Autoregulation of GPCR signalling through the third intracellular loop – Nature

Reagents

Human A₁R, human β₁AR, HA-tagged human β₂AR, murine CB₁R, human D₁R, human M₁R, human PTH₁R, and human V_1AR constructs were cloned into a pcDNA5/FRT backbone following standard cloning procedures. SPASM sensor constructs were assembled as previously described with Gly-Ser-Gly repeats between domains (Receptor-4×GSG-mCitrine-4×GSG-10 nm ER/K linker-4×GSG-mCerulean-4×GSG-G peptide)²⁸. Receptor-fluorescent protein fusion constructs were separated by a 2×GSG linker. PTH₁R luciferase complementation reporter constructs had the same basic topology as SPASM sensor constructs with lgBiT/smBiT luciferase fragments (Promega N2014) in place of the FRET acceptor/donor and with an inserted fluorescent protein to track expression level (PTH₁R-lgBiT-4×GSG-10 nm ER/K linker-4×GSG-TagRFP-3×GSG-smBiT-4×GSG-G peptide). Point mutations to various receptor constructs were made to various constructs using a modified site-directed mutagenesis procedure⁴⁴. Large ICL3 deletions were introduced into receptor constructs using the Q5 site-directed mutagenesis method (New England Biolabs E0554). PTH₁R insertions were introduced using a BsmBI-v2 based vector assembly method (New England Biolabs E1602). Nb6B9 (Nb6B9-2×GSG-SNAP tag-Flag-2×GSG-6×His) was cloned into a pBiex1 backbone following standard cloning procedures. tRNA/Synthetase plasmid (pIRE4-Azi) was a gift from Irene Coin (Addgene plasmid #105829, RRID:Addgene_105829). Carbamoylcholine chloride (14486) was purchased from Cayman Chemical. Synthetic Arginine 8 Vasopressin (CYFQNCPRG-NH₂), Spep (DTENIRRVFNDCRDIIQRMHLRQYELL), Qpep (DTENIRFVFAAVKDTILQLNLKEYNLV), Bio-Qpep (N-Biotin-DTENIRFVFAAVKDTILQLNLKEYNLV), Bio-Spep (N-Biotin-DTENIRRVFNDCRDIIQRMHLRQYELL) and PTH_1–34 (SVSEIQLMHNLGKHLNSMERVEWLRKKLQDVHNF) were purchased from Genscript. 3-Isobutyl-1-methylxanthine (IBMX), alprenolol, ascorbic acid, bovine serum albumin, copper II sulfate, forskolin, isoproterenol (+)-bitartrate salt, lysozyme from chicken egg white, metoprolol tartarate, sodium ascorbate, and X-tremeGENE HP transfection reagent were purchased from Sigma-Aldrich. [¹²⁵I](±)-cyanopindolol (NEX174) was purchased from PerkinElmer. 4-azido-l-phenylalanine (1406), AZDye488-Alkyne (1277), AZDye546-Alkyne (1285), 2-(4-((bis((1-(tert-butyl)-1H-1,2,3-triazol-4-yl)methyl)amino)methyl)-1H-1,2,3-triazol-1-yl)acetic acid (BTTAA; 1236), Cy3-Alkyne (TA117), and OG 488-Alkyne (1397) were purchased from Click Chemistry Tools. ATTO488-Alkyne (AD 488-141) was purchased from ATTO-TEC. Polyethyleneimine (PEI) 25 kDa linear polymer was purchased from Polysciences (23966). A 1 mg ml⁻¹ solution of PEI was prepared by reconstituting the polymer in endotoxin-free water at 80 °C and passing the solution through a 0.2 µM syringe filter; sterile PEI solution was stored at −20 °C. Ni-NTA agarose (QIA30210) was purchased from Qiagen. PNGase F (P0704) SNAP-Surface 488 (BG-488, S9124S), and Streptavidin-coated magnetic beads (S1420S) were purchased from New England Biolabs. DNAse I (04716728001), fibronectin (F1141) and normal goat serum (G9023) were purchased from Millipore Sigma. anti-HA-Alexa488 (A-21287), formaldehyde (cat# PI28908), and ProLong Diamond Anti-fade Mountant (P36970) were purchased from ThermoFisher. 2-Arachidonylglycerol (1298), dopamine hydrochloride (3548), and N⁶-cyclopentyladenosine (1702) were purchased from Tocris. All reagents were reconstituted and stored according to the manufacturer’s specifications.

Cell culture

HEK 293T Flp-In T-Rex Cells (ThermoFisher, R78007) were cultured in Dulbecco’s modified eagle medium supplemented with 10% (v/v) fetal bovine serum, 20 mM HEPES pH 7.4, and GlutaMAX at 37 °C with 5% CO₂. Cells were not authenticated, nor were they tested for mycoplasma contamination.

Transfection procedures

For membrane preparations, cells were passaged onto 15 cm dishes at ~50% confluence. For each 15 cm dish, 18 µg of DNA, 63 µl of PEI, and 900 µl Opti-MEM were combined, incubated for 15–30 min, and combined with resuspended cells. 4 h after transfection and passaging, the medium was replaced. Receptor transfections (β₂AR–mCerulean) were incubated for 20 h, G_s peptide sensor transfections were incubated for 23 h, and G_q peptide sensor transfections were incubated for 26 h.

For second messenger assays, cells were seeded the day prior to transfection at 30–40% confluence. For each well of a 6-well plate, 1 µg DNA, 3 µl X-tremeGENE HP, and 100 µl Opti-MEM were combined, incubated for 15–20 min, and added to seeded wells. As optimal expression time varied for each construct and with cell passage number, transfections were performed at multiple time points between 16–28 h for each second messenger assay.

