ANALYSISdoi:10.1038/nature14663

Universal allosteric mechanism for Gaactivation by GPCRsTilman Flock1, Charles N. J. Ravarani1*, Dawei Sun2,3*, A. J. Venkatakrishnan1{, Melis Kayikci1, Christopher G. Tate1,Dmitry B. Veprintsev2,3 & M. Madan Babu1

G protein-coupled receptors (GPCRs) allosterically activate heterotrimeric G proteins and trigger GDP release. Giventhat there are 800 human GPCRs and 16 different Ga genes, this raises the question of whether a universal allostericmechanism governs Ga activation. Here we show that different GPCRs interact with and activate Ga proteins through ahighly conserved mechanism. Comparison of Ga with the small G protein Ras reveals how the evolution of shortsegments that undergo disorder-to-order transitions can decouple regions important for allosteric activation fromreceptor binding specificity. This might explain how the GPCR–Ga system diversified rapidly, while conserving theallosteric activation mechanism.

G proteins bind guanine nucleotides and act as molecular switchesin a number of signalling pathways by interconverting betweena GDP-bound inactive and a GTP-bound active state1,2. They

consist of two major classes: monomeric small G proteins3 and hetero-trimeric G proteins4. While small G proteins and the a-subunit (Ga) ofheterotrimeric G proteins both contain a GTPase domain (G-domain),Ga contains an additional helical domain (H-domain) and also forms acomplex with the Gb and Gc subunits. Although they undergo a similarsignalling cycle (Fig. 1), their activation differs in one important aspect.The guanine nucleotide exchange factors (GEFs) of small G proteins arelargely cytosolic proteins, whereas the GEFs of Ga proteins are usuallymembrane-bound GPCRs. While GEFs of small G proteins interactdirectly with the GDP binding region1,3, GPCRs bind to Ga at a site

almost 30 A away from the GDP binding region5 and allosterically trig-ger GDP release to activate them.

The high-resolution structure of the Gas-bound b2-adrenergic recep-tor (b2AR)5 provided crucial insights into the receptor–G protein inter-face and conformational changes in Ga upon receptor binding6,7. Recentstudies described dynamic regions in Gas

6 and Gai8, the importance of

displacement of helix 5 (H5) of Gas and Gat by up to 6 A into thereceptor5, the extent of helical domain opening during GDP release9,10,and identified residues that contribute to Gai activation7,10. These stud-ies focused on single, specific Ga proteins; however, in humans there are16 different Ga genes, with at least 21 isoforms4 that can be grouped intofour functional subfamilies (Gas, Gai, Gaq, Ga12), which each regulatedifferent signalling pathways11. Although they belong to the same proteinfold, they have diverged significantly in their sequence such that each Gaprotein can be specifically activated by one or several of the ,800 humanGPCRs4. Thus, a fundamental question is whether there is a universalmechanism of allosteric activation that is conserved across all Ga proteintypes10. Allosteric communication in proteins is mediated through con-formational changes, which are facilitated by the re-organization of non-covalent contacts between residues. Thus, studying these contacts canprovide detailed insights into the mechanism of allostery12–16. On the basisof a comprehensive analysis, here we propose that GPCRs interact andactivate Ga subunits through a conserved mechanism. We describemolecular details of the key structural transitions and pinpoint residuesthat constitute the ‘common core’ of Ga activation.

Common Ga numbering and residue contact networksWe created a structural and sequence alignment of 80 Ga structuresfrom diverse organisms and 973 sequences from 66 species that have aGPCR–G protein system (auto-activating plant Ga proteins were notconsidered; Methods). To enable the comparison of any residue/posi-tion between different Ga proteins, we devised the common Ga num-bering (CGN) system (Fig. 2a). The CGN provides an ‘address’ for everyresidue in the DSP format, referring to: (1) the domain (D); (2) theconsensus secondary structure (S); and (3) the position (P) within thesecondary structure element. For instance, phenylalanine 336 in Gai1 isdenoted as Phe336G.H5.8 as it is the eighth residue within the consensushelix H5 of the G-domain. The corresponding position in Gas2 is

αGAP

GAP-bound state

40 structures

Effectorα

Active state

25 structures

GPCR

βγ α

1 structure

GEF-bound state

Inactive state

11 structures

βγ α

GDP

GTP

Pi

GTPase domain(G-domain)

α-helical domain(H-domain)

Intracellular

Extracellular

GPCR

Gα

Figure 1 | Ga signalling states and activation. Heterotrimeric G proteins (1)release GDP upon binding to a guanine nucleotide exchange factor (GEF),which are G-protein-coupled receptors, (2) bind GTP and recruit downstreameffectors, and (3) hydrolyse GTP, promoted by a GTPase activating protein(GAP), leading to (4) the inactive, GDP-bound state. Structures of the Gasubunit (blue) bound to GDP (Protein Data Bank (PDB) accession 1got;inactive state; top) and bound to the b2-AR (grey) (PDB 3sn6; active state;bottom) are shown.

*These authors contributed equally to this work.

1MRC Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge CB2 0QH, UK. 2Laboratory of Biomolecular Research, Paul Scherrer Institut, 5232 Villigen, Switzerland. 3Department of Biology,ETH Zurich, 8039 Zurich, Switzerland. {Present address: Department of Molecular and Cellular Physiology, Stanford University School of Medicine, Stanford, California 94305, USA.

0 0 M O N T H 2 0 1 5 | V O L 0 0 0 | N A T U R E | 1

G2015 Macmillan Publishers Limited. All rights reserved

Phe376G.H5.8. Loops are labelled in lowercase letters of their flankingsecondary structure elements (SSE); for example, s6h5 refers to the loopconnecting strand S6 with helix H5 (see Extended Data Fig. 1, Methodsand Supplementary Note). A CGN mapping webserver is available athttp://www.mrc-lmb.cam.ac.uk/CGN.

The Ga structures were assigned to the four major signalling states(Fig. 1) and the non-covalent contacts between residues were calculatedfor each structure. We computed the consensus non-covalent residuecontacts from all Ga structures of the same signalling state. Using theCGN, we integrated information on evolutionary conservation for everyposition and derived the consensus contacts mediated by universallyconserved residues for each signalling state (Fig. 2b). We find that eachstep of the signalling cycle undergoes contact re-organization to variableextents. Since these conformational changes involve conserved residues,the observed contact re-organization is likely to be universal for all Gaproteins. A description and additional interpretations are provided in

the Supplementary Note and Supplementary Information. Below, wedescribe the major findings pertaining to Ga activation. We first focuson the GPCR–Ga interface and then describe the molecular details ofhow the non-covalent contacts are re-organized and propagated to theGDP binding pocket, leading to GDP release.

GPCR–Ga protein interfaceAnalysis of the buried surface area (BSA) and residue contacts betweenthe b2AR–Gas interface5 shows that H5 contributes ,70% (845 A2;15 residues) of the total BSA. Other SSEs (s2s3, h4hg, H4, h4s6, S6)cover ,20% (289 A2; 14 residues), and the amino-terminal membrane-anchored helix HN and its loop with strand S1 contribute ,10%(120 A2; 5 residues) of the total BSA (Fig. 3a). H5 is the key interfaceelement that contacts residues in transmembrane helices (TM3, TM5and TM6) and intracellular loops 2 and 3 (ICL2 and ICL3) of the b2AR5.Contacts from the other Ga regions are mainly restricted to ICL3 of the

PDB-1: 3sn6 (Gαs2)

PDB-25: 4gnk (Gαq)

PDB-2: 1gia (Gαi1)

PDB-1: 1azt (Gαs2)( )

...

PDB-3: 3cx7 (Gα13)

PDB-2: 4ekc (Gαq)

PDB-1: 2gtp (Gαi1)

...

...

PDB-11: 1zcb (Gα13)

PDB-2: 1gp2 (Gαi1)

PDB-1: 1got (Gαt1)

GE

F-b

ou

nd

on

e s

tru

ctu

re,

Gs

11

str

uctu

res,

Gs,i,q

,13

Inactive

Active

25

str

uctu

res,

Gs,i,q

GA

P-b

ou

nd

40

str

uctu

res,

Gi,q

,13

His

PheVal

IleAla

PDB-11: 1got (Gαt1)

1,246 contacts - 326 residues

...

1,292 contacts - 318 residues

His

PheGlu

IlePro

PDB-1: 1zcb (Gα13)Phe359

His72

Use C

GN

to

esta

blis

h

top

olo

gic

al eq

uiv

ale

nce

His

PheXaa

IleXaa

G.H5.8

G.H1.12

His

PheXaa

IleXaa

G.H5.8

G.H1.12

State consensusresidue contact network

Gαs2

Gαi2

Gα13

...

...e.g. G.H5.8e.g. G.H1.12

Identify conserved

residues in all Gα(561 complete sequences)

260 long-range contacts (535 total) between 188 residues (289 total)

72 highly conserved residues

State consensus network between

universially conserved residues43 contacts44 residues

>100,000 (2,436 unique) residue contactsbetween 386 positions

80 structures79 unique long-range residue contacts

between 70 universally conserved residues

Phe332

His53

16 human Gα genes

4 Gα subfamilies

Human Gα alignmentGαs2Gαi2

Gα13

...

Gαs2-chimp

Gαs2-cow

Gαs2-rat...

...

