Main

A thermostabilized, detergent-resistant mutant of β1AR10,11,12 (TS-β1AR; see Methods) was selectively labelled with [15N]valine and produced in insect cells without further chemical modifications. Its 28 valine residues are homogeneously distributed across the receptor (Extended Data Fig. 1) at locations suitable to sense ligand binding and receptor activation. Although resonances of main chain atoms are considerably more difficult to observe than those of mobile side chains of surface residues, they are expected to be better reporters of functional, long-range backbone motions. We succeeded to obtain well-resolved TROSY (transverse relaxation-optimized spectroscopy) spectra of the valine 1H–15N backbone resonances of detergent-solubilized TS-β1AR in its apo form and in complexes with six ligands (Extended Data Fig. 1) ranging in their efficacy from antagonists to agonists (Extended Data Table 1). Despite the absence of deuteration and very short T2 relaxation times (~4 ms for 1HN), 26 valine resonances could be observed with sufficient sensitivity and resolution. Distinct and reversible chemical shift changes were detected for many valines after ligand exchange. 16 valines were assigned unambiguously and 5 tentatively using spectra from 18 point mutants, as well as further spectral and structural information (Extended Data Table 2).

Many valine residues in the vicinity of the ligand binding pocket could be assigned, showing chemical shift changes that report on the ligand functional groups (Fig. 1). Remarkably, residue V172(4.56) (the number in parenthesis corresponds to the Ballesteros–Weinstein numbering system13), which is located close to the ligand aromatic head group, exhibits an unusual 15N chemical shift of ~105–110 p.p.m. (Extended Data Figs 1, 2 and Fig. 1c). This anomaly seems caused by a distorted backbone geometry, which is presumably conserved among adrenergic receptors and results from a missing hydrogen bond to the proline at position 176(4.60) (Extended Data Fig. 2). Instead, the carbonyl of V172(4.56) participates in a water-mediated hydrogen bond network, which connects the ligand binding site, TM3, TM4, TM5, and TM612,14. Seemingly as a result of these interactions, the V172(4.56) 1H–15N resonances cluster according to the substitution patterns of the ligand head group (Fig. 1c): one cluster is observed for the partial agonists/antagonists cyanopindolol, alprenolol and carvedilol, which have larger head groups with ortho- and/or meta-substitutions; a second cluster is observed for the agonists isoprenaline and dobutamine, which bear a meta- and para-substituted catechol ring. We attribute the distinct chemical shifts for isoprenaline or dobutamine to the loss of a coordinated water caused by specific hydrogen bond interactions between their catechol moieties and the side chain of S215(5.461) (Extended Data Fig. 2d).

Figure 1: Ligand-induced 1H–15N chemical shift changes in the vicinity of the ligand binding pocket of β1AR
figure 1

a, Partial view of the β1AR–carvedilol crystal structure (4AMJ) showing valine residues (blue spheres) in the vicinity (<8.5 Å) of the ligand (magenta sticks) binding site. b, Chemical structures of the β1AR ligands used in this study. Ligand affinities derived from whole-cell binding assays on the thermostabilized β36-m23 β1AR construct17 are indicated as pKD values. Similar pK values were measured for the TS-β1AR construct (Extended Data Table 1). c, Left, ligand-induced response of V172(4.56) 1H–15N resonances. The black bar represents a scale of 0.1 p.p.m. (1H) and 1 p.p.m. (15N). The labels o, m, p indicate ligands with ortho, meta, and para substitutions at the head group, respectively. c, Right, partial view of the β1AR–carvedilol structure (4AMJ) showing the interaction network connecting V172(4.56) to S215(5.461), P219(5.50), I129(3.40), and F299(6.44). d, Left, representation as c, left, for the 1H–15N resonances of V314(6.59). Centres of resonances are indicated by circles. d, Middle, correlation of a best-fit linear combination of the V314(6.59) chemical shifts (−49.3 δ1H + 2.02 δ15N + 133) to the ligand affinity pKD (Extended Data Table 1). d, Right, partial view of the carvedilol (red, 4AMJ), dobutamine (orange, 2Y01), cyanopindolol (green, 2VT4 B), and isoprenaline (blue, 2Y03) complex structures showing the ligand-induced movement of V314(6.59). e, Left, representation as d, left, for the 1H–15N resonances of V125(3.36) and V103(2.65). e, Middle, correlations of best-fit linear combinations of chemical shifts for V125(3.36) (−0.402 δ1H + 1.17 δ15N − 132) and V103(2.65) (−336 δ1H − 634 δ15N + 7.64·104) to the ligand insertion depths and tail volumes (Extended Data Table 1), respectively. e, Right, partial view of the carvedilol, dobutamine, cyanopindolol, and isoprenaline crystal structures (representation as d, right) showing the ligand-induced movement of V125(3.36), D121(3.32), and V103(2.65).