For PTH₁R luciferase complementation reporter assays, cells were seeded the day prior to transfection at 25–30% confluence on 12-well plates. For each transfection, 0.5 µg DNA, 1.5 µl X-tremeGENE HP, and 100 µl Opti-MEM were combined, incubated for 15–20 min, and added to seeded wells. For each mutant, PTH₁R–G_s peptide, PTH₁R–Gq peptide and PTH₁R–(no peptide) constructs were transfected in parallel. Spep and Qpep constructs were transfected 24 h prior to collection, and no-peptide constructs were transfected 18 h before collection.

Stop codon replacement

β₂AR- nonsense mutants were created by replacing single amino acids with amber (TAG) stop codons. To minimize the influence of conformational changes in the unstructured β₂AR C-tail on FRET measurements, all β₂AR constructs used a truncated receptor C terminus (∆350–413). Stop codon replacement transfections were performed following previously described procedures²⁰. In brief, 0.5 M Azi was prepared as a fresh stock in 0.5 M NaOH and filtered through a 0.2 µM syringe filter. The fresh Azi stock was added to ~50% confluent 15 cm dishes and incubated for 1–2 h at a final concentration of 0.5 mM. Dishes were co-transfected with β₂AR nonsense mutant plasmid and pIRE4-Azi at a 1:2 ratio, following the same transfection procedure as membrane preparations. 0.5 mM Azi was added to the changed culture medium. Transfections were incubated 40–44 h prior to collection. Wild-type controls (Extended Data Figs. 2 and  3a–d) were performed in absence of Azi and pIRE4-Azi. Labelling controls (Extended Data Fig. 3e–g) were performed in presence of Azi and pIRE4-Azi.

Cell imaging

Fibronectin-coated coverslips were prepared by incubating coverslips in 0.01 mg ml⁻¹ fibronectin diluted in PBS on a parafilm surface for 1 h at room temperature. After incubation, coverslips were transferred to a cell culture dish. HEK293 cells were seeded at 30% confluency and incubated for 24 h. Cells were transfected following the procedure in ‘Stop codon replacement’, scaled down to a 6-well dish (1 µg DNA, 3.5 µl PEI, and 100 µl Opti-MEM). Following 24 h of expression, culture medium was aspirated from the culture dish, and coverslips were washed 3 times with PBS. Cells were fixed using a solution of 4% formaldehyde diluted in PBS for 10 min at room temperature. Coverslips were washed three times with PBS to remove remaining formaldehyde.

Coverslips mounted using ProLong Diamond Anti-fade Mountant and left at room temperature overnight to cure. For imaging, coverslips were sealed with vaseline/lanolin/paraffin. Images were acquired on a Nikon A1Rsi laser scanning confocal microscope using a 60× oil immersion objective (Nikon).

Crude cell membrane extracts

Procedure

Cells were collected by scraping and pipetting with media and washed twice in phosphate buffered saline (PBS; 10 ml per 15 cm dish) by centrifugation (300g, 5 min, room temperature). Cell pellets were resuspended in chilled hypotonic buffer (10 mM HEPES pH 7.4, 1 mM EGTA, 1 mM DTT, 1.5 µg ml⁻¹ aprotinin, 1.5 µg ml⁻¹ leupeptin, 5 µg ml⁻¹ PMSF; 5 ml per confluent 15 cm dish) and incubated for 30 min on ice. Solutions were lysed gently in a chilled Dounce homogenizer (40 strokes). Nuclei and intact cells were separated from the lysate by centrifugation (1,000g, 2 min, 4 °C). Lysates were centrifuged (135,000g, 25 min, 4 °C). Spun-down lysate was resuspended in assay buffer (20 mM HEPES, 150 mM NaCl, 10 mM KCl, 5 mM MgCl₂; 1 ml per 15 cm dish of cells). Pellets were then homogenized, first by passing through a 1 ml micropipette 10 times, then passing through a 26-gauge needle 10 times. Lysate was centrifuged again at 135,000g following the same procedure and resuspended in assay buffer supplemented with 12% sucrose (w/v; 0.5 ml per 15 cm dish). Pellet homogenization was repeated as before. One-hundred microlitres of a 1:20 dilution of sample in assay buffer was used for analytical fluorescence spectra (Fluoromax 4, Horiba Scientific). Spectra were used to confirm expression level (mCerulean peak emission (excitation 430 nm, 475 nm):optical density emission (excitation 430 nm, emission 450 nm) ratio of 1.0 ± 0.2) and sensor integrity (mCitrine peak emission (excitation 490 nm, emission 525 nm):mCerulean peak emission (excitation 430 nm, emission 475 nm) ratio of 2.0 ± 0.2). Resuspended lysates were aliquoted, flash frozen in liquid nitrogen, and stored at −80 °C.

Fluorescent dye labelling

Click chemistry for radioligand-binding assays

Cell membranes (Extended Data Fig. 3j–m) were prepared following the above procedure, but at a higher cell concentration (1.5 ml hypotonic buffer per 15 cm dish), and without EGTA or DTT in the hypotonic buffer to prevent decreased efficiency of the click reaction. After Dounce homogenization, the following reagents were added to the membrane mixture (final concentrations in parentheses): Cell membranes were prepared following the above procedure, but at a higher cell concentration (1.5 ml hypotonic buffer per 15 cm dish), and without EGTA or DTT in the hypotonic buffer to prevent decreased efficiency of the click reaction. After Dounce homogenization, the following reagents were added to the membrane mixture (final concentrations in parentheses): NaCl (250 mM), KCl (10 mM), MgCl₂ (5 mM), bovine serum albumin (1 mg ml⁻¹) (2 ml final). The dye and labelling reagents were pre-mixed and incubated on ice for 1 min before being added to the lysate mixture at the following final concentrations: 20 µM AZDye488-Alkyne, 250 µM BTTAA, 50 µM copper (ii) sulfate, and 2.5 mM sodium ascorbate. Reactions were incubated with mild shaking (500 rpm) for 30 min at 25 °C.

