STRUCTURAL BIOLOGY

Structures of the M1 and M2muscarinic acetylcholinereceptor/G-protein complexesShoji Maeda1*, Qianhui Qu1,2*, Michael J. Robertson1,2,Georgios Skiniotis1,2†, Brian K. Kobilka1†

Muscarinic acetylcholine receptors are G protein–coupled receptors that respond toacetylcholine and play important signaling roles in the nervous system.There are fivemuscarinic receptor subtypes (M1R to M5R), which, despite sharing a high degree ofsequence identity in the transmembrane region, couple to different heterotrimericGTP-binding proteins (G proteins) to transmit signals. M1R, M3R, and M5R couple to theGq/11 family, whereas M2R and M4R couple to the Gi/o family. Here, we present andcompare the cryo–electron microscopy structures of M1R in complex with G11 and M2R incomplex with GoA. The M1R-G11 complex exhibits distinct features, including an extendedtransmembrane helix 5 and carboxyl-terminal receptor tail that interacts with G protein.Detailed analysis of these structures provides a framework for understanding themolecular determinants of G-protein coupling selectivity.

Muscarinic acetylcholine receptors (mAChRs)are family A G protein–coupled receptors(GPCRs) activated by the neurotransmitteracetylcholine. Members of this familyplay key roles in a variety of physiolog-

ical functions, including regulation of heart rate,smooth muscle contraction, glandular secretion,and memory formation (1). The mAChR familyconsists of five highly conserved subtypes (M1Rto M5R) that share 64 to 82% sequence identityand 82 to 92% sequence similarity in the trans-membrane region. However, mAChR subtypesdiffer in their tissue distribution and the type ofheterotrimeric GTP-binding protein (G protein)that they engage to instigate signaling. Hetero-trimeric G proteins consist of Ga, Gb, and Ggsubunits and are classified by the a subunit, whichdetermines GPCR coupling specificity. mAChRsubtypes M2R and M4R preferentially signalthrough Gi/o proteins. The Gi/o-protein family in-cludes Gi1, Gi2, Gi3, GoA, and GoB. Activation of Gi/o

proteins inhibit adenylyl cyclase and decreaseintracellular concentrations of adenosine 3′,5′-cyclicmonophosphate (cAMP). By contrast,M1R,M3R, and M5R predominantly couple to Gq/11

proteins (Gq and G11), leading to the activation ofphospholipase C and the increase of cytosolic Ca2+

(Fig. 1A) (2). None of the muscarinic receptorscouple efficiently to Gs proteins that activateadenylyl cyclase.M1R is abundantly expressed inthe central nervous system (CNS), whereas M2Ris more abundant in the peripheral tissues, in-cluding heart and colon (3). Studies of knockoutmice show that the M1R plays a role in memory

formation, whereas theM2R regulates heart rate(2, 4). TheM2R is also a presynaptic autoreceptorin the peripheral parasympathetic nervous sys-tem and therefore influences responses of othermuscarinic receptor subtypes.Owing to the high sequence homology, yet

distinct G-protein preferences, the mAChR sub-family has been used as a model system to studythe G-protein selectivity of GPCRs. So far, high-resolution structural information on mAChRshas mostly been limited to antagonist-boundinactive states with the exception of an active-state M2R structure stabilized by a G-proteinmimetic nanobody bound to a high-affinity agonistand a positive allosteric modulator (PAM) (5–8).The structure of rhodopsin bound to the trans-ducin peptide first revealed interactions betweenthe a5 helix of a G protein and an active GPCR(9). The first intact GPCR–G-protein structure tobe reported was that of the b2-adrenergic re-ceptor (b2AR)–Gs complex, determined by x-raycrystallography (10). The structure of the A2Areceptor in complex with a modified Gs protein(mini-Gs) was more recently also determined byx-ray crystallography (11). Efforts to understandsubtype selectivity on the basis of these struc-tures did not reveal a clear consensus sequenceon receptors that recognize the same G proteinbut concluded that receptors from differentsubfamilies evolved different mechanisms toactivate the same G protein (12). Recently, sev-eral family B and family A GPCR structures incomplex with either Gs or Gi/o proteins havebeen determined by single-particle cryo–electronmicroscopy (cryo-EM). These structures pro-vided further structural insights into G-proteinactivation and have revealed that the trans-membrane helix 6 (TM6) in GPCR-Gs complexesgenerally undergo a larger outward displace-ment than in GPCR-Gi/o complexes, resulting ina wider G-protein binding pocket that can ac-

commodate the bulkier C terminus of Gs (13–18).To extend our understanding of coupling speci-ficity to GPCR-Gq/11 complexes, we used single-particle cryo-EM to obtain the structures of activeM1R and M2R engaged with heterotrimeric G11

and GoA proteins, two distinct G proteins be-longing to the Gq/11 and Gi/o family, respectively.Comparison of these complexes provides struc-tural insights into the activation of the Gq/11

family of G proteins and the basis for the dis-tinct coupling preference of mAChRs for Gq/11

and Gi/o.