Position in the consensus secondary structure element (SSE) mapped

to the human reference alignment

G12

Gq

Gi

Gs

GNAS2

a

GNAL

GNAI2

GNAI1

GNAI3

GNAT2

GNAT1

GNAT3

GNAO

GNAZ

GNAQ

GNA11

GNA14

GNA15

GNA12

GNA13

HN

H5

H1HF

HA

HE HD

HCHB

HG

H4

H3

S1

S2S3 S

4S

5

S6

H2

Secondary structure element (consensus) (e.g., H5)

Do

main

(e.g

., G

)

G-d

om

ain

(G

)H

-do

main

(H

)

+

e.g.

Domain

SS elem

ent

Position

Gαs2: f Gαi2:

Gα13:

...f

Common Gαnumbering

www.mrc-lmb.cam.ac.uk/CGN

Gαs orthologue alignment

...

973 sequences

66 organisms

16 orthologue alignments

b

For example, inactive state consensus network For example, inactive state consensus network

between residues conserved across all Gα subfamilies

For example, residue interaction networks of 11 Gαstructures of the inactive state

for each of the 11 structures

Residue contact network

457 unique long-range residue contactsbetween 247 shared positions

Figure 2 | The common Ga numbering (CGN) system and Ga conservedcontact networks. a, Every position in a Ga is denoted by its domain (D),the consensus secondary structure element where it is present (S), and position(P) within the consensus SSE. Names of SSEs are shown in the cartoon; loopsare named with lowercase letters of the flanking SSEs (for example, h1ha;see Extended Data Fig. 1 for all SSEs). An alignment of all 973 Ga proteins

belonging to all the 4 subfamilies from 66 species allowed the identification ofequivalent residues. An online webserver allows mapping of any Ga sequenceor structure to the CGN system (http://www.mrc-lmb.cam.ac.uk/CGN).b, Computation of consensus residue contact networks between conservedresidues for different Ga signalling states. See Methods, SupplementaryNote and Supplementary Data.

2 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 5

RESEARCH ANALYSIS

G2015 Macmillan Publishers Limited. All rights reserved

receptor. An analysis of the contacts of the Gat carboxy-terminal pep-tides bound to rhodopsin17–20 shows that conserved residues in H5 tendto interact with the corresponding, topologically equivalent residues inrhodopsin (Supplementary Data).

We mapped evolutionary conservation onto the b2AR–Gas interfaceand found that H5 is the only interface region that harbours residuesthat are highly conserved across species and Ga protein types (,27% ofH5; 7 residues; Fig. 3b). These residues have significantly higher BSAcompared to the non-conserved H5 interface residues. Several highlyconserved Ga interface residues interact with conserved interface resi-dues on the GPCR (Extended Data Fig. 2a). Computational energycalculations show that these residues make the highest interface energycontribution, suggesting that they are important for complex formation(Extended Data Fig. 2a). Thus, the universally conserved residues on H5might form the conserved ‘interaction hotspots’ for different Ga pro-teins to interact with their cognate receptors in a similar binding mode.We also found that two-thirds of the H5 residues are variable but half ofthese (8 residues) still contact the receptor. Thus, H5 harbours distinctsets of interface residues that are either conserved or variable across thedifferent Ga proteins. The variable positions on H5 together with theother interface regions (Fig. 3b) could be important for selective coup-ling to different receptors, as shown for individual Ga proteins21,22. Thissuggests that the conserved Ga interface positions provide the basis for acommon mode of receptor binding, while the variable positions mightconfer selectivity in receptor coupling (Supplementary Note).

The role of helix H5 in Ga activationIn the 79 structures of Ga not bound to a GPCR, the C-terminal residuesof H5 are characterized by missing electron density (Extended DataFig. 2b). This region undergoes a disorder-to-order transition andextends H5 upon receptor binding as shown for Gas/t/i

5,6,8,20,23.Analysis of H5 from 561 full-length Ga homologues suggests that thehigher disorder propensity of the last eight residues compared to the restof H5 is a universal feature (Extended Data Fig. 2b). Within this dis-ordered region, hydrophobic positions IleH5.15, LeuH5.20 and LeuH5.25

contact the GPCR, fold into a helix and are conserved between humanproteins and the yeast homologue Gpa1 (,1,200 million years (Myr)ago). This highly conserved peptide motif24 is involved in the disorder-to-order transition upon receptor binding and suggests that this struc-tural transition mediated by the three key conserved interface residues islikely to be a universal feature of all G proteins.

To understand the effect of GPCR binding on contact re-organizationwithin Ga, we analysed the consensus contacts in the inactive state (11structures) and the active state (b2AR–Gas structure). Although we hadonly one structure of the receptor–Ga protein complex, identifying thedifferences between the contacts of the active state and the consensuscontacts of the inactive state allowed us to focus on residues that arere-organized upon receptor binding (and hence likely to be relevant forall Ga types; Supplementary Note). In all inactive state structures, theN-terminal part of H5 makes extensive contacts with a number ofSSEs within the G-domain of Ga. Two universally conserved positions

0

0.2

0.4

0.6

0.8

1.0

0 0.2 0.4 0.6 0.8 1.0

H5.8

hns1.3

S1.2

S3.1

S3.3

H4.2

6

h4s6.3

H5.1

1

H5.1

2

H5.1

3

H5.1

5

H5.1

6

H5.1

7

H5.1

9

H5.2

0

H5.2

3

H5.2

5

H5.2

6

Relative BSA (%)

Seq

uence id

entity

Exposed Buried

Co

nserv

ed

Variab

le

SSE element

H4

H5

HN

Others

a

GPCR interface

G protein interface

GPCR

H4

ICL3

TM6

ICL2

H5HN

TM3

TM5

Gα

BS

A (%

)

20

0

60

40

ICL2 TM3 TM5 TM6ICL3

Total BSA GPCR:

1,264 Å2

Total BSA Gα:

1,255.4 Å2

~ 7

0%

HN Other H4 H5

BS

A (%

)

20

0

60

40

b

cInactive state GPCR−bound state

Intra

Inter

2 4 6 8 10 12 14 16 18 20 22 24 26 2 4 6 8 10 12 14 16 18 20 22 24 26G.H5.1 G.H5.1

H1

S2

S3

H5

H1

ICL3

TM6

ICL2

H5

TM3

TM5

S2

S3H5

Disorder-to-order transition

H5

Position on helix H5 Position on helix H5

Interface moduleTransmission module Interface moduleTransmission module

Hig

hly

co

nserv

ed

inte

rface re

sid

ues

Sp

ecifi

city

co

nfe

rring

inte

rface re

sid

ues

5

4

1

2

3

Num

ber

of

co

nta

cts

GPCR

βγ α

βγ α

GPCR interface

G protein interface

G

HN

H5HN

H4

ICL2

ICL3

TM5

TM6TM3

H5HN

H4

ICL2

ICL3

TM5

TM6TM3

Figure 3 | Helix 5 contains the conserved interface region and is comprisedof two modules. a, Inter Ga–GPCR residue contact network (inset) and buriedsurface area (BSA) analysis of the heterotrimeric Gas–b2AR structure (3sn6).The line width between the nodes (SSE) denotes the number of consensusresidue contacts. GPCR positions are denoted by extending the Ballesteros–Weinstein numbering system (note parts of ICL3 become extended TM5 in theactive state of the receptor). b, Scatter-plot of Ga sequence conservation and

normalized BSA highlights the conserved and variable interface residues.c, Consensus contact rewiring between the inactive and the GPCR-bound stateby H5 residues. Positions mediating intra Ga protein contacts (blue) andreceptor-mediated contacts (red) are shown. The circle size represents thenumber of contacts. H5 can be divided into transmission and interface module.The disorder-to-order transition of the H5 C-terminal region upon receptorbinding and SSE contact rewiring are shown.

0 0 M O N T H 2 0 1 5 | V O L 0 0 0 | N A T U R E | 3

ANALYSIS RESEARCH

G2015 Macmillan Publishers Limited. All rights reserved

(PheH5.8 and ValH5.7) on H5 contact conserved positions in H1, S2 andS3 (left lobe), and S5 and S6 (right lobe), respectively (Extended DataFig. 3). In the GPCR-bound state, H5 loses 20% of its intra-Ga contacts(primarily with H1), and gains 27 intermolecular residue contacts withthe GPCR (Fig. 3c). Upon receptor binding, the H5 contacts with theright lobe are not lost, but are re-organized to accommodate the struc-tural changes and might be important for the stability of the receptor-bound complex (for discussion, see Supplementary Note and ref. 25).Thus, H5 is composed of two highly conserved modules with distinctfunctions; that is, an interface module important for receptor binding,and a transmission module that harbours intra-G-protein contacts,which are re-organized upon receptor binding (Fig. 3c).

In contrast to H5, the non-conserved interface regions from H4, h4s6and h4hg undergo less marked re-organization of intra-Ga contactsupon receptor binding (Extended Data Fig. 3). This suggests that theconserved mechanism of allosteric activation is primarily mediated bythe movement of H5, thereby breaking the contacts between H5 and H1.As the residues that form these contacts are conserved in all 16 Gaproteins, the described contact re-organization is likely to be universalfor all Ga proteins.

