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. enabling a high-resolution structure (2). 1073/pnas.1113645108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1113645108 PNAS Early Edition | 1of6 Downloaded by guest on September 30, 2021 ABnucleotide-free T4L-β2AR-Gs C

63% 37% T4L T4L β2AR β2AR m m m m

Gαs Gαs Gs-βγ Gs-βγ AH

T4L T4L T4L T4L β2AR β2AR β2AR β2 AR m m m m mmmm

Ras Ras Gαs Gαs α βγ Gαs Gs-βγ βγ G s Gs- Gs- Gs-βγ AH AH

D nucleotide-free T4L-β2AR-Gs + Nb37

Fig. 1. Two-dimensional projection analysis of the T4L-β2AR-Gs complex in the nucleotide-free state. (A) Raw EM image of detergent-solubilized T4L-β2AR-Gs complex embedded in negative stain. (Scale bar, 50 nm.) (B) Representative EM class averages of the nucleotide-free complex with the projection profile of the AH domain not visible (Left), or visible on the Ras domain (Right, AH indicated by arrow). The cartoon models represent the conformations reflected by the EM averages, with the one on the left depicting the variable positioning of the AH domain, suggesting flexibility or multiple conformations (the position of the detergent micelle is indicated by gray shaded arcs and labeled with “m”). (Scale bar, 10 nm.) (C) Reprojections (Upper) of the crystal structure (2) (Lower) in the same overall orientation as B reveal the identity of each EM density component. The crystal structure on the Right shows the AH domain in the same position (relative to the Ras-like domain) as the one determined in the crystal structure of Gαs-GTPγS alone (4). (D) Representative class averages of nucleotide-free complex with nanobody Nb37 bound on the AH domain (arrows) reveal its flexibility. (Scale bar, 10 nm.)

the receptor stems from the large micelle formed by the de- to the T4L domain, the projection profile of the AH domain in tergent (3). Diametrically opposite to the T4L domain, two main this position (SI Appendix, Fig. S3) is not sensitive to the rela- interacting densities representing the Gs trimer appear in close tively limited out-of-plane tilts (± 20°) of the preferred particle proximity to the receptor on its intracellular surface. One of the orientation on the carbon support (SI Appendix, Fig. S5). This two domains appears to extensively interact with the receptor EM analysis provided the initial evidence for a high degree of density, suggesting it corresponds to the Ras-like domain of Gαs, mobility of the AH domain relative to the Ras domain in the fi while its neighboring domain has a pro le consistent with the nucleotide-free β2AR-Gs complex. Furthermore, the structural side view of Gβγ. Interestingly however, several class averages heterogeneity observed provided insights to the challenges in revealed an additional small globular density bound on the Ras- obtaining 3D crystals of the complex. like domain of Gαs (Fig. 1B, Right, and SI Appendix, Fig. S2). In To promote complex stabilization for high-resolution struc- this location, the additional density could only be attributed to tural studies, we generated and screened llama antibodies the α-helical (AH) domain of Gαs, occupying a position expected (nanobodies) to the purified complex (2). Nanobodies are small from the crystal structure of Gαs-GTPγS alone (4) and the (∼15 kDa), clonable variable domains of a heavy chain-only structure of the Gi heterotrimer (5) (Fig. 1C, and SI Appendix, antibody, obtained by immunizing a llama with purified de- Fig. S6), but in entirely different location from that observed in tergent-solubilized β2AR-Gs complex stabilized with a short the crystal structure of the β2AR-Gs complex (2). To assess the homobifunctional crosslinker (2). By screening samples with fraction of particles displaying the AH domain in this location, negative-stain EM, we identified two nanobodies (Nb35 and we selected and classified only projections clearly displaying the Nb37) that bound to the complex, but not the receptor alone. profiles of Ras-like, Gβγ, β2AR, and T4L domain densities in the Class averages of particles incubated with Nb35 indicated a ho- same position, thereby restricting the range of particle projection mogeneous protein complex displaying increased density be- orientations. The classification revealed that the AH domain was tween Gαs and Gβγ. Even though the nanobody projection ordered on the Ras-like domain in ∼35% of the particles, but in profile was not clearly distinguished in the preferred particle most other particle projections this density was absent (Fig. 1B, orientations, its presence appeared to enhance the uniformity in and SI Appendix, Fig. S3). It should also be noted that in contrast the disposition of Gαs-ras and Gβγ domains. The use of Nb35