Sample preparation and cryo-EM

For this work, we usedG11iN andGoAiN, which aremodified forms of G11 and GoA, respectively. G11iN

is a chimeric G11 in which most of the aN helixis replaced with the equivalent region of Gi1 toimpart the ability to bind to scFv16 and stabilizethe nucleotide-free GPCR-G protein complex (19)(fig. S1). GoAiN is a modified GoA that has fourmutations in the region of the aNhelix that bindsscFv16 tomake this homologous stretch identicalto that of Gi1 (fig. S1). We made these mutationsin GoA because scFv16 was originally developedagainst Gi1, even though wild-type GoA also bindsscFv16 (19). Hereafter, we refer to them as G11

and GoA, because these modifications do not af-fect interactions with their respective receptors.M1R-G11-scFv16 and M2R-GoA-scFv16 complexeswere formed from purified components andstabilized by the addition of apyrase to removeresidual guanosine 5′-diphosphate (GDP). Thefinal size-exclusion chromatography profiles andthe SDS–polyacrylamide gel electrophoresis showpure and monodisperse samples before beingloaded onto grids, blotted, and frozen (figs. S2, Ato E, and S3, A to E). The M2R-GoA-scFv16 com-plex showed severe dissociation upon cryo-EMgrid preparation, resulting in 3 to 6% complexparticles out of all autopicked projections, com-pared with ~10% for the M1R-G11-scFv16 com-plex. Addition of the nonionic detergent octylb-D-glucopyranoside (b-OG) at a concentration of0.05% (~0.1 critical micelle concentration) to theM2R-GoA-scFv16 sample before grid applicationresulted in a modest increase in the number ofvitrified intact complexes (7.5~11%) (fig. S3E).Single-particle cryo-EM analysis of these samplesenabled us to obtain maps at nominal resolutionsof 3.3 and 3.6 Å for M1R-G11-scFv16 and M2R-GoA-scFv16, respectively (Fig. 1, B and C; figs. S2to S4; and table S1).

Comparison of overall structures

The higher resolution attained for the M1R-G11-scFv16 complex may be the result of more-extensive interactions between M1R and G11,including those involving intracellular loop 2(ICL2) and ICL3 and the C terminus of M1R, asdiscussed later. The overall structure of the ac-tive M1R is similar to that of the active confor-mation of M2R (fig. S5A), with root mean squaredeviation (RMSD) values of 1.55 Å for the wholecomplex, 0.94 Å when comparing receptorsalone, and 0.84 Å when comparing G proteinsalone. The conformations of critical residues for

RESEARCH

Maeda et al., Science 364, 552–557 (2019) 10 May 2019 1 of 6

1Department of Molecular and Cellular Physiology, StanfordUniversity School of Medicine, Stanford, CA, USA.2Department of Structural Biology, Stanford UniversitySchool of Medicine, Stanford, CA, USA.*These authors contributed equally to this work.†Corresponding author. Email: yiorgo@stanford.edu (G.S.);kobilka@stanford.edu (B.K.K.)

on Novem

ber 20, 2020

http://science.sciencemag.org/

Dow

nloaded from

the receptor activation such as D3.49R3.50Y3.51,N7.49P7.50xxY7.53, and P5.50I/V3.40F6.44 motifs [super-scripts indicate Ballesteros-Weinstein number-ing for GPCRs (20)] are also similar betweenactive conformations of M1R and M2R (fig. S5B),suggesting that the activation mechanism isshared between these muscarinic receptors thatprefer different G-protein partners. When com-paring inactive (7) and active states of M1R, weobserve the outward displacement of TM6 thatis characteristic of receptor activation. The TM6movement is accompanied by a small rotation ofthe helix, as well as a tilt of TM5 toward TM6.These structural changes allow the C-terminalhelix of the a subunit of the G protein to engagethe receptor core (Fig. 2).On the extracellular side of theM1R, the extra-

cellular loop 2 (ECL2), TM5, and TM6 reorga-nize, with TM6 coming closer to ECL2 by 3.9 Å(Tyr179-Ser388) and causing a contraction of theextracellular vestibule (Fig. 2B). The same reor-ganization takes place in M2R, as seen previ-ously in the nanobody-stabilized active state(5, 8), although it was not observed in activeb2AR or m-opioid receptor (mOR) (10, 15). Theextracellular vestibule in muscarinic receptorsis a well-documented binding site for allostericmodulators (21, 22), and its contraction may bea key feature for cooperativity with PAMs (23).Indeed, LY2119620, a positive allosteric modu-lator forM2R, was found to bind to this contract-ed vestibule (8). Our M2R-GoA-scFv16 cryo-EMmap shows a distinct density corresponding toLY2119620 (figs. S4D and S5C). Although weincluded VU0357017, a PAM for M1R (24), in theM1R-G11-scFv16 sample for cryo-EM, no obviousdensity corresponding to this PAMwas observedat the commonallosteric vestibule or other regionsof the receptor.