The role of helix H1 in Ga activationIn the inactive state, H1 acts as a structural ‘hub’ by linking differentfunctional regions of Ga. H1 contacts the N-terminal part of H5 (trans-mission module), H-domain and GDP through universally conservedresidues (Fig. 4a). In this manner, H1 links the H-domain and the GDPbinding site with the conserved residues in the H5 transmission module,which in turn is physically linked to the H5 interface module that bindsthe receptor. The conserved consensus contacts between H5 and H1seem essential for the structural integrity of H1. Computational calcula-tions of the per-residue contribution to protein stability for the 79 non-receptor-bound structures are consistent with their role in stabilizing H1(Extended Data Fig. 4). Upon GPCR binding, the H5–H1 contacts arelost, affecting the structural integrity of H1. The contact-mediatingpositions in H1 have missing electron density in the b2AR–Ga structure5

and hydrogen/deuterium exchange experiments have shown that thisregion is dynamic in Gas upon receptor binding6. Upon becoming

flexible, the conserved consensus contacts between H1, GDP and theH-domain hinge region are lost; this results in the loss of a significantfraction of all contacts made with GDP, thereby weakening its bindingaffinity (Fig. 4b), and results in increasing the likelihood of H-domainopening. Since the entire sequence of H1, the contacting residues on H5,and the H-domain hinge positions are highly conserved across all Gaproteins, the mechanism of GDP release is likely to be universal for allGa proteins.

Universal mechanism of Ga activationWhile variable interface residues in H5 and elsewhere allow specificbinding to distinct GPCRs, we find that H5 primarily harbours con-served positions that might allow a common mode of receptor bindingand a conserved mechanism of allosteric activation. The contact re-organization between conserved residues links the disorder-to-ordertransition of H5 upon receptor binding to a change in structural stabilityof H1, ultimately leading to GDP release. More specifically, H5 isdivided into: (1) The N-terminal transmission module, which forms ap–p cluster linking H5, S2, S3 and H1 via universally conserved residuesPheH5.8, HisH1.12, PheS2.6 and PheS3.3 in the inactive state; and (2) theC-terminal interface module, which undergoes a disorder-to-ordertransition in the intracellular cavity of the receptor via universally con-served positions. This structural transition results in a displacement ofthe H5 transmission module, thereby interrupting the p–p cluster. There-organized residues in the cluster (PheH5.8 and PheS3.3) contact con-served residues within ICL2 of the receptor (extrapolated Ballesteros–Weinstein numbering: 3.58 and 3.57 of b2AR5), as confirmed recentlyfor the CB2 receptor–Gai complex26. Since the conserved p–p cluster isimportant for the structural integrity of H1, its disruption leads to anincreased flexibility of H1. H1 has a central role in the inactive state byforming contacts both to GDP (Extended Data Fig. 4) via the Walker Amotif2,27 and to the H-domain (through a ‘cation–phinge’ motif; ExtendedData Fig. 5). Thus, the increased flexibility due to the partial unfolding ofH1 facilitates GDP release and H-domain opening. The only other con-served inter-domain contact is an ‘ionic latch’ between the C-terminal loopof helix HG of the G-domain and the hChD loop of the H-domain. Thiscontact is broken upon receptor binding, which might be a result of thereorganization of the right Ga lobe (Extended Data Fig. 3).

In addition to H1, residues around HG and within the s6h5 loop(TCAT motif) contact the GDP (Extended Data Fig. 3). The conservedguanine-contacting TCAT motif preserves many of its contacts withinGa upon receptor binding, although TCAT-to-H1 contacts are lost andnew contacts are formed between H5, S5, S6 and the TCAT motif (thatis, the re-organized right lobe). Likewise, residues that contact the gua-nosine moiety, including the interaction between the TCAT motif andHG, s1h1 (P-loop), S5 and S6, are not extensively re-organized duringGa activation. Whether this arrangement poises Ga for GTP binding(which differs from GDP by a single phosphate and whose physiologicalconcentration exceeds GDP several-fold22), and whether GTP has thecapacity to stabilize H1 on its own and trigger Ga release from thereceptor, remains to be addressed. An analysis of the GTP-bound Gareveals that the presence of the third phosphate facilitates additionalcontacts with the switch regions (Extended Data Fig. 4c).

Conceptually, the GPCR-bound Ga conformation can be consideredas a metastable Ga transition state that is stable only due to interactionswith the GPCR. Thus, the lost intra-Ga contacts between H1 and thetransmission module of H5 are compensated by the helix extension ofH5, the gained receptor interface contacts, and some re-organized con-tacts in the right lobe of Ga (H5 with S5, S6; see role of conservedY320G.S6.2 in ref. 25). In this manner, H5 and H1 act as the primaryconduits of information transfer between the receptor interaction inter-face (input) and the GDP binding site (output). The residues of thestructural motifs and functional elements described here are conservedin all the different families of Ga proteins (Extended Data Fig. 5a). Thus,the mechanism described here is likely to be universal for activation ofall cognate GPCR and Ga protein pairs.

Fra

ctio

n o

f to

tal co

nsensus c

onta

cts

with G

DP

H1

HF

hfs2

s1h1

HG

S5

s6h5

s5hg

TCAT H

G

regi

on

Hinge

regi

on

0

0.4

0.2

15 conservedconsensuscontacts

H1

HF

hfs2

S2

s6h5

h1ha

S3

H5

HG

H4

GPCR bound state

H1

HF

hfs2

S2

s6h5

HG

h1ha

GDP

S3

H5

Inactive state

H4

Loss of structuralintegrity of H1

a

βγ α

GPCR

Gα

b

GPCR

βγ α

H1

Figure 4 | Helix H1 is the key SSE that contacts H5, GDP and the H-domain.a, Consensus SSE contacts involving H1. The line width between the nodes(SSE) denotes the number of consensus residue contacts. Upon receptorbinding, H5 is displaced and crucial contacts between H1 and H5 (dark blue)are lost. This might explain the increased flexibility of H1 in the GPCR-boundstate, which results in the loss of GDP contacts (green) and the H-domainhinge region (formed by H1, h1ha and HF; light blue). b, The extent of GDPconsensus contacts mediated by the different SSEs. See Extended Data Fig. 4.

4 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 5

RESEARCH ANALYSIS

G2015 Macmillan Publishers Limited. All rights reserved

Disease and engineered Ga mutationsWe analysed disease-causing mutations in the human population andfound three key positions that are mutated (Supplementary Table 3),resulting in constitutive Ga activity: (1) a variant of the transmissionmodule residue PheH5.8 in Ga11 causes autosomal dominant hypocal-caemia type 2 by becoming constitutively active28, possibly by desta-bilizing the contact between H5 and H1; (2) Gai variant Alas6h5.3Ser, aposition important for H1 stabilization and GDP binding, causes tes-totoxicosis by constitutively activating adenylate cyclase29; and (3)Ga11 variant ArgH1.9Cys causes autosomal dominant hypo-parathyr-oidism30, possibly by affecting H-domain opening and GDP release.We also analysed previously performed perturbation experiments ondifferent Ga types using the CGN system and can explain how thesemutations might affect the activation mechanism (Fig. 5a andSupplementary Note). Furthermore, comprehensive alanine-scanningmutagenesis performed on Gai1, coupled with thermostability assaysof the mutants in the GDP-bound or nucleotide-free state coupled torhodopsin25, is also consistent with the analysis performed here(Fig. 5b). For example, mutations in the H5 interface module mainlyaffect Gai1–rhodopsin complex stability, whereas mutations in the H5transmission module primarily affect the nucleotide-bound stateof Gai1. Alanine mutations in H1 highly destabilize the GDP-boundstate, but not the Gai1–rhodopsin complex. In addition, mutatingresidues in the conserved p–p cluster that interact with PheH5.8

significantly affects Gai1 stability in the GDP-bound form, but notin the receptor-bound form (see ref. 25). Taken together, our analysis,the CGN numbering system and the universal activation mechanismdescribed here provides a unified framework to relate and interpret anumber of independent experimental and disease mutations in differ-ent Ga subfamilies.

Evolution of Ga activation mechanismTo understand how allosteric regulation might have evolved in Ga, wecompared the crystal structures of each equivalent signalling state of Gaand Ras. We found that H5 and H1 have significantly changed theirfunctional role. In Ras, H5 has a disordered extension (hyper-variableregion) that is post-translationally modified for membrane anchoring31.In Ga, the equivalent region forms the GPCR interface module(the N-terminal disordered region of HN is the membrane anchor32).The nucleotide-binding N-terminal part of H1 is conserved in itssequence and its structural orientation in both Ras and Ga (Fig. 6 andExtended Data Fig. 6). In contrast, the central part of H1, which contactsthe H-domain hinge, is only conserved in Ga. The C-terminal part of H1has a different role in Ras and Ga. While the last turn of H1 in Ras foldsback to bind the guanine moiety via a conserved p–p stack, the equival-ent region in Ga forms the metastable part of H1 that remains helical inthe GDP-bound inactive state due to contacts with the N-terminaltransmission module of H5. These contacts are missing in Ras, as theC terminus of H1 and the N terminus of H5 are each three residuesshorter, and H1 in Ras is stable without the interactions with H5. Thismeans that although Ras and Ga are evolutionarily related and share thesame architecture, minor but crucial differences in the number andpattern of non-covalent contacts between H5 and H1 have allowedthe emergence of an allosteric mechanism for GDP release in Ga.Small extensions in H1 and H5 permit H1 to sense whether H5 is boundto the GPCR. The disordered C-terminal tail of H5 provides both con-served and variable interface residues that allow for a conserved Gaactivation mechanism and yet permit the evolution of receptor-bindingspecificity.