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1113645108 Westfield et al. Downloaded by guest on September 30, 2021 indeed allowed us to obtain the crystal structure of the T4L- A T4L β2AR-Gs complex, which showed that the nanobody binds at the interface of the Gαs-Ras and Gβγ. In this location, Nb35 would β2AR not be predicted to interact with or stabilize the AH domain (2). Accordingly, the classification of Nb35-bound complexes revealed Gαs Ras a similar distribution of particles with an ordered AH domain on Gs-βγ the Ras-like domain as in the absence of Nb35 (SI Appendix, AH Figs. S7–S9). In contrast to Nb35, Nb37 appears bound directly to the AH domain [as also determined by deuterium-exchange MS 4L (DXMS)]) (SI Appendix, Fig. S10) and could be distinguished in single-particle EM class averages of the β AR-Gs complex as an 2 m m β extension of the AH domain. Using Nb37 as a domain marker 2AR allowed us to track the variable positioning of the small AH region of Gαs(SI Appendix, Figs. S11 and 12). The 2D class av- Ras erages from this preparation reveal an enhanced and elongated density adopting different orientations around the Ras-like do- Gαs main, ranging from close proximity to the Gβγ module to ex- Gs-βγ tending much further out of the complex in the opposite AH direction (Fig. 1D and SI Appendix, Fig. S12). Collectively, these findings suggest that in the absence of nucleotide, the AH domain is flexible, thereby sampling different positions around the Ras-like domain. Deuterium-exchange studies are consistent with a dynamic interface between the Gαs Ras and AH domains (6). Therefore, the unexpected position of the “open” AH domain in the crystal structure (2) represents just one of the possible conformations.