Comparison of the iperoxo bindingpocket in M1R and M2R

Even though iperoxo is a highly efficacious agonistfor both M1R and M2R, its affinity for M2R is10- to 100-fold higher than for M1R, dependingon how it is measured (25). Although the aminoacids that form the iperoxo binding pocket arewell resolved in theM1R cryo-EMmap (fig. S6A),the density for the ligand does not allow for itsunambiguous docking into the binding pocket.This suggests some degree of flexibility in theligand or more than one binding pose. Our com-putational docking studies identified a rangeof closely related poses that both fit well to theexperimental map and also received the highestdocking scores (fig. S6B). Every pose formed thestrong cation–p interactions observed in thecrystal structure of M2R; however, there were arange of rotations around the alkyne axis andseveral positions for the heterocyclic group. No-tably, the heterocyclic group is unable to occupya geometry that makes ideal hydrogen bonds tothe protein, which is likely a major contributingfactor to the range of docked structures withsimilar scores. While there may be additionalreasons for the poorly resolved density for theligand in the experimental map, these docking

results are consistent with a flexible and dy-namic ligand. Although the residues coordinatingiperoxo are identical and the side-chain confor-mations between M1R and M2R are similar (8)(fig. S6C), the surface-accessible volume of the

orthosteric binding pocket in M1R is signifi-cantly larger than that of the M2R, with 73.7 A3

in the M1R and 55.0 A3 in the M2R [calculatedby CASTp server (26)] (table S2). This might con-tribute to the lower affinity of iperoxo forM1R as

Maeda et al., Science 364, 552–557 (2019) 10 May 2019 2 of 6

Fig. 1. Overall architectures of M1R-G11iN-scFv16 and M2R-GoAiN-scFv16 complexes. (A) Signalingselectivity among muscarinic receptors. M1R, M3R, and M5R predominantly signal throughGq/11-type G protein, whereas M2R and M4R couple to Gi/o (B and C). Cryo-EM structures of M1R-G11iN-scFv16 and M2R-GoAiN-scFv16 complex. Color code for the proteins is as follows: M1R (green), M2R(orange), Ga11iN (gold), GaoAiN (blue), Gb1 (cyan), Gg2 (magenta), scFv16 (pink).

Fig. 2. Comparison between inactive and active M1R. (A) Comparison of M1R between inactive(PDB code 5CXV) and active form viewed from the side. (B) Extracellular view (top), and intracellularview (bottom) of superposed M1Rs. Active and inactive M1R are colored in green and gray,respectively. Conformational changes are shown with blue arrows.

RESEARCH | RESEARCH ARTICLEon N

ovember 20, 2020

http://science.sciencem

ag.org/D

ownloaded from

compared toM2R (25). In addition, ourM1R con-struct has an ICL3 truncation and an uninten-tionalN110Qmutation, but thesemodifications donot affect the binding affinity of iperoxo (table S3).

Structure of the nucleotide-free G11

Nucleotide-free G11 in M1R-G11 shares commonstructural changes with other nucleotide-freeG-protein subtypes characteristic of the nucleotide-free GPCR-engaged state (10, 15), including sep-aration of the a-helical domain (AHD) from theRas-like domain (23, 27, 28) (fig. S7A).As observed for other GPCR-engaged G pro-

teins, the a5 helix of Ga11 undergoes rotationaland translational movement upon GDP release,as highlighted by the conserved side-chain po-sitions of Phe341, Phe336, and Phe376 in Gaq/11,Gai1/oA, and Gas, respectively (Fig. 3). This con-formational change accompanies the reorgani-zation of the b6-a5 loop, the b1-a1 loop (P-loop),and the a1 helix (Fig. 3). The displacement ofthe a1 helix toward the GDP binding site and theb6-a5 loop away from the pocket likely contrib-utes to GDP dissociation. The side chain of Gln58in the a1 helix forms a hydrogen bond with thebackbone carbonyl of Ala331 in the b6-a5 loop(fig. S7B). The interaction between the a1 helixand the b6-a5 loop is retained in both GDP-bound and GDP/AlF4-bound states of Gq (29, 30);thus, it is likely that the conformational changesof these elements and the a5 helix take placein a concerted manner. The importance of thisinteraction is revealed in a mutational study ofGi1, where alanine substitution of the equiva-lent residue showed a high basal nucleotide ex-change (31).

