Structural prerequisites for G-protein activation by the neurotensin receptor

We previously determined the structure of neurotensin receptor NTSR1 in an active-like conformation with six thermostabilizing mutations bound to the peptide agonist neurotensin. This receptor was unable to activate G proteins, indicating that the mutations restricted NTSR1 to relate agonist binding to G-protein activation. Here we analyse the effect of three of those mutations (E166A3.49, L310A6.37, F358A7.42) and present two structures of NTSR1 able to catalyse nucleotide exchange at Gα. The presence of F3587.42 causes the conserved W3216.48 to adopt a side chain orientation parallel to the lipid bilayer sealing the collapsed Na+ ion pocket and linking the agonist with residues in the lower receptor part implicated in GPCR activation. In the intracellular receptor half, the bulkier L3106.37 side chain dictates the position of R1673.50 of the highly conserved D/ERY motif. These residues, together with the presence of E1663.49 provide determinants for G-protein activation by NTSR1.

G PCRs are highly versatile signalling molecules that modulate second messenger responses in the cell. Binding of an extracellular agonist leads to conformational changes in the receptor, triggering activation of associated signalling partners on the intracellular side of the membrane. GPCRs are no longer thought to be two-state switches (inactive or active, although rhodopsin may come close to this definition) but are able to sample many conformational states depending on the bound ligand, associated signalling partner and membrane environment 1 . Recent advances in the structural biology of GPCRs have resulted in high-resolution snapshots of inactive 2 , active-like 3 and active receptor conformations [4][5][6][7] .
Our research focuses on understanding the structural and functional requirements for the activation of the neurotensin receptor 1 (NTSR1, ref. 8). Its agonist ligand neurotensin (NTS) is a 13-amino acid peptide that functions as a neurotransmitter and a hormone in the nervous system and in peripheral tissues 9 . NTS has a wide range of biological activities with important aspects in antinociception, cancer cell growth and the pathogenesis of schizophrenia 10,11 . Most of the known agonist effects of NTS are mediated through NTSR1 (refs 8,11).
Previously, we determined the structure of NTSR1 bound to NTS [8][9][10][11][12][13] (Arg-Arg-Pro-Tyr-Ile-Leu) in an active-like conformation with six thermostabilizing mutations providing insight into the binding mode of a peptide agonist. This receptor was unable to catalyse nucleotide exchange at the Ga subunit, indicating that some of the stabilizing mutations may have restricted the ability of NTSR1 to relate agonist binding to the activation of G protein. Here we analyse the effect of three of those six mutations (E166A 3.49 , L310A 6.37 , F358A 7.42 ) on G-protein activation and present structures of active-like NTSR1 that are able to activate G protein.
Unique to these structures is the presence of a phenylalanine at position 7.42 causing the conserved W321 6.48 to adopt a side chain rotamer conformation parallel to the lipid bilayer, which has not been seen in any GPCR structures to date. The W321 6.48 residue seals the top of the collapsed Na þ ion-binding pocket and along with the F358 7.42 residue links the agonist peptide, bound closer to the extracellular surface, with residues in the lower part of NTSR1 that are implicated in conformational changes for GPCR activation. In the intracellular receptor half, the bulkier L310 6.37 side chain dictates the position of R167 3.50 of the highly conserved D/ERY motif. This, together with the presence of the neighbouring E166 3.49 provides determinants for G-protein activation.

Results
NTSR1 constructs used in this study. Here, we describe structural, biochemical and pharmacological data of several NTSR1 mutants (see Supplementary Tables 1 and 2) with either wild-type intracellular loop 3 (ICL3), or with most of ICL3 replaced by T4 lysozyme (T4L). In the Methods section, we distinguish between NTSR1 constructs containing the wild-type ICL3 sequence or T4L. In the main text, we use only one name for a particular construct for ease of reading. For example, NTSR1-ELF refers interchangeably to NTSR1-ELF-T4L and NTSR1-ELF, the latter containing the wild-type ICL3, not T4L. The identity of the respective construct is evident from the context of writing. Protein crystals were obtained with receptors where most of ICL3 was replaced by T4L. Pharmacological and biochemical experiments were conducted with receptors containing the wild-type ICL3 sequence, but also included T4L variants for comparison. In Figures and Tables relating to biochemical and pharmacological data, the identity of constructs is unambiguously specified.