Fluorescence lifetime assays and labelling controls

Cell membranes were prepared following the procedure in ‘Crude cell membrane extracts’ through the first ultracentrifugation step (135,000g, 25 min, 4 °C). Pellets were rinsed 3 times with 1 ml assay buffer, and then resuspended in (final concentrations in parentheses): HEPES (20 mM), NaCl (250 mM), KCl (10 mM), MgCl₂ (5 mM), bovine serum albumin (1 mg ml⁻¹), dye (ATTO488-Alkyne, AZDye488-Alkyne, AZDye546-Alkyne, Cy3-Alkyne, and/or OG 488-Alkyne) (5 µM), BTTAA (5 mM), and copper (ii) sulfate (50 mM). Solution was homogenized, first by passing through a 1 ml micropipette 10 times, then passing through a 20-gauge needle 10 times. To initiate the reaction, sodium ascorbate (50 mM) was added to the mixture and mixed by pipetting (2 ml final volume). Reactions were incubated with mild shaking (500 rpm) for 30 min at 25 °C.

Azi incorporation controls

Membranes (Extended Data Fig. 3e, SNAP-Tag labelling) were prepared following the procedure in ‘Fluorescence lifetime assay and labelling controls’, with a modified labelling buffer recipe (final concentrations in parentheses): BG-488 (10 µM), DTT (1 mM). No BTTAA or copper (ii) sulfate were added to the mixture. Solution was homogenized, first by passing through a 1 ml micropipette 10 times, then passing through a 20-gauge needle 10 times (500 µl final volume). Membranes were incubated for 30 min at 37 °C.

Sample processing

For the three labelling procedures above, aggregated lysate was removed using centrifugation (1,000g, 2 min). Lysates were ultracentrifuged as described above for crude cell extracts. After the first ultracentrifugation step, lysate pellets were rinsed 10 times with 1 ml assay buffer. After rinsing, pellets were processed as described above.

To assess expression and labelling efficiency, lysates were diluted 1:10 in optical quartz cuvettes (3-3.30-SOG-3, Starna Cells), and assessed by fluorescence spectroscopy Horiba Fluoromax 4 Fluorometer using the following parameters for the indicated fluorophores: AZDye488/ATTO488/Oregon Green: excitation at 470 nm, emission scan from 500 nm–650 nm; mNeonGreen: samples in optical quartz cuvettes, excitation at 470 nm, emission scan from 495 nm–600 nm; AZDye546: samples in optical quartz cuvettes, excitation at 540 nm, emission scan from 555 nm–620 nm; Cy3: samples in optical quartz cuvettes, excitation at 535 nm, emission scan from 550 nm–620 nm and TagRFP: excitation at 540 nm, emission scan from 565–600 nm. A 1:10 dilution of lysate was also used as an analytical lifetime measurement. Lysates were aliquoted, flash frozen, and stored at −80 °C.

Radioligand-binding assays

Protein content estimation

Membranes were diluted 1:5 in assay buffer. 15 µl of sample was assayed for protein content against a BSA standard (0, 0.4, 0.8, 1.2 mg ml⁻¹) in technical triplicates using a DC assay, following the manufacturer’s protocol (BioRad 5000111). DC assay was detected using absorbance (Tecan Spark Plate reader, 750 nm, 9 nm bandwidth, 25 flashes). This protein content estimate was used to calculate membrane B_max values in fmol mg⁻¹.

Radiolabelled antagonist-binding assay

Membranes containing expressed β₂AR constructs were diluted (equivalent of 10,000 mCerulean counts (excitation 430 nm/emission 350 nm) per 200 µl reaction, or ~0.7 µg) in assay buffer supplemented with 30 µM of either Qpep or Spep (depending on condition), 1 mg ml⁻¹ Bovine Serum Albumin, 10 mM GTPγS, and 1 mM ascorbic acid and sonicated briefly. Increasing concentrations of [¹²⁵I](±)-cyanopindolol were added to the membrane samples. Non-specific [¹²⁵I](±)-cyanopindolol binding was assessed by repeating the same assay conditions listed above in the presence of 1 mM alprenolol. Reactions were equilibrated on ice for 90 min in 96-well deep well plates. Reactions were transferred to Multiscreen plates (Millipore Sigma MAFCN0B50). Equilibrated reactions were passed through GF/C filters using a vacuum manifold (Millipore Sigma MSVMHTS00) and washed 5 times with 200 µl ice-cold tris-buffered saline (50 mM Tris pH 7.4, 150 mM NaCl). Filters were air dried, removed from plates, and transferred to 75 mm glass tubes. Filter-bound ¹²⁵I was measured by automatic gamma counting (PerkinElmer Wizard²). Due to the high concentration of receptor used in the assays, bound [¹²⁵I](±)-cyanopindolol and non-specific [¹²⁵I](±)-cyanopindolol signal for each reaction condition were fit to a saturation-binding curve to obtain the equilibrium dissociation constant (K_d) values for [¹²⁵I](±)-cyanopindolol⁴⁵. Binding assays were performed in technical duplicates. Three biological replicates were collected, with different membrane preparations for each replicate.