Comparison of the receptor G-proteinbinding interfaces of the M1R-G11

and M2R-GoA complexes

Although the overall structure of M1R-G11 is sim-ilar to those ofM2R-Go and other GPCR–G-proteincomplexes, we observe several distinct features.Comparison of the structures of M1R-G11 andM2R-GoA shows a clear difference in the orien-tation of G11 and GoA relative to the receptor(Fig. 4A). The a5 helix and Ras domain of G11 arerotated ~15° away from receptor TM5 comparedto those of GoA. Other GPCR-Gi/o complexes havea similar rotational placement to GoA in M2R-GoA (fig. S8). This rotational shift of the a5 helixin M1R-G11 may be due to more-extensive inter-actions between the ICL2 of M1R and G11 com-pared with the ICL2 of M2R and Go (and otherGPCR-Gi/o complexes) (Fig. 4B).In the M1R-G11 structure, Arg134

34.54 in ICL2interacts with Arg37 located at the junction be-tween helix aN and strand b1 of G11 (Fig. 4B). Theside chain of Arg13434.54 in ICL2 forms a hydro-gen bond with the backbone carbonyl of Arg37in the aN-b1 junction, whereas the side chain ofArg37 forms a hydrogen bond with the back-bone carbonyl of Arg13434.54 in ICL2. Arg13434.54

in ICL2 is conserved among Gq/11-coupling mus-carinic receptors and has been shown to con-tribute to the Gq/11 coupling (32). The equivalentposition is proline in Gi/o-coupling muscarinic

receptors (fig. S9). In the chimeric G11iN used inthis study, the junction ismade between Lys29 ofGai1 and Ala36 of Ga11, so that observed inter-actions are between theM1R andG11 amino acids,and not Gai1 amino acids.Leu131 in ICL2 of M1R is buried in a hydro-

phobic groove formed by the aN-b1 junction,the b2-b3 loop, and the a5 helix of Ga11 (Fig. 4B).The same hydrophobic interaction is conservedin other GPCR/G-protein complexes and has beenreported to be crucial for the efficient nucleotideexchange reaction in multiple GPCRs, includingM1R, M3R, and b2AR (33, 34). Leu131 is buried inthis hydrophobic pocket of G11 as deeply asPhe139 of b2AR is buried in a similar pocket inGs. Reflecting its importance, mutation of L131in M1R to Ala, Asn, or Asp leads a loss of Gq cou-pling, whereas coupling is preserved in L131Mand L131F mutations (34). This interaction doesnot appear to be as important for Gi/o coupling.The CB1, which has Leu at the homologous po-sition, couples promiscuously to Gi/o and Gs. Mu-tation of this Leu to Ala leads to loss of coupling

to Gs, but not to Gi (35). Consistent with thesestudies, the homologous L129 in the M2R formsonly weak hydrophobic interactions with L195in the b2-b3 loop and T340 in the a5 helix ofGoA. Indeed, mutation of L131A in M1R reducescoupling with G11 to 17% of wild type, whereasthe equivalent mutant L129A in M2R reducescoupling to GoA to 49% of wild type (fig. S10).Recent structural studies have revealed that

the outward displacement of TM6 is generallysmaller in Gi/o-coupled GPCRs (e.g., rhodopsin,mOR) (15, 16) than in Gs-coupled GPCRs (e.g.,b2AR, GLP1R) (10, 13, 14). The C terminus of thea5 helix of Gas has larger side chains comparedto Gi/o, and the large displacement of TM6 maybe required to accommodate these side chains.This suggests that the extent of TM6 displace-ment is one of the determinants of G proteincoupling preference (36). When compared to theb2AR-Gs complex, the a5 helix of Ga11 adopts astraight conformation similar to that of Gas(Fig. 3), but is translationally shifted ~3 Å closerto TM1-TM2. As a result, the entire G11 protein

Maeda et al., Science 364, 552–557 (2019) 10 May 2019 3 of 6

Fig. 3. Structural comparison of G proteins. Structural changes upon receptor engagement andnucleotide release in Gaq/11, Gai1, and Gas. The GDP-bound structure is superimposed onto thenucleotide-free state for each family member. Rotation and translation of the a5 helix is shown in theupper panels, and a1 helix and b6-a5 loop displacements are shown in the lower panels. The PDBcode for each structure is as follows: GDP-bound Gaq (3AH8), GDP-bound Gai1 (1GP2), nucleotide-freeGai1 (6DDE), GDP-bound Gs (6EG8), nucleotide-free Gs (3SN6).

RESEARCH | RESEARCH ARTICLEon N

ovember 20, 2020

http://science.sciencem

ag.org/D

ownloaded from

is shifted in the same direction (Fig. 4C). LikeGas, Ga11 harbors bulky side chains at the ter-minal residues. The translational shift, however,enables M1R to accommodate Ga11 despite thesmaller outward displacement of TM6 relative toGs-coupled GPCRs.One of the most distinct features of the M1R-