Active-like NTSR1 mutants which activate Gq protein. We previously reported the crystal structure of NTSR1-GW5 (ref. 3) containing six stabilizing mutations (A86L 1.54 , E166A 3.49 , G215A ECL2 , L310A 6.37 , F358A 7.42 and V360A 7.44 ; ref. 12; superscripts are the Ballesteros-Weinstein numbers 13 ). NTSR1-GW5 displayed features of an active-like receptor such as an outward-tilted transmembrane helix (TM) 6 at the cytoplasmic surface and key conserved residues in positions characteristic for active and/or active-like but not for inactive GPCRs. NTSR1-GW5 did not catalyse nucleotide exchange at Gaq in response to NTS in G-protein-coupling assays ( Fig. 1) 3 , suggesting that some of the stabilizing mutations have limited the ability of NTSR1-GW5 to activate G protein. On the basif of their location, we assumed that the mutations E166A 3.49 , L310A 6.37 and F358A 7.42 affected the NTSR1 activation state, whereas the involvement of the mutations A86L 1.54 , G215A ECL2 and V360A 7.44 was not obvious 3 . We therefore reverted the three mutations E166A 3.49 , L310A 6.37 and F358A 7.42 to wild-type residues, alone or in combination (Supplementary Table 1), to analyse their role in G-protein activation. In contrast to NTSR1-GW5, the triple revertant NTSR1-ELF (with E166, L310, F358) was able to stimulate nucleotide exchange at Gaq to almost wild-type level (Fig. 1, Supplementary Fig. 1, Supplementary Table 2). The double revertant NTSR1-EL (with E166 and L310) showed reduced activity in nucleotide exchange assays, but highlights the importance of E166 3.49 , of the highly conserved D/ERY motif, along with the neighbouring L310 6.37 for G-protein activation (Fig. 1).
Pharmacological characterization of NTSR1-LF and NTSR1-ELF. Table 2) showed that the apparent affinity of NTSR1-LF and NTSR1-ELF for the agonist NTS was comparable to that of the wild-type receptor and NTSR1-GW5. The agonist-binding experiments using wild-type NTSR1 were conducted at equilibrium. In contrast, binding of [ 3 H]NTS to the NTSR1 mutants did not reach equilibrium within the incubation time because of the slow agonist off-rates. The apparent IC 50 values for the antagonist SR48692 (ref. 14) were 3-5-fold higher than the corresponding wild-type value, but more than 20-fold lower than the value for NTSR1-GW5. The shift in IC 50 Table 2) possibly indicating a high-affinity agonist conformation of the NTSR1 mutants. In contrast to wild-type NTSR1, the presence of NaCl did not increase the dissociation of [ 3 H]NTS from NTSR1-LF, as was also observed with NTSR1-GW5 (ref. 3). However, NaCl did increase the dissociation of [ 3 H]NTS from NTSR1-ELF, albeit not as pronounced as that seen with the wild-type receptor ( Supplementary Fig. 1). The crystal structures of NTSR1-LF-T4L and NTSR1-ELF-T4L cannot explain the observed differences in the Na þ ion-dependent kinetics of NTS dissociation from NTSR1-LF and NTSR1-ELF. However, NTSR1-ELF may be more dynamic than NTSR1-LF, indicated by its lower thermal stability (Supplementary Table 3) and thus higher flexibility, accounting for the observed differences of the kinetic properties of both receptor mutants.

Ligand-binding experiments (Supplementary
In G-protein-coupling assays, NTSR1-LF had moderate ability to activate the G protein. NTSR1-ELF was able to stimulate nucleotide exchange at Gaq to almost wild-type level (  Table 2). The doseresponse curves for the determination of the half maximal effective concentration of NTS on the exchange of GDP for GTPgS on Gq (EC 50 values) were multiphasic (Hill slope o1) for NTSR1-ELF and wild-type NTSR1, indicating high and low affinity agonist-binding sites at non-saturating G-protein concentrations. In contrast, NTSR1-LF showed monophasic curves indicating a single class of binding sites. The EC 50 value for wildtype NTSR1 is 20-fold higher than the K i value determined in homologous [ 3 H]NTS/NTS competition experiments, possibly because the affinity of NTS for wild-type NTSR1 is affected by the presence of Na þ ions (discussed above) reducing the efficacy of G-protein activation at the NaCl concentrations used in the nucleotide exchange reactions. The differences between EC 50 and K i values persist for NTSR1-LF and NTSR1-ELF, although they become smaller (15-fold and 4-fold for NTSR1-LF and NTSR1-ELF, respectively). As the NTSR1 mutants show reduced sensitivity of agonist binding in the presence of Na þ ions, the above explanation(s) may not suffice, but, in addition, possibly reflect the consequence of non-saturating G-protein concentrations in the GDP/GTPgS exchange assays (which may also apply to wild-type NTSR1).