To obtain a more detailed view of the complex architecture, AH PHARMACOLOGY we used the random conical tilt approach (7) to calculate initial 3D reconstructions of complexes with and without ordered AH domain on the Ras-like domain (SI Appendix, Fig. S13). These AH initial 3D models were subsequently used for multireference supervised alignment (8, 9) to separate particle projections from our entire dataset according to the AH positioning (see SI Ap- B pendix). This approach allowed us to obtain quality 3D reconstructions from particle projections with and without density corresponding to AH domain on the Ras-like domain. The reconstructions are in excellent agreement with the corresponding 2D averages (Fig. 2A and SI Appendix, Fig. S14). In 3D reconstructions of particles where density for AH domain is observed, its orientation relative to the Ras-like domain appears to be similar to that found in the crystal structure of Gαs-GTPγS (4) (Fig. 2A and SI Appendix, Fig. S6). In addition, we obtained AH+Nb37A 3D reconstructions of nucleotide-free complexes with bound β Nb37 marking the positioning of the AH domain. The 3D maps Fig. 2. Three-dimensional reconstructions of the T4L- 2AR-Gs complex in clearly reveal that the Nb37-enhanced density of the AH domain the nucleotide-free state. (A) Representative class averages and corre- can adopt different conformations around the Ras-like domain, sponding 3D reconstructions of particles in each category show the vari- indicative of a relative flexibility in the interaction between the ability in the positioning of the AH domain in the nucleotide-free complex. two domains (Fig. 2B and SI Appendix, Figs. S15 and S16). This In the reconstruction to the left, the AH domain (orange ribbon) is shown in the same position as found in the docked crystal structure (2). Absence of variability in the 3D conformation of the AH domain or the AH/ fi fi suf cient density to accommodate this domain indicates that its position is Nb37 module around the Ras-like domain was further con rmed highly variable in this particle population. In the reconstruction to the right, by cross-validating 3D reconstructions (SI Appendix, Fig. S17). the AH domain is modeled within the available EM density right below the As noted above, previous crystal structures of G proteins show Ras-like domain of Gαs, as also suggested by the 2D averages. (B) Three-di- that bound nucleotides contribute to the stability of interactions mensional reconstructions of distinct conformations of nucleotide-free T4L- β between the Ras and AH domains. We therefore investigated 2AR-Gs complex with bound nanobody Nb37. The Nb37-enhanced density the positioning of the AH domain in the presence of guanine of the AH domain (marked with an oval) shows variable positioning around α nucleotides and nucleotide fragments. Pyrophosphate (PPi), the Ras-like domain of G s. (Scale bars, 5 nm.) representing two phosphates in GTP or GDP, has been shown 2+ β to bind Ras in a Mg -dependent manner, presumably at the - Gβγ dissociation and dissolution of the complex (12). However, and γ-phosphate positions (10). PPi and its chemically more PPi (with or without Mg2+) does not disrupt the receptor-G stable analog, foscarnet, more known for its antiviral properties fi (11), also bind to heterotrimeric Gαs with an apparent affinity of protein complex. This nding is in contrast to dissociation of the γ ∼0.5 and 1.6 mM, respectively, as determined by competition complex observed with the GTP mimetic GTP S (2). binding with a fluorescent GTPγS probe (Bodipy-GTPγS) (SI Although the presence of PPi does not appear to affect the AH 2+· Appendix, Fig. S18). Binding of PPi and foscarnet most likely domain positioning, in the presence of Mg foscarnet we observe substitutes for the α- and β-phosphates of GDP rather than the β- asignificantly higher proportion (∼70%) of complexes with an and γ-phosphates of GTP. Binding of β- and γ-phosphates would ordered AH region on the Ras-like domain (Fig. 3A and SI Ap- result in modification of the switch II domain with subsequent pendix,Figs.S19–S28). The observations in 2D class averages were

Westfield et al. PNAS Early Edition | 3of6 Downloaded by guest on September 30, 2021 A mediately fixed the sample by negative stain embedding. Addi- α−Helical domain stabilization on Ras tion of either of these nucleotides at concentrations above 10 μM 80% T4L resulted in significant amounts of partially dissociated complexes 69% 70% β2AR noitalupop 62% (SI Appendix, Fig. S29). This result is expected because a large 60% α Ras excess of either of these nucleotides would uncouple the G G s βγ Gs- AH 50% protein from the receptor. However, short incubation with lower 40% GDP concentrations (1 μM) and immediate sample fixation for 40% 37%

34% 33% EM allowed to us to examine intact complexes, revealing that the e lcitra 30% AH region was ordered in ∼60% of the intact particles (SI Ap- 20% pendix, Figs. S30–S32). In contrast, even low concentrations of γ μ ﬁ p 10% GTP S(1 M) showed a signi cant amount of destabilized complexes, and we were able to capture an array of intermediate 0% Nuc. - free Nuc. - free 1 mM PPi 10 mM PPi 10 mM 1 μM dissociation states (Fig. 3B and SI Appendix, Figs. S30 and S33). +Nb35 Foscarnet GDP Collectively, these data strongly suggest that the presence of B complex in the presence of 1 μM GTPγS nucleotide, or nucleotide fragments such as foscarnet, results in T4L AH domain stabilization against the Ras-like domain of Gαs. In the absence of nucleotide, the position of the AH domain is β 2AR highly variable (Figs. 3 and 4). Our results are in agreement with a recent study of the com-