DiscussionOur analysis suggests that GPCRs interact and activate Ga subunitsthrough a highly conserved mechanism in which the interruption ofthe contacts between H1 and H5 is a key step for GDP release. In thissense, while H1 is the molecular switch for GDP release, H5 is the distaltrigger that is ‘pulled’ upon receptor binding. Given that Ga proteinsbelong to the same fold, the existence of evolutionarily conserved resi-dues per se is not surprising. However, the observation that (1) theconserved residues form a network of non-covalent contacts that linksthe GPCR-binding site with the GDP-binding pocket and that (2) thisnetwork of contacts is consistently re-organized upon receptor bindingsuggests that this mechanism might constitute the common conservedset of structural changes for the allosteric release of GDP (Supple-mentary Note). While the conserved residue contacts are crucial forGa activation, non-conserved positions can still be important for allos-teric activation in distinct Ga proteins10. Thus, the identified residues arenecessary but not sufficient for G-protein activation. The variable inter-face residues, as well as the bc subunits, could have important roles inreceptor binding specificity for individual proteins. Thus, the conserveduniversal mechanism probably represents the ‘skeleton’ that can beincorporated into different contexts in different Ga proteins to maintaina conserved mechanism of allosteric activation and yet permit specificbinding to the receptor.

A comparison to small G proteins revealed how Ga evolved to bindGPCRs at a site that is distal to the GDP binding pocket. Emergence ofshort regions in H5 and H1 that can undergo structural transitionsseem to have been co-opted to make a new GEF (receptor) interfaceand link it to the GDP binding site. In such a system, displacementof a secondary structure element (H5) upon receptor binding cantransmit information by re-organizing key non-covalent contacts

Gαi1–GPCR complex stability

–50 –25 0 25

G.H1.1G.H1.2G.H1.3G.H1.4G.H1.5G.H1.6G.H1.7G.H1.8G.H1.9G.H1.10G.H1.11G.H1.12

G.H5.1G.H5.2G.H5.3G.H5.4G.H5.5G.H5.6G.H5.7G.H5.8G.H5.9G.H5.10G.H5.11G.H5.12G.H5.13G.H5.14G.H5.15G.H5.16G.H5.17G.H5.18G.H5.19G.H5.20G.H5.21G.H5.22G.H5.23G.H5.24G.H5.25G.H5.26

Change in relative complex stability (%)

Gαi1–GDP stability

ΔTm (ºC)

–15 –10 –5 0

Inte

rface m

od

ule

Tra

nsm

issio

n m

od

ule

Inte

rface m

od

ule

Tra

nsm

issio

n m

od

ule

H5 H

5

H1

*

βγ α

H1

H5 35

References

Gi

ThrG.H5.1 38Ala mutation causes constitutive GDP/GTP exchangeGt

ValG.H5.6 38Ala mutation causes constitutive GDP/GTP exchangeGt

Gα type

PheG.H5.8 10,3828

Gt,i,1

Ala and Cys mutations accelerate GDP exchangeAutosomal dominant hypocalcaemia type 2

H5 modules 36,37Gt

Gly5-spacer or Ala4-spacer between H5 modulesdecouples Gα activation from GPCR binding

Changing H5 helicity induces GDP release

PheG.S3.3 10Gi Cys mutation accelerates GDP exchange rate

AlaG.s6h5.3 29GsSer mutation results in constitutive activationand causes testotoxicosis

ArgG.H1.9 30G11

Leu or Cys mutations cause autosomal dominanthypoparathyroidism

EffectPositiona

b

R*

βγ α

Figure 5 | Mutational studies support the universal Ga activationmechanism. a, Disease and engineered mutations can be explained by themodel of Ga activation in different Ga subfamilies. References10, 28–30, 35–38 arecited in this figure. b, Comparison of the stability of Ga–GDP (DTm; uC) andthe Gabc–GPCR complex (change in relative complex stability; %) by Gai1

alanine mutagenesis of every position in H1 and H5. Asterisk indicates thatmutant FH5.8A is not stable in the receptor-free state but can still form thecomplex with the receptor.

0 0 M O N T H 2 0 1 5 | V O L 0 0 0 | N A T U R E | 5

ANALYSIS RESEARCH

G2015 Macmillan Publishers Limited. All rights reserved

between conserved residues that connect different secondary structureelements. Thus, a common ancestor of the GTPase fold might haveprovided the structural framework that can be perturbed by GPCRbinding through the interruption of H5–H1 contacts. This ‘new’ allos-teric site for GEFs is physically separated from the Ga effector andregulator binding interfaces (Fig. 6a), and could have provided the basisfor the expansion of the GPCR family without affecting the down-stream signalling factors.

Our findings suggest that in addition to evolving extensive inter-faces and allosteric ‘wires’ as observed in other proteins13,33, anothersolution for allosteric communication is evolving short segments thatundergo disorder-to-order transitions upon a trigger (for example,receptor binding). This mechanism involves the re-organization of anetwork of existing contacts to induce conformational changes thataffect a distal site without the requirement of directly contacting resi-dues but linked by the same secondary structure (for example, theinterface and transmission module of H5 make distinct sets of contactsbut are linked through the protein backbone; like a puppet on a string).Since disordered and loop regions tolerate more sequence changesthan structured regions34, an important implication is that such seg-ments allow for the independent evolution of regions that are import-ant for binding specificity but still maintain a conserved allostericactivation mechanism. Generalizing this principle, we suggest thatdisordered segments that can undergo structural transitions (regulatedfolding or unfolding) and thereby re-organize existing networks ofcontacts within structured regions of proteins could have an importantrole in other protein families and may be exploited in protein engin-eering applications.

Online Content Methods, along with any additional Extended Data display itemsandSourceData, are available in the online version of the paper; references uniqueto these sections appear only in the online paper.

Received 5 November 2014; accepted 16 June 2015.

Published online 6 July 2015.

1. Vetter, I. R. & Wittinghofer, A. The guanine nucleotide-binding switch in threedimensions. Science 294, 1299–1304 (2001).

2. Leipe, D. D., Wolf, Y. I., Koonin, E. V. & Aravind, L. Classification and evolution ofP-loop GTPases and related ATPases. J. Mol. Biol. 317, 41–72 (2002).

3. Rojas, A. M., Fuentes, G., Rausell, A. & Valencia, A. The Ras protein superfamily:evolutionary tree and role of conserved amino acids. J. Cell Biol. 196, 189–201(2012).

4. Anantharaman, V., Abhiman, S., de Souza, R. F. & Aravind, L. Comparativegenomics uncovers novel structural and functional features of the heterotrimericGTPase signaling system. Gene 475, 63–78 (2011).

5. Rasmussen, S. G. et al. Crystal structure of the b2 adrenergic receptor–Gs proteincomplex. Nature 477, 549–555 (2011).

6. Chung, K. Y. et al. Conformational changes in the G protein Gs induced by the b2adrenergic receptor. Nature 477, 611–615 (2011).

7. Preininger, A. M., Meiler, J. & Hamm, H. E. Conformational flexibility and structuraldynamics in GPCR-mediated G protein activation: a perspective. J. Mol. Biol. 425,2288–2298 (2013).

8. Oldham, W. M., Van Eps, N., Preininger, A. M., Hubbell, W. L. & Hamm, H. E.Mechanism of the receptor-catalyzed activation of heterotrimeric G proteins.Nature Struct. Mol. Biol. 13, 772–777 (2006).

9. Westfield, G. H. et al. Structural flexibility of the Gas a-helical domain in the b2-adrenoceptor Gs complex. Proc. Natl Acad. Sci. USA 108, 16086–16091 (2011).

10. Alexander, N. S. et al. Energetic analysis of the rhodopsin-G-protein complex linksthe a5 helix to GDP release. Nature Struct. Mol. Biol. 21, 56–63 (2014).

11. Neves, S. R., Ram, P. T. & Iyengar, R. G protein pathways. Science 296, 1636–1639(2002).

12. Chothia, C. & Lesk, A. M. Helix movements and the reconstruction of the haempocket during the evolution of the cytochrome c family. J. Mol. Biol. 182, 151–158(1985).

13. Suel, G. M., Lockless, S. W., Wall, M. A. & Ranganathan, R. Evolutionarily conservednetworks of residues mediate allosteric communication in proteins. Nature Struct.Biol. 10, 59–69 (2003).

14. del Sol, A., Fujihashi, H., Amoros, D. & Nussinov,R. Residuescrucial for maintainingshort paths in network communication mediate signaling in proteins. Mol. Syst.Biol. 2 (2006).

15. Kornev, A. P., Haste, N. M., Taylor, S. S. & Eyck, L. F. Surface comparison of activeand inactive protein kinases identifies a conserved activation mechanism. Proc.Natl Acad. Sci. USA 103, 17783–17788 (2006).

16. Zhang, X., Perica, T. & Teichmann, S. A. Evolution of protein structures andinteractions from the perspective of residue contact networks. Curr. Opin. Struct.Biol. 23, 954–963 (2013).

17. Deupi, X. et al. Stabilized G protein binding site in the structure of constitutivelyactive metarhodopsin-II. Proc. Natl Acad. Sci. USA 109, 119–124 (2012).

18. Standfuss, J. et al. The structural basis of agonist-induced activation inconstitutively active rhodopsin. Nature 471, 656–660 (2011).