PowerPoint slide

Full size image

Remarkably, the resonance positions of V172(4.56) in complex with the antagonist atenolol strongly differ from the already described complexes and the apo form: considerable 1HN (>0.4 p.p.m.) and 15N (>4 p.p.m.) upfield shifts indicate, respectively, a further weakening of the main chain hydrogen bond V172(4.56)HN···I168(4.52)O and a stronger kink of the backbone. This rearrangement seems caused by the insertion of the para-acetamide group of the ligand head between residues S215(5.461) and V172(4.56) (Extended Data Fig. 2d). This will lead to a substantial disruption of the TM3–TM4–TM5 interface, thereby precluding receptor activation, in agreement with atenolol’s inverse agonist pharmacology. Thus, the amide chemical shifts of V172(4.56) constitute a very sensitive readout for the state of this water-mediated, inter-helical activation switch.

The 1H–15N chemical shifts of further residues in the vicinity of the binding pocket report on additional characteristics of the ligands. V314(6.59) and V202(ECL2) are located at the extracellular surface of the receptor in a “vestibule” next to the entry/exit pathway of the orthosteric binding site15,16. The resonances of these residues are either severely broadened or undetectable in the absence of ligands, whereas they are observable in the presence of ligands (Fig. 1d and Extended Data Fig. 3). A line shape analysis for V314(6.59) (Extended Data Fig. 4) indicates that this extracellular part of the receptor undergoes micro- to millisecond motions in the apo form, which are quenched by ligand binding. This is consistent with results on β2AR, which suggest that high-affinity ligands stabilize the conformation of ECL2 and ECL35. The ligand-induced shifts of the V314 resonance correlate strongly (r2 = 0.95) with the reported ligand affinity17 (Fig. 1d). Interestingly, the resonances of V314(6.59) in the apo form and in high-affinity ligand complexes are very close. This may indicate that the high-affinity ligand complexes mimic the average apo conformation. Finally, the 1H–15N chemical shifts of V125(3.36) at the bottom of the binding site and of V103(2.65) close to the ligand tail reveal additional trends (Fig. 1e): the chemical shifts of V125(3.36) correlate with the depth of ligand insertion towards the central part of TM3 (r2 = 0.81), whereas those of V103 correlate with the volume of the ligand tail (r2 = 0.90).

Compared to inactive β2AR, complexes of activated β2AR with either G protein2 or the G protein-mimicking nanobody NB809,18 show large movements at the intracellular sides of TM5, TM6 and their intervening loop ICL3, which form the binding site for the G protein. These conformational changes are expected to be conserved throughout the GPCR family19. Four valine residues could be assigned in this region of TS-β1AR: V226(5.57), V230(5.61), V280(6.25), and V298(6.43) (Fig. 2). In contrast to the chemical shift changes in the vicinity of the ligand binding pocket, which depend strongly on the ligand chemistry, the shifts of the TM5 residues observed in this region report on ligand efficacy. This effect is most prominent for residue V226(5.57), for which the 1H–15N resonances fall on one line from antagonists to agonists (Fig. 2a). The chemical shifts for the different ligands correlate very strongly (r2 = 0.89) with their reported17 efficacies for Gs signalling (Extended Data Table 1). This highly linear effect suggests that the receptor filters the diverse input signals from the various ligands to a unified and precise structural response on TM5, which can be read out by the chemical shifts of V226(5.57). Interestingly, the V226(5.57) atenolol peak is situated at a position corresponding to lower efficacy than for the apo receptor. This gives direct structural evidence of atenolol’s inverse agonist action, which reduces the activation relative to the basal level of the apo receptor.