G11 complex is the additional 2.5 helical turns ofTM5 extending inward toward the b6 strandand the a4 helix of Ga11 (Fig. 4D and fig. S5A). Thishelical extension is not observed in theM2R-Go

complex, even though the M2R construct usedin this study has a longer ICL3, and has also notbeen observed in other GPCR-Gi/o complexes re-ported to date. An extended TM5 has been ob-served in the b2AR-Gs complex, where it runs ina straight conformation and makes mostly polarcontacts with Gs (10) (fig. S11A). By contrast, inter-actions between the TM5 extension of M1R andG11 are largely formed by hydrophobic interac-tions made between the inner surface of the TM5andahydrophobic patch inGa11, but include a salt-bridge (Fig. 4D). The equivalent region in Gs andGi/o is eithermodestly or highly negatively charged(fig. S11B), and these charged or polar properties inGs orGi/owoulddisfavor the interactionwithhydro-phobic residues in the TM5 extension of M1R.Although this extended TM5 cannot be formed

in all Gq/11-coupled receptors owing to shorterICL3s (37), it is notable that an extended TM5is also found in the structure of squid rhodopsinin its ground state. Unlike vertebrate rhodopsinsthat couple with transducin, a Gi/o-family mem-

ber, invertebrate rhodopsins predominantly cou-ple with Gq/11-type G protein to transduce thephotosensing signal (38). Those invertebraterhodopsins have a longer ICL3 compared to theirvertebrate counterparts. The crystal structure ofa squid rhodopsin shows an extended TM5 thatbends inward (39), and structural superpositionwith the active M1R shows similarity betweenthese two Gq/11-coupling GPCRs (fig. S11C). Theinward bend begins at the same position and theinward surface of the extended TM5 shares hydro-phobic properties similar to those of the extendedTM5 in the M1R-G11 complex.

Interactions between the M1RC terminus and G11

In the M1R-G11 complex, we observe 11 residuesof the C terminus of M1R after H8 that extendinto a groove formed by the Ras domain and theGb. A similar C-terminal extension has not beenobserved in other GPCR–G protein complexesreported thus far. Although the density for theextension is weaker than for other regions of thecomplex, a continuous backbone density runsfrom H8 helix along the Gb subunit at the Ga/Gb interface, suggesting the presence of a weakinteraction between the C terminus of M1Rand the Gb subunit (Fig. 5A). It has been shownthat in cells, M3R forms a precoupling assemblyspecifically with Gq/11, and this complex forma-tion is independent of the ligand but dependenton the polybasic cluster at the C terminus ofM3R(40). This polybasic cluster is conserved among

Gq/11-coupling muscarinic receptors (fig. S12A).Although the Gb interface with the Ga subunit isnegatively charged across all G-protein subtypes,the Ga subunit of Gs and Gi1 subtypes have ahighly positively charged interface, whereas theGaq/11 surface is neutral (fig. S12B). This charac-teristic surface property may enable GPCRs thathave a polybasic C-terminal cluster to engage Gq/11-type G proteins more efficiently (fig. S12B). Con-sistent with a functional role for this interaction,a mutant M1R in which most of the polybasicresidues in the C-tail were replacedwith alaninesshowed lower affinity for G11 (Fig. 5B). In thisassay, G11 coupling to wild-type and mutant M1Rwas monitored by observing the ability of G11 toenhance the affinity of the M1R for the agonistiperoxo, which then displaces the antagonist N-methyl scopolamine (NMS) from the M1R (Fig.5B). A C-terminal polybasic sequence is widelyfound among Gq/11-coupling GPCRs, as previ-ously reported (40) and illustrated in fig. S13,suggesting that this mechanism is conserved inGq/11-type GPCRs and might contribute to cou-pling specificity.

Contributions of receptorTM5-TM6 and the Ga helix a5 toG protein–coupling selectivity

Muscarinic receptors have been used as a modelsystem to study the structural basis of GPCR/G protein coupling selectivity because they sharehigh sequence similarity in the transmembraneregion but still show distinct preference for their

Maeda et al., Science 364, 552–557 (2019) 10 May 2019 4 of 6

Fig. 4. Comparison of the structures of M1R-G11 and M2R-GoA.(A) Superposition of M1R-G11 and M2R-GoA complexes with thealignment based on the receptor. Rotational shift from GoA to G11 isdepicted with curved arrows. Enlarged area in the panel (B) is shownas a broken rectangle. (B) View of ICL2 interface between the G protein

on M1R-G11 (left) and M2R-GoA (right). (C) Superposition of b2AR-Gs

and M1R-G11 complexes with the alignment based on the receptor.A translational shift is shown by straight arrows. (D) The extendedhelical structure from TM5 interacts with G11; interface residuesare depicted as sticks.