Architecture of NTSR1-ELF and NTSR1-LF. To understand the structural implications of E166 3.49 , L310 6.37 and F358 7.42 on the activation of G protein, we determined the structure of NTSR1-ELF to 2.9 Å resolution (Table 1). In addition, we determined the structure of NTSR1-LF (with L310 and F358) to 2.6 Å resolution. This latter mutant has moderate ability to activate G protein ( Fig. 1) likely because of the absence of the E166 3.49 side chain. The comparison with NTSR1-ELF is thus expected to provide insight into the structural role of E166 3.49 , a residue critical for governing receptor conformation and G-protein recognition 15 . For both constructs, we obtained crystals using the lipidic cubic phase crystallization method in combination with the chimeric T4 lysozyme approach to improve the probability of obtaining well-diffracting crystals.
Superposition of NTSR1-ELF or NTSR1-LF with NTSR1-GW5 reveals that the structures are overall similar (root mean squared deviation with values of 0.7 Å for Ca atoms, excluding T4 lysozyme) indicating active-like NTSR1 conformations (Fig. 2). Despite the overall similar architecture of the three NTSR1 structures, NTSR1-ELF and NTSR1-LF differ from NTSR1-GW5 in a number of regions. The amino (N) terminus adopts a short helix (S53-L55) and is, compared with NTSR1-GW5, extended by two residues (G50 and P51) in NTSR1-ELF and three residues (A49, G50 and P51) in NTSR1-LF, providing additional contacts to extracellular loop (ECL) 2. ECL3 (residues I334-T340) of NTSR1-ELF and NTSR1-LF is shifted slightly towards the receptor core by B2.5 Å (Fig. 2). Subtle, yet distinct differences exist in the NTS 8-13 binding mode (Supplementary Tables 4-7). Its R8 side chain is in strong hydrogen bond-mediated contact with D54 and D56 of the receptor N terminus. The R8 side chain in NTSR1-LF (but not in NTSR1-ELF) is also connected to TM7 through water-mediated hydrogen bonds to D345 7.29 . The R9 side chain also forms a strong hydrogen bond with the main chain oxygen of I334 of ECL3. Overall, NTS 8-13 engages in more hydrogen bond-mediated interactions with NTSR1-LF and NTSR1-ELF compared with NTSR1-GW5 (Supplementary  Tables 4-7).
On the intracellular side, the ends of TM3 of NTSR1-LF and NTSR1-ELF have shifted outward, whereas the ends of TM5 and TM6 have moved towards the receptor core (Fig. 2, Supplementary Table 8) compared with NTSR1-GW5. In addition, the relative positions between TMs within a given receptor have changed. TM3 and TM6 are positioned closer to each other, whereas TM5 and TM6 have shifted away from each other compared with NTSR1-GW5. Of the intracellular loops, ICL1 showed electron density for the main chain atoms in NTSR1-LF but not in NTSR1-ELF. ICL2, thought to be important for G-protein coupling, adopts a two-turn a helix in a slightly different position compared with the p helix seen in NTSR1-GW5 (Fig. 2). The structural significance of this observation for the ability of NTSR1 to activate G protein is unclear, as ICL2 adopts an a-helical structure in the active b 2 -adrenergic receptor-Gs complex 7 and in the inactive b 1 AR structure 16 but is an extended loop in the inactive b 2 AR structure 17 .
Transmembrane helix 7 and helix 8. In contrast to NTSR1-GW5, TM7 shows partial 'unwinding' after the conserved NPxxY motif around N370 7.54 (Fig. 3), similar to that seen in the structure of the active muscarinic acetylcholine M2 receptor 5 , but not in the structure of active b 2 AR 7 . This region (L371-N375) is followed by a short helix 8 (H8) in both NTSR1-LF and NTSR1-ELF structures. The aromatic ring of F376 8.50 , a conserved residue of the H8 motif, is well resolved in NTSR1-LF but is only weakly anchored in a hydrophobic pocket between TM1 and TM7. In NTSR1-ELF, F376 8.50 is no longer anchored between TM1 and TM7 but is rotated outward forming hydrophobic interactions with L371. These features of NTSR1-LF and NTSR1-ELF are distinct from the NTSR1 mutant TM86V-DIC3A, which adopts an apparent inactive receptor conformation at the inner side in the crystal structure, for example, lacking the outward movement of TM6 (ref. 18). In TM86V-DIC3A, TM7 does not unwind as seen in NTSR1-ELF. The residue F376 8.50 is partially inserted into the pocket between TM1, TM2 and TM7, whereas F376 8.50 of NTSR1-ELF is not anchored into the receptor core (Fig. 3). In addition, H8 of NTSR1-ELF and NTSR1-LF is shorter than that of TM86V-DIC3A by two and three residues from the carboxy (C) terminus, respectively (Supplementary Table 9). Whether this reflects a general instability of H8 (ref. 18), or its dynamic nature in the active-like receptor conformation, remains to be explored. Recent molecular dynamics simulations suggested that unravelling of H8 is related to the agonist-occupied state of NTSR1 (ref. 19).