Intact - 47% Receptor only - 15% plex formed by Gi and rhodopsin. Using double electron-elec- Partially dissociated - 38% tron resonance spectroscopy, Hamm, Hubbell, and colleagues Fig. 3. Nucleotide-dependent positioning of the Gαs AH domain. (A) Dis- documented large (up to 20 Å) changes in distance between tribution of particles with a distinct projection profile of the AH domain nitroxide probes positioned on the Ras and AH domains of Gi stabilized on the Ras-like domain across different conditions (Inset Right, upon formation of a complex with light-activated rhodopsin (13). marked with a white dot). A class average of a particle with a nonvisible AH The broad distance distributions observed for several labeling domain is shown for comparison (Inset Left). The presence of foscarnet and pairs are compatible with multiple conformations in dynamic GDP significantly increases the number of particles with stabilized AH do- equilibrium. Our findings are also consistent with results from main. (B) Representative class averages of the T4L-β2AR-Gs complex after rapid mixing with GTPγS(1μM) and immediate stain embedding reveal both DXMS that show increased deuterium exchange at both the nu- intact as well as partially or fully dissociated complexes. The fraction of cleotide binding pocket and at sites of interaction between the Ras corresponding subpopulations is indicated. (Scale bars, 10 nm.) and AH domains upon formation of the β2AR-Gs complex (6). Support for the open conformation in vivo may come from studies on the action mechanism of the cholera toxin, the en- also reproduced by individual 3D reconstructions for the different terotoxin secreted by the pathogen Vibrio cholerae. Cholera toxin, conformers in each state (SI Appendix, Figs. S24 and S28). These together with ADP ribosylation factor (ARF), ADP ribosylates fi results further con rm that the variability in the visibility of the AH Gαs at R201, rendering the residue catalytically ineffective and domain is indeed because of its variable positioning and not be- the G protein GTPase-deficient (14). The ADP ribosylated and cause of negative stain artifacts, such as incomplete embedding. constitutively active Gαs will continue to stimulate adenylyl cy- The ability of foscarnet to stabilize the AH domain on the Ras clase, cAMP production, and protein kinase A (PKA) activation. − domain suggests it is acting as a ligand fragment that binds to the Activated PKA opens intestinal Cl channels and leads to in- nucleotide binding pocket. Given its low affinity, it is not sur- creased water secretion that results in diarrhea (15). Crystallo- prising that the stabilization is incomplete. graphic studies of G proteins in GDP and GTPγS-bound forms In contrast to PPi and foscarnet, addition of GDP or the indicate that the catalytic arginine (R201 in Gs) is buried and nonhydrolyzable GTP analog GTPγS leads to dissociation of the likely inaccessible to cholera toxin and ARF (4). However, for- β2AR-Gs complex (2). To examine the effect of GDP and mation of the nucleotide-free form of Gαs and opening of the Ras GTPγS, we rapidly mixed the complex with nucleotide and im- and AH domains would facilitate accessibility to R201.

β2AR extracellular

βγ Gs- Gα-ras

Gα-H

1 2 3 4 Nucleotide free GTP uncoupling dissociation

Fig. 4. Model of conformational transitions in the β2AR-Gs complex. The nucleotide-free β2AR-Gs complex is characterized by a highly ﬂexible AH domain (model 1). GTP binding promotes stabilization of the AH domain on the Ras-like domain of Gαs (model 2). In this model, the AH domain has the same position

as the one observed in the crystal structure of Gαs-GTPγS (4). GTP binding results in subsequent uncoupling of Gs and β2AR (model 3), with eventual dissociation of Gαs and Gβγ (model 4).