RESEARCH | RESEARCH ARTICLEon N

ovember 20, 2020

http://science.sciencem

ag.org/D

ownloaded from

G-protein signaling pathway. Our structures re-veal substantial differences in the interactionsbetween receptors and their cognate G proteinfor M1R-G11 and M2R-Go complexes. Moreover,of the 19 amino acids involved in interactionsbetween M1R and G11, only 12 are conserved inM2R (fig. S9); however, the structures do notreveal which of these interactions are the mostimportant for coupling specificity. The G-proteinpreference for muscarinic receptors has beenextensively studied by using chimeric receptorsand site-directed mutagenesis (32, 41–53). Al-though most of these comparisons have beenmade with M2R and M3R, of the 19 amino acidsinvolved in interactions betweenM1R and G11, 18are identical in the M3R (fig. S9). Notably, manyof these studies used chimeric G proteins to fa-cilitate signaling studies. For example, it has pre-viously been shown that the last five amino acidsof G-protein a subunits are the major determi-nants of coupling specificity for some GPCRs(51, 54, 55). Thus, replacing the last five aminoacids of Gq with the last five amino acids of Go,Gi/o-coupled M2R receptor could stimulate ino-sitol phosphate turnover (48, 52), which is a morerobust assay than inhibition of adenylyl cyclaseby Gi/o.Chimeric receptors studies showed that the

major determinants of Gi/o and Gq/11 couplingspecificity were found in ICL3 (41, 42, 44, 49),but ICL3 chimeras that included ICL2 fromM3R could enhance coupling of M2R to Gq/11

(32). Through more refined chimeric and site-directed mutagenesis, it was possible to iden-tify specific residues in ICL2, TM5, and TM6 thatwere important for M3R coupling to Gq (shownwith an asterisk in fig. S9) (46–48, 51). These res-idues are identical inM1R (fig. S9). Among them,Tyr2555.62 in M3R (equivalent to Tyr2125.62 inM1R) has been identified as a critical residuefor efficient Gq/11 coupling (46). Mutagenesis ofthis position to alanine abolishes Gq/11 coupling,whereas substitution to other aromatic residueshows efficient signaling indistinguishable fromthat of the wild-type (46, 47). Site-directed mu-tagenesis showed that A6.33A6.34xxL6.37S6.37 arealso critical determinants for Gq/11 coupling toM3R, whereas V6.33T6.34xxI6.37L6.37 are importantfor coupling of M2R to Gi/o (x denotes any aminoacid residue) (figs. S9 and S14) (32, 48).Unexpectedly, of the five residues in TM5 and

TM6 identified as being critical for Gq/11 cou-pling, only two interact directly with G11: A

6.33

and L6.37 form van der Waals interactions withthe highly conserved L(-2) (a5 helix numberingstarts with –1 from the terminal residue) of G11.Tyr2125.62 does not interact with G11, but ratherwith Ala6.33 and Ala6.34 (Fig. 6). This interactionlikely stabilizes TM6 at a position that is optimalfor engaging G11 (Fig. 6 and fig. S14). In M2R,S5.62 forms van der Waals interactions with T6.34

and I6.37, whereas only V6.33 forms interactionswith the highly conserved L(–2) and L(–7) ofGoA (Fig. 6 and fig. S14). Thus, among the fiveresidues shown by mutagenesis studies to beimportant for coupling specificity, most of theinteractions occur between TM5 and TM6 and

Maeda et al., Science 364, 552–557 (2019) 10 May 2019 5 of 6

Fig. 5. Interaction of the M1R C terminus with G-protein a/b interface. (A) Density of the Cterminus of M1R positioned at the interface between Ga11 and Gb subunit. (B) Titration of G11 in thecompetition ligand-binding assay using iperoxo and [3H]NMS as probes. Coupling of M1R to G11

increases the affinity for iperoxo, which displaces [3H]NMS. G11 couples to the poly(A) mutant of theC-terminal basic region of M1R with lower affinity. Data points represent the mean ± SEM of threeexperiments performed in triplicate.

Fig. 6. Comparison of a5/TM5/TM6 interactions in M1R-G11 and M2R-GoA complexes.The rightpanels show a rotated view of the a5/TM5/TM6 interface for M1R-G11 and M2R-GoA complexesindicated in the red box on the left panel. Residues in M1R and M2R that have been indicated bymutagenesis to be important for G-protein coupling specificity are shown as sticks on the receptorstructure and are highlighted in blue in the sequence alignments of TM5 and TM6 (lower left).

RESEARCH | RESEARCH ARTICLEon N

ovember 20, 2020

http://science.sciencem

ag.org/D

ownloaded from

may be critical for stabilizing these TM seg-ments for optimal interactions with the C ter-minus of the G protein (Fig. 6 and fig. S14).The interactions between M2R and GoA and

between M1R and G11 provide insights into thecoupling specificity in muscarinic receptors. Al-though the M1R amino acids that interact withG11 are conserved in M3R and M5R, they arenot conserved in other Gq/11 coupled receptors(fig. S15). Similarly, M2R amino acids that in-teract with GoA are conserved in M4R but notin other GPCR–Gi/o complexes (fig. S15). Theseobservations suggest a possible intermediatecomplex that might play a more important rolein G-protein coupling specificity. Evidence forsuch an intermediate state has been observed insingle-molecule studies for the b2AR-Gs complex(56). A transient intermediate complex will bedifficult to capture by crystallography or cryo-EMand may require the creative application of otherbiophysicalmethods. Furthermore, the Gproteinsused in our study are chimeric proteins, andalthough these chimeric junctions are not involvedin interactions with the receptor, we cannot ex-clude the possibility that these protein modifica-tions could have a functional or structural impact.