The conserved residue W321 6.48 of the CWxP motif. Of significance is the position of W321 6.48 within the CWxP motif, a highly conserved amino acid in class A GPCRs. Spectroscopic evidence suggested changes in the environment of W 6.48 upon rhodopsin activation 20,21 . However, a rotamer change of W 6.48 is not observed in any crystal structure of active rhodopsin 22 , the b 2 AR 6 or the M2 receptor 5 suggesting that changes in the W 6.48 rotamer orientation might not be an essential part of the GPCR activation mechanism 6 . In contrast to all the previously determined GPCR structures, the W321 6.48 side chain in NTSR1-LF and NTSR1-ELF is oriented parallel to the lipid bilayer ( Fig. 4; Supplementary Fig. 2). This orientation results from the presence of F358 7.42 whose phenyl side chain prevents W321 6.48 from adopting the side chain conformation found in NTSR1-GW5, which contained the stabilizing F358A 7.42 mutation. Consequently, the W321 6.48 indole side chain makes additional van der Waals interactions with residues of TM3, TM5 and TM6 (Supplementary Fig. 3). It is worth noting that 73% of class A GPCRs, including rhodopsin, have small residues (G, A, S) at position 7.42; bulky tyrosine and phenylalanine residues are rare and comprise only 4% of class A GPCRs. For example, neuromedin U receptors and NTSR2 have a phenylalanine at position 7.42 and a tryptophan residue at position 6.48. The luteinizing hormone/choriogonadotropin and thyroidstimulating hormone receptors have a tyrosine at position 7.42, but a methionine at position 6.48 in lieu of a tryptophan residue. All class A receptors, for which crystal structures have been determined to date, have a small residue at position 7.42 (muscarinic receptors have a cysteine 7.42 residue, the P2Y 12 receptor has a threonine 7.42 residue, the orexin OX2 receptor has a valine 7.42 residue) except NTSR1, possibly explaining in part why the W 6.48 rotamer conformation seen here in NTSR1-LF and NTSR1-ELF has not been observed in other receptor structures.
A network of interactions links NTS to the hydrophobic core. NTSR1-LF and NTSR1-ELF structures help explain how the agonist peptide transmits its extracellular signal to the intracellular portion of the receptor. Hydrogen bond and van der Waals interactions link NTS [8][9][10][11][12][13] with residues of the hydrophobic core associated with helical rearrangements seen in active-state structures 23 (Fig. 4). NTS [8][9][10][11][12][13] is connected to Y324 6.51 via a hydrogen bond network from the carboxylate of its L13 residue through R328 6.55 . The aromatic ring of Y324 6.51 , in turn, is engaged in hydrophobic stacking interactions with F358 7.42 that is in contact with W321 6.48 , as previously discussed. The hydrophobic network results in the packing of W321 6.48 against the hydrophobic F317 6.44 that has been implicated in the reorganization of transmembrane segments upon agonist binding in b 2 AR 6 . The rotamer position of F317 6.44 is almost the same in NTSR1-GW5, NTSR1-LF and NTSR1-ELF    [8][9][10][11][12][13] (yellow) is connected to R327 6.54 , R328 6.55 and Y324 6.51 via a hydrogen bond network. Water molecule W9 is labelled. A close-by water molecule W10 that is bonded to L13 of NTS [8][9][10][11][12][13] in NTSR1-LF-T4L has been omitted for clarity. No corresponding water molecules were detected in NTSR1-ELF-T4L, likely because of the lower resolution of the structure. A series of stacking interactions relate Y324 6.51 to F358 7.42 , W321 6.48 and the hydrophobic core 23 residue F317 6.44 . (d) Position of the F 6.44 side chain in NTSR1-ELF-T4L (blue), the NTSR1 mutant TM86V-DIC3A (cyan, PDB code 3ZEV) and the inactive (purple, PDB code 2RH1) and active b 2 AR (orange, PDB code 3SN6), viewed from the intracellular side. (Fig. 4a-c). This is perhaps not surprising as all the three structures represent active-like receptor states (the differences in the ability to activate G protein may, in part, be attributed to the positioning of R167 3.50 by L310 6.37 , and the presence of E166 3.49 , as discussed). However, comparison of NTSR1-ELF with the mutant TM86V-DIC3A 18 , which represents an inactive NTSR1 at the inner side, suggests a rearrangement of the F317 6.44 side chain (Fig. 4d). The extent of the F317 6.44 repacking upon activation of NTSR1 remains to be determined as F317 6.44 in TM86V-DIC3A is in a receptor region, which may not represent a fully inactive conformation. The network of hydrophobic interactions identified in NTSR1-LF and NTSR1-ELF were not seen in the NTSR1-GW5 structure owing to the F358A 7.42 stabilizing mutation.