Figure 2: Correlation of ligand-induced chemical shift changes at the TS-β1AR intracellular side with Gs efficacy.
figure 2

a, Left, response of the V226(5.57) 1H–15N resonance to various ligands (colour coding as in Fig. 1). The centres of resonances are indicated by circles. The 1H–15N resonances fall on one line from atenolol (antagonist) over apo to alprenolol (partial agonist), carvedilol (antagonist), cyanopindolol (partial agonist), dobutamine (full agonist) and isoprenaline (full agonist). a, Right, correlation of a best-fit linear combination of the V226(5.57) chemical shifts (−515 δ1H −31.7 δ15N +8.41·103) in different ligand complexes to their efficacy for the Gs signalling pathway17. b, Left, overlay of TM5 and TM6 backbones of thermostabilized β1AR in antagonist- (blue, PDB code 4AMJ) and agonist-bound (green, PDB code 2Y03) form. The agonist does not induce detectable helix movements. b, Right, TM5 and TM6 backbone movements upon activation in human β2AR. The overlay of inactive (blue, PDB code 2RH1) and G protein-bound β2AR (magenta, PDB code 3SN6) structures shows the large bend of TM6 along with the smaller conformational change of TM5 upon activation. Hydrogen bonds 5.57-HN···5.53-O and 5.61-HN···5.57-O are indicated by dashes. According to the behaviour of the 1HN chemical shifts of V226(5.57) and V230(5.61) in TS-β1AR, these hydrogen bonds expand in an efficacy-dependent manner during agonist binding. c, Response of the 1H–15N resonances for V226(5.57), V230(5.61), V298(6.43), and V280(6.25) to various ligands. For clarity, only the centres of resonances are depicted as circles with colour coding as in Fig. 1. The black bar represents a scale of 0.1 p.p.m. and 1 p.p.m. for the 1H and 15N chemical shifts, respectively. The schematic representation of the receptor indicates the locations of the respective valine residues at the cytoplasmic sides of TM5 and TM6 within the helical bundle of β1AR.

PowerPoint slide

Full size image

Current high-resolution structures of β1AR do not show significant changes between antagonist- and agonist-bound forms (Fig. 2b). The decrease of the V226(5.57) 1HN chemical shift by about 0.2 p.p.m. from the agonist isoprenaline to the antagonist atenolol indicates a lengthening of the V226(5.57)-HN···I222(5.53)-O hydrogen bond by about 0.05 Å (ref. 20). This small, but clearly NMR-detectable length variation is below the resolution limit of current GPCR structures (Supplementary Text 1), but may indicate the start of TM5 bending towards the active conformation as observed in the G protein-bound form of β2AR (Fig. 2b). Remarkably, this response to agonists occurs even in the thermostabilized receptor TS-β1AR. Albeit reduced in absolute size compared to V226(5.57), V230(5.61) displays similar linear chemical shift changes as a function of ligand efficacy for the Gs pathway (Fig. 2c). As this residue is located one helical turn further towards the cytoplasm, the detected conformational change is not just local, but spans a certain length in TM5.

Compared to V226(5.57) and V230(5.61) in TM5, the chemical shift response to ligands is much less pronounced for V298(6.43) and in particular V280(6.25) at the intracellular side of TM6 (Fig. 2c). This suggests that agonist binding to the TS-β1AR does not induce the large conformational change in TM6 observed in the activated β2AR–G protein2 or β2AR–NB8018 complexes. However, G protein activation upon agonist binding has been reported for other less thermostabilized β1AR constructs17, indicating that they can still be activated, albeit at low levels. Thus we reverted the mutations most likely to interfere with the activation mechanism in TS-β1AR, that is, I129(3.40)V in the connector switch15, Y227(5.58)A in TM521,22, and Y343(7.53)L in the NPxxY motif of TM714,22 to the native residues. These reverse mutants were then tested for G protein activation and the NMR response in TM6. A summary of the results is given in Extended Data Table 3. The original TS-β1AR, the single mutants TS-β1AR(V129I), TS-β1AR(A227Y), and TS-β1AR(L343Y) as well as the double mutant TS-β1AR(V129I/A227Y) showed no detectable G protein activation upon isoprenaline binding (Extended Data Fig. 5). However, G protein activation was detectable for the least thermostable TS-β1AR(A227Y/L343Y) double mutant (Tm reduced by 11 °C relative to TS-β1AR), which recovers the conserved tyrosines in TM5 and TM7 that are known to stabilize the active state of rhodopsin22. None of the reverse mutants showed major changes in the NMR spectra of various ligand complexes compared to the original TS-β1AR (Extended Data Fig. 6). In particular, residues V298(6.43) and V280(6.25) at the intracellular side of TM6 did not show an increased response to agonists. This is in agreement with recent DEER (double electron–electron resonance) and 19F-NMR data showing that agonists alone do not fully stabilize the active state of TM6 in β2AR9. Interestingly, the V129(3.40)I and the V129(3.40)I/A227(5.58)Y mutations shifted the 1H–15N resonances of V226(5.57) towards a more active (that is, bent) state of TM5 in both the atenolol- and isoprenaline-bound forms (Extended Data Fig. 6b), thereby given direct experimental evidence for an allosteric activation pathway spanning about 13 Å from I129(3.40) on TM3 to V226(5.57) on TM5.