Conclusions

Here, we report the structures of M1R-G11 andM2R-GoA complexes obtained by cryo-EM. Al-though the overall architecture is similar betweenM1R-G11 and M2R-GoA, the orientation of the Gprotein relative to the receptor is distinct forthese complexes. Notable features in theM1R-G11

complex are an extended helix from TM5 thatmakes more extensive interactions with the Gprotein, and a polybasic cluster in the C terminusof M1R that interacts with Ga/Gb subunit inter-face. This C-terminal interaction is likely involvedin preassembly of the complex prior to agonistbinding for Gq/11-coupled receptors. These struc-tures highlight the importance of the localorganization of the TM5/TM6/a5 interface inG-protein selectivity determination in mAChRs.

REFERENCES AND NOTES

1. J. Wess, Annu. Rev. Pharmacol. Toxicol. 44, 423–450 (2004).

2. C. J. Langmead, J. Watson, C. Reavill, Pharmacol. Ther. 117,232–243 (2008).

3. E. P. Lebois, C. Thorn, J. R. Edgerton, M. Popiolek, S. Xi,Neuropharmacology 136, 362–373 (2018).

4. M. Thomsen, G. Sørensen, D. Dencker, Neuropharmacology136, 411–420 (2018).

5. K. Haga et al., Nature 482, 547–551 (2012).6. A. C. Kruse et al., Nature 482, 552–556 (2012).7. D. M. Thal et al., Nature 531, 335–340 (2016).8. A. C. Kruse et al., Nature 504, 101–106 (2013).9. H.-W. Choe et al., Nature 471, 651–655 (2011).10. S. G. F. Rasmussen et al., Nature 477, 549–555 (2011).11. B. Carpenter, R. Nehmé, T. Warne, A. G. W. Leslie, C. G. Tate,

Nature 536, 104–107 (2016).12. T. Flock et al., Nature 545, 317–322 (2017).13. Y. Zhang et al., Nature 546, 248–253 (2017).14. Y.-L. Liang et al., Nature 546, 118–123 (2017).15. A. Koehl et al., Nature 558, 547–552 (2018).16. Y. Kang et al., Nature 558, 553–558 (2018).17. C. J. Draper-Joyce et al., Nature 558, 559–563 (2018).18. J. García-Nafría, R. Nehmé, P. C. Edwards, C. G. Tate, Nature

558, 620–623 (2018).19. S. Maeda et al., Nat. Commun. 9, 3712 (2018).20. J. A. Ballesteros, H. Weinstein, in Receptor Molecular Biology,

S. C. Sealfon, Ed. (Academic Press, 1995), vol. 25 ofMethods in Neurosciences, pp. 366–428.

21. A. A. Jensen, T. A. Spalding, Eur. J. Pharm. Sci. 21, 407–420(2004).

22. R. O. Dror et al., Nature 503, 295–299 (2013).23. R. O. Dror et al., Science 348, 1361–1365 (2015).24. G. J. Digby et al., J. Neurosci. 32, 8532–8544 (2012).25. R. Schrage et al., Biochem. Pharmacol. 90, 307–319 (2014).26. W. Tian, C. Chen, X. Lei, J. Zhao, J. Liang, Nucleic Acids Res.

46, W363–W367 (2018).27. N. Van Eps et al., Proc. Natl. Acad. Sci. U.S.A. 115, 2383–2388

(2018).28. A. I. Kaya et al., J. Biol. Chem. 289, 24475–24487 (2014).29. A. Nishimura et al., Proc. Natl. Acad. Sci. U.S.A. 107,

13666–13671 (2010).30. V. M. Tesmer, T. Kawano, A. Shankaranarayanan, T. Kozasa,

J. J. G. Tesmer, Science 310, 1686–1690 (2005).31. N. Kapoor, S. T. Menon, R. Chauhan, P. Sachdev, T. P. Sakmar,

J. Mol. Biol. 393, 882–897 (2009).32. N. Blin, J. Yun, J. Wess, J. Biol. Chem. 270, 17741–17748 (1995).33. O. Moro, J. Lameh, P. Högger, W. Sadée, J. Biol. Chem. 268,

22273–22276 (1993).34. O. Moro, M. S. Shockley, J. Lameh, W. Sadée, J. Biol. Chem.

269, 6651–6655 (1994).35. X. P. Chen et al., Br. J. Pharmacol. 161, 1817–1834 (2010).36. A. S. Rose et al., J. Am. Chem. Soc. 136, 11244–11247 (2014).37. G. Pándy-Szekeres et al., Nucleic Acids Res. 46 (D1),

D440–D446 (2018).38. M. Koyanagi, A. Terakita, Photochem. Photobiol. 84,

1024–1030 (2008).39. M. Murakami, T. Kouyama, Nature 453, 363–367 (2008).40. K. Qin, C. Dong, G. Wu, N. A. Lambert, Nat. Chem. Biol. 7,

740–747 (2011).41. T. Kubo et al., FEBS Lett. 241, 119–125 (1988).