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1113645108 Westfield et al. Downloaded by guest on September 30, 2021 In conclusion, single-particle EM analysis of the β2AR-Gs the percentage of projections from each condition was determined complex has provided novel structural insights into the dynamic according to the number of projections contributing to the assigned class “ ” nature of the assembly (Fig. 4). EM visualization of nanobody averages (SI Appendix, Fig. S34). The results of this blind test showed very good agreement with our assignments from individual classifications. bound complexes allowed us to clearly reveal the variable posi- fi tioning of the AH domain under nucleotide-free conditions and, For 3D reconstructions, in a rst step we used the random conical tilt technique (7) to determine initial 3D maps by back-projection of tilted additionally, to identify conditions that were key for the suc- particle images belonging to individual classes. After a first round of an- cessful characterization of the complex by X-ray crystallography. gular refinement, corresponding particles from the images of the untilted Furthermore, single-particle EM examination of the β2AR-Gs specimen were added, and the images were subjected to another cycle of complexes in varying concentrations of GDP and GTPγS en- refinement. We thus generated reliable initial models for complexes with abled us to capture transient intermediate dissociation states variability in the positioning of the ΑΗ domain of Gαs(SI Appendix,Fig. S13). After contrast transfer function (CTF) correction according to local between β2AR and Gs. This approach should prove useful for studying other signaling complexes involving GPCRs and other defocus values obtained by CTFTILT (18), the full dataset from each condition was subjected to multiple reference-supervised alignment (8, 9) with membrane proteins. The combination and integration of these “ fi ” technologies will be crucial for studying structural aspects of the multire ne routine in EMAN (1.9) by using our initial models as reference maps. This approach allowed us to separate particles from the challenging macromolecular complexes at large. entiredataset(ofeachcondition)accordingtothepositioningoftheΑΗ domain of Gαs. The number of contributing particles in each condition and Experimental Procedures conformation is provided in SI Appendix, Table S3.Forfinal maps, we used Specimen Preparation and EM Imaging of Negative-Stained Samples. The T4L- the separated datasets, as provided by the multiple reference-supervised β 2AR-Gs complex and nanobodies (Nb) were prepared as described in Ras- alignment, and used FREALIGN (19) for further refinement of the orien- mussen et al. (2). Specimens were visualized by EM in the following con- tation parameters and reconstruction (SI Appendix,Fig.S14–S16, S24, and β ditions: (i) T4L- 2AR-Gs complex alone (nucleotide-free) (SI Appendix, Fig. S28). The resolution for each map was determined at FSC = 0.5 and is β β S1); (ii) T4L- 2AR-Gs in presence of Nb35 (SI Appendix, Fig. S7); (ii) T4L- 2AR- provided in SI Appendix,TableS3. Gs in presence of Nb37 (SI Appendix, Fig. S11); (iv) T4L-β2AR-Gs in presence of 1 mM PPi (SI Appendix, Fig. S19 A and B); (v) T4L-β AR-Gs in the presence 2 Molecular Modeling. The crystal structure of T4L-β2AR-Gs (2) was fit in the EM β of 10 mM PPi (SI Appendix, Fig. S19 C and D); (vi) T4L- 2AR-Gs in the pres- density as a rigid body. Because of the presence of the detergent micelle, ence of 10 mM foscarnet and 10 mM MgCl (SI Appendix, Fig. S25); (vii) T4L- 2 which accounted for significant density surrounding β2AR, all docking β μ 2AR-Gs in the presence of 1 M GDP and 10 mM MgCl2 (SI Appendix, Fig. operations were performed manually with visual inspection of the best fit. S30 A and B); (viii) T4L-β AR-Gs in presence of 1 μM GTPγs and 10 mM MgCl 2 2 The docking of the T4L-β2AR-Gs complex revealed that the EM density cor- (SI Appendix, Fig. S30 C and D). All samples were prepared for EM using the βγ responding to G was shifted further away from the receptor in all 3D maps PHARMACOLOGY conventional negative staining protocol (1). For nanobody labeling, the T4L- compared with the crystal structure. The most likely cause for this variation is β 2AR-Gs complex was incubated for 15 min at room temperature with ap- the presence of the planar lipid bilayer provided by the cubic lipid phase in proximately equimolar concentrations of Nb35 or Nb37, and subsequently the crystals, which most likely maintains a different interaction with Gβγ prepared by negative staining. For nucleotide (GDP or GTPγS) and nucleotide compared with the detergent micelle. However, it is also possible that this fragment (PPi or foscarnet) incubations, these components were rapidly difference is attributed to the crystal packing, or a limited deformation of mixed with the complex and the sample was immediately fixed by negative- the complex because of the presence of the carbon support on the EM grid, stain embedding. or both. Accordingly, Gβγ was translated manually by 9 Å to best fit its Specimens were imaged at room temperature with a Tecnai T12 electron density yet retain all interactions with Gαs. This final model is shown un- microscope operated at 120 kV using low-dose procedures. Images were modified for all fittings in 3D reconstructions (Fig. 2, and SI Appendix, Figs. recorded at a magnification of 71,138× and a defocus value of ∼1.5 μmon S14–S16, S24, and S28). For maps showing the ΑΗ domain of Gαs on the Ras- a Gatan US4000 CCD camera (SI Appendix, Figs. S1, S7, S11, S19, S25, S29, like domain, we manually modeled the ΑΗ domain in its corresponding and S30). All images were binned (2 × 2 pixels) to obtain a pixel size of 4.16 Å position by taking into account steric constraints. In this conformation, the on the specimen level. Tilt-pair particles from 60° and 0° images were se- position of the ΑΗ domain is very similar to the one observed in the crystal lected using WEB (16). Particles for only 2D classification of 0° projections structure of Gαs-GTPγS(4)(SI Appendix, Fig. S6). were excised using Boxer (part of the EMAN 1.9 software suite) (17). The number of particles or tilt-pair particle projections per condition is provided Deuterium-Exchange MS. For DXMS, 1.5 mL of R:G complex or 1.5 mL of R:G:NB