19. Choe, H. W. et al. Crystal structure of metarhodopsin II. Nature 471, 651–655(2011).

20. Scheerer, P. et al. Crystal structure of opsin in its G-protein-interactingconformation. Nature 455, 497–502 (2008).

21. Slessareva, J. E. et al. Closely related G-protein-coupled receptors use multiple anddistinct domains on G-proteina-subunits for selective coupling. J. Biol. Chem. 278,50530–50536 (2003).

22. Oldham, W. M. & Hamm, H. E. Heterotrimeric G protein activation by G-protein-coupled receptors. Nature Rev. Mol. Cell Biol. 9, 60–71 (2008).

23. Dratz, E. A. et al. NMR structure of a receptor-bound G-protein peptide. Nature363, 276–281 (1993).

24. Tompa, P., Davey, N. E., Gibson, T. J. & Babu, M. M. A million peptide motifs for themolecular biologist. Mol. Cell 55, 161–169 (2014).

25. Sun, D. et al. Probing Gai1 protein activation at single amino acid resolution. NatureStruct. Mol. Biol. (in the press).

H1

H5

H1

H5

Gα–GEF interface

Ras–GEF interface

Gα

Inactive state GEF-bound state

H5

H1

H5

H1

RasH5

H1

H5

H1

GAP

EffectorRas

GEF

GEF

GAP

βγ

Effectorα

Disorder-to-order

Order-to-disorder

No significant change

No significant change

Displa

cem

ent

GαRas

H1GαRas

H5

a

c

bN-terminal lipid anchor

C-terminal

lipid anchor

Helical region

Disordered region

Variable/missing position

Figure 6 | H5–H1 interaction permits the allosteric activation mechanism.a, GEF interaction surfaces (red) for Ga (3sn6) and the small G protein humanHRas (PDB 1bkd) in the same orientation. b, H1 and H5 of the inactive(1got, 4q21) and GEF-bound state (3sn6, 1bkd) for Ga (blue) and HRas (grey).c, Consensus sequences of equivalent residues of H5 and H1 in Ga and Ras. The

helical region is highlighted in a grey background and the H5 disorderedregion is shown in a dashed grey box. Conserved residues that contact GDP(green), the H-domain (light blue) or form contacts between H5 and H1 (darkblue) are highlighted. See Extended Data Fig. 6.

6 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 5

RESEARCH ANALYSIS

G2015 Macmillan Publishers Limited. All rights reserved

26. Mnpotra, J. S. et al. Structural basis of G protein-coupled receptor-Gi proteininteraction: formation of the cannabinoid CB2 receptor-Gi protein complex. J. Biol.Chem. 289, 20259–20272 (2014).

27. Hanson, P. I. & Whiteheart, S. W. AAA1 proteins: have engine, will work. Nature Rev.Mol. Cell Biol. 6, 519–529 (2005).

28. Nesbit, M. A. et al. Mutations affecting G-protein subunit a11 in hypercalcemia andhypocalcemia. N. Engl. J. Med. 368, 2476–2486 (2013).

29. Iiri, T., Herzmark,P., Nakamoto, J.M., van Dop,C.& Bourne,H.R.Rapid GDP releasefromGsa inpatientswithgainand lossofendocrine function.Nature371, 164–168(1994).

30. Li, D. et al. Autosomal dominant hypoparathyroidism caused by germlinemutation in GNA11: phenotypic and molecular characterization. J. Clin.Endocrinol. Metab. 99, E1774–E1783 (2014).

31. de la Vega, M., Burrows, J. F. & Johnston, J. A. Ubiquitination: Added complexity inRas and Rho family GTPase function. Small GTPases 2, 192–201 (2011).

32. Oldham, W. M. & Hamm, H. E. Structural basis of function in heterotrimeric Gproteins. Q. Rev. Biophys. 39, 117–166 (2006).

33. Cui, Q. & Karplus, M. Allostery and cooperativity revisited. Protein Sci. 17,1295–1307 (2008).

34. Brown, C. J., Johnson, A. K., Dunker, A. K.& Daughdrill, G.W. Evolution and disorder.Curr. Opin. Struct. Biol. 21, 441–446 (2011).

35. Tanaka, T. et al. ahelix content of G protein or subunit is decreased upon activationby receptor mimetics. J. Biol. Chem. 273, 3247–3252 (1998).

36. Natochin, M., Moussaif,M.&Artemyev,N.O.Probing themechanism of rhodopsin-catalyzed transducin activation. J. Neurochem. 77, 202–210 (2001).

37. Marin, E. P., Krishna, A. G. & Sakmar, T. P. Disruption of the a5 helix of transducinimpairs rhodopsin-catalyzed nucleotide exchange. Biochemistry 41, 6988–6994(2002).

38. Marin, E. P., Krishna, A. G. & Sakmar, T. P. Rapid activation of transducin bymutations distant from the nucleotide-binding site: evidence for a mechanisticmodel of receptor-catalyzed nucleotide exchangebyG proteins. J. Biol. Chem. 276,27400–27405 (2001).

Supplementary Information is available in the online version of the paper.

Acknowledgements We thank U. F. Lang, C. Chothia, R. Weatheritt, S. Balaji,H. Harbrecht, A. Morgunov, G. Murshudov and N. S. Latysheva for their commentson this work. This work was supported by the Medical Research Council(MC_U105185859; M.M.B., T.F., M.K., A.J.V. and C.N.J.R.; MC_U105197215; C.G.T.),the Swiss National Science Foundation grants 141898, 133810, 31-135754 (D.B.V.),the MRC Centenary Award (A.J.V.), the AFR scholarship from the Luxembourg NationalResearch Fund (C.N.J.R.) and the Boehringer Ingelheim Fond (T.F.). M.M.B. is also aLister Institute Research Prize Fellow.

Author Contributions T.F. collected data, wrote scripts and performed all the analysis.C.N.J.R. helped with data analysis and writing the scripts for structure analysis with T.F.A.J.V. helped T.F.with data analysis. D.S. and D.B.V. contributed experimental results onalanine scanning. M.K. prepared the CGN webserver. C.G.T. helped with aspects of datainterpretation. T.F. and M.M.B. designed the project, analysed the results and wrote themanuscript. All authors read and provided their comments on the draft. M.M.B.supervised the project.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. The authors declare no competing financial interests.Readers are welcome to comment on the online version of the paper. Correspondenceand requests for materials should be addressed to T.F. (tflock@mrc-lmb.cam.ac.uk) orM.M.B. (madanm@mrc-lmb.cam.ac.uk).

0 0 M O N T H 2 0 1 5 | V O L 0 0 0 | N A T U R E | 7

ANALYSIS RESEARCH

G2015 Macmillan Publishers Limited. All rights reserved

METHODSNo statistical methods were used to predetermine sample size.Generation of sequence and structure data sets. Identification of relevant Gaprotein structures. All structures related to the Ga protein family (Pfamfamily: PF00503) were collected from Pfam (release 27.0) and Ensembl39

using the R BioMart interface40. In addition, the identified 973 Gahomologue sequences (see below) were scanned against the entireProtein Data Bank (PDB) database using the BLAST algorithm41 toensure all Ga-containing structures were identified. 91 PDB entries(146 Ga chains) were identified, of which two were obsolete (2pz3retracted, 2ebc superseded by 3umr). Crystallographic data and coord-inate files were retrieved from the RCSB PDB API (Tuesday 4 February2014 at 16.00 PST). Ga structures from the parasite Entamoeba histo-lytica (4fid) and Arabidopsis thaliana (2xtz), as well as non-full-lengthGa (1aqg and 1lvz are solution NMR studies of the C-terminal helix ofGa, 3rbq contains the 11 amino acid N-terminal part of transducinbound to UNC119), were excluded from the analyses. Four structuresof the last 10 C-terminal amino acid residues of Gat bound to rhodopsin(2x72, 3dqb and 3pqr) or meta-rhodopsin (4a4m) were used for theGPCR–Ga interface analysis. Five PDB entries had no publicationassociated and were manually traced back to their original articles:3umr was published in Johnston et al.42 and 4g5o, 4g5q, 4g5r, 4g5s werediscussed in a study by Jia et al.43. The final set of structures in ouranalyses span orthologues from human, mouse, rat and cow and encom-pass twelve different Ga genes from eight different Ga subfamilies(GNAI1, GNAI3, GNAO, GNAS2, GNAT1, GNAQ, GNA12,GNA13), thereby representing all Ga families (Gas, Gai, Gaq andGa12). A full list of all retrieved PDBs is provided in SupplementaryTable 1.Identification of canonical human Ga protein sequences and paraloguealignment. All relevant human Ga protein isoforms and variants wereobtained from Ensembl39 using R (full list in Supplementary Table 1).The ‘canonical’ protein sequences for each of the 16 human Ga genes, asdefined by Uniprot44, were used as representative sequences for eachhuman Ga gene throughout this work. The sequences were alignedusing MUSCLE45 and were manually refined using the consensus sec-ondary structure as a guide (see below). Phylogenetic relationships of Gawere obtained from Treefam46 (family TF300673). The cladogram of the16 canonical human Ga protein alignment was built with thePhylogeny.fr web service47 using the PhyML v3.0 algorithm48 with theSH-like Approximate Likelihood-Ratio Test using the Jones–Taylor–Thornton substitution matrix and TreeDyn49 for visualization.Orthologue alignments of one-to-one Ga orthologues of 16 human Gagenes. Phylogenetic relationships of Ga sequences were collected fromTreeFam46, the Orthologous MAtrix (OMA) database50 andEnsemblCompara GeneTrees (Compara)51 using R scripts. Comparahad the highest fraction of complete Ga sequences for each humanGa gene, except for Gas, for which OMA had a better sequence coverage.In total, 973 genes from 66 organisms were used, of which 773 were one-to-one orthologues. To build an accurate, low-gap alignment of such anumber of sequences, 16 independent orthologous alignments for eachhuman Ga gene were first created by aligning one-to-one orthologuegroups using the PCMA algorithm52 followed by manual refinement.Subsequently, each orthologue alignment was cross-referenced to theCGN (see below) by referencing its respective human sequence to thehuman paralogue alignment. Conservation scores of each CGN positionwere calculated using both sequence identity and sequence similarity,based on the BLOSUM62 substitution matrix (Supple-mentary Note) using all complete sequences of the cross-referencedalignments (561 sequences). Sequence conservation was mapped ontoPDB structures (Supplementary Data) and visualized by generatingPDB files with B-factors substituted by conservation scores.Phylogenetic distances. The evolutionary distance of the retrievedsequences relative to human was evaluated with TimeTree53. Ga one-to-one orthologues extend back to chordate (sea squirts; Ciona savignyiand Ciona intestinalis for Ga15), separated around 722.5 million years