Competition binding assay

Membrane containing expressed β₂AR sensors were resuspended following the same procedure as the radiolabelled antagonist-binding assay. Sub-saturating [¹²⁵I](±)-cyanopindolol (50 pM final) and increasing concentrations of isoproterenol were added to the resuspended membrane samples. Non-specific [¹²⁵I](±)-cyanopindolol binding was assessed by measuring 50 pM [¹²⁵I](±)-cyanopindolol in the presence of 1 mM alprenolol. Maximum [¹²⁵I](±)-cyanopindolol signal was assessed using 50 pM [¹²⁵I](±)-cyanopindolol without competing unlabelled ligand. Reactions were equilibrated, washed, and measured following the same procedure listed above for the radioligand antagonist-binding assay. Using the mean dissociation constant (K_d) values for [¹²⁵I](±)-cyanopindolol obtained in the radioligand antagonist-binding assay, bound [¹²⁵I](±)-cyanopindolol and non-specific [¹²⁵I](±)-cyanopindolol signal for each reaction condition were fit to a displacement curve to obtain the equilibrium dissociation constant (IC₅₀) values for isoproterenol:

Where Y is the percentage of [¹²⁵I](±)-cyanopindolol bound to the receptor, B_max is the maximum percentage [¹²⁵I](±)-cyanopindolol bound to the receptor, B_min is the minimum [¹²⁵I](±)-cyanopindolol bound to the receptor, and A is the concentration of isoproterenol. For two-site binding, (WT + Qpep condition), the following model was used:

Where F₁ is the fraction of sites for affinity site 1, IC_50,1 is the IC₅₀ of affinity site 1, and IC_50,2 is the IC₅₀ of site 2. Weighted averages of these IC₅₀ values were used for comparison purposes and inhibition constant (K_i) calculations. IC₅₀ values were converted to K_i values using the Cheng–Prusoff correction⁴⁶ as follows:

Where L is the concentration of [¹²⁵I](±)-cyanopindolol, and K_D is the equilibrium dissociation constant for [¹²⁵I](±)-cyanopindolol determined in the antagonist-binding assays. Binding assays were performed in technical duplicates. Three biological replicates were collected with different membrane preparations for each replicate.

Nanobody expression and purification

Five-hundred millilitres of terrific broth with 100 µg ml⁻¹ carbenicillin was inoculated with Escherichia coli (JM109 strain) transformed with the Nb6B9 vector at A₆₀₀ = 0.05. The culture was grown to A₆₀₀ = 0.8 at 37 °C with shaking (180 rpm). The culture was then induced with 0.4 mM IPTG and incubated for 16 h at 16 °C with shaking (180 rpm). The culture was pelleted at 3,000g for 15 min. The pellet was then resuspended in 30 ml lysis buffer (20 mM HEPES pH 7.4, 150 mM NaCl, 10 mM imidazole, 10 mM MgCl₂, 5 mM CaCl₂, 1 mg ml⁻¹ lysozyme, 1 µg ml⁻¹ DNAse I, 1 mM DTT, 1 µg ml⁻¹ aprotinin/leupeptin, 0.1 µg ml⁻¹ PMSF, 1% v/v Triton X-100) and incubated for 30 min at 4 °C on an orbital shaker (100 rpm). The lysate was then sonicated for 10 min (10 s on, 10 s off) and clarified by centrifugation (18,000g, 20 min, 4 °C).

A column containing 3 ml Ni-NTA agarose was equilibrated with 15 ml wash buffer (20 mM HEPES pH 7.4, 150 mM NaCl, 10 mM imidazole). Clarified lysate was flowed over the column. Column was washed with 15 ml wash buffer, 15 ml high-salt wash buffer (20 mM HEPES pH 7.4, 500 mM NaCl, 10 mM imidazole), and 15 ml wash buffer. Four total 3 ml elution fractions were collected in wash buffer containing 60 mM imidazole, 120 mM imidazole, 180 mM imidazole, and 240 mM imidazole. All fractions except the 60 mM Imidazole fraction were pooled, concentrated using a 10,000 kDA molecular weight cut-off centrifugal filter (Amicon Ultra-15), and further purified over a Superdex 200 Increase 10/300 GL gel filtration column (GE Healthcare) in size exclusion buffer (20 mM HEPES pH 7.4, 400 mM NaCl). Purity of size exclusion eluate was confirmed via SDS–PAGE. Eluate was concentrated and rebuffered into assay buffer containing 10% w/v Glycerol. Concentration was determined by 280 nm absorbance (Thermo Scientific Nanodrop One). Aliquots were flash frozen and stored at −80 °C.

Fluorescence measurement assays

Fluorescence gel scanning

Ten micrograms (Extended Data Fig. 3e) or 20 µg (Extended Data Fig. 4n) (concentration determined using the procedure in ‘Radioligand-binding assays’, ‘Protein content estimation’) of membranes prepared as described in ‘Crude membrane extracts’ were denatured and deglycosylated with PNGase F. Samples were prepared following the manufacturer’s recommended protocol for denaturing conditions, with the denaturation step scaled to 25 µl and the deglycosylation step scaled to 35 µl. 50 mM DTT and 1× LDS sample buffer were added to the reaction mixture at the end of the procedure. Samples were then separated using 7.5% (Extended Data Fig. 3e) or 10% (Extended Data Fig. 4n) polyacrylamide gels. Gels were scanned for fluorescence (GE Healthcare Typhoon FLA 9500) of AZDye488/ATTO488/Snap Surface Block 488 (excitation 473 nm, long-pass emission filter 510 nm, gain 1,000 V) or AZDye546 (excitation 532 nm, long-pass emission filter 570 nm, gain 1,000V).

Time-correlated single photon counting FRET assay

Dye-only controls or membranes containing ICL3 FRET sensors were resuspended in assay buffer containing 1 mM ascorbic acid to an equivalent of ~1.5% of excitation counts/second (0.12 MHz for emission pulses for an 8 MHz excitation pulse) (a final concentration of ~3–4 nM green fluorophore). Ligand (100 µM isoproterenol), G peptide (10 µM), and/or Nb6B9 (0.5 µM) were added for a final reaction volume of 110 µl. For conditions without drug, peptide or nanobody, an equivalent volume of assay buffer was added. Reactions were equilibrated for 5 min prior to reading. One-hundred and five microlitres of each reaction was loaded into an optical quartz cuvette. Measurements were taken by time-correlated single photon counting (DeltaPro, Horiba Scientific) using a 479 nm pulse diode laser and a 515 nm long-pass emission filter. Time-resolved fluorescence decay data were fit to the equation:

The three-exponential fit was optimized empirically (χ² ≈ 1.25, where two-exponential fit χ² ≈ 1.4) (Extended Data Fig. 4a–h). Amplitude-weighted average lifetimes (τ_avg) were calculated from the three-exponential decay equation:

Each condition was performed in technical duplicate or triplicate, depending on sensor yield.