The collapsed Na þ ion-binding pocket. Na þ ions have a negative allosteric effect on agonist binding to wild-type NTSR1 and the highly conserved D113 2.50 in the middle of TM2 has been assigned a pivotal role in the Na þ ion sensitivity of agonist binding and G-protein activation 24 . Recent high-resolution structures of GPCRs in the inactive state have revealed a conserved Na þ ion-binding pocket within the receptor transmembrane bundle [25][26][27][28] , providing a structural explanation for the allosteric effect of Na þ ions on agonist binding. In each of those inactive structures, the Na þ ion is coordinated by a salt bridge to the highly conserved D 2.50 and by four additional contacts with receptor side chains and water molecules (Fig. 5d). In the active-like NTSR1-LF and NTSR1-ELF structures (and in NTSR1-GW5), the Na þ ion-binding pocket has collapsed (Fig. 5), which explains the reduced Na þ ion sensitivity of agonist binding (Supplementary Table 2). The D113 2.50 side chain atoms form an extensive hydrogen bond network with T156 3.39 , S362 7.46 and N365 7.49 of the NPxxY motif, preventing the coordination of a Na þ ion. Absent in the collapsed NTSR1 Na þ ion pocket are any water molecules, which fill the cavity in inactive-state receptors 26 . In NTSR1-LF and NTSR1-ELF, W321 6.48 forms van der Waals interaction with residues of the Na þ ion-binding pocket, effectively sealing off the top of the collapsed Na þ ion pocket and disrupting a vertical cavity seen in NTSR1-GW5 ( Supplementary Fig. 4).
The structure of the NTSR1 mutant TM86V-DIC3A 18 is similar to our active-like NTSR1 structures in the extracellular half, which is responsible for ligand binding, but dissimilar in the intracellular half, adopting an apparent inactive receptor conformation in the crystal structure. The Na þ ion-binding region is located underneath the ligand binding pocket between the intracellular and extracellular receptor halves. In TM86V-DIC3A, the D113 2.50 side chain is also in contact with neighbouring residues, albeit the interactions are different when compared with NTSR1-GW5, NTSR1-LF and NTSR1-ELF (Fig. 5). In addition, no electron density for water molecules or a Na þ ion has been reported in the TM86V-DIC3A structure 18 . Note that S362 7.46 , which contacts D113 2.50 in NTSR1-LF and NTSR1-ELF, is mutated to an alanine residue in TM86V-DIC3A (Fig. 5).
The residue L310 6.37 positions the R167 3.50 side chain. The residue at position 6.37 (L310 6.37 in NTSR1) is highly conserved among class A GPCRs; 80% of receptors have hydrophobic residues (L, I, V) at this position. The significance of this residue becomes apparent in the NTSR1-LF and NTSR1-ELF structures (Fig. 6). L310 6.37 is central to the positioning of the R167 3.50 side chain such as to allow a bona fide productive interaction with the G protein. In the signalling incompetent, active-like NTSR1-GW5 structure, R167 3.50 is linked to the conserved N257 5.58 , S164 3.47 and G306 6.33 by a hydrogen bond network, likely facilitated by the decreased side chain size of the L310A 6.37 stabilizing mutation. Those interactions stabilize R167 3   unfavourable for contacting the Ga subunit. The presence of the larger L310 6.37 side chain prevents R167 3.50 from adopting the conformation seen in NTSR1-GW5 (Fig. 6). In NTSR1-LF and NTSR1-ELF, the R167 3.50 side chain has an orientation that is similar to that found in the b 2 AR-Gs complex 7 , the active M2 receptor 5 and opsin in its G-protein-interacting conformation 29 .
The current NTSR1 structures represent snapshots of an activelike receptor bound to agonist, but not to a G protein, possibly explaining subtle differences of the R167 3.50 side-chain conformations compared with receptor structures in complex with G protein or G protein mimetics (Fig. 6).