from Homo sapiens, and the most ancestral one-to-many orthologueextends back to Opisthokonta (yeast; Saccharomyces cerevisiae), sepa-rated by 1,215 million years from human. In this work, we only inves-tigated G proteins from organisms that have a GPCR–G-protein system.Since plants do not encode GPCRs and the heterotrimeric G proteins areknown to be auto-activated, we did not consider the plant G proteins inour analysis.Development of a common Ga numbering (CGN) system. Common Ganumbering system. Comparative analyses of different protein structuresand sequences to infer general principles of a protein family require away of relating structural, genomic, or experimental data from differentstudies to each topologically equivalent position on homologous pro-teins. For GPCRs, the Ballesteros–Weinstein numbering scheme54

enables the referencing of positions in the transmembrane helices ofdifferent GPCRs, not considering loop regions. We sought to developa common G protein numbering (CGN) system that includes loopregions and describes Ga residues in three levels of detail (DSP), similarto a postal address. D refers to the structural domain and is optional(catalytic GTPase domain: G; helical domain: H); S stands for one of the37 consensus secondary structure elements (including loops) of theconserved Ga topology; and P relates to the corresponding residueposition within the consensus secondary structure element mapped toan alignment of all ‘canonical’ human Ga sequences (Extended Data Fig.1). For a detailed guide of how to use the CGN and map any Ga protein,please refer to the CGN webserver (http://www.mrc-lmb.cam.ac.uk/CGN) and Supplementary Note.Mapping structures to Uniprot sequences. Since several Ga protein struc-tures represent chimaeric G proteins, have peptide tags, or contain pointmutations, each residue/position in a PDB structure was mapped to itsUniprot sequence(s) using the Structure Integration with Function,Taxonomy and Sequence (SIFTS)55 webserver followed by a manualvalidation for missing positions. This allowed assigning residue posi-tions of each Ga structure to their equivalent positions in the humanparalogue alignment and the orthologue alignments.Determination of domain D and consensus secondary structure S.Secondary structure assignments were made for each Ga structure usingthe STRIDE algorithm56. The consensus secondary structure elements(SSEs) were determined by considering the most prominent secondarystructure type at each topologically equivalent Ga position when com-paring the secondary structure assignment of all 80 Ga structures (meanand standard deviation of secondary structure type at each CGN posi-tion were calculated). Topologically equivalent positions had a highagreement in their SSE assignment and showed well-defined flankingregions (Supplementary Note). In addition, the assigned consensus SSEswere manually confirmed through a 3D-structure alignment usingMUSTANG57, from which the domains (G-domain and H-domain)were defined. The Ga SSE nomenclature follows a standardized expan-sion of the previously defined nomenclature58: uppercase letters H and Srepresent helices or sheets, respectively. SSEs of the G domain follow anumerical identifier (H1, H2, …, H5 and S1, S2, …, S6 with the excep-tion of HG and HN); SSEs of the H domain have an alphabetical identi-fier (HA, HB, …, HF), starting from the N- to the C terminus of Ga. TheN-terminal region that forms a membrane-anchored helix is defined asHN. Systematic identifiers for historical names of some loop regions(switch regions, P-loop, etc.) were derived by concatenating the flankingSSE names using lowercase letters; for instance s6h5 refers to the loopbetween S6 and H5. A reference table including the historical loopnames is provided in Extended Data Fig. 1 and Supplementary Table 2.Determination of position P. P describes the amino acid position withinan SSE, as determined by mapping the consensus secondary structure tothe human paralogue alignment (Extended Data Fig. 1 andSupplementary Table 2). Insertions in orthologues are annotated P-i,where i stands for the number of inserted residues after position P, forinstance Arg334H4.27-2 for the second amino acid of an insertion afterposition 27 of helix H4, found in pufferfish (Tetraodon nigroviridis) Gas

(Supplementary Note).

RESEARCH ANALYSIS

G2015 Macmillan Publishers Limited. All rights reserved

Consensus non-covalent contact networks between conserved residues. Non-covalent residue contact networks. Non-covalent contacts between resi-dues of a protein define its topology, conformation and stability. Foreach of the 80 Ga protein structures, a local version of the RINerator 0.5package from 201459 was used to calculate H-bonds and van der Waalsinteractions between residues. Matrices of the all-against-all atomicdistances of all residue contacts within each structure were computedusing R and the bio3d package60. Non-canonical interaction such as p–pstacking were identified with NCI61. All other calculations, analysis andprocessing were performed using custom written scripts in R.Assignment of Ga structures to signalling states. Structural differencesbetween Ga seem to arise from a convolution of the conformationalstate, binding partner, and Ga protein type and species(Supplementary Note). To identify the non-covalent contacts of a struc-ture that are crucial for each signalling state, and independent of the Gaprotein type and species, all Ga structures were assigned to one of thefour different Ga signalling states depending on (1) the bound ligand,and (2) the interaction partner (Supplementary Table 1). The four statesare (1) heterotrimeric GDP-bound state (inactive state), (2) nucleotide-free heterotrimeric receptor bound complex (GEF-bound state), (3)GTPcS and potentially ‘effector’-bound state (active state), and (4)RGS-bound GDP1ALF hydrolysis transition state (GAP-bound state).Eleven structures are in the inactive state, one full-length structure (andfour structures of the C-terminal Ga peptide in complex with a GPCR)in the GPCR-bound state, 25 have GTPcS bound or/and are co-crystal-lized with their downstream effectors, and 40 structures have Ga in theGTP-hydrolysis transition state with GDP and aluminium fluoride(ALF) bound (GDP1ALF) and/or are co-crystalized with their RGSor a GTP-hydrolysis promoting peptide mimicking the RGS bindinginterface (for example, Go-Loco motif). Two structures (1cip, 1svs) hadnon-standard Ga ligands bound, and 2zjz62 did not have a detaileddescription of its biochemical relevance, and thus were not assigned toany signalling state. Eleven structures were identified as chimaeras and21 included mutations (Supplementary Note). The publications describ-ing the protein structure of each PDB entry were checked to confirm therelevance of the assigned signalling state.Consensus contacts between conserved residues. To compare residuecontact networks (RCNs) from different structures, topologically equi-valent positions were cross-referenced with the CGN system. All RCNanalyses, consensus RCN calculation, and conservation analysis wereconducted using customized R scripts: matrices representing theabsence or presence of non-covalent contacts between each possible pairof CGN residues in each PDB structure were computed (SupplementaryFig. 1). The consensus contacts of each signalling state were computed asthe probability to find a contact in all structures of the state. Structuremodels can differ in their number of equivalent residues due to missingelectron density, not fully fitted models, or truncations for crystal-lographic purposes. Thus, each consensus contact probability was nor-malized by the number of structures of the state that have the respectiveresidue pair, in order to distinguish the absence of a contact from theabsence of an equivalent position in a single PDB. To expand the struc-tural analysis to other Ga proteins for which only sequence data wasavailable, sequence conservation was mapped to each CGN residue (seeabove).Visualization of consensus contacts and identification of universal struc-tural motifs. The consensus contacts between conserved residues in thedifferent signalling states were visualized to investigate the contact re-organization in detail. For 2D visualization, the respective consensusRCNs were exported to Cytoscape63 using the RCytoscape interface64.For 3D visualization, R was used to create consensus RCNs in PyMol(The PyMOL Molecular Graphics System, Version 1.5.0.4 Schrodinger,LLC.) by creating pseudo PDB structure coordinate files that showresidues as spheres from their C-alpha atoms and lines/edges betweenthem using CONECT entries. Information on sequence conservationwas mapped via the B-factor field of the pseudo PDB structures. Forsimplification, only contacts present in more than 90% of all structures

with a sequence identity .90% were shown as ‘consensus contacts’between conserved residues—this threshold was chosen based on thebimodal distribution of contact occurrence (Supplementary Fig. 2). Inaddition, only long-range interactions (.i 1 4) are shown for the con-sensus RCNs. It is important to note that these cut-offs were appliedonly for visualization purposes, while for the analysis, no cut-off wasneeded. All relevant consensus contacts were additionally visuallyinspected for each of the 80 PDB structures by creating automatedPyMol sessions from R that superimpose all the 80 structures. To gen-erate RCNs between SSEs, the sum of all contacts of the respective SSE asdefined by the consensus SSE of the CGN were computed. Chimera65