∆FRET assay

Membranes containing β₂AR-SPASM sensors were resuspended in assay buffer based on mCerulean fluorescence (1 × 10⁶ mCerulean counts at 475 nm) and sonicated briefly. For conditions containing Qpep, a final concentration of 10 µM Qpep was added to the membrane mixture following sonication. A final concentration of 100 µM isoproterenol or an equivalent amount of assay buffer was added to each reaction (100 µl final). Reactions were equilibrated for 5 min at 25 °C with shaking (500 rpm). 110 µl of each reaction was loaded into an optical quartz cuvette. Fluorescence spectra (Horiba Fluoromax 4) for mCerulean were acquired for each sample (excitation 430 nm, emission scan 450 nm–600 nm, bandpass 4 nm). The mCitrine (emission 525 nm): mCerulean (emission 475 nm) ratio (FRET ratio) was calculated from each acquired spectra. Each drug–peptide condition was performed in quintuplicate. For each experiment, the ∆FRET metric was calculated by subtracting the average FRET ratio of the buffer only conditions from the average FRET ratio of the isoproterenol-treated conditions.

BioSp, BioQp pulldown assay

Membranes containing β₂AR–mCerulean or β₂AR-TagRFP (1:10 dilution of frozen membrane aliquots) were resuspended in assay buffer containing 10 mg ml⁻¹ Bovine Serum Albumin, 1 mM ascorbic acid and 100 µM isoproterenol and sonicated briefly. Bio-Spep/Bio-Qpep (10 µM final) was added to the reaction mixture (300 µl final) and equilibrated on ice for 30 min. 100 µl of the reaction mixture was removed; an analytical fluorescence spectra of this sample was acquired (mCerulean: Horiba Fluoromax 4, excitation 430 nm, emission scan 450 nm–600 nm, bandpass 4 nm; TagRFP: Tecan Spark Plate Reader, 96-well clear bottom plate, bottom read, excitation at 521 nm, emission scan from 560 nm–650 nm, gain 232) as a measure of the total β₂AR for a given condition. To the remaining 200 µl, 20 µl of 0.4 mg ml⁻¹ streptavidin-coated magnetic beads were added to the reaction mixture, equilibrated for 5 min at ambient temperature, and precipitated using a neodymium disc magnet N52 (20 × 40 mm). Fluorescence spectra was taken of the remaining supernatant in duplicate. Percent mCerulean bound was calculated as the average peak fluorescence count (mCerulean:emission 475 nm, TagRFP: emission 584 nm) of the remaining supernatant samples subtracted from the peak fluorescence counts of the total receptor sample, divided by the peak fluorescence count of the total receptor sample.

Second messenger assays

General procedure

One millilitre of medium was removed from each well containing transfected cells. The remaining volume was used to gently shear and resuspend the cells by pipetting. The cell mixture was centrifuged (3 min, 300g) and media was removed using a vacuum manifold. The cell pellet was resuspended in 1 ml of either cAMP assay buffer (PBS with 0.5 mM ascorbic acid, 0.2% (w/v) glucose) or InsP₁ assay buffer (10 mM HEPES pH 7.4, 1 mM CaCl₂, 4.2 mM KCl, 145 mM NaCl, 5.5 mM glucose, 50 mM LiCl). The cells were washed once by repeating this procedure.

For Extended Data Fig. 2b, cells were diluted to 2 × 10⁶ ± 5 × 10⁵ cells per ml. Expression level was estimated through TagRFP fluorescence (Horiba Fluoromax 4): excitation 540 nm, emission scan 565–600 nm, bandpass 4 nm.

For all other second messenger assays, expression level and cell density of each condition was estimated through fluorescence. Fluorescence spectra for mCerulean (Horiba Fluoromax 4, excitation 430 nm, emission scan 450–600 nm, bandpass 4 nm) and mCitrine (Horiba Fluoromax 4, excitation 490 nm, emission scan 500–600 nm, bandpass 4 nm) were acquired for each condition. Cells were resuspended at 350,000 ± 30,000 fluorescence counts at a wavelength representing optical density of the sample (excitation 430 nm/emission 450 nm), corresponding with 2 × 10⁶ ± 5 × 10⁵ cells per m^l. This was confirmed by counting the cells on a haemocytometer (Countess II). The following metrics were assessed for optimal expression: mCerulean peak emission (excitation 430 nm/475 nm):optical density emission (excitation 430 nm/emission 450 nm) of 2.0 ± 0.3 and mCitrine peak emission (excitation 490 nm/emission 525 nm):mCerulean peak emission (excitation 430 nm/emission 475 nm) of 2.0 ± 0.2.