42. J. Wess, M. R. Brann, T. I. Bonner, FEBS Lett. 258, 133–136(1989).

43. J. Lechleiter et al., EMBO J. 9, 4381–4390 (1990).44. J. Wess, T. I. Bonner, F. Dörje, M. R. Brann, Mol. Pharmacol. 38,

517–523 (1990).45. K. Blüml, E. Mutschler, J. Wess, Proc. Natl. Acad. Sci. U.S.A. 91,

7980–7984 (1994).46. K. Blüml, E. Mutschler, J. Wess, J. Biol. Chem. 269, 402–405

(1994).47. K. Blüml, E. Mutschler, J. Wess, J. Biol. Chem. 269,

11537–11541 (1994).48. J. Liu, B. R. Conklin, N. Blin, J. Yun, J. Wess, Proc. Natl. Acad.

Sci. U.S.A. 92, 11642–11646 (1995).49. D. Singer-Lahat, J. Liu, J. Wess, C. C. Felder, FEBS Lett. 386,

51–54 (1996).50. J. Liu, N. Blin, B. R. Conklin, J. Wess, J. Biol. Chem. 271,

6172–6178 (1996).51. E. Kostenis, J. Gomeza, C. Lerche, J. Wess, J. Biol. Chem. 272,

23675–23681 (1997).52. E. Kostenis, B. R. Conklin, J. Wess, Biochemistry 36, 1487–1495

(1997).53. J. Wess et al., Life Sci. 60, 1007–1014 (1997).54. B. R. Conklin, Z. Farfel, K. D. Lustig, D. Julius, H. R. Bourne,

Nature 363, 274–276 (1993).55. B. R. Conklin et al., Mol. Pharmacol. 50, 885–890 (1996).56. G. G. Gregorio et al., Nature 547, 68–73 (2017).

ACKNOWLEDGMENTS

Funding: The work was supported by NIH grant R01GM083118(B.K.K.) and the Mathers Foundation (G.S. and B.K.K.). B.K.K. is aChan Zuckerberg Biohub investigator. Author contributions: S.M.cloned, and purified each protein, made M1R/G11iN/scFv16 andM2R/GoAiN/scFv16 complexes, and performed radioligand-bindingassay and GTP turnover assay. Q.Q. prepared cryo-EM grids,collected images, analyzed and processed the data, andreconstructed the final map. S.M. built and refined the model withinput from Q.Q. M.J.R. performed molecular docking studies.S.M. and Q.Q. analyzed the structures and prepared the draftmanuscript and figures. G.S. and B.K.K. further wrote themanuscript. G.S. and B.K.K. supervised the project. Competinginterests: B.K.K. is a cofounder of and consultant for ConfometRx.The remaining authors declare no competing interests. Data andmaterials availability: The atomic coordinates for M1R/G11iN/scFv16 and M2R/GoAiN/scFv16 have been deposited in theProtein Data Bank with the accession codes 6OIJ and 6OIK,respectively. The EM maps for M1R/G11iN/scFv16 and M2R/GoAiN/scFv16 have been deposited in EMDB with the codes EMD-20078and EMD-20079, respectively.

SUPPLEMENTARY MATERIALS

science.sciencemag.org/content/364/6440/552/suppl/DC1Materials and MethodsFigs. S1 to S16Tables S1 to S3References (57–69)

31 December 2018; accepted 15 April 201910.1126/science.aaw5188

Maeda et al., Science 364, 552–557 (2019) 10 May 2019 6 of 6

RESEARCH | RESEARCH ARTICLEon N

ovember 20, 2020

http://science.sciencem

ag.org/D

ownloaded from

Structures of the M1 and M2 muscarinic acetylcholine receptor/G-protein complexesShoji Maeda, Qianhui Qu, Michael J. Robertson, Georgios Skiniotis and Brian K. Kobilka

DOI: 10.1126/science.aaw5188 (6440), 552-557.364Science 

, this issue p. 552Sciencecoupling selectivity.−respective G proteins. A side-by-side comparison provided a molecular understanding of G protein

electron microscopy structures of M1R and M2R bound to their− determined cryoet al.different G proteins. Maeda subtypes of muscarinic acetylcholine receptors (M1R to M5R) play different roles in the nervous system by binding tobinding and activating specific G proteins. The selectivity occurs even among highly related GPCRs. For example, five

coupled receptors (GPCRs) bind ligands outside the cell and trigger events inside the cell by selectively−G proteinChoosing a partner

ARTICLE TOOLS http://science.sciencemag.org/content/364/6440/552

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2019/05/08/364.6440.552.DC1

REFERENCES

http://science.sciencemag.org/content/364/6440/552#BIBLThis article cites 68 articles, 17 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Science. No claim to original U.S. Government WorksCopyright © 2019 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of

on Novem

ber 20, 2020

http://science.sciencemag.org/

Dow

nloaded from