With the exception of rhodopsin23, the stabilization of fully active GPCR conformations seems to require binding of an agonist and an intracellular partner8,9. Indeed, when both the agonist isoprenaline and the G protein-mimicking nanobody NB8018 were added to TS-β1AR(A227Y/L343Y), very large chemical shift responses for many valine residues in TM3–TM6 were observed, whereas no change was observed for several valines in TM1, 2, and 7 (Extended Data Fig. 7). This very strong response extends even to the extracellular residue V314(6.59), providing evidence of a long-distance connection from the G protein binding site to the ligand entry site. The strong chemical shift changes are reverted when the partial agonist/antagonist cyanopindolol is added to the isoprenaline-TS-β1AR(A227Y/L343Y)–NB80 complex. The spectrum then becomes identical to that of the ‘pure’ cyanopindolol–TS-β1AR(A227Y/L343Y) complex (Extended Data Fig. 7a), indicating that cyanopindolol replaces isoprenaline and causes the release of NB80. In agreement with the G protein activation data, the isoprenaline-bound original TS-β1AR and the mutants TS-β1AR(A227Y) and TS-β1AR(L343Y) did not show binding of NB80 in the NMR spectra. Moreover, supplementing NB80 to the ultrastable TS-β1AR did not change its affinity for isoprenaline (Extended Data Fig. 7b), whereas it caused a hundred-fold affinity increase in the case of TS-β1AR(A227Y/L343Y) and the truncated native turkey β1AR receptor (tβtrunc)17. This increase is identical to data for β2AR18 and shows the energetic coupling between the NB80 and agonist binding also for β1AR.

In combination, these data prove that agonist binding, even in the absence of a G protein mimic, induces initial changes in the conformational equilibrium of TM5 towards the conformation observed in the G protein complex of β2AR. Remarkably, these rearrangements occur in all thermostabilized forms of β1AR. However, a full shift of the equilibrium towards such an active conformation, including allosteric changes at the extracellular side, occurs only when G protein or its mimetic NB80 is bound. This process requires the presence of both Y227(5.58) and Y343(7.53), which significantly reduce the thermal stability. Different active conformations may be reached for non-G protein effectors such as β-arrestin.

The possibility to detect NMR signals at many receptor sites in response to ligand binding and point mutations provides an experimental method to trace allosteric signalling paths. Figure 3 shows examples of these pathways, derived from the response to the ligands atenolol and isoprenaline and the single point mutations V129(3.40)I, A227(5.58)Y, and L343(7.53)Y. Choosing a cutoff of 0.05 p.p.m. for the resulting combined 1H, 15N chemical shift change (Fig. 3a, red line), long-range (>10 Å) connections become evident throughout the receptor (Fig. 3b, c). Whereas detected ligand signals radiate broadly to almost all helices, the point mutants give evidence of smaller interaction networks connecting TM3 to TM4/5, TM5 to TM3/4/6, as well as TM7 to TM2/3 (Fig. 3b, c). Interestingly, the TM2/TM7 network seems to be only weakly connected to the TM3–TM6 network. Together, these data provide experimental evidence at high resolution of an extensive signal transduction network that connects the ligand binding site to the intracellular sides of TM5, TM6, and TM7. Such a network of loosely coupled allosteric connections has been postulated previously for β2AR on the basis of molecular dynamics simulations24.