cAMP accumulation

Resuspended cells were added 1:1 with a 2× concentration of ligand (10 µl final) into an opaque 384-well flat bottom plate (Greiner Bio-One). For single-concentration experiments, a saturating amount of agonist (10 µM PTH_1–34 (PTH₁R), 10 µM 2-arachidonoylglycerol (CB₁R), 10 µM carbachol (M₁R), 10 µM dopamine (D₁R), 10 µM N⁶-adenosine (A_1AR), 100 nM arginine-vasopressin (V₁R)) was used. For cAMP accumulation experiments with β-ARs, 10 µM isoproterenol was used; for FSK inhibition assays with β-ARs, 100 µM metoprolol was used. For dose–response curves, a saturating concentration of forskolin (10 µM) was included as a control to measure cAMP stimulation independent of the transfected receptor. For experiments comparing multiple receptors measuring non-cognate or secondary G_s signalling (Fig. 4e, Extended Data Fig. 8i–k), 500 µM 3-isobutyl-1-methylxanthine (IBMX) was used to inhibit phosphodiesterase activity. For all other experiments, no IBMX was used to minimize cAMP accumulation from endogenous receptors (Extended Data Fig. 6a). For experiments in Fig. 2g and Extended Data Fig. 6b,e, plates were incubated at 37 °C for 10 min to stimulate cAMP production. For all other experiments, plates were incubated at room temperature for 10 min. We found that the room temperature incubation maintained cAMP accumulation levels to equivalent levels as 37 °C incubation and decreased well-to-well variability in the experiments. Reactions were quenched and processed for the cAMP-Glo Assay (Promega) following the manufacturer’s instructions. Luminescence was measured on a Tecan Spark plate reader (500 ms integration, one measurement per well). Data were either normalized to β₂AR-WT (Figs. 2g and 4e and Extended Data Figs. 6 and  8j), β₂AR-WT-Nopep (Extended Data Fig. 7d), or to maximum forskolin stimulation (Figs. 3e,f and 5c and Extended Data Figs. 6 and  9k). Dose–response curves were fit to the equation:

Where E is the response, L is the concentration of isoproterenol, E_max is the response at saturating concentrations of isoproterenol, E_min is the response in the absence of isoproterenol, EC₅₀ is the isoproterenol concentration that gives 50% of the E_max, and N_H is the Hill coefficient of the curve.

Forskolin inhibition

Experiments were set up as described above. A 1 µM forskolin treatment for each receptor was compared to 1 µM forskolin with saturating concentrations of agonist. All conditions were supplemented with 500 µM IBMX, and a 10 min room temperature stimulation condition was used. Reactions were quenched and processed for the cAMP-Glo Assay (Promega) following the manufacturer’s instructions.

InsP₁

Resuspended cells were added 1:1 with a 2× concentration of ligand (70 µl final) into an opaque 96-well U-bottom plate (Greiner Bio-One). For single-concentration experiments, a saturating amount of ligand (10 µM isoproterenol, 10 µM PTH_1–34) was used. Plates were incubated at 37 °C for 2 h to stimulate InsP₁ production. Reactions were quenched and processed for the InsP₁ HTRF Assay (CisBio), following a protocol modified to achieve a higher signal to noise ratio. Fifteen microlitres of D2-conjugated InsP₁ resuspended in lysis buffer (Cisbio) and 15 µl of terbium cryptate conjugated anti-InsP₁ antibody resuspended in lysis buffer (Cisbio) were added to the stimulated cell mixture. Cell lysate was equilibrated for 1 h at ambient temperature with shaking (500 rpm). Reactions were transferred (4 × 20 µl) to a 384-well plate for technical replicates. Fluorescence readings (Flexstation3, Molecular Devices) of acceptor D2 (excitation 340 nm, emission 665 nm, cut-off 630 nm) and donor terbium cryptate (excitation 343 nm, emission 620 nm, cut-off 570 nm) were acquired with a delay of 50 µs and an integration time of 300 µs. FRET ratio for each reading was calculated as the ratio of acceptor emission to donor emission. InsP₁ signal for a drug and transfection combination was calculated as the average FRET ratio of a given transfection condition without drug treatment subtracted from the average FRET ratio of a of a given transfection condition with drug treatment. Dose–response curves were fit to the same equation as cAMP dose–response curves.

Outlier handling

Biological samples with poorly matched cell density or receptor expression levels (desired parameters for collection are indicated in ‘General procedure’) were flagged as potential failed sample replicates. Flagged samples that were outliers (absolute Z-score greater than 3) in comparison to other biological replicates were omitted.

PTH₁R-luciferase complementation reporter assay

Expressed PTH₁R-luciferase complementation constructs were vesiculated as previously described, with modifications for smaller sample volume²⁵. Media in each well was used to gently shear and resuspend the cells by pipetting. The cell mixture was centrifuged (3 min, 300g) and the media was aspirated. Cell pellet was resuspended in 1 ml PBS and centrifuged again as above. Cells were resuspended again in 0.6 ml vesiculation buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 20 mM CaCl₂, 2 µg ml⁻¹ aprotinin, 2 µg ml⁻¹ leupeptin, 2 mM N-ethylmaleimide) and incubated at 30 °C with shaking at 180 rpm for 2 h. Cellular debris was removed by centrifugation (1,000g for 1 min). To collect additional vesicles, debris was resuspended in 0.3 ml PBS, briefly vortexed, and centrifuged again (1,000g for 1 min). The ~0.9 ml of combined decanted supernatant was centrifuged one additional time (1,000g for 2 min) to better remove cellular debris. Vesicles were collected by centrifugation (3,200g for 40 min at 4 °C) and washed in 0.5 ml assay buffer. Centrifugation step was repeated, and vesicles were resuspended in 0.1 ml assay buffer.

Vesicle samples were collected in a 96-well clear bottom plate and assayed for TagRFP fluorescence (Tecan Spark, excitation 521 nm, emission 585 nm, gain 150). For each set of constructs for a given ICL3 insertion (Spep, Qpep, and control), samples were diluted to the lowest TagRFP counts.

Four ICL3 insertion constructs were assayed at a given time (12 total). 45 µl of each sample was transferred to an opaque 96-well flat bottom plate. 45 µl of Nano-Glo Substrate (Promega N1110) diluted 1:50 in assay buffer was added to each well in the new plate. After tapping the plate to collect liquid at the bottom of the well, a kinetic luminescence read was started (500 ms integration, continuous for 40 min) and luminescence signal was tracked. When luminescence signal equilibrated (plateau between 300–350 s), the kinetic read was paused and 10 µM of PTH_1–34 was added to each well. The plate was tapped 2–3 times to mix, and the kinetic read was resumed. Moving averages were computed for each kinetic trace (3-point averages for 5 min equilibration, 8-point moving averages post-drug treatment). Each kinetic trace was normalized to the last point of the pre-drug equilibration. The maximum luminescence value that appeared stable over time was used for further analysis. Specific Spep and Qpep signals were calculated by subtracting the control value from the Spep value and the Qpep value for a given experiment.