Discussion
Stabilization of detergent-solubilized receptor-ligand complexes is one of the key factors for successful crystallization and structure determination of membrane proteins. Wild-type NTSR1 is not particularly stable in detergent solution 30 ; thus the use of stabilized NTSR1 mutants has resulted in the successful production of well-diffracting crystals (Supplementary Table 3). Our previously reported active-like NTSR1-GW5 mutant 3 was obtained by conformational thermostabilization 31 in the presence of the agonist neurotensin 12,30 . An alternative approach, directed evolution [32][33][34] , has resulted in structures of NTSR1 (ref. 18), which are similar to our active-like NTSR1 structures in the extracellular half but have adopted an apparent inactive receptor conformation in the intracellular half. The structures of NTSR1-LF and NTSR1-ELF, presented here, are very similar (root mean squared deviation values of 0.3 Å for Ca atoms, excluding T4 lysozyme), yet their ability to activate the G protein in response to NTS differs: NTSR1-LF mediates moderate nucleotide exchange at Gaq, whereas NTSR1-ELF has almost wild-type receptor properties (Supplementary Table 2). The glutamic acid 3.49 of the highly conserved D/ERY motif, absent in NTSR1-LF, but present in NTSR1-ELF, has been deemed critical for G-protein coupling 15 . Thus the mutation E166A 3.49 alone may explain the pharmacological behaviour of NTSR1-LF, highlighting the importance of E166 3.49 for G-protein activation. In the active M2 receptor, D120 3.49 is stabilized by a hydrogen bond with N58 2.39 (T68 2.39 in b 2 AR); and it has been suggested that N58 2.39 either directly stabilizes the active receptor conformation, or engages in direct interactions with G protein 5 (Supplementary Fig. 5).
In the b 2 AR-Gs complex 7 , T68 2.39 and D130 3.49 interact with the ICL2 helix via Y141 ICL2 positioning the helix such that a phenylalanine docks into a hydrophobic pocket on the G protein surface. ICL2 has been found essential for the G-protein activation pathway, especially for the dissociation of the receptor-G protein complex in the presence of GTP 35 . Thus E166 3.49 may optimally position ICL2 in the presence of G protein allowing efficient G-protein binding and release. In b 2 AR, Y141 ICL2 links the receptor-G protein interactions of ICL2 with the D/ERY motif. The equivalent interaction in NTSR1 may come by M181 ICL2 as the NTSR1 residue, equivalent to Y141 ICL2 of b 2 AR, is an alanine (A177).
In conclusion, our current NTSR1 structures provide insight into mechanistic details of an active-like, agonist-occupied peptide GPCR. The conserved W321 6.48 , oriented parallel to the membrane plane, seals the top of a collapsed Na þ ion-binding pocket; W321 6.48 in combination with F358 7.42 link the agonist peptide, bound near the receptor surface, with hydrophobic core residues in the inner half of NTSR1. The highly conserved residue L310 6.37 in the vicinity of the D/ERY motif is central to the sidechain orientation of R167 3.50 to promote the productive interaction with Gq protein. The neighbouring E166 3.49 residue is vital for G-protein activation, possibly by coupling receptor-G protein interactions with the D/ERY motif.
Methods NTSR1 constructs. The baculovirus construct NTSR1-LF-T4L consisted of the hemagglutinin signal peptide and the Flag tag 36 , followed by the thermostabilized rat NTSR1 (T43-K396 containing the mutations A86L, E166A, G215A, V360A) with the ICL3 residues H269-E296 replaced by the cysteine-free bacteriophage T4 lysozyme (N2-Y161 with the mutations C54T and C97A) and a GSGS linker. A deca-histidine tag was placed at the C terminus. NTSR1-LF contained the wild-type ICL3 sequence. NTSR1-ELF-T4L and NTSR1-ELF were like NTSR1-LF-T4L and NTSR1-LF, respectively, but had only three mutations (A86L, G215A, V360A). The wild-type NTSR1 used here was like NTSR1-LF but did not have the four mutations. Additional NTSR1 mutants, used for pharmacological analyses, are listed in Supplementary Table 1. In the Methods sections, we distinguish between NTSR1 constructs containing T4L or the wild-type ICL3 sequence. In the main text, we use only one name for a particular construct; for example, NTSR1-ELF refers interchangeably to NTSR1-ELF-T4L and NTSR1-ELF, the latter containing the wild-type ICL3, not T4L. The identity of the respective construct is evident from the context of writing.