Figure 3: Experimental detection of allosteric signalling pathways using the NMR response to ligand binding and point mutations at different backbone sites.
figure 3

a, Combined 1H, 15N chemical shift deviations [Δδ = (Δδ1H2/2 + Δδ15N2/50)1/2] of valine resonances observed upon ligand binding or induced by the indicated point mutations. For ligand binding, the three pairwise deviations Δδ were calculated between the apo, atenolol-, and isoprenaline-bound forms of TS-β1AR. The maximum of these deviations is shown. For the reverse mutants, deviations Δδ are shown relative to TS-β1AR for their apo (black), atenolol- (cyan), and isoprenaline-bound (blue) bound forms. Valines within 10 Å from the Cα atom of the mutated amino acid are shown on a grey background. Distances were calculated using the coordinates of the thermostabilized β1AR (PDB code 4BVN). A red line marks a cut-off value Δδ of 0.05 p.p.m. for significant chemical shift deviations. b, Topology of the signalling network determined from point mutations (left) and ligand binding (right). Signal paths were identified by chemical shift deviations Δδ larger than 0.05 p.p.m. induced by these two perturbations (a). Signal paths to valines within 10 Å from the ligand or point mutation (that is, localized conformational changes) are indicated as dashed lines, and those beyond 10 Å (long-range conformational changes) as solid lines. The ligand signals broadly towards all helices but TM1. In contrast, the network determined by the point mutations is more localized and connects TM3 to TM4/5, TM5 to TM3/4/6, and TM7 to TM2/3. The latter network seems to be divided into two subnetworks involving TM3/4/5/6 and TM2/7. c, Long-range allosteric signal paths identified from ligand binding or point mutations (a and b) indicated on schematic β1AR representations showing the involved TMs. Helices are colour-coded according to b.

PowerPoint slide

Full size image

In summary, we have shown that highly resolved solution NMR backbone spectra can be obtained for a eukaryotic GPCR. The NMR observations delineate the allosteric signalling pathways and comprehensively connect many previous experimental and theoretical observations, which may ultimately allow to understand the dynamic mechanisms of GPCRs at the atomic level.

Methods

β1AR constructs

The TS-β1AR mutant was derived from the turkey β1AR44-m23 mutant used in crystallographic studies25 by adding three additional thermostabilizing mutations (I129V, D322K, and Y343L) and a neutral mutation (D200E) from the ultra-stable β1AR-JM3 mutant11. As compared to the wild type, TS-β1AR contains truncations at the amino and carboxy termini and intracellular loop (ICL3), a total of nine thermostabilizing point mutations, three further point mutations as well as a C-terminal hexahistidine tag (Extended Data Fig. 1). The final TS-β1AR sequence is MGAELLSQQWEAGMSLLMALVVLLIVAGNVLVIAAIGSTQRLQTLTNLFITSLACADLVVGLLVVPFGA TLVVRGTWLWGSFLCELWTSLDVLCVTASVETLCVIAIDRYLAITSPFRYQSLMTRARAKVIICTVWAI SALVSFLPIMMHWWRDEDPQALKCYQDPGCCEFVTNRAYAIASSIISFYIPLLIMIFVALRVYREAKEQ IRKIDRASKRKTSRVMLMREHKALKTLGIIMGVFTLCWLPFFLVNIVNVFNRDLVPKWLFVAFNWLGYA NSAMNPIILCRSPDFRKAFKRLLAFPRKADRRLHHHHHH.

Additional valine-to-alanine or isoleucine point mutations were introduced into TS-β1AR for NMR assignment purposes. All constructs were made using the QuikChange site-directed mutagenesis method (Agilent). Baculovirus for insect cell expression was generated using the Bac-to-Bac system (Invitrogen).

β1AR expression and purification

All β1AR constructs were expressed in baculovirus-infected insect cells as described26. Selective labelling by [15N]valine was achieved by growing cells on unlabelled serum-free insect cell medium (InsectXpress, Lonza) and then exchanging into custom-made serum-free medium (SF4, BioConcept) devoid of valine and yeast extract, to which 100 mg l−1 [15N]valine were supplemented. Virus was added immediately after the medium exchange. The culture was harvested at 48 or 72 h post infection.