Molecular dynamics simulations

For maximum sampling of the conformational heterogeneity exhibited by the third intracellular loop of β₂AR, we used multiple runs of all-atom molecular dynamics simulations using the agonist isoproterenol, as detailed below.

Initial state structural models of β₂AR

β₂AR with a truncated N terminus (∆1–34) and truncated C terminus (∆341–413) was used for all simulations. We built the ICL3 sequence (228-RQLQKIDKSEGRFHVQNLSQVEQDGRTGHGLRRSSK-263) as an unstructured loop into distinct models of the receptor derived from known structures of β₁AR and β₂AR, with the following rationale (Supplementary Table 2, Supplementary Fig. 4):

1. (1)

   Model A. Published inactive state crystal structures of β₂AR lack atomic coordinates in the ICL3 region. However, the inactive state crystal structure of thermostabilized wild turkey β₁AR (PDB ID: 2YCX)⁴⁷ has a truncated, but structurally resolved ICL3 that folds into the receptor’s intracellular cavity. Alignment of this structure with the with the agonist- and G protein bound structure of β₂AR (PDB ID: 3SN6)¹⁷ showed that a C-terminal portion of β₁AR’s ICL3 aligns with the Cα5 helix of the G protein bound β₂AR (Supplementary Fig. 5). We posited that this is a possible ‘autoregulated’ state of GPCR activity, wherein the ICL3 competitively inhibits G protein binding. Thus, we used the β₁AR inactive state structure (2YCX) as a template to derive a homology model of β₂AR in the autoregulated inactive state using SWISS-MODEL software⁴⁸.

2. (2)

   Model B. Complementing the agonist- and G protein bound structure of β₂AR is a structure of β₂AR in complex with a 14-amino acid peptide derived from the Cα5 helix of the G protein (PDB ID: 6E67)²⁴. The orientation of the 14-amino-acid peptide is distinct from that of the Cα5 helix of the G protein in the 3SN6 structure, and as posited by Kobilka and colleagues, represents an intermediate state in the G protein activation mechanism.

3. (3)

   Model C. In our previous work²⁵, we performed molecular dynamics simulations of β₂AR using the atomic coordinates of PDB 6E67 as an initial state. In these simulations, we observed that the C-terminal cap of transmembrane helix 5 transitioning into the ICL3 region unravels. This is distinct from the helical conformation of this region observed in models A and B. Since this could be a part of the transition from the intermediate to the active state, we used this structural model as a starting point.

4. (4)

   Model D. In our previous work²⁵, we also observed that upon removing both a fused T4 lysozyme and an engineered disulfide bond between the receptor and the 14-amino-acid G peptide from the 6E67 structure, the peptide unravelled in our simulations, leaving just one turn of the Gs peptide capping the receptor’s G protein-binding site. As this could represent the movement of ICL3 out of the autoregulated state, we built the C-terminal portion of ICL3 mimicking this structure, with the rest of the loop modelled in an extended conformation.

Cell membrane mimicking multi-lipid bilayer

To study the effect of multiple lipids on the GPCR conformation ensemble, we used a mixed lipid bilayer to mimic the cell membrane (Supplementary Fig. 3). The outer leaflet of the membrane bilayer consists of POPC, DOPC, POPE, DOPE, sphingomyelin (Sph), ganglioside (GM3) and cholesterol in the ratio of 20:20:5:5:15:10:25, while the inner leaflet contains POPC, DOPC, POPE, DOPE, POPS, DOPS, phosphatidylinositol 4,5-bisphosphate, and cholesterol in the ratio of 5:5:20:20:8:7:10:2²². To further mimick the cell membrane conditions, we neutralized the composite lipid bilayer with 150 mM of CaCl₂. To obtain a random distribution of lipids in coarse grain simulations, the lipid bilayer was built three times in the same composition given above. After equilibration of each of the three simulation boxes, we performed 10 μs of coarse grain molecular dynamics (CGMD) simulations with Martini2.2 forcefield⁴⁹. The coarse grain lipid bilayer models were converted to all-atom models using the script backward.py from the Martini website⁵⁰. We extracted five different cell membrane lipid configurations, described in detail in our previous work²². We then inserted our four initial state models of β₂AR into these lipid configurations. After elimination of steric clash between the receptor and lipids, we found with one GPCR–lipid bilayer complex for each model A to C and five GPCR–lipid bilayer complexes for model D.

All-atom molecular dynamics simulation protocol

Each GPCR–lipid bilayer complex was solvated with water and neutralized with 150 mM of CaCl₂. The disulfide bonds were built according to the disulfide bonds listed in the 6E67 structure’s template PDB file. The minimization–heating–equilibration–production was carried out as previously described^51,52,53,54. Each GPCR–lipid bilayer complex was minimized and equilibrated using a 50 ns NPT equilibration simulation protocol (constant number of particles, pressure and temperature). Equilibration was performed starting with position restraints placed on the receptor, heavy atoms (C, N, O, S and P) in the ligand, and in the head group of the lipids. The force constant on the position restraints were reduced from 5 to 0 kcal mol⁻¹ by a 1 kcal mol⁻¹ interval per 10 ns simulation window. The last 10 ns of equilibration simulations were performed with no constraints. Starting from the last frame of the equilibration protocol, we performed 400 ns all-atom molecular dynamics simulation runs with NPT ensemble at 310 K with 2 fs time step using GROMACS with CHARMM36mFF⁵⁵. We stored molecular dynamics snapshots during the molecular dynamics simulations at 20 ps intervals. The non-bond interactions in each simulation were calculated with a cut-off of 12 Å. The particle mesh Ewald method was applied to calculate van der Waals interactions⁵⁶. The temperature was maintained at 310 K using Nose-Hoover thermostat⁵⁷ and pressure at 1 atmosphere using Parrinello–Rahman method⁵⁸.