Expression of NTSR1 in insect cells. Recombinant baculoviruses were generated using a modified pFastBac1 transfer plasmid (Invitrogen). Trichoplusia ni cells were infected with recombinant virus, and the temperature was lowered from 27 to 21°C. Cells were collected by centrifugation 48 h post infection, resuspended in Expression of Gq protein in insect cells. The baculovirus construct His-Tev-Gaq consisted of a hexa-histidine tag, followed by a tobacco etch virus (TEV) protease recognition site and the human Gaq sequence (M7-V359). Human Gb 1 was unmodified. Human Gg 1 was preceded by a hexa-histidine tag. We refer to His-Tev-Gaq Gb 1 His-Gg 1 as Gq protein. Trichoplusia ni cells were triple-infected with the recombinant viruses at 27°C. Cells were collected by centrifugation 48 h post infection, resuspended in hypotonic buffer (10 mM Hepes pH 7.5, 10 mM MgCl 2 , 20 mM KCl), flash frozen in liquid nitrogen and stored at À 80°C until use.
Preparation of urea-washed P2 insect cell membranes. NTSR1-enriched membranes were obtained as a P2 fraction from insect cells 30,37 . Before G-proteincoupling assays and ligand binding experiments, the P2 membranes were treated with urea to remove peripherally bound membrane proteins 38,39 . The receptor density in urea-washed P2 membranes was determined by [ 3 H]NTS saturation binding analysis 30 .
Ligand-binding experiments. All radioligand binding assays were conducted with urea-washed P2 insect cell membranes containing the indicated NTSR1 constructs. Independent experiments were carried out in single data points. Separation of bound from free ligand was achieved by rapid filtration through GF/B glass fibre filters (Whatman) pretreated with polyethylenimine (0.5% w/v). The amount of radioactivity was quantified by liquid scintillation counting (Beckman LS 6500). Data were analysed by nonlinear regression using the GraphPad Prism software and best fit to a one-site binding equation to determine the dissociation constants (K d ). Note that the saturation binding experiments using wild-type NTSR1 were conducted at equilibrium. In contrast, binding of [ 3 H]NTS to the NTSR1 mutants did not reach equilibrium within the incubation time because of the slow agonist off-rates (discussed previously).
Homologous competition assays with NTS were performed in the presence of GTPcS assays. Before G-protein-coupling assays, the P2 membranes were treated with urea to remove peripherally bound membrane proteins 38,39 . GDP/[ 35 S]GTPgS exchange assays were performed 30,41 with 1 nM receptor, 140 nM Gq protein and specified amounts of NTS, nonpeptide antagonist SR48692 (ref. 14)  Reactions were terminated by diluting the reaction mixture with 2 ml of ice-cold stop buffer (20 mM TrisHCl pH 8.0, 100 mM NaCl and 25 mM MgCl 2 ) and were filtered over nitrocellulose membranes on a vacuum manifold. Filters were then washed six times with 2 ml of ice-cold stop buffer. The nitrocellulose membranes were dried, and the radioactivity was quantified by liquid scintillation in a Beckman Coulter LS 6500 scintillation counter. Data from dose-response experiments were fit to equations with a Hill slope of 1 or variable slope.
The Gq protein used for exchange assays was purified in buffer containing CHAPS. However, the Gq protein addition in the exchange assays did not exceed one-fifth of the reaction volume, thus limiting the free CHAPS concentration to one-fourth of its critical micellar concentration 42 .
For some Gq protein preparations, the sample was subjected to ion exchange chromatography before the final size exclusion chromatography step. For this, protein containing Ni-NTA fractions were pooled, supplemented with 10 mM GDP, concentrated using a 30,000 MWCO concentrator (Amicon Ultra, Millipore) and passed over PD10 columns (GE Healthcare) equilibrated in IEX binding buffer (20 mM Hepes pH 8.0, 5 mM MgCl 2 , 1 mM DTT, 0.5% (w/v) CHAPS) supplemented with 10 mM GDP. Then the sample was concentrated and loaded onto a 1 ml HiTrap DEAE FF column (GE Healthcare) equilibrated in IEX binding buffer. After washing with IEX binding buffer, the Gq protein was eluted using a gradient to 100% elution buffer (20 mM Hepes pH 8.0, 5 mM MgCl 2 , 1 mM DTT, 500 mM NaCl, 0.5% (w/v) CHAPS) over 10 column volumes.