was used to manually re-evaluate atomic contacts, and PyMol was usedto create publication-quality images.Interface analysis. Buried surface area and inter-Ga–GPCR residue con-tact networks. Inter-chain RCNs between Ga and the receptor (Gas andb2AR chains A and R in 3sn6, Gat C-terminal peptide and rhodopsinfrom chains B and A in 2x72, 3dqb, 3pqr, 4a4m) were calculated asdescribed above. The buried surface area (BSA) was obtained from thePDBe PISA (Proteins, Interfaces, Structures and Assemblies)66 XMLrepository and normalized by the accessible surface area for each residueposition. BSA and Ga–GPCR RCNs were mapped to the CGN and theBallesteros–Weinstein numbering, respectively. Sequence conservationfrom 561 complete Ga homologue sequences and 249 human non-olfactory class A GPCRs was mapped onto the interface to determinethe conserved ‘hotspot’ residues in the interface and visualized in PyMol(Supplementary Data). The BSA histogram, the visualization of theresidue interaction network per secondary structure elements, and thecorrelation of BSA per residue versus conservation were produced in Rand ggplot2.Force-field-based energy estimations. The per-residue energy contribu-tions to Ga monomer and Ga–GPCR complex stability were calculatedusing FoldX 3.0, which uses energy terms weighted by empirical datafrom protein engineering experiments to provide a quantitative estima-tion of each residue’s contribution to protein stability and protein com-plex stability (http://foldx.crg.es/). For the interface analysis, the 3sn6structure was energy minimized with the FoldX ‘repair pdb’ functionand subsequently, the per-residue energy contributions for both theGas–b2AR complex and the monomers in isolation were calculatedusing the FoldX ‘sequence detail’, ‘analyse complex’, and ‘stability’ func-tions at 298K, pH 7.0, and 0.05M ionic strength. The per-residue energycontributions to complex stability were calculated as the differencebetween the energy contributions of each residue in the monomer andcomplex (DDGinterface) and visualized with R (Extended Data Fig. 2a).For energy contributions of each residue within Ga monomers(Extended Data Fig. 4b), the average energy contribution and standarddeviation for each Ga position was computed after running the FoldX‘stability’ and ‘sequence detail’ functions at 298K, pH 7.0, and 0.05Mionic strength for each of the 79 non-complex structures.Disorder propensity calculations for all Ga homologue sequences andstructures. The disorder propensity of each of the 561 complete Gahomologue sequences was calculated with IUPred67 (prediction-type-setting: ‘short disorder’). The missing structure positions were identifiedwith bio3d package60 (Extended Data Fig. 2b).New and published mutational studies of different Ga classes. Identificationof mutations, mutant structures and chimaeras. Additional literature onGa mutations was retrieved with the text mining tool MutationMapper68

and manually validated and filtered for correct hits/search results.Disease mutations were retrieved from the Database of SingleNucleotide Polymorphisms (dbSNP) and the Catalogue of SomaticMutations in Cancer (COSMIC)69 with biomaRt40, and from theHuman Gene Mutation Database (HGDM)70. Mutations, chimaerasand peptide tags in the analysed structures were identified by comparingtheir Uniprot sequence to their PDB sequence using SIFTS55 and miningPDBe annotations. All mutation data were mapped back to the CGNand visualized on their respective human Ga structure.

ANALYSIS RESEARCH

G2015 Macmillan Publishers Limited. All rights reserved

Alanine scanning and stability of Gai. The alanine scanning expressionlibrary of Gai1 was prepared as reported before71. The recombinant Gai1

alanine mutants were expressed in 24 well plates, purified by standardNi-NTA affinity chromatography followed by buffer exchange using 96-well filter plates. The bovine rhodopsin and bc subunits were preparedfrom bovine retinas72. The melting temperature of each alanine mutantupon addition of GDP or GTPcS was measured by differential scanningfluorimetry assay. The effect of each alanine Gai1 mutant on R*–Gi

complex formation and complex stability were measured by a high-throughput assay based on native gel electrophoresis. Detailed methodsand protocols are provided in ref. 25.Ga versus Ras comparison. The Ras conformational cycle was featured inthe RSCB PDB73 April 2012 PDB-101 Molecule of the Month by DavidGoodsell (http://dx.doi.org/10.2210/rcsb_pdb/mom_2012_4), withhigh-resolution structures showing human HRas in its active GTPcS-bound state (PDB 5p2174) and the GDP-bound inactive state (PDB4q2175). 1bdk76 was used as representative of the HRas GEF-bound state.These Ras representative structures were combined with an alignment ofall human HRas paralogues identified from the OMA database50. Astructural alignment between the identified active and inactive Rasstructures and the corresponding active and inactive Ga (1got and3ums) was used to accurately map Ras positions to the CGN(Supplementary Data) despite the low sequence identity (,6%) betweenRas and Ga. The number of atomic non-covalent contacts between helixH5 and helix H1 in the Ras and Ga structures was manually compared inChimera for the structures 1got and 4q21.

39. Flicek, P. et al. Ensembl 2014. Nucleic Acids Res. 42, D749–D755 (2014).40. Kasprzyk, A. BioMart: driving a paradigm change in biological data management.

Database 2011, bar049 (2011).41. Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J. Basic local alignment

search tool. J. Mol. Biol. 215, 403–410 (1990).42. Johnston, C. A. et al. Structural determinants underlying the temperature-sensitive

nature of a Ga mutant in asymmetric cell division of Caenorhabditis elegans. J. Biol.Chem. 283, 21550–21558 (2008).

43. Jia, M. et al. Crystal structures of the scaffolding protein LGN reveal the generalmechanism by which GoLoco binding motifs inhibit the release of GDP from Gai.J. Biol. Chem. 287, 36766–36776 (2012).

44. Magrane,M.&Consortium,U.UniProtKnowledgebase: ahub of integratedproteindata. Database (Oxford) 2011, bar009 (2011).

45. Edgar, R. C. MUSCLE: multiple sequence alignment with high accuracy and highthroughput. Nucleic Acids Res. 32, 1792–1797 (2004).

46. Ruan, J. et al. TreeFam: 2008 Update. Nucleic Acids Res. 36, D735–D740 (2008).47. Dereeper, A. et al.Phylogeny.fr: robustphylogenetic analysis for the non-specialist.

Nucleic Acids Res. 36, W465–W469 (2008).48. Guindon, S. et al. New algorithms and methods to estimate maximum-likelihood

phylogenies: assessing the performance of PhyML 3.0. Syst. Biol. 59, 307–321(2010).

49. Chevenet, F., Brun, C., Banuls, A. L., Jacq, B. & Christen, R. TreeDyn: towardsdynamic graphics and annotations for analyses of trees. BMC Bioinform. 7, 439(2006).

50. Altenhoff, A. M., Schneider, A., Gonnet, G. H. & Dessimoz, C. OMA 2011: orthologyinference among 1000 complete genomes. Nucleic Acids Res. 39, D289–D294(2011).

51. Vilella, A. J. et al. EnsemblCompara GeneTrees: Complete, duplication-awarephylogenetic trees in vertebrates. Genome Res. 19, 327–335 (2009).

52. Pei, J., Sadreyev, R. & Grishin, N. V. PCMA: fast and accurate multiple sequencealignment based on profile consistency. Bioinformatics 19, 427–428 (2003).

53. Hedges, S. B., Dudley, J. & Kumar, S. TimeTree: a public knowledge-base ofdivergence times among organisms. Bioinformatics 22, 2971–2972 (2006).

54. Ballesteros, J. A.&Weinstein, H.ReceptorMolecularBiologyVol. 25 (Elsevier, 1995).55. Velankar, S. et al. SIFTS: Structure Integration with Function, Taxonomy and

Sequences resource. Nucleic Acids Res. 41, D483–D489 (2013).56. Heinig, M. & Frishman, D. STRIDE: a web server for secondary structure

assignment from known atomic coordinates of proteins. Nucleic Acids Res. 32,W500–W502 (2004).

57. Konagurthu, A. S., Whisstock, J. C., Stuckey,P. J.& Lesk, A. M. MUSTANG: a multiplestructural alignment algorithm. Proteins 64, 559–574 (2006).

58. Noel, J. P., Hamm, H. E. & Sigler, P. B. The 2.2 A crystal structure of transducin-acomplexed with GTPcS. Nature 366, 654–663 (1993).

59. Doncheva, N. T., Klein, K., Domingues, F. S. & Albrecht, M. Analyzing and visualizingresidue networks of protein structures. Trends Biochem. Sci. 36, 179–182 (2011).

60. Grant, B. J., Rodrigues, A.P., ElSawy, K.M., McCammon, J. A. &Caves, L. S.Bio3d: anR package for the comparative analysis of protein structures. Bioinformatics 22,2695–2696 (2006).

61. Babu, M. M. NCI: A server to identify non-canonical interactions in proteinstructures. Nucleic Acids Res. 31, 3345–3348 (2003).

62. Morikawa, T.et al.Crystallization and preliminary X-ray crystallographicanalysis ofthe receptor-uncoupled mutant of Ga1. Acta Crystallogr. F. 63, 139–141 (2007).