In one of the simulation runs starting from model D, we observed that the cap of the helix that was blocking the G protein site (Supplementary Fig. 4) left the G protein-binding pocket and transitioned to the fully open state of ICL3. However, when we generated a free energy surface of our combined simulations (see ‘Free energy landscape’), there was no connection between these two states. To enrich the sampling of this rare event, we generated a swarm of simulation trajectories. We extracted three snapshots from the original simulation with transition event at 50 ns, 100 ns and 150 ns (Supplementary Table 2). We then performed a production run for 2 μs from each of these snapshots. Thus, we generated a total of 22 μs of molecular dynamics simulations to analyse the heterogenous conformation ensemble of ICL3.

Free energy landscape

In order to describe the global motion of ICL3 in our simulations, we mapped our simulation trajectories onto a free energy landscape using the Markov state model in the software MSMBuilder2 (Version 3.8.0)⁵⁹. The backbone dihedral angles (phi and psi) of the whole GPCR were chosen as order parameters to describe the motion of ICL3. The phi and psi angle matrix was projected into two-dimensional space using time-correlated independent component analysis (tICA) with a lag time of 2 ns, and free energy landscape constructed based on the inverse of the population density. Four major free energy basins were observed on the free energy surface, which were mostly distinguished by the conformation of ICL3. The MinibatchKMean clustering method was applied on all sampling points to distinguish them into distinct clusters⁶⁰. 5 total clusters were generated: one for each free energy basin, and two subclusters for one of the intermediate free energy basins (Fig. 2a,b).

Centre of mass distance measurement

To quantify putative structural constraints between ICL3 and other intracellular regions of the receptor, we calculated the distance between the centre of mass of ICL3 (S236–G257) and the centre of mass of either ICL2 (T136–T146) or ICL1 (F61–T66). Distance calculations were performed for each of the five conformation clusters extracted from the free energy landscape. Based on the tight distance distribution observed in cluster 1, we performed additional distance calculations comparing the centre of mass of five individual ICL3 sequence segments (241-HVQ/NLS/QVE/QDG/RT-254) and the centre of mass of either ICL2 or ICL1 for cluster 1.

Bioinformatics analyses

Meta-analysis of ICL3 mutation data

As ICL3 is highly variable in sequence length, we opted to use an N- and C- terminal numbering scheme to keep track of the locations of mutagenized sites, where the N-terminal half of an ICL3 sequence is N1-Nn, and the C-terminal half is Cn-C1, where n is one-half the length of a receptor’s ICL3 sequence. We used TM5.56 as a starting point for the N-terminal sequence numbering to demarcate cytoplasmic exposure of TM5. The same logic applied to TM6.37 for the C-terminal sequence numbering (Extended Data Fig. 1 and Supplementary Tables 1–4).

We included all mutational data (pK_d, pEC₅₀, and E_max) that we could find with a wild-type reference point. For pK_d plots, we only included agonist binding data. To plot the effect of location of mutation versus functional effect, we normalized each ICL3 length to the shortest ICL3 in the dataset (22 amino acids). Each position mutated was assigned the effect of the mutation.

ICL3 length versus G protein interface conservation

G protein interface conservation (Fig. 5a and Extended Data Fig. 9a–j) was calculated as the sequence similarity of all amino acids composing the GPCR’s G protein-binding interface. The residues composing this interface were inferred from previous structural alignment and interface mapping^33,34. Sequence similarity was calculated from four separate GPCR interface alignments, in which receptors were separated based on their primary G protein signalling transduction pathway in the IUPHAR/BPS Guide to Pharmacology database^34,61.

Interface composition was compared to ICL3 sequence length. The starting position and ending position of ICL3 for each GPCR was determined based on the generic residue numbers of the first and last cytosol-exposed residue in the ICL3 region of B2AR, as determined from crystal structures (PDB ID: 3SN6).

We repeated the analysis for two other datasets assessing GPCR G protein subtype specificity using parallelized high-throughput screening techniques^37,38. These datasets allow for quantitative comparison different G proteins coupling to a given receptor. To assess if there were high-level differences in G protein selectivity for the short-ICL3 and long-ICL3 groups in these datasets, we compared the highest log(E_max/EC₅₀) value (considered cognate) with the second-highest value (considered secondary) for each receptor. For this analysis, we did not include the receptors that only had a log(E_max/EC₅₀) for a single receptor (Extended Data Fig. 9g,j).

Statistics

Statistical analyses were performed in RStudio (version 2022.12.0). For experiments comparing two conditions, an unpaired two-sided t-test was used. For experiments comparing more than two conditions, analysis of variance was used. One-way ANOVA was used for single level comparisons (for example, effects of mutations), and two-way ANOVA was used for two-level comparisons (for example, effects of mutations and effects of G peptides). To compare between conditions, Tukey’s post hoc test was used. For two-way ANOVA comparisons where the interaction effect was not significant (P > 0.05), we did not make individual post hoc comparisons between levels (for example, we would still compare mutation A to mutation B, peptide A to peptide B, but not mutation A compared to peptide B). Statistics were not used to pre-determine sample size for any experiments. Conditions for biological samples (membranes, cells, vesicles) were plated and/or assayed in random order between experimental replicates for all datasets. Investigators were not blinded to group allocation during data collection or analysis, as all data presented are quantitative and no subjective metrics were assessed.

Software

Fluorescence lifetime data were fit in DAS6 (Horiba). Curve fits were performed in Excel using the Solver add-in. Figures were generated in RStudio (version 2022.12.0) using the ggplot2 package⁶². Image processing was performed in Fiji^63,64. Molecular structure representations were created using VMD (version 1.9.3)⁶⁵ and Pymol (version 2.0.6)⁶⁶.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.