Finally, the protein containing Ni-NTA or IEX fractions were pooled and supplemented with 10 mM GDP, concentrated using a 30,000 MWCO concentrator (Amicon Ultra, Millipore) and run on a Superose 6 10/300 column (GE Healthcare) equilibrated in GF buffer (10 mM MOPS pH 7.5, 100 mM NaCl, 1 mM DTT, 0.5% (w/v) CHAPS). The fractions containing Gq protein were pooled, supplemented with 10 mM GDP, aliquoted, flash frozen and stored at À 80°C until use for pharmacological experiments.
Purification of NTSR1-LF-T4L and NTSR1-ELF-T4L. The cells from 3 l of insect cell culture were thawed and the volume was brought to B240 ml with hypotonic buffer (10 mM Hepes pH 7.5, 10 mM MgCl 2 , 20 mM KCl). The cells were then resuspended using a Turrax T-25 (IKA) homogenizer at 8,000 r.p.m. for 2 min. After centrifugation (45Ti rotor, 125,000g, 20 min, 4°C, Optima L90K, Beckman), the membranes were resuspended (Turrax T-25) in B180 ml of high-salt buffer (10 mM Hepes pH 7.5, 1 M NaCl, 10 mM MgCl 2 , 20 mM KCl) supplemented with AEBSF (100 mM) and centrifuged again. The high-salt buffer wash was repeated one more time resulting in B12 g of wet membrane pellet. All the subsequent steps were performed at 4°C or on ice, and AEBSF (100 mM final concentration) was repeatedly added throughout the procedure. The washed membranes were resuspended in 122 ml of buffer (80 mM TrisHCl pH 7.4, 48% (v/v) glycerol) containing 16 mM NTS 8-13 (Arg-Arg-Pro-Tyr-Ile-Leu) and stirred for 45 min to allow agonist binding to membrane-inserted NTSR1. The receptor was extracted by drop-wise Data collection and structure determination. Data collection was performed using the JBluIce-EPICS data acquisition software at the GM/CA-CAT (23-ID-D) beamline at the Advanced Photon Source of the Argonne National Laboratory using a 10-20 mm minibeam at a wavelength of 1.0332 Å and a Pilatus 3 6 M detector. Crystals within the loops were located by diffraction using the automated rastering module of JBluIce-EPICS 46,47 . Partial data sets (wedges of 20-30 degrees) were collected from crystals exposed to the non-attenuated minibeam for 0.3 s and 0.3-degree oscillation per exposure.
For NTSR1-LF-T4L, a 99.4% complete data set at 2.6 Å resolution was obtained by indexing, integrating, scaling and merging partial data sets from six crystals using XDS 48 and Aimless 49 of the CCP4 Suite 50 . For NTSR1-ELF-T4L, a 90.2% complete data set at 2.9 Å resolution was obtained from partial data sets from five crystals.
Structure determination was performed by molecular replacement using the Phaser module of the CCP4 Suite. Two search models were created using the structure of NTSR1-GW5-T4L 3 (PDB code 4GRV) with one containing the T4 lysozyme domain and one containing the receptor seven-helix bundle. One copy of each search model was found, producing a single solution. Subsequent refinement was performed using the MR solution with rounds of PHENIX.AutoBuild 51 followed by PHENIX.Refine with simulated annealing, and Refmac5 using translation, libration and screw-rotation (TLS) parameters 52 . TLS displacement groups used in the refinement were defined by the TLSMD server 53 . Manual examination and rebuilding of refined coordinates was accomplished using COOT 54 . The structures were refined with final R/R free values of 0.23/0.28 for both NTSR1-LF-T4L and NTSR1-ELF-T4L. The quality of the model was checked using the Molprobity server 55 . The Ramachandran statistics for NTSR1-LF-T4L are: 95.3% favoured regions, 4.7% allowed regions, 0% outliers. The Ramachandran statistics for NTSR1-ELF-T4L are: 94.4% favoured regions, 5.6% allowed regions, 0% outliers. The signal-to-noise ratio of I/s(I)Z2.0 was at 2.8 and 3.1 Å for NTSR1-LF-T4L and NTSR1-ELF-T4L, respectively. A summary of data collection and refinement statistics is reported in Table 1. Simulated annealing omit maps for H8 were generated using the Composite Omit Map function of the Phenix program suite (Supplementary Fig. 6). A stereo figure of representative electron density for TM7 of NTSR1-LF-T4L and NTSR1-ELF-T4L is given in Supplementary Fig. 7.
Figures were prepared in PyMOL (Schrödinger). Structural alignments were done with the 'align' command of PyMOL. Sigma-A weighted maps imported into PyMOL were generated by using the SigmaA program of the CCP4 suite and then converted to maps using the Phenix.mtz2map command line tool.