After cell lysis, the membrane fraction was separated from the lysate via ultracentrifugation and subsequently solubilized with 2% n-decyl-β-D-maltopyranoside (DM, Anatrace). The solubilized membrane fraction was then purified by nickel ion affinity chromatography followed by alprenolol ligand affinity chromatography. The active receptor was eluted with buffer (20 mM Tris, 350 mM NaCl, 0.1% DM, pH 7.5) containing either atenolol (1 mM) or alprenolol (0.1 mM). Final yields of detergent-solubilized receptor were 1.5 mg l−1 of cell culture. The molecular weight of the receptor-detergent complex was estimated as ~100 kDa by static light scattering.

Thermal shift assays of mutant receptors

Detergent-solubilized, purified apo TS-β1AR and reverse-mutation receptors for thermal stability assays were obtained from their atenolol-bound form by washing with buffer devoid of ligand on a HiTrap SP HP (GE Healthcare) column. Their thermal stability was determined by the microscale fluorescent stability assay for binding of the thiol-specific fluorochrome N-[4-(7-diethylamino-4-methyl-3-coumarinyl)phenyl]maleimide (CPM)27 in a Rotor-Gene Q (QIAGEN) real-time PCR cycler using 1 μg of receptor in 20 mM Tris, 350 mM NaCl, 0.1% DM, pH 7.5 and a heating rate of 2 K min−1.

NMR experiments

NMR samples were prepared in Shigemi tubes as 250 μl volumes of typically 100 μM receptor, 1 mM ligand (except for apo form), 20 mM TRIS, 100 mM NaCl, 0.1% DM, 5% D2O, pH 7.5. For isoprenaline or dobutamine, 2 mM of sodium L-ascorbate were supplemented as anti-oxidant. All solution NMR measurements were carried out at 304 K on a 800 MHz or a 900 MHz Bruker Avance III spectrometer equipped with a cryogenic probe. 2D 1H,15N TROSY spectra were recorded with total acquisition periods of 16 ms (15N) and 43 ms (1H) with typical total experimental times of 24–48 h. As compared to a standard TROSY pulse sequence, the 1H–15N INEPT delays were set to 3 ms to reduce magnetization losses from relaxation.

Assignment procedure

To obtain sequence-specific assignment information, we initially attempted to detect HNCO and HNCA correlations on samples additionally labelled with 13C at specific backbone sites28. However, due to low sensitivity, only very few correlations were observable. Therefore, assignments were obtained from a combination of information from TROSY spectra recorded on 18 TS-β1AR valine point mutants with different ligands, four HN(CO) correlations and five distinct structure-based chemical shift predictions (Extended Data Table 2).

Ligand exchange experiments

Receptor complexes with different ligands were generated by sequential exchange according to increasing ligand affinity, that is, in the sequence atenolol-isoprenaline-dobutamine-alprenolol or alprenolol-carvedilol-cyanopindolol. For exchange, the sample was washed three times with buffer devoid of ligand at tenfold dilution in Amicon Ultra 50 kDa cutoff concentrators. Subsequently, the sample was washed again twice with buffer containing 100 μM new ligand, separated by a period of 1 h incubation. Final concentrations of the ligands were adjusted to 1 mM. Apo receptor was generated from the atenolol complex by six washing steps of tenfold dilution in ligand-free buffer using a 1 h incubation period for the last three steps.

NMR NB80 binding experiment

Binding of NB80 to β1AR mutants was assessed using TROSY and 1D proton NMR spectra. These spectra were recorded on the β1AR mutants (TS-β1AR: 132 μM, TS-β1AR(A227Y): 120 μM, TS-β1AR(L343Y): 110 μM, and TS-β1AR(A227Y/L343Y): 120 μM) in the presence of saturating amounts (1 mM) of the agonist isoprenaline before and immediately after addition of an equimolar (relative to the receptor) amount of NB80. For TS-β1AR(A227Y/L343Y) additional spectra were recorded after a further addition of the partial agonist cyanopindolol (1 mM) to the already present isoprenaline and NB80.