in SI Appendix, Table S1. was mixed with 4.5 mL of D2O buffer (20 mM Hepes, pH 7.5, 100 mM NaCl, 10 mM BI-167107, 100 mM TCEP, 0.0015% MNG-3 in D2O) and incubated for 10, Two-Dimensional Classiﬁcations and 3D Reconstructions of T4L-β2AR-Gs. The 100, 1,000, and 10,000 s on ice. At the indicated times, the sample was 2D reference-free alignment and classiﬁcation of particle projections were quenched by 15 mL of ice-cold Quench solution (0.1 M NaH2PO4,20mM performed using SPIDER (16). For all conditions, the 0° particle projections TCEP, 16.6% glycerol, pH 2.4), immediately frozen on dry ice, and stored at –

were iteratively classified into multiple classes for 10 cycles (SI Appendix, 80 °C. Nondeuterated control was prepared in H2O buffer (20 mM Hepes, pH Figs. S2, S8, S12, S20, S22, S26, S31, and S33). SI Appendix, Table S1 provides 7.5, 100 mM NaCl, 10 μM BI-167107, 100 μM TCEP, 0.0015% MNG-3 in H2O), the number of classes for each condition. For AH conformation assignments, mixed with Quench solution, and snap-frozen on dry ice. Samples were we used the first classification to select only the particles from averages thawed and immediately passed through an immobilized porcine pepsin fl clearly displaying the profiles of Ras-like, Gβγ, β2AR, and T4L domain den- column (16 mL bed volume) at a ow rate of 20 mL/min of 0.05% tri- sities in the same position, thereby restricting the range of particle pro- fluoroacetic acid. Peptide fragments were collected contemporaneously on jection orientations. These projections were pulled together and subjected a C18 trap column for desalting and separated by a Magic C18AQ column to a second iterative classification (referred to as the “secondary” classifi- (Michrom BioResources Inc.) using a linear gradient of acetonitrile from 6.4% cation) (SI Appendix, Figs. S3, S9, S21, S23, S27, and S32). SI Appendix, Table to 38.4% over 30 min. MS analysis was performed using LCQ Classic mass S2 provides the number of classes and particle projections for each condition spectrometer from Thermo Finnigan, with capillary temperature of 20 °C. in the secondary classification. For counting the numbers of particles with Deuterium quantification data were collected in MS1 profile mode, and and without stabilized ΑΗ domain on the Ras-like domain, three different peptide identification data were collected in data-dependent MS/MS mode. operators examined each secondary classification and assigned each class Recovered peptide identification and analysis were carried out using DXMS average according to the projection profile of the specific region (SI Ap- Explorer (Sierra Analytics Inc.), a software specialized in processing DXMS pendix, Figs. S3, S9, S21, S23, S27, S32, and S33). The assignment from the data (SI Appendix, Fig. S10). different operators was in good agreement, and the particle numbers belonging to individual classes were added to calculate percentages for each Bodipy-GTPγS Binding. The effect of foscarnet and PPi was measured using 100 conformation. Assignments for each full individual dataset were done in nM bodipy-GTPγS-FL (Invitrogen). Fluorescence intensity of bodipy-GTPγS-FL fi addition to the secondary classi cation, and the results were in agreement. (lex ∼470 nm) increases upon G-protein binding, as demonstrated by McE- To test any bias, the particles from nucleotide-free, 1 mM PPi, 10 mM PPi, wan et al. (20). A wavelength scan of bodipy-GTPγS-FL (100 nM) in the ab- and foscarnet conditions were combined into a single dataset of 15,753 sence (dotted) or presence (solid) of a molar excess of purified Gαs (1 mM) particles and were classified into 200 classes. The individual class averages was determined to assess optimal spectroscopy conditions. The capacity of were assigned as before according to the visibility of the AH domain, and PPi and the chemically stable pyrophosphate analog of PPi, foscarnet, to