63. Lopes, C. T. et al. Cytoscape Web: an interactive web-based network browser.Bioinformatics 26, 2347–2348 (2010).

64. Shannon, P. T., Grimes, M., Kutlu, B., Bot, J. J. & Galas, D. J. RCytoscape: tools forexploratory network analysis. BMC Bioinform. 14, 217 (2013).

65. Pettersen, E. F.et al.UCSF Chimera–a visualization systemfor exploratory researchand analysis. J. Comput. Chem. 25, 1605–1612 (2004).

66. Krissinel, E. & Henrick, K. Inference ofmacromolecular assemblies from crystallinestate. J. Mol. Biol. 372, 774–797 (2007).

67. Dosztanyi, Z., Csizmok, V., Tompa, P. & Simon, I. IUPred: web server for theprediction of intrinsically unstructured regions of proteins based on estimatedenergy content. Bioinformatics 21, 3433–3434 (2005).

68. Vohra, S. & Biggin, P. C. Mutationmapper: a tool to aid the mapping of proteinmutation data. PLoS ONE 8, e71711 (2013).

69. Forbes, S. A. et al. COSMIC: exploring the world’s knowledge of somatic mutationsin human cancer. Nucleic Acids Res. 43, D805–D811 (2015).

70. Stenson, P. D. et al. The Human Gene Mutation Database: 2008 update. GenomeMed 1, 13 (2009).

71. Sun, D. et al. AAscan, PCRdesign and MutantChecker: a suite of programs forprimer design and sequence analysis for high-throughput scanning mutagenesis.PLoS ONE 8, e78878 (2013).

72. Maeda, S. et al. Crystallization scale preparation of a stable GPCR signalingcomplex between constitutively active rhodopsin and G-protein. PLoS ONE 9,e98714 (2014).

73. Berman,H.M.et al.The Protein DataBank.Nucleic AcidsRes.28, 235–242 (2000).74. Pai, E. F. et al. Refined crystal structure of the triphosphate conformation of H-ras

p21 at1.35 A resolution: implications for the mechanism of GTP hydrolysis. EMBOJ. 9, 2351–2359 (1990).

75. Milburn, M. V. et al. Molecular switch for signal transduction: structural differencesbetween active and inactive forms of protooncogenic ras proteins. Science 247,939–945 (1990).

76. Boriack-Sjodin, P. A., Margarit, S. M., Bar-Sagi, D. & Kuriyan, J. The structural basisof the activation of Ras by Sos. Nature 394, 337–343 (1998).

RESEARCH ANALYSIS

G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 1 | Human paralogue reference alignment forcommon Ga numbering system. a, Reference alignment of all canonicalhuman Ga paralogues. The domain (D), consensus secondary structure (S) and

position in the SSE of the human reference alignment (P) are shown on top ofthe alignment. b, Reference table of the definitions of SSEs used in the CGNnomenclature.

ANALYSIS RESEARCH

G2015 Macmillan Publishers Limited. All rights reserved

RESEARCH ANALYSIS

G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 2 | Energy estimation of the GPCR–Ga residuecontributions and Ga disorder propensity. a, Energy contribution of singleinterface residues to the Gas–b2AR complex calculated with FoldX (T 5 298K,pH 5 7.0, ionic strength 5 0.05M). Conserved Ga residues (blue sequencelogo) that were identified to form receptor–Ga inter-protein contacts withconserved GPCR residues (red sequence logo) are shown. The contact networkbetween residues of the b2AR and Gas is shown (red, conserved receptorresidue; blue, conserved Ga residue; grey, variable residues; spheres representCa positions and links represent non-covalent contact). b, Consensus disorder

plot for all Ga proteins. The mean value of the disorder propensity of all full-length Ga sequences (561 sequences; homologous to all 16 human Ga proteins)is shown as a black line; the standard deviation at each position is shown aslight red ribbon. The colour tone of the line indicates the number of gaps atan aligned position (black, no gaps). The left inset shows the disorderpropensity of H1. The right inset highlights that H5 is highly structured in itsN terminus, and has increased disorder propensity towards the C terminus,which is in agreement with the missing electron density in the 79 structures.

ANALYSIS RESEARCH

G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 3 | Rewiring of consensus contacts betweenconserved Ga residues upon receptor binding. CGN numbers and sequencelogo for consensus contacts within Ga in the inactive state (left) and GPCR-bound state (right) are shown. Receptor residues are shown in red; H5 residuesin dark blue; H1 residues in light blue; and GDP in green. The domains are

highlighted with a blue background (G-domain darker blue, H-domainlight blue). This figure highlights the most important consensus residuecontacts between conserved residues. Additional contacts in the right lobe arediscussed in the Supplementary Note and in ref. 25. For a full list of residuecontacts, please refer to Supplementary Data.

RESEARCH ANALYSIS

G2015 Macmillan Publishers Limited. All rights reserved

ANALYSIS RESEARCH

G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 4 | Details of helix H1 linking H5, GDP and theH-domain. a, This figure expands Fig. 4 from the main text to provide residue-level details of the role of helix H1. Residues forming contacts with H5 areshown in blue, with the H-domain in light blue and with GDP in green. Non-covalent consensus contacts between universally conserved residues at the SSElevel (left) and per residue-level (centre). Lines denote non-covalent contactsbetween residues. The degree of conservation is shown as sequence logo.Residues are numbered according to the CGN. Helix H1 is almost 100%conserved across all 16 Ga types and forms three structural motifs forinteractions with H5, the H-domain and GDP (right). b, Average per residueenergy contribution to Ga protein stability as calculated from 79 structures

from all four Ga subfamilies in the non-receptor-bound signalling states usingFoldX (T 5 298K, pH 5 7.0, ionic strength 5 0.05M). The average energycontribution is shown as dots, the standard deviation as bars. c, Per residuedetail of Ga-GDP and Ga-GSP (non-hydrolysable GTP analogue) consensuscontacts. The bar-plot shows the frequency of finding a contact mediated bytopologically equivalent positions with GDP/GSP. Number of side-chainand main-chain contacts are shown as dark grey and light grey bars,respectively. The degree of conservation of contacting residues (calculatedfrom the 561 complete Ga sequences) is represented in the right panel andthe consensus sequence for each position is shown.

RESEARCH ANALYSIS

G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 5 | Conserved structural motifs of Ga and knowndisease and engineered mutations. a, A universally conserved cluster ofp–p and hydrophobic interactions between S2 (PheG.S2.6) and S3 (PheG.S3.3), H1(MetG.H1.8 and HisG.H1.12) and H5 (PheG.H5.8) links H5 and H1 in the absence ofthe receptor. Upon receptor binding, residues within this motif (PheG.H5.8

and PheG.S3.3) interact with the conserved Pro and Leu of ICL2 of the receptoras has been shown for Gas (3sn6) and Gai (ref. 26). Interrupting the contactsbetween H5 and H1 seems to be the trigger for transmitting the signal ofGPCR binding to helix H1 (which interacts with GDP and the H-domain.) Theonly conserved residue contact between the H-domain and the G-domainthat is not in the hinge region is formed by a universally conserved salt bridge(H-domain ionic latch) between the very N-terminal end of HG of theG-domain (LysG.s5hg.1) and the loop connecting HD and HE in the H domain

(AspH.hdhe.5). The hinge region is formed by H1, the loop between H1 and HA,and HF. H1 interacts via (1) a cation–p interaction mediated by a universallyconserved residue with the loop connecting H1 and HA (LysG.H1.6 andTyrG.h1ha.4) and (2) a hydrophobic interaction with HF (LysG.H1.9 andLeuH.HF.5). b, Disease and engineered mutations that can be explained by theuniversal Ga activation mechanism mapped on a Ga protein. Ca positionof residues are shown as spheres; mutations at green positions causespontaneous GDP release by interrupting consensus contacts betweenconserved residues, thereby ‘mimicking’ the effect of receptor binding to Ga.Pink positions have also been reported to cause disease by constitutivelyactivating Ga. Insertion of an Ala4 or Gly5 after the yellow position separate theH5 transmission and interface module, thereby allowing GPCR bindingwithout triggering GDP release.

ANALYSIS RESEARCH

G2015 Macmillan Publishers Limited. All rights reserved

RESEARCH ANALYSIS

G2015 Macmillan Publishers Limited. All rights reserved

Extended Data Figure 6 | Helix H5–H1 interaction in Ga provides theallosteric GEF activation mechanism. a, Schematic representation ofstructural motifs on H1 that are shared or unique to Ga and Ras. While the partof H1 with the phosphate-binding motif is conserved across both proteinfamilies, the C-terminal part is conserved only in Ga. H1 in Ga has threeadditional residues that allow for extensive residue contacts between H1 andH5. In Ras, these interactions are missing and H5 and H1 are both 3 residuesshorter. The consensus sequence and secondary structure of equivalent

residues of H1 in Ga and Ras is also depicted. b, Comparison of the residuecontact network between topologically equivalent residues in H5 and H1 in thecorresponding inactive GDP-bound states of Ga (1got) and Ras (4q21).The weight of the link between SSEs denotes the number of atomic contacts.c, Sequence alignments of H1 and H5 of human Ga and Ras paralogues. Thesequence alignment was obtained based on cross-referencing the alignmentsusing the structures of Ga and Ras.

ANALYSIS RESEARCH

G2015 Macmillan Publishers Limited. All rights reserved