Scintillation proximity assay with 3H-dihydroalprenolol

For pharmacological binding assays membranes were prepared from SF9 insect cells as described previously26. The total protein content of the membranes was estimated by A280 measurements using an average extinction coefficient of 1.0 per mg ml−1. All assays were carried out in 96-well plates at 200 μg ml−1 total protein in membranes and 2 mg ml−1 WGA-YSi beads (Perkin-Elmer) in a 100 μl total volume per well. Samples were equilibrated at room temperature for at least 16 h. KD values for the radioactive ligand [3H]dihydroalprenolol (3H-DHA) were determined by titrating 3H-DHA from 0.032 to 100 nM. Non-specific binding was determined in presence of 1 μM S-propanolol to block the ligand binding site. Competition assays were performed in the presence of 20 nM 3H-DHA (hot ligand) and increasing concentrations of the competitor (cold ligand). Dilutions of alprenolol, atenolol, cyanopindolol, dobutamine and isoprenaline were made with phosphate buffered saline (PBS, Biochrom, Germany). Due to the limited solubility of carvedilol in water, stock dilutions of carvedilol were prepared in DMSO. The final concentration of DMSO in the samples was 5%. To test the effect of NB80 binding on the affinities (IC50) of isoprenaline for various receptor mutants, the competition assays were also carried out in the presence of saturating concentrations of NB80 (10 μM). IC50 values were determined by fitting the measured radioactive counts per minute CPM(X) at a specific concentration X of the competitor to the equation  , where CPMmax and CPMmin are maximal and minimal counts of the assay, respectively. The fits were carried in MATLAB (MathWorks, http://www.mathworks.com) with Monte-Carlo estimation of errors. Ki values were calculated from the obtained IC50 values according to the formula where A is the concentration of the radioactive ligand and KD is its affinity for the receptor determined in the direct binding experiment.

G protein activation assay

G protein activation was measured on purified β1AR mutants reconstituted with MSP1E3D129 into POPC/POPG nanodiscs. MSP1E3D1 was expressed and purified as described29 and cleaved with TEV protease. 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC, Avanti Polar Lipids) and 1-palmitoyl-2-oleoyl-sn-glycerol-3-phospho-(1′-rac-glycerol) sodium salt (POPG, Avanti Polar Lipids) were solubilized at a ratio of 1:1.5 (w/w) POPG/POPC in ND buffer (20 mM HEPES pH 8, 100 mM NaCl, 1 mM EDTA) with 50 mM sodium cholate (Sigma-Aldrich) at 4 °C. 133.3 μM MSP1E3D1 was incubated with 8 mM solubilized POPC/POPG and 10 μM purified β1AR in ND buffer with a final concentration of 24 mM sodium cholate for 1 h at 4 °C. Nanodiscs containing the receptor were separated from empty nanodiscs using a cobalt-chelating resin. The heterotrimeric G protein was prepared by incubating 10 μM recombinant Gαi1 and 10 μM native Gβγt in activation buffer (25 mM HEPES pH 7.5, 150 mM NaCl, 2 mM MgCl2, 1 mM DTT) for 30 min at 4 °C.

G protein activation was detected by the change in tryptophan fluorescence caused by the exchange of GDP for GTPγS, associated conformational changes in the Gα subunit and its dissociation from the Gβγ subunit of the heterotrimeric G protein30. All measurements were carried out on a Varian Cary Eclipse fluorescence spectrophotometer (λex = 295 nm, λem = 340 nm, 1.5 nm excitation slit, 20 nm emission slit, 2 s averaging time, 15 s cycle time) using final sample volumes of 1 ml in 10 × 4 mm cuvettes (Hellma, CH) and magnetic stirrers at 20 °C. Prior to activation, the fluorescence intensity baseline was recorded with 100 nM heterotrimeric G protein for approximately 500 s. The activation was started by adding 6 nM β1AR and 10 μM GTPγS, and the fluorescence intensity was monitored for a further 1 h. For experiments in the presence of an agonist, the concentrated receptor stock solution (1.5 μM) was pre-incubated for 30 min at 4 °C with 40 μM isoprenaline, and the buffer during the measurements contained 2 μM of isoprenaline to maintain the saturation conditions for the receptor.