Westfield et al. PNAS Early Edition | 5of6 Downloaded by guest on September 30, 2021 inhibit bodipy-GTPγS-FL (lex∼470 nm, lem∼515 nm) was measured as de- Fluorescence was measured in a 96-well microtiter plate format on a M5 scribed in SI Appendix, Fig. S18 B and C in both Gαs(B and C) and hetero- fluorescence plate reader (Molecular Precision). trimeric Gαsβγ (C). The fluorescence of 100 nM bodipy-GTPγS-FL was measured in the presence of 1 mM G protein. PPi or foscarnet were added ACKNOWLEDGMENTS. We thank J. Tesmer for suggestions. This work was together with bodipy-GTPγS-FL and initiated by the addition of G protein (1 supported by the Lundbeck Foundation Junior Group Leader Fellowship (to S.G.F.R.); the Fund for Scientific Research of Flanders (Fonds Wetenschappe- fi mM) in 20 mM Tris-HCl, pH 8.0, 3 mM MgCl2, 1 mM DTT in a nal volume of lijk Onderzoek-Vlaanderen) and the Institute for the encouragement of 200 mL bodipy-GTPγS-FL binding to heterotrimeric G protein included 0.1% Scientific Research and Innovation of Brussels (to E.P. and J.S.); the National dodecylmaltoside (final). Fluorescence was measured on a short time scale Institute of General Medical Sciences (NIGMS) Molecular Biophysics Training (600 s) to minimize the accumulation of hydrolysis product bodipy-phos- Grant GM008270 (to G.H.W. and B.T.D.); the National Institute of Neural fl Disorders and Stroke Grant R01-NS28471 (to B.K.K.); the Mather Charitable phate (21). Hydrolysis, which also appears as an increase in uorescence, was Foundation (to B.K.K.); NIGMS Grants R01-GM083118 (to B.K.K. and R.K.S.) 2+ determined simply by chelating Mg with 10 mM EDTA following the 600-s and R01-GM068603 (to R.K.S.); National Institute of Diabetes and Digestive incubation. Inhibition of bodipy-GTPγS-FL by PPi and foscarnet can be re- and Kidney Diseases (NIDDK) Grant R01-DK090165 (to G.S.); National versed with the subsequent addition of high Mg2+ (25 mM), which enhances Institutes of Health Grants AI076961, AI08192, AI2008031, CA118595, GM20501, GM066170, GM093325, and RR029388 (to V.L.W.); Michigan bodipy-GTPγS-FL binding, indicating that PPi and foscarnet are not irre- Diabetes Research and Training Center Grant, NIDDK, P60DK-20572 (to versibly binding or denaturing the G protein. Gαs was purified as described R.K.S.); and the University of Michigan Biological Sciences Scholars Program in Sunahara et al. (4). Gαsβγ was purified as described by Rasmussen et al. (2). (R.K.S. and G.S). G.S. is a Pew Scholar of Biomedical Sciences.

1. Ohi M, Li Y, Cheng Y, Walz T (2004) Negative staining and image classification— 12. Sprang SR (1997) G protein mechanisms: Insights from structural analysis. Annu Rev Powerful tools in modern electron microscopy. Biol Proced Online 6(1):23–34. Biochem 66:639–678. 2. Rasmussen SG, et al. (2011) Crystal structure of the β(2) adrenergic receptor-Gs protein 13. Van Eps N, et al. (2011) Interaction of a G protein with an activated receptor opens complex. Nature, 10.1038/nature10361. the interdomain interface in the alpha subunit. Proc Natl Acad Sci USA 108: 3. Rubinstein JL (2007) Structural analysis of membrane protein complexes by single 9420–9424. particle electron microscopy. Methods 41:409–416. 14. Freissmuth M, Gilman AG (1989) Mutations of GS alpha designed to alter the re- 4. Sunahara RK, Tesmer JJ, Gilman AG, Sprang SR (1997) Crystal structure of the adenylyl activity of the protein with bacterial toxins. Substitutions at ARG187 result in loss of cyclase activator Gsalpha. Science 278:1943–1947. GTPase activity. J Biol Chem 264:21907–21914. 5. Wall MA, et al. (1995) The structure of the G protein heterotrimer Gi alpha 1 beta 1 15. Gabriel SE, Brigman KN, Koller BH, Boucher RC, Stutts MJ (1994) Cystic fibrosis het- gamma 2. Cell 83:1047–1058. erozygote resistance to cholera toxin in the cystic fibrosis mouse model. Science 266 6. Chung KY, et al. (2011) beta(2) adrenergic receptor-induced conformational changes (5182):107–109. in the heterotrimeric G protein Gs. Nature, 10.1038/nature10488. 16. Frank J, et al. (1996) SPIDER and WEB: Processing and visualization of images in 3D 7. Radermacher M, Wagenknecht T, Verschoor A, Frank J (1987) Three-dimensional re- electron microscopy and related fields. J Struct Biol 116:190–199. construction from a single-exposure, random conical tilt series applied to the 50S ri- 17. Ludtke SJ, Baldwin PR, Chiu W (1999) EMAN: Semiautomated software for high-res- bosomal subunit of Escherichia coli. J Microsc 146(1):113–136. olution single-particle reconstructions. J Struct Biol 128(1):82–97. 8. Brink J, et al. (2004) Experimental verification of conformational variation of human 18. Mindell JA, Grigorieff N (2003) Accurate determination of local defocus and specimen fatty acid synthase as predicted by normal mode analysis. Structure 12(2):185–191. tilt in electron microscopy. J Struct Biol 142:334–347. 9. Ménétret JF, et al. (2005) Architecture of the ribosome-channel complex derived from 19. Grigorieff N (2007) FREALIGN: High-resolution refinement of single particle struc- native membranes. J Mol Biol 348:445–457. tures. J Struct Biol 157(1):117–125. 10. Zhang B, Zhang Y, Shacter E, Zheng Y (2005) Mechanism of the guanine nucleotide 20. McEwen DP, Gee KR, Kang HC, Neubig RR (2001) Fluorescent BODIPY-GTP analogs: exchange reaction of Ras GTPase—Evidence for a GTP/GDP displacement model. Real-time measurement of nucleotide binding to G proteins. Anal Biochem 291(1): Biochemistry 44:2566–2576. 109–117. 11. Sundquist B, Oberg B (1979) Phosphonoformate inhibits reverse transcriptase. J Gen 21. Jameson EE, et al. (2005) Real-time detection of basal and stimulated G protein Virol 45:273–281. GTPase activity using fluorescent GTP analogues. J Biol Chem 280:7712–7719.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1113645108 Westﬁeld et al. Downloaded by guest on September 